Citation
Is there an epidemiological paradox for birth outcomes among Colorado women of Mexican origin?

Material Information

Title:
Is there an epidemiological paradox for birth outcomes among Colorado women of Mexican origin? sí y no : it depends on the outcome
Creator:
Devine, Sharon Jean
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
xvii, 186 leaves : ; 28 cm.

Subjects

Subjects / Keywords:
Hispanic American women -- Colorado ( lcsh )
Birth weight, Low -- Cross-cultural studies -- Colorado ( lcsh )
Premature infants -- Cross-cultural studies -- Colorado ( lcsh )
Mothers -- Nutrition -- Colorado ( lcsh )
Newborn infants -- Health and hygiene -- Colorado ( lcsh )
Birth weight, Low ( fast )
Hispanic American women ( fast )
Mothers -- Nutrition ( fast )
Newborn infants -- Health and hygiene ( fast )
Premature infants ( fast )
Colorado ( fast )
Genre:
Cross-cultural studies. ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )
Cross-cultural studies ( fast )

Notes

Thesis:
Thesis (Ph. D.)--University of Colorado Denver, 2009. Health and behavioral sciences
Bibliography:
Includes bibliographical references (leaves 171-186).
General Note:
Department of Health and Behavioral Sciences
Statement of Responsibility:
by Sharon Jean Devine.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
436076614 ( OCLC )
ocn436076614

Downloads

This item has the following downloads:


Full Text

PAGE 1

IS THERE AN EPIDEMIOLOGICAL PARADOX FOR BIRTH OUTCOMES AMONG COLORADO WOMEN OF MEXICAN ORIGIN? Sl Y NO: IT DEPENDS ON THE OUTCOME by Sharon Jean Devine B.S. Linguistics, Georgetown University, 1970 J.D Boston University School of Law, 1975 M.A Anthropology, University of Colorado Denver, 2005 A thesis submitted to the University of Colorado Denver in partial fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Science 2009

PAGE 2

2009 by Sharon Jean Devine All rights reserved.

PAGE 3

This thesis for the Doctor of Philosophy degree by Sharon Jean Devine has been approved Date

PAGE 4

Devine, Sharon Jean (Ph.D Health and Behavioral Science) Is There An Epidemiological Paradox For Birth Outcomes Among Colorado Women Of Mexican Origin? Si Y No: It Depends On The Outcome Thesis directed by Professor Susan Niermeyer ABSTRACT This study examines whether an epidemiological paradox exists for low birth weight, preterm birth, small for gestational age, and large for gestational age among Hispanic mothers in Colorado. It compares birth outcomes by race/ethnicity and place of birth (Mexico or U.S.) to identify individualand neighborhood-level contributors and to contextualize quantitative findings The study analyses two retrospective cohorts all mothers (N=356,389) and mothers of Mexican origin (N=85,755) delivering singletons in Colorado during 2000 2005, using multiple logistic regression to test the social gradient of health by race/ethnicity and by nativity, to identify any paradoxical outcomes and to explore the healthy migrant and healthy immigrant explanations for better outcomes among Mexican-born mothers. General liRear regression analyzes the association of neighborhood deprivation and immigrant orientation for mothers of Mexican origin in Adams (N=16, 1 07) and Denver (N=23,332) Counties Five interviews with key informants and ten interviews with mothers of Mexican origin, half of whom were born in Mexico, are analyzed using directed content analysis Four key findings emerge. First, an epidemiological paradox exists for Hispanics for all four birth outcomes, despite having worse social and medical profiles than non-Hispanic White mothers. Second, the paradox exists for Mexican-born mothers for low birth weight preterm birth and small for gestational age. No paradox exists for large for gestational age. Third, neither the healthy migrant nor healthy immigrant explanation is supported. Finally neighborhood measures of immigrant orientation and neighborhood deprivation do not influence the likelihood of outcomes in Adams and Denver Counties. The public health importance centers on the identification of a hidden epidemic of large for gestational age among Mexican-born mothers and insight into the structure of health

PAGE 5

disparities. Any paradox at the low-weight end of the spectrum of birth outcomes no longer obscures the existence of negative high-weight outcomes, an important finding in Colorado where Hispanics represent 30% of singleton births The broader political economic perspective suggests that reliance on individual-level interventions alone is insufficient to reduce LGA disparities because Mexican-born immigrants are constrained by structural barriers to better health outcomes including poverty, lack of access to healthy foods, and social and linguist i c isolation. This abstract accurately represents the content of the candidate's thesis. I recommend its publication

PAGE 6

DEDICATION I dedicate this dissertation with gratitude to my parents Ethel and George, both scholars and especially to Mom, who showed me how to be a life-long learner; to my children Devin and Katharine, may you be life-long learners too; to my sister, Josie, with thanks for all the flowers that brightened my journey

PAGE 7

ACKNOWLEDGMENTS This research was supported in part through a grant from the Craig R. Janes Fund for Graduate Research at the University of Colorado Denver I wish to acknowledge Dr. Susan Niermeyer who took a chance on mentoring me; Dr. John Brett, who convinced Susan to take that chance; Dr. Jean Scandlyn whose every conversation opened up a new thought; Dr. Richard Miech, who encouraged me to fly the plane; Dr Allison Sabei-Soteres for her statistics advice and Dr. Lorna Moore, who encouraged me to pursue a doctorate before it was even on my radar screen. To my cohort go many thanks for their support and friendship, especially Maria de Jesus Diaz-Perez who made all things possible at Salud Family Health Centers iMil gracias to Salud and the women who told me their stories! And finally, heartfelt thanks to Dr. Susan Dreisbach who employed me during this journey and to Brenda Beaty who helped fill in the holes in my knowledge of SAS.

PAGE 8

TABLE OF CONTENTS FIGURES ..... ................................................... . . . .. ................. ............. x i ii TABLES . ... ...... .................... .. . .... .. ............................... .. ... ........... ..... xv CHAPTER 1. INTRODUCTION .................................... . .. .. .... . .... ................................ 1 Epidemiological Paradoxes ......................... .................... .. ... .. .. .... ..... 1 Birth Outcomes .... ........................... .......... .. . .. ..................... ... .. .. . 2 Low Birth Wei ght. ... .......... .. ..... . .. ..... .. .. ..................... ... ..... 3 Preterm Birth .. .. . ... .. . .... ... ... .. ..... ..... .. .. .. ... . ... .. ........ .... . 5 Small for Gestational Age .... .... ... .. .......... .. .. ... . .... ... ... .. .... ... 6 Large for Gestational Age . ... .. . .. ............ ...................... .. .... 6 Fetal Programming ... .. .. .. .......... .......... ........ ..... ........... ... . .. ? Population Characteristics of Colorado .................... . . .. ... .. ............ ..... 8 Why Study Paradox i cal Health Outcomes? .. ... . . .... . .......... .... . . . ........... . . 9 Research Design and Specific Aims .. .... .. . ..... . ............ .. .. . . ... ..... . 11 Overview of Research Methods ... .. .. ............................ ... ................ .. 11 Description of the Study Population ................. ... . ......... ... .. . .......... .... 13 Organization of the Dissertation .. .... .. .... . ... .......... .. .... ... ... ... .. ..... ... 13 2. EPIDEMIOLOGICAL PARADOXES .......... ..... ..................................... .... . 14 The Social Gradient of Health . .. .... . . ... .......... .. . .. ................. .. ..... 14 The Social Gradient of Health at the Individual Level. ... ....... .. ..... 15 The Social Gradient of Health at the Area Level. ...... . . .. ... .... .... 15 Unexpected Deviances from the Social Gradient of HealthThe Paradox .... 17 viii

PAGE 9

Traditional Explanations of the Hispanic Immigrant Paradox .. .................. 19 Healthy Migrant Hypothesis ... ..................... .. .......................... 20 Healthy Immigrant Hypothesis ... .. ....... .... .. .................. ......... 21 Acculturation/Assimilation .................................... .......... ....... 22 Social Support . .... .............. ....... .......................................... 24 An Alternative TheoryPolitical Economy ................ ........... .. .. ........... 25 3. RESEARCH DESIGN AND METHODS .... ................... .................. ........... 29 Nested Mixed Method Design .................................... ... ... .. ................ 30 Quantitative Research Component. ............................ . ...... .. .. . ; ..... .... 31 Levels of Analysis ... ......... ............................. .... ................... 32 Sample Study Data and Variables ................................ ......... 33 Dependent Variables .......... ... .. ..... ......... ............................... 34 Independent Variables for Aims 1 and 2 ................................... 35 Demographic and Socioeconomic Risk Factors ............ .... 36 Medical Risk Factors ........................ .... ...... .. .. ............. 38 Behavioral Risk Factors . . ........ ... .... .. ......................... 39 Race/Ethnicity/Nativity ................. ....... .. .... ................ .40 Missing Data and Size of Study Population for Aim 1 ...... ............ .41 Missing Data and Size of Study Population for Aim 2 .. ................ .44 Contextual Variables for Aim 3 .. .. ............... .................. .. ....... .47 Missing Data and Size of Population for Aim 3 ... ... . ... ... ... ........ .49 Methods for Aims 1 and 2 ... .... .. ... .. ...... ..... ..... ... . .... ... ..... ............... 50 Model Building for Aims 1 and 2 .. ......................................... 51 Model Building for Aim 3 .............. .... ....... ............. . ....... ... ... 60 Qualitative Research Component. ........ ... ....... . .......... ........ ... ........... 61 ix

PAGE 10

Sample and Study Data ... .... .... ... ........... . ..... .. .... .. ... . .. ... 61 Methods for Aim 4 .. .. ... ................ . .. ......... ... .. ........... ... ... .. .... .. ... ... 63 4 QUANTITATIVEANALYSIS ............... .. ... .. ... .............. . . ..... ... ..... ...... .. 65 A i m 1 .. ................ ... ............ .... . .......................... .... ............. ..... 65 Comparison of Risk Factors by Race/Ethnicity ......... .. . ..... ........ . 65 Frequency of Adverse Birth Outcomes by Race/Ethnicity ..... . .... .. 68 Odds Ratios of Birth Outcomes by Race/Ethnic i ty .. ........... ... .. .... 69 Low Birth Weight. ................... ............ ..................... 70 Preterm Birth . . ... ... .. ... .... .. .. . . ........ .. ............. .... . 73 Small for Gestational Age .... .. ............ . .... .... ... .. . . ..... 77 Large for Gestational Age ... .. ............ ....... . ......... ...... .. 81 Discussion of A i m 1 ......................... ... ..... .... .. .... .... ... .. . . 84 Aim 2 .... .. ................... ... ..... ...... . ... .... . ......... ................ ..... ...... 87 Comparison of Risk Factors of Mothers of Mex i can Or i g i n by Nativity ... . ... 87 Frequency of Adverse Birth outcomes by Nativity .. .. . ... .. .... . .. . 90 Odds Ratios of Birth Outcomes by Nat i vity ............. . .. ...... . ..... 91 Low Birth Weight. .. .. ... ............. .... . ... .................... ..... 91 Preterm Birth ... ....... . ... . ... .................. . ......... ......... 94 Small for Gestat i ona l Age ............. ..... . .................... .... 98 Large for Gestational Age .... . ... . ............... ........... ... 1 00 Discussion of Aim 2 ... .. ..... .. ... .. .. .. . .. . .. ............................ 1 05 A i m 3 ...... ... .. ..... ..... ........ ............... ... ..... ...... .... .. .. .......... ..... ..... 110 Influence of Neighborhood Deprivation and Immigrant Or i entation on Outcomes ... ... ........................................ : ... .... ... .. ............ .... .. .... 111 D i scussion of Quantitative Analysis .... ... .. ................... ... .................... 114 X

PAGE 11

5. QUALITATIVE RESULTS ... ... .. .... ... ... ..... ................................... ... ........ 117 Recent Mothers . .......... ............. ............ ...................................... 117 Key Informants .... ... ... .. . .............. ... . .. .... ... ................................. 118 Diet and Exercise During Pregnancy ... . ..... . .......................... 118 Diet. ........... .. ......... .. ... .................... .. .. ..... ... ......... 118 Exercise/Energy Expenditure . ... . .............................. 121 Maternal Weight and Weight Gain . . .. .... .. .... ... ................. ..... 122 Body lmage ............. .............. ................................ 123 Smoking and Drinking .... ..... .................. ...... . .............. ....... 124 Other Cultural Beliefs about Pregnancy .... . ............... .... ......... 125 Sources of Social Support .. .. ..... ... .................................. ... 126 Political Economy and Birth Outcomes ........................................ ..... 127 Discussion of Qualitative interviews .................. ...... ... ... ......... 130 6. DISCUSSION ............ ... .................................................................... 133 Limitations ............................. .... .... .... .. ... ......... .............. .. ....... 134 Is There an Epidemiological Paradox in Weight-Related Birth Outcomes? ........ ......... .... ................... . ........ ......... ................. .... 134 Do the Hypotheses in the Literature Explain the Paradox? ............. .... . ... . 137 How Should Health be Measured? . ......... .......... .... .... ... .............. . 137 Healthy Migrant Hypothesis .. . ........... .. ......... ....................... 138 Healthy Immigrant Hypothesis ....... ... .. .. ... .. ... ....................... 139 Ne i ghborhood Effects ............... .. ... .... ... .. .... .. ..... .. .. ... ... ... .... 140 Political Economy-A Broader Perspective ........................ .......... . ..... 140 Significance .. ... ..... ............... ..... ... ............ .. ........ ....................... 141 xi

PAGE 12

APPENDIX ... ... .. ... ............ . . .......... ... . ... . .. ......... ..................... . ... . . .. ... .... 143 A. Summary of Selected Population Studies ........................................ ........ 144 B. Human Subjects Approvals .... ................................. . ..... ........ ... . . .. ..... 146 C Solicitation Guide Interv i ew Guides, Consents ... . ... . ... ............. .. ...... . . 152 BIBLIOGRAPHY .... ................................. ...................... . .... ................. . .... .. ..... 171 xii

PAGE 13

LIST OF FIGURES F i gure 2 1 PATHWAYS TO OBESITY AND DIABETES ... ................... .. .. .... ... .... ... . . .. ... . 26 3 1 SCHEMATIC OF RESEARCH DESIGN ........ ......... . ... ........... . ... .... ......... .. .. ... .. 31 3 2 AIM 1 LBW DRIFT OF ODDS .... ... .. ...... ... .................. ... . .. ............................ .. 44 3 3 AIM 1 PRETERM DRIFT OF ODDS .. .. ........ ............. .... ................. .. ....... .. . .. .44 3.4 AIM 1 SGA DRIFT OF ODDS .... .. .... ........ . . . ................... .... ... ................... .. . .44 3.5 AIM 1 LGA DRIFT OF ODDS .............. .. .... . ... . ....................................... ... .. ... 44 3.6 AIM 2 LBW DRIFT OF ODDS .. .... ........ ... . ........... .... .. .. .... .. ............... ... ...... .46 3 7 AIM 2 PRETERM DRIFT OF ODDS ......... . . ........... .. ........ ... ... ... ... ......... ...... .. 46 3 8 AIM 2 SGA DRIFT OF ODDS ............ .. ......................... . .... ........... . .. .. .. ....... 47 3 9 AIM 2 LGA DRIFT OF ODDS ... ... ...... ................... . .. ..... .... .. ........ .. .. . ........... 47 3.10 SCHEMATIC OF MODEL BUILDING FOR AIMS 1 AND 2 ......... ... .. ... .............. ..... 53 4 1 DISTRIBUTION OF RISK FACTORS BY RACEJETHNICITY . .. .. .. . ...................... 67 4.2 UNADJUSTED FREQUENCIES OF BIRTH OUTCOMES BY RACEJETHNICITY ...... 69 4 3 LBW ODDS RATIOS BY RACEJETHNICITY AND MODEL. ... . .. .... . ... . ......... ..... 70 4.4 PRETERM BIRTH ODDS RATIOS BY RACEJETHNICITY AND MODEL. .... . .. ... . ... 74 4 5 SGA ODDS RATIOS BY RACEJETHNICITY AND MODEL. ....... ....... ............ ......... 78 4 6 LGA ODDS RATIOS BY RACEJETHNICITY AND MODEL. . ........ ............ ... .. .. ... 82 4 7 FULLY ADJUSTED ODDS RATIOS BY RACEJETHNICITY .............. .... ............... 85 4.8 DISTRIBUTION OF RISKS BY NATIVITY ...................... . ... ... .. ... .. .................. 89 4 9 UNADJUSTED FREQUENCIES BY NATIVITY ......................... ......... ............... 91 4 10 LBW ODDS RATIOS BY NATIVITY AND MODEL. .. . .... ................ .. ......... ..... .... 92 xiii

PAGE 14

4 .11 PRETERM BIRTH ODDS RATIOS BY NATIVITY AND MODEL. . ...... .............. ....... 95 4.12 SGA ODDS RATIOS BY NATIVITY AND MODEL. .. ........ .... ... .... ... .. ... ... . .......... . 98 4 13 LGA ODDS RATIOS BY NATIVITY AND MODEL. . .................. ..... .................... 101 4 14 FREQUENCY OF LGA BY NATIVITY BY YEAR. .. .... .... .. . ... . . .............. . . ... .... 1 04 4 15 FREQUENCY OF LGA BY WEIGHT GAIN . ....... ... ......... ................. ............. . 104 4 16 FREQUENCY OF LGA BY AGE ....... ........ .. ... ......... .... .. ...... ... .. ..... . ... .. ....... 1 04 4.17 FREQUENCY OF LGA BY PARITY .. ... ... ... .... . .... .... .. ... ......... ...................... 105 4.18 FREQUENCY OF LGA-SPECIFIC RISKS ... . .... ..... .. ... . .... .... ...... . ........... ..... 105 4.19 FULLY ADJUSTED ODDS RATIOS BY NATIVITY .. ..... ..... ....... ... ... .. ..... ... ...... 106 4 20 ODDS OF OUTCOMES FOR MEXICAN -BORN MOTHERS IN ADAMS AND DENVER COUNTIES AND STATEWIDE ..... .. ...... ........... ............... . ..... ............ 11 0 6 1 BIRTH WEIGHT IN GRAMS OF MEXICAN-BORN MOTHERS (ABOVE) WITH U.S.-BORN MOTHERS ............. ............ .... .. ...... .. . .. ........................ 137 x i v

PAGE 15

LIST OF TABLES Table 1 1 Principal individual risk factors for LBW .............................................................. .4 1.2 LBW and preterm birth rates by race/ethnicity in U.S. and Colorado ....................... . ... 5 1.3 Population of Mexican origin as% of Hispanic population in Colorado-2000 .... . ........ 8 1.4 Comparison of socioeconomic posit ion of Hispanic, Mexican, and non-Hispanic White populationsU.S. Census 2000 and 2002 ........ ... .................... 9 1 5 Study population by race/ethnicity of mother-Aim 1 ..... ... ...... ................. ............ 13 1.6 Study population by nativity of mothers of Mexican origin -Aim 2 ... ... ...................... 13 1.7 Study population of mothers of Mexican origin in Denver and Adams Counties-Aim 3 ...... ...... ....... .... ... ... ............ .............. ...................... ........................ ......... ............ 13 2.1 Measures of socioeconomic position ..... .... .... .................. .......... ...... . ...... ..... .... 17 3 1 Individual-level variables for Aims 1 and 2 .... ......... ............... ... .. ....... ... .............. 35 3.2 Original dataset by race/ethnicity ... .............................. .... ... .... ............. ............ .42 3.3 Number and percentage of missing cases by variable for Aim 1 .................. .......... .42 3.4 Number and percentage of missing variables by race/ethnicity ... .... .......... .............. .43 3.5 Final study population by race/ethnicity of mother-Aim 1 ..................................... .44 3 6 Original dataset by place of nat i vity of mother ...... .... .... .......... .. ........ ................... .45 3.7 Number and percentage of missing cases by variable for Aim 2 .... ... ........... ............ .45 3.8 Number and percentage of m issing variables by nativity ...... .... .......... .... ... ... .... ...... .46 3.9 Final study population by nativity of mothers of Mexican orig i n by year-Aim 2 ... ..... .47 3.10 Values of contextual scales for Adams and Denver Counties 2000 .... ... ............ ..... .49 3 .11 Population of mothers of Mex i can origin by nativity 2000-2005 in Adams and Denver Counties .......... .. .......... .. ............. ..... ...... ................................. 50 4 1 Percent frequency distribution of risk factors by race / ethnicity 2000-2005 ... .......... .... 66 XV

PAGE 16

4 2 Percent frequency of LBW preterm birth, SGA and LGA by race/ethnicity 2000-2005 .......... .. . ..................... . . .... ................... .... ... .. ....................... 68 4 3 Unadjusted and adjusted odds ratios (95% Cl) of LBW and race/ethnicity ...... . ...... ... 70 4 4 Estimated coefficients and odds ratios for LBW by race/ethnicity ..... . ..... ... ............ 71 4.5 Unadjusted and adjusted odds ratios of preterm birth by race/ethnicity .................... 74 4 6 Estimated coefficients and odds ratios for preterm birth and race/ethnic it y ..... ... .. ...... . 75 4.7 Unadjusted and adjusted odds ratios (95% Cl) of SGA by race/ethnicity ................ .... 78 4.8 Est i mated coefficients and odds ratios for SGA and race/ethnicity ... ...... .................. 79 4.9 Unadjusted and adjusted odds ratios (95% Cl) of LGA by race/ethnicity .................... 82 4.1 0 Estimated coefficients and odds ratios for LGA and race/ethnicity ........ .. ....... .. ..... ... 83 4.11 Comparison of fully adjusted odds ratios of birth outcomes by race/ethnicity ... ... ........ 85 4.12 Comparison of contributing risk factors by birth outcome and race/ethnicity ... ............. 86 4 13 Comparison of fully adjusted odds ratios of AGA by race/ethnicity ............................ 87 4 14 Percent frequency distribution of risk factors of mothers of Mexican origin by nativity 2000 2005 ............ . .. ..... ............................ ............ ... ..... .............. 88 4.15 Percent frequency of LBW preterm birth SGA and LGA by nativity 2000-2005 .. .. .. .......... ...... ............................................................................ 90 4.16 Unadjusted and adjusted odds ratios (95% Cl) of LBW by nativity ........................ ... 92 4.17 Estimated coefficients and odds ratios for LBW and nativity ....... .............. ... ... ... ..... 93 4.18 Unadjusted and adjusted odds ratios (95% Cl) of preterm birth by nat i v i ty ...... .. ......... 95 4.19 Estimated coefficients and odds ratios for preterm birth and nativity ......................... 96 4 20 Unadjusted and adjusted odds ratios (95% Cl) of SGA by nativity .... .. .. ...... .............. 98 4.21 Estimated coefficients and odds ratios for SGA and nativity .......................... .......... 99 4.22 Unadjusted and adjusted odds ratios (95% Cl) of LGA by nativity .......................... 1 00 4.23 Estimated coefficients and odds ratios for LGA and nativity ...... ... .. ... ..... . ............. 101 4 24 Model 4 adjusted odds ratios (95% Cl) of LGA by nativity ....... . .. ... .. ..... ...... .. ..... 1 03 xvi

PAGE 17

4 25 Fully adjusted odds ratios of birth outcomes by nativity ........................................ 1 06 4.26 Comparison of fully adjusted odds ratios of AGA by nativity .................................. 1 07 4.27 Fully adjusted odds ratios (95% Cl) by nativity and model .................... .... ............. 1 08 4.28 Fully adjusted odds ratios (95% Cl) by nativity for LGA using LGA-specific medical risks ..... .................................... . .... . ..... .... ..... ........... .. .. ... .. ....... 1 08 4.29 Comparison of fully adjusted odds ratios of birth outcomes of Mexican-born mothers in Adams and Denver Counties and statewide ................ .................. .... .. 11 0 4.30 Outcomes in Adams County .. ............ ................................. .. .. ............. ... ....... 111 4.31 Estimated coefficients for neighborhood deprivation and immigrant orientation in Adams County .......... .. .. .. .... .. . ..... .. . . .. .................................................. 112 4 32 Outcomes in Denver County .... ............................ .. ... ...................................... 113 4.33 Estimated coefficients for neighborhood deprivation and immigrant orientation in Denver County ................................. .............................. ..................... ..... 113 5 1 Selected characteristics of women interviewed .... ... ..... ... .. .................................. 118 5.2 Weight gain of mothers interviewed and birth weight of infants ........ .. ........ .. .......... 122 6 1 Comparison of fully adjusted odds ratios by race/ethnicity ..................................... 134 6.2 Comparison of fully adjusted odds ratios by nativity .... .............................. ...... ..... 135 xvii

PAGE 18

CHAPTER 1 INTRODUCTION Only the paradox comes anywhere near to comprehending the fullness of life."1 Epidemiological Paradoxes Since at least the Middle Ages, observers have noted a general social gradient of health that associates health outcomes with certain compositional aspects of populations. Better outcomes are associated with an individual's increasing income/wealth, social status, and education; worse health outcomes are associated with lower income, status, and education (Lynch & Kaplan 2000:13-35). In public health terms, these differences are usually called health disparities, and much effort is expended to understand the causes of poorer health and to try to reduce those disparities (Berkman & Kawachi 2000). The expected social gradient is not invariant, however. For example, it does not appear to hold for all diseases or conditions, for all subgroups within a racial or ethnic population, or for some foreign-born individuals who immigrate to the Unites States (Gorman 1999; Singh & Yu 1996). In addition to a social gradient related to the compositional aspects of populations, public health researchers have identified a similar social gradient of health at the area, or contextual level, where better health is often associated with wealthier, better-endowed neighborhoods (Diez-Roux 1998). As is the case with compositional characteristics of a population, research on neighborhood contextual factors affecting health is usually focused on poorer neighborhoods and their correlation with poorer health outcomes (Reagan & Salsberry 2005; Pearl eta/. 2001; Gorman 1999; O'Campo eta/. 1997). One exception is the recent work by Finch et a/. (2007), who associated the predicted probability of low birth weight among the offspring of native and foreign-born Hispanics in Los Angeles County, California with measures of neighborhood disadvantage and "immigrant-orientation" and found a protective effect of immigrant orientation at the neighborhood level 1 Although this quotation from Carl Jung refers to the richness of paradoxes in religion, this and his further observation that non-ambiguity and lack of contradiction are one-sided, can apply equally, if more prosaically, to understanding unexpected occurrences in population health (Jung 1980:18). 1

PAGE 19

Studies from the 1970s and 1980s identified a "paradox" from the general social gradient of health among Hispanics as a population; that is, despite socioeconomic and demographic risk profiles that are less advantageous than those of the majority reference population, Hispanics statistically showed a number of unexpectedly better health outcomes (Markides & Coreil1986; Teller & Clyburn 1974). Thereafter researchers noted that the lack of health disparity appeared even more pronounced for certain birth outcomes, usually low birth weight and premature birth, among foreign-born immigrants, whether Hispanic or other racial/ethnic groups such as immigrants from Africa and India These immigrant populations typically come to the U.S. with little wealth low education, and low social status, and they usually live in poorer neighborhoods (Hummer et a/. 1999 ; Singh & Yu 1996), thus putting them at the low end of the social grad i ent. Birth Outcomes Countries adopt various measures of population health, including life expectancy measures of the burden of disease on the functioning of the population and birth outcomes. Birth outcomes are considered signal measures of health i nternationally (Un i ted Nations 2008; UNICEF 2004) and in the U.S. (USDHHS 2000) because they represent the future health of the population. Although low birth weight (LBW) and preterm birth have long been identified as birth outcomes of interest, small for gestational age status (SGA) is increasingly recognized as an adverse birth outcome (Lee eta/. 2003). In addition, large tor gestational age status (LGA) is gaining recognition as an adverse birth outcome that should be better studied (Dyer eta/. 2007 ; Dollberg eta/. 2000). Weight-related birth outcomes, specifically low birth weight, preterm birth, small tor gestational age and large for gestational age are i mportant indicators of well-being both with respect to the infant's immediate health and over the life course (Oken & Gillman 2003; Barker 2002; Eriksson et a/. 2001; Basch 1999 : 79-81 ). All four measures are key indicators of health; understanding their patterns in different populations can help guide public health policies and allocation of resources. Rates of LBW preterm birth SGA and LGA vary among countries and among population groups within countries (Frisbie & Song 2003; Gould eta/. 2003; Hummer et al. 1999; Singh & Yu 1996). In its initiative Healthy People 2010, the U .S. Department of Health and Human Services has adopted aggressive objectives to reduce rates of LBW and preterm birth, both in the aggregate and with regard to disparities in their rates across race and ethnicity by 2010 (2000). Although the highest risk of morbidity and 2

PAGE 20

mortality is associated with low-weight births, higher morbidity and mortality occur among LGA infants compared with those who are appropriate for gestational age (AGA) (Lubchenco et a/. 1972; Lubchenco & Bard 1971 ). Moreover, considering AGA and LGA births together may obscure significant outcomes, especially when the rates of LGA births are rising (Barbour, L., personal interview, November 10, 2008). Low Birth Weight Low birth weight is defined as an infant weighing less than 2500 grams2 at birth (approximately 5.5 pounds) (WHO 2006:36). LBW is an important measure of public health because LBW infants are almost 40 times more likely to die during their first month of life than infants weighing more than 2500 grams at birth (Collins & Schulte 2003:223; 10M 1985). In addition, LBW infants comprise two-thirds of deaths in the first 28 days of life and 20% of deaths from 28 days until one year of age (10M 1985). Moreover, LBW infants who survive have elevated risks of morbidity from conditions such as cardiovascular disease, diabetes, and obesity during their life course (Barker 2002; Merson eta/. 2001: 122). LBW results proximately from premature birth (<37 completed weeks of gestation), poor weight gain of the fetus, or both. Although poor weight gain and preterm birth often occur together, they have independent and multiple causes (March of Dimes 2007; 10M 1985). Because fundamental (or more distal) causes of LBW are multiple and poorly understood, most medical research into LBW has focused on identifying and addressing risk factors that are associated with the increased probability of delivering an LBW infant. Table 1 .1, adapted from the Institute of Medicine (1985:7), lists the principal individual level risk factors for LBW. These factors sort into demographic and socioeconomic factors, medical risks, and environmental/behavioral factors. 2 Interestingly, some experts define low birth weight as 2500 grams or less (Merson eta/. 2001 :213; 10M 1985), while others use less than 2500 grams (March of Dimes 2007). This study follows the definition used by the WHO (2006) and the Colorado Department of Public Health and Environment (2000a), each of which defines LBW as less than 2500 grams at birth. 3

PAGE 21

Table 1.1. Principal individual risk factors for LBW Demographic/Socioeconomic Age (<17; >34) Race (Black) Low socioeconomic status Medical Risks Genitourinary anomalies/surgery Diabetes Parity (0 or >4) Low weight for height Chronic hypertension Non-immune for selected infections Previous LBW infant Low birth weight of mother Multiple parity Poor weight gain of mother Short inter-pregnancy interval BehavioraVEnvironmental Maternal smoking Poor nutritional status of mother Alcohol or other substance abuse Unmarried Low level of education Hypotension Hypertension/preeclampsia/eclampsia Infections Placenta previa/abruptio placenta Hyperemesis gravidarum Oligohydramnios/polyhydramnios Anemia Isoimmunization (Rh incompatibility) Fetal anomalies Incompetent cervix Iatrogenic preterm birth Toxic exposures High altitude/hypoxia In 2000, the rate3 of LBW in the United States overall was 7.6% and the disparity among races and ethnicities was large; the rate among non-Hispanic Blacks was 13.1 %, among non-Hispanic Whites it was 6.6%, and among Hispanics of all races it was 6.4% (CDC 2002 : Table 44) All of Colorado's rates exceeded the national rates by race/ethnicity. Colorado's statewide LBW rate was 8.4% in 2000, while the rate among non-Hispanic Blacks was 15.0%, among non-Hispanic Whites it was 8 0%, and among Hispanics of all races the rate was 8 1% (CDC 2002: Table 46). According to the Colorado Department of Public Health & Environment (CDPHE), the major contributing factors to LBW in the state during 1995-1997 were multiple births (1 in 5 LBW in Colorado was a multiple birth during that period); inadequate maternal weight gain; smoking; and premature rupture of membranes (CDPHE 2000a). The state identified inadequate maternal weight gain as the largest contributor to Colorado's LBW rate. In response to this finding the state adopted a marketing campaign entitled "A Healthy Baby is Worth the Weighf' (CDPHE 2005b). One objective of Healthy People 2010 is to reduce LBW in the U.S to 5.0% and to eliminate disparities by race and ethnicity (USDHHS 2000). In this regard, the National 3 The use of rate" refers to the frequency or percentage of all live births rather than a rate per 1000 births, the usual epidemiological use of rate." 4

PAGE 22

Institutes of Health has called for research to better understand mechanisms that underlie racial and ethnic disparities in LBW. Preterm Birth The Inst i tute of Medicine's recent report on preterm birth considers a birth to be premature if it occurs before 37 completed weeks of gestation (10M 2006) Colorado uses the same definition (CDPHE 2000d). The Institute of Medicine describes preterm birth as a cluster of problems with a set of overlapping factors of influence" associated with risks that i nclude individual-level factors, neighborhood characteristics medical conditions genetics and environmental exposures (2006:1). As with LBW, the causes of preterm birth are complex, and rates vary by race and ethnicity. The costs of preterm birth include increased infant mortality and morbidity, such as ophthalmic, neurologic and psychomotor development difficulties, as well as deficits in pulmonary health over the life course (Kramer 2003). In 2000, the overall U .S. preterm birth rate was 11.6% ; the rate for non-Hispanic Blacks was 17.4% for non-Hispanic Whites it was 1 0.4% and for Hispan ics of all races the rate was 11. 2% (CDC 2002a : Table 44). Rates of preterm birth in the U.S. increased in 2004 by 0.5% to 1.1 % depending on the race/ethnicity category. In contrast to Colorado s relatively high LBW rate, its rate of preterm birth is lower than the national average. In 2000 the rate of preterm births for all races in Colorado was 9.0%, while the U .S. average was 11.6% (CDPHE 2000b). The rate of preterm births is rising, however with the U .S. average for all races/ethnicities reaching 12.5% in 2004 (10M 2006), and the Colorado average reaching 9.8% (CDC 200Gb; CDPHE 2004). Healthy People 2010 has set a target to reduce preterm birth to 7.6% for all races and ethnicities by 2010 (USDHHS 2000). Table 1.2. LBW and preterm birth rates by race/ethnicity in U.S. and Colorado Race/Ethnicity Low Birth Weight % Preterm Birth % 2000 2004 2000 2004 us CO" us co C0 All races/ethnicities 7 6 8 4 8 1 9 0 11. 6 9 0 12 5 9 8 White (non Hispanic) 6 6 8 0 7.2 8 7 10 4 8 8 11.5 9 9 Black (non-Hispanic) 13. 1 15.0 13.7 14. 5 17.4 13 5 17 9 13 6 Hispanic (all races) 6 4 8.1 6 8 8.6 11. 2 8.7 12 0 9 2 1 CDC 2002a : Table 44 2. CDC 2002a : Tab l e 46 3 CDC 2006b 4 CDPHE 2004b 5 CDPHE 2000b 6 CDPHE 2004a 5

PAGE 23

Small for Gestational Age Small for gestational age is generally defined as birth weight of an infant below the 1oth percentile of weight for gestational age (Kiiegman & Das 2002), although some researchers use the 15 \ 3rd, or 5th percentile (Gould et a/. 2003). SGA measures size in relation to gestational age to adjust for lower weight due to preterm birth and is therefore a more targeted measure of intrauterine growth. Its accuracy depends on the clinical estimate of gestation or gestation reported by the mother, neither of which is as accurate as birth weight. Some studies calculate SGA from data on birth weight and gestation in the birth record; more studies use LBW because it requires only one measure. Chung eta/. (2003) argue that there are baseline differences in birth weight by gestational age among different races/ethnicities. Their work suggests that ethnic differences in weight by gestation may contribute to the differences in SGA rates. There are no nationally-adopted tables of the distribution of birth weight by gestation. For the reasons described in Chapter 3 at pages 3435, this study uses the tables created by Alexander eta/., which report weight by weeks of gestation by race and ethnicity (1999). Factors associated with SGA include underlying medical risks of the type reported in the birth record (hypertension, renal disease, diabetes, chronic pulmonary disease, poor placentation), low pregnancy weight, smoking, very young maternal age, older maternal age, and first births (Lee et a/. 2003). SGA has been associated with the development of non insulin-dependent diabetes mellitus (Barker 1998), hypertension (Barker 2002}, and coronary heart disease (Barker 2001) later in life Large for Gestational Age LGA is defined as birth weight of an infant above the 90th percentile of weight for gestational age (Alexander et a/. 1999). LGA results in babies who are not only large, but whose excess weight is found in adipose issue around the trunk of the baby Some babies are too large to fit through the birth canal, leading to shoulder dystocia and increased incidence of delivery by c-section (Casey eta/. 1997 ; Modanlou eta/. 1982). An LGA infant often has reduced ability to regulate glucose immediately after birth and higher risk of developing metabolic syndrome and type II diabetes during the life course (Dyer eta/. 2007; Hediger eta/. 1998). Maternal diabetes gestational diabetes, high weight gain during pregnancy, and high maternal BMI before pregnancy are associated with increased odds of delivering an LGA 6

PAGE 24

baby In addition, mothers who were themselves born LGA are more likely to give birth to an LGA infant (Ahlsson eta/. 2007 ; Ehrenberg eta/. 2004) It is estimated that 7 7% of Mexican Americans had diagnosed diabetes and 5.4% had undiagnosed diabetes in 2000 (Martorell 2005; see also Mainous et a/. 2008) The rate of overweight and obesity is extremely high in Mexico 62% of adults in 2000 and the rate of diabetes among the Mexican population continues to increase (Jimenez-Cruz & Barcardi-Gascon 2004). While genetics most likely plays a part in obesity, environmental influences also play a role, such as lower levels of physical activity, lower total daily energy expenditure, and increasing availability of high-fat, energy-dense foods and larger portion sizes (Hill et a/ 2000). Moreover people are products of their life histories, which in turn are products of the regional and global economic environments in which they live (di Leonardo 1984:12} Fetal Programming Studies of health over the life course of babies born small (LBW, some preterm b i rths, and SGA) have suggested a theory of "thrifty phenotype or ''f etal programming to explain why infants born small have higher odds of obesity and chronic conditions associated with obesity, such as type II diabetes and cardiovascular disease, than babies born appropriate for gestational age (Barker 1998, 2001 2002). In brief the health of an infant is partly programmed by the environment in utero. Some LBW, preterm, and SGA infants experience a fetal environment in which the available nutrition is limited perhaps by poor perfusion of the placenta, undernutrition in the mother maternal smoking or other causes of inadequate nutrition reaching the growing fetus Scarcity causes the fetus to take in as much nutrition as it can obtain and programs its phenotype to be 'thrifty' and save it (Breier et a/. 2001; Hales & Barker 2001 ) When the infant is born, if the environment is one of abundance, the programming nonetheless remains in place and raises the odds that the infant, and later adult will continue to gather nutrients as if still in a nutritionally-deprived environment. Catch-up growth often results in overweight and obesity in adolescence and adulthood. Fetal programming may also account for adverse health conditions over the life course for LGA babies. Mothers of LGA babies often supply excess glucose to their fetuses either because the mothers have habitually high and continuous carbohydrate intake or because they have metabolic syndrome or gestational diabetes At birth an LGA infant has higher odds of reduced ability to regulate glucose within the first few days after birth (Dyer et a/ 2007). In response to glucose challenge, even asymptomatic infants display early 7

PAGE 25

marke r s of insulin resistance. It i s bel i eved by a number of researchers that a fetal environment that includes excess glucose programs the fetus with a predispos i tion for i nsulin resistance, which later in life may manifest in obesity type II diabetes and cardiovascular disease Women who are born LGA have higher odds of delivering LGA i nfants of their own suggesting that fetal programming can persist from one generation to the next (Drake & Walker 2004 ; Barbour L., personal interview, November 10 2008) Population Characteristics of Colorado Based on the U .S. Census the percentage of Hispanics or Lat i nos "4 i n Colorado has r i sen from 17.1% in 2000 to 19.5% i n 2005 (2007a) Hispanics represent one of the fastest growing ethnic groups in the U.S.; by 2050, Hispanics are expected to account for 22.5% of the U S population (U .S. Census 2001 ). Notwithstanding current U.S. polic i es intended to reduce i mmigration from Mexico the Hispanic population would double by 2050 even i f i mm i grat i on stopped (Smith & Edmonston 1997). Table 1.3 presents selected information about the Hispanic population in Colorado (U.S Census 2000d) Ind i v i duals of Mexican ori g i n are the second largest racial / ethnic group in Colorado mak i ng up at least 10.5% of the state s population (Census 2000c). Persons of Mexican origin are more likely to l i ve i n poverty not speak English at home and be foreign-born than the Hispanic/Lat i na populat i on as a whole. Similarly Coloradans of Mexican origin are less l i kely to have completed h i gh school or college than the Hispanic / Lat ina population in the state Table 1.3. Population of Mexican origin as% of Hispanic population in Colorado 2000 Characteristic Population Total Female Population Average Family Size High School Graduate Bachelor Degree + Fore i gn Born Non-Engl i sh at home Famil i es < poverty Individuals < poverty Hispanic 735 ,601 350 795 4 218 902 39 335 201, 072 366 528 25 903 135,421 Mexican 450 760 205,156 4 113,129 19,478 170 356 259 294 16 613 90 575 61. 28 % 58.48 % 51. 68 % 49 52 % 84.72 % 70 74 % 64 89 % 66 68 % 4 The federal government classifies those of Spanish Central Amer i can and Latin Ameri can or i g i n as Hispan i c or Latino. Many i n th i s group prefer one ethn i c label over another. For th i s study, persons i n this category as referred to generally as Hispan ic." H i spanics of Mex i can origin are referred to as Mexican born or U.S.-born of Mex i can orig in. 8

PAGE 26

National estimates by the U.S. Bureau of the Census, as reported in Table 1.4, confirm that Hispanics and Mexicans as a particular subpopulation, have worse economic and socioeconomic position profiles than non Hispanic Whites in 2000. Based on Colorado s Pregnancy Risk Assessment Monitoring System (PRAMS) data from 19982002, Hispanic women in Colorado were less likely to receive adequate prenatal care than their counterparts a result that is influenced primarily by the inability of Hispanics who are not born in the U.S. and are undocumented to obtain prenatal care on the recommended schedule (CDPHE 2005b). In most categories Mexicans fare worse than Hispanics as a whole (U.S. Census 2000d). In 2000 31. 7% of Denver County's population identified as Hispanic/Latino and 28.2% of Adams County identified as Hispanic/Latino (U.S. Census 2000b). CDPHE reports that the proportion of linguistically isolated Spanish speaking households increased by 171% statewide between 1990 and 2000 based on census data. The proportion of linguistically i solated households in Adams County increased by 416% and by 158% in Denver County during this same time period (CDPHE 2005a). Table 1.4. Comparison of socioeconomic position of Hispanic, Mexican and non Hispanic White populationsU.S. Census 2000 Measure Mexican HI Non-Hispanic spamc White 2000 2000 2000 H i gh school graduate 51. 0 % 57 0 % 88.4 % Unemployment 7 0 % 6.8 % 3 4 % Employment in service occupations N / A 1 9.4 % 1 1 8 % Professional / management employment 11. 9 % 14 0 % 33 2% Income of $35,000 or more 20 6 % 23.3 % 49 3 % Living below poverty level 24 1 % 22 8 % 7.7% By most measures people of Mexican or i gin in the U.S and Colorado are at the low end of the spectrum for education, employment, and income and at the high end for being class i fied as poor. Why Study Paradoxical Health Outcomes? It is important to study paradoxical birth outcomes f i rst to learn i f they exist i n Colorado Previous studies have found some paradox i cal results in California New York and national samples (Rosenberg et a/. 2005; Zambrana et a/. 1997; S i ngh & Yu 1996) Second it is i mportant to unpack the apparent ep i demiological paradox within the general grouping of Hispanics because of the heterogeneity of subpopulations (Hummer eta/. 1999 ; Frisb i e et a/. 1998 ; S i ngh & Yu 1996) and to disentangle the effect of nat i v i ty (Gould et a/. 9

PAGE 27

2003; Singh & Yu 1996). Evidence that the paradox is short-lived points to a need to improve outcomes for all populations. Third, it is beneficial to examine "non-small" infants for evidence of previously ignored adverse outcomes. Finally population studies alone are insufficient to answer complex questions about outcomes and their underlying causes. Several hypotheses have been proposed to explain the unexpected deviance from the general social gradient of health among Hispanic infants all of which suggest that the incidence of paradoxically good birth outcomes among infants of foreign-born immigrants improves the rates for Hispanics as a whole. Three of the most common are (1) the healthy migrant hypothesis [people who choose to migrate to the U.S are healthier than those who stay in their native countries] (Cho eta/. 2004; Marmot eta/. 1984) (2) the healthy immigrant hypothesis [immigrants engage in healthier behaviors, such as better diet, less smoking, less alcohol consumption, at least immediately after they immigrate to the U.S ] (Flores & Brotanek 2005; Kasirye eta/. 2005 ; Abrafdo eta/. 1999; Guendelman eta/. 1999) and (3) a social support hypothesis [immigrants enjoy increased social support, social network i ng and/or social capital, probably founded in retained cultural beliefs, which compensate for less education income, and access to health care] (Landsale et a/. 1999; Rumbaut & Weeks 1996}. Other researchers argue that examining the degree of assimilation or acculturation explains differing health outcomes among immigrants. Finch eta/. suggest that segmented assimilation theory explains divergent health trajectories among immigrant women and that immigrants who live in ethnic enclaves may engage in healthy behaviors brought from their native countries or continue to live in accordance with cultural practices or values, which may be protective (2007). Finally, some researchers argue that there is no paradox; rather it is merely a statistical artifact (Palloni & Arias 2004; but see Hummer eta/. 2007). Amaro and de Ia Torre (2002) warn that the assumption of some public health agencies that Hispanic mothers are not at risk for poor birth outcomes because of the paradox'' may not in fact be accurate for all subgroups or all conditions. In add i tion if Mexican-born immigrants in Colorado are shown to enjoy unexpectedly better birth outcomes than the social gradient would predict and if their advantage decays over time as shown by some, it would be beneficial to generate initiatives to promote conditions that might extend the longevity of those better outcomes (Ricketts eta/. 2005; McGlade eta/. 2004). If there is no paradox, then it will be important to direct attention to the poor outcomes. 10

PAGE 28

Research Design and Specific Aims Originally this study sought to determine whether an epidemiological paradox exists for each of four birth outcomes: LBW, preterm birth, SGA, and infant mortality, among Hispanics, primarily Mexican-born, in Colorado during the years 2000 2005 at both the individual and neighborhood level. During the analysis of data, when it became clear that the paradox does exist for LBW, preterm birth, and SGA among Colorado's Hispanic population and even more clearly among Mexican-born mothers, curiosity about large for gestational age infants led to the addition of LGA as a birth outcome. The data on infant mortality proved less reliable than was hoped and the rates of infant mortality were very low resulting in small measurable variances among the population groups of interest. Accordingly, infant mortality was dropped. This multi-level, nested, mixed method study has four aims: to test the expected social gradient of health on birth outcomes by racelethnicity and to identify any paradoxical outcomes (Aim 1 ); to test the expected social gradient of health on birth outcomes by nativity of Hispanic mothers of Mexican origin and to explore the healthy migrant and healthy immigrant hypotheses as explanations for any paradoxically better birth outcomes experienced by Mexican-born mothers (Aim 2); to examine the association of neighborhood deprivation and immigrant orientation on birth outcomes of mothers of Mexican origin in two large Colorado counties (Aim 3); and to explore potential reasons for differential outcomes among Mexican-born mothers and U.S.-born mothers of Mexican origin (Aim 4). Overview of Research Methods Most literature on the subject of the paradox is either quantitative or qualitative. The quantitative studies that have identified a paradox in health outcomes tend to explain their results by referencing suggested explanations from the qualitative literature with little contextual analysis. The qualitative literature tends to report small contextual studies that do not include quantitative observations of outcomes. Use of both approaches, in the form of a mixed-method study, provides a fuller interpretation and understanding of the social epidemiology of birth outcomes through corroboration of findings using multiple methods and enhancement and clarification of results (Greene et at. 1989). In addition, existing birth 11

PAGE 29

outcome studies focus on the "lower weight" end of the spectrum of outcomes. No study was found that included birth outcomes at both the lowand the high-weight ends of the birth weight spectrum Thus, the use of mixed methods and the selection of birth outcomes relevant to the current clinical reality among the largest ethnic population in Colorado permit more nuanced conclusions and reveal potential connections that have been unintentionally masked by less comprehensive designs. The quantitative component of the study is a retrospective cohort design intended to determine the existence of any epidemiological paradoxes and to explore the validity of the healthy migrant and healthy immigrant hypotheses and the contribution of neighborhood deprivation and immigrant orientation to outcomes The qualitative portion consists of semi structured interviews with key informants and recent Mexican-born mothers and U.S.-born mothers of Mexican origin to explore the results of the quantitative analysis and to identify potential explanations for the outcomes. Based on the expected social gradient of health and census data, the ranking of social and health profiles should be: Mexican-born mothers poorest, U S -born mothers of Mexican descent-poor but not as poor as Mexican-born mothers, and White non-Hispanic mothers (wherever born)best. However both the healthy migrant and the healthy immigrant hypotheses suggest that the health profiles of Mexican-born immigrants should be better than U.S.-born women of Mexican descent. Multiple logistic regression is used to test the likelihood of each outcome by race/ethnicity among all Colorado mothers and by nativity for Mexican-born and U.S.-born mothers of Mexican origin (Aims 1 and 2}. By testing successive regression models adjusting for the main effect alone (race/ethnicity or nativity) (Model 1 ), adding demographic and social economic factors (Model 2), adding medical risk factors from the birth record (Model 3), and finally adding health behavior factors during pregnancy (smoking, consumption of alcohol, and weight gain) (Model 4), it is possible to compare births by race/ethnicity and then to explore the healthy migrant and immigrant hypotheses. The influence of neighborhood effects at the census tract level in Adams and Denver Counties is tested using general linear modeling. The final stage of this mixed method study includes qualitative interviews of five key informants and ten recent mothers of Mexican origin recruited from Salud Family Health Clinics, a provider of health services to the poor and underserved in Colorado. Five mothers were born in Mexico and five were born in the U.S. 12

PAGE 30

Description of the Study Population The Division of Vital Records in the Colorado Department of Public Health & Env i ronment provided a CD-ROM containing a de identified state census of all recorded singleton births, with flags for infant mortality from its linked death records, for the years 2000 2005. This dataset consists of 392 ,881 recorded births. After removal of cases for which any study variable was missing the dataset for Aim 1 consists of 356 389 births and the dataset for Aim 2 consists of 85, 755 births .. Table 1.5. Study population by race/ethnicity of mother-Aim 1 Study Population 356,389 White Hispanic Black Other Study Population by Race/Ethnicity 219 029 (61.46%) 106,291 (29 82%) 15,448 (4 33%) 15,621 (4 38%) Table 1.6. Study population by nativity of mothers of Mexican origin Aim 2 Study Population 85,755 Study Population by Nativity U.S.-Bom Mexican Origin 32,484 (37 88%) Mexican-Born 53,271 (62 12%) The study population for the contextual analysis in Aim 3 is reported in Table 1.3. Table 1.7. Study population of mothers of Mexican origin in Denver and Adams CountiesAim3 County Adams Denver Study Population 16,107 23,332 U.S.-Born Mothers Mexican Origin 5 733 (35 59%) 4,914 (21.06%) Mexican-Born Mothers 10,374 (64 41%) 18,418 (78 94%) Organization of the Dissertation Number of Census Tracts 79 100 This chapter introduces the issues, briefly reviews the rationale for this type of study states its specific aims describes the overall research methods, and enumerates the study population Chapter 2 reviews the literature of the epidemiological paradox especially as applied to Hispanics and immigrants in the U.S., and the various hypotheses and related theories that have been suggested to explain it. Chapter 3 describes the research design and methods for the quantitative and qualitative analyses. Chapter 4 discusses the quantitative analys i s and results. Chapter 5 discusses the qualitative analysis and results. Chapter 6 synthesizes the results and discusses the importance of the findings. 13

PAGE 31

CHAPTER 2 EPIDEMIOLOGICAL PARADOXES The Social Gradient of Health The biomedical tradition has historically approached health and disease from an individual perspective; it is the individual, after all, who gets sick. In this paradigm, treatment focuses on the individual and often emphasizes perceived unifying, biological causes of disease. The biomedical approach emphasizes personal agency in the form of theories of individual behavior change that assign the individual control of health promotion, such as smoking cessation, eating a healthier diet, and avoidance of unhealthy exposures. In contrast, public health approaches rely more heavily on structure, seeking causes of health and disease in conditions that affect a population, such as access to clean water, and pursuing health promotion in the form of enhancing social conditions that facilitate access to health benefits and eliminating health disparities (Hamlin & Sheard 1998). These different perspectives are important, because they inform the questions one asks about health outcomes and the potential approaches to improve them. One of the foundations of public health is belief in a general social gradient of health, in which there is a positive relationship of better health with increasing income/wealth, social status, and education (Lynch & Kaplan 2000:13-35; Cassel 1964). Epidemiological studies have identified the social gradient in various historical time periods, among different population groups, for multiple health conditions, and for both morbidity and mortality (Lynch & Kaplan 2000:13). Two well-known demonstrations of the social gradient are the Chadwick Study from the mid-1800s in the U.K. (Chadwick 1842) and the more recent Whitehall Studies showing a social gradient of coronary heart disease by U.K. occupational classification (Rose & Marmot 2001 ). Differences in health corresponding to socioeconomic position are thought to result from structural societal factors that lead to differential access to health resources and damaging exposures (Miech et at. 2006; Link & Phelan 1995; Marmot et a/ 1987; Syme & Berkman 1976). Socioeconomic position, also variously referred to as SES social class, and social status, is defined by Lynch & Kaplan (2000:14) as [t]he social and economic factors that influence what position(s) individuals and groups hold within the structure of society, i.e., what social and economic factors 14

PAGE 32

are the best indicators of location in the social structure that may have i nfluences on health. The concept of socioeconomic position is inf l uenced by three schools of thought: Marx s analysis of class based on relations of production; Weber's strat i f i cation of society by class, status and power leading to unequal distribution of resources; and the Functionalist stratification of society as a necessary outgrowth of the complex workings of modern society (Lynch & Kaplan 2000). Each of these formulations is based in the broader theoretical perspective of political economy which suggests that much of what happens i n the wor l d i s driven by access to resources or lack thereto The Social Gradient of Health at the Individual Level Belief in the soc i al gradient of health has led to identification of epidem i ologic exposures" that translate into measures of position and health risk factors at the level of the individual. Some soc i al epidemiologists look at socioeconomic posit i on as reflective of access to resources (Link & Phelan 1995) Thus income and wealth are each resources in themselves as well as i nstrumental vehicles for access to resources of other k i nds, i nclud i ng healthier food safer neighborhoods, and health care Education is viewed as instrumental for access to information and knowledge (such as the value of good nutrition, exercise or appropriate prenatal care) (Winkleby eta/. 1992) as well as correlated w i th h i gher paying and safer jobs and perhaps w i th employer assistance i n obtain i ng and paying for health insurance (although employer-provided health insurance is d i minishing in frequency i n the U.S.) (Economic Policy Institute 2006) Marr i age i s generally seen also as a veh i cle for access to more resources w i th the potential for a married pregnant woman to share more resources than she might have on her own In general having low income being unmarr i ed being unemployed and reach i ng lower educational levels are associated w i th h i gher rates of the adverse birth outcomes of interest in this study The Social Gradient of Health at the Area Level The basis of anthropological and sociological theory that s t rat i fication of soc i ety leads to differential material political symbolic psychosocial and behavioral effects ultimately reflects aspects of power and its distribution that affect not only individual characteristics but also contextual characteristics of a population based on the area in wh ich people live (Kawachi & Berkman 2003 ; Macintyre & Ellaway 2000 ; D i ez-Roux 1998) Contextual characteristics of neighborhoods have been assoc i ated w i th all-cause mortality, 15

PAGE 33

infectious disease, birth outcomes, obesity and lack of physical activity (Kawachi & Berkman 2003; Krieger eta/. 2003). Research identifying negative exposures and their accumulation and exacerbation has been especially useful for elucidating effects of poverty, discrimination residential segregation and the differential distribution of health-promoting resources in poor communities and among disadvantaged populations (Collins & Schulte 2003; Fang et a/. 1999; Collins & David 1990; Kleinman & Kessel 1987). Pathways by which area characteristics affect health are not always clear (Collins & Schulte 2003:228-230), although there are numerous hypotheses Positive relationships between undesirable area characteristics (such as high percentages of poverty unemployment or female heads of household) and adverse birth outcomes are generally thought to be related to neighborhood stressors (Gorman 2005) Stressors might include residence in a violent community (Zapata et a/. 1992) subjective psychophysiological reactions to a ne i ghborhood (Collins eta/. 1998) and racial or ethnic discrimination (Krieger 2000; Collins eta/. 2000; LaVeist 1989). Undesirable neighborhoods may also fail to provide convenient access to healthy food sources (larger grocery stores versus convenience stores and fast food outlets) safe and inexpensive public transportation safe places to walk and exercise employment, and health care related resources such as clinics hospitals and medical profess i onals. In general, neighborhood disadvantage increases the risk of LBW independent of the individual s fam i ly situation (Sellstrom & Bremberg 2006 ; Coll i ns & Schulte 2003 ; Robert 1999). But Finch et a/. (2007) demonstrate that after controlling for individual r i sk factors the degree of immigrant orientation in Los Angeles neighborhoods moderates the expected increase in LBW associated with neighborhood deprivation, especially for foreign born mothers Social support (resources available to indiv i duals and groups through connections in soc ial networks) i s sometimes suggested as a moderator of effects of individua l or a r ea characteristics Social support can be an indiv i dual attribute as well as an area or even a 'v i rtual characteristic.1 Social networks provide various aspects of support includ i ng information and adv i ce emotional support and access to material resources (Berkman & 1 Most studies focus on neighborhoods as real estate," and networks as defined and confined by physical space (Berkman & Clark 2003:288). But soc i al networks need not be geographic in today's social environment, where one may live in one neighborhood but f i nd soc ial support and networks through non-neighborhood work relat i onsh i ps church affiliations relat i onships w i th family located in another state or country and even Internet-based socia l networks ( dileonardo 1984 : 129-157) 16

PAGE 34

Clark 2003:289). One example of positive social support may be found in examining gay and lesbian networks that provide both safe places to get together and access to resources, especially in urban areas. In this same vein, it is possible that Mexican-born immigrants seek to live in neighborhoods with other recent immigrants from Mexico from whom they receive helpful social support (Finch et a/. 2007) or with whom they share cultural values, including language and practical solidarity that are protective of pregnancy Several studies also have noted that the density of ethnic populations and extended kin networks is associated with better health outcomes such as child health status in Mexico (Kana iaupuni et a/. 2005) and mental health in older Mexican Americans (Ostir eta/. 2003). Table 2.1, adapted from Lynch & Kaplan (2000:18-19), lists some measures of exposure that have been used in social epidemiological studies of health. Table 2.1. Measures of socioeconomic position Individual Level Measures Income Education Wealth Occupation Area Level Measures Occupational structure Educational structure Housing characteristics Poverty area Deprivation Population characteristics Access to resources Absolute income Income measured as % of official poverty level income Years of education Milestones of education, such as degrees Value of total assets Self-reported using population-specific scales of wealth Census classifications Score based on occupational prestige ratings Categorization based on managerial hierarchy of job % population in various occupational categories in area % population at various educational levels in area Age of housing, density of inhabitants per room, access to facilities, segregation % households below poverty-level income Unemployment, car/home ownership, overcrowding, crime % population subgroups {racial/ethnic) in area Resources such as hospitals, clinics, parks, private services Unexpected Deviances from the Social Gradient of HealthThe Paradox In the United States, the expected social gradient of health reflects a kind of "majority minority'' population axis, along which poorer, less educated populations, which are also usually made up of racial or ethnic minorities, are assumed to (and often do) have poorer health outcomes (Collins & Schulte 2003:225). Given studies that associate lower income, lower educational attainment, lower social position, and residential segregation by income 17

PAGE 35

with poorer health outcomes (see Lynch & Kaplan 2000 for a general review), researchers note with interest when such patterns do not hold and when some groups appear to have better outcomes than would be predicted based on socioeconomic position outcomes that approach the reference (usually "majority'') group (Chung eta/. 2003; Buekens eta/. 2000; Hummer eta/. 1999; Cobas eta/. 1996; Markides & Coreil 1986). It is useful to clarify the nature of this apparent paradox. It is not that a socioeconomically disadvantaged population necessarily has better health outcomes compared to the majority more advantaged population (although some studies have found such results). It is rather that some key health outcomes are more like those of the majority population than those of other disadvantaged populations, when their socioeconomic positions would suggest that their outcomes should be more similar (Hummer eta/. 2007), which then raises the question of what accounts for the paradox Although evidence of an epidemiological paradox may have been i dentified as early as 1974 for a Hispanic population in the U.S. (Teller & Clyburn), most researchers credit Markides and Coreil with coining the term "Hispanic epidemiological paradox" in 1986. They reviewed studies of the health status of Hispanics from the southwestern United States that were conducted in the 1960s and 1970s and concluded that Hispanics enjoyed some health outcomes that were similar to non-Hispanic Whites in categories such as infant mortality, life expectancy, mortality from cardiovascular disease, mortality from major types of cancer, and measures of functional health even while their socioeconomic position was more like that of most Blacks who, as a group, had worse health outcomes than Whites. But for indicators such as incidence of diabetes and infectious and parasitic diseases, their review showed that Hispanics had poorer health outcomes than Whites. Since then there has been an explosion of work on the paradox in general and with respect to specific outcomes, subpopulations, and hypotheses to try to explain any deviance from the expected social gradient. Some studies have used national population databases (Hummer eta/. 1999; Singh & Yu 1996); others have examined locations with large numbers of immigrants from Puerto Rico (New York City) (Rosenberg eta/. 2005) and from Mexico and Central America (California) (Keleher & Jessop 2002; Fuentes-Afflick et a/. 1999; Zambrana et a/ 1997) An important result of these studies has been recognition of the heterogeneity of people classified as "Hispanic." Rates of LBW vary depending on the country of origin, with Cubans and Mexicans usually having lower rates of LBW than Hispanic mothers from other countries (Hummer eta/. 1999; Reichman & Kenney 1998). 18

PAGE 36

Even more interesting, the paradox has also been shown among some first generat ion (foreign-born) immigrants who have better outcomes than immigrants from the same country in their second and later generations in the U S The advantage of foreign birth seems to decay rather rapidly with the time spent in the United States (Madan eta/. 2006) These findings have led to a plethora of studies on i mmigration assimilation and acculturation (Acevedo-Garcia eta/. 2005; Kasirye eta/. 2005; Rosenberg eta/. 2005 ; Cho et a/. 2004; Gould eta/. 2003; Weigers & Sherraden 2001 ; Fuentes Afflick eta/. 1999; Hummer eta/. 1999 ; Lansdale eta/. 1999 ; Cobras eta/. 1996; Singh & Yu 1996}. Seve r al of the studies reviewed by Markides and Coreil i n 1986 are somewhat limited. Many looked only at individual-level characteristics and early studies were constrained by ambiguous ethn i c designations Until recently the U.S. Census i dentified H i span i c ethnic ity using Spanish surname rather than self identificat i on as Hispanic/Latina or various sub-classifications with i n H i spanic. Vital records contained a paucity of i nformat i on (a problem that still exists although to a lesser extent). Also many of the early studies used unadjusted incidence rates of health outcomes or relied on conven i ence samples Since then more researchers routinely i nclude adjusted odds ratios and use census databases to avoid the need for sampling. Even today however many studies of the paradox focus only on the population s compositional attributes by including only individual-level characteristics, such as race / ethn i city gender age marital status health conditions and health behav i ors ( Rosenberg et a/ 2005 ; Hummer et a/. 1999; Singh & Yu 1996) and most do not report i nteractions among var i ab l es used i n the analysis Fewer studies approach the problem from the perspective of area characteristics seeking to explore whether a population s contextual characteristics i nf l uence health outcomes (Pickett et a/ 2005). Population studies of birth outcomes address ing Hispanics and in some i nstances fore i gn-born mothers from 1990 2007 are described i n Append i x A. Few of these studies i nclude SGA as an outcome none i ncludes LGA and none addresses Colorado mothers Trad i tional Explanations of the Hispan i c Immigrant Paradox Several hypotheses have been proposed to explain unexpectedly better birth outcomes of Mexican-born mothers and U .S.born women of Mexican origin. Among the most common explanations of these unexpected results are: ( 1} select i ve m i gration by healthy Mex i cans [healthy (Cho eta/. 2004 ; Marmot et a/. 1984) 19

PAGE 37

(2) healthier diet and behaviors, especially lower rates of smoking and alcohol consumption among Hispanics generally, and especially among first-generation immigrants [healthy (Flores & Brotanek 2005; Kasirye eta/. 2005; Abraido eta/. 1999; Guendelman eta/. 1999), (3) acculturation/assimilation as a way of explaining the decline in better outcomes over time as acculturation increases, women of Mexican origin increase their rates of smoking and drinking and consume a less healthy diet, and (4) social support among Hispanics during pregnancy (Landsale et a/ 1999 ; Rumbaut & Weeks 1996). Healthy Migrant Hypothesis The healthy migrant hypothesis may include elements from both ends of the agency -structure axis. Although migration to the United States from Mexico is motivated primarily by Mexico's poor economy, which is a structural rationale, self-selection for migration may reflect a form of agency. A number of researchers suggest that those who choose to migrate will have experienced better childhood health than those who remain in Mexico (Crimmins et a/ 2005) and that better health outcomes among first generation immigrants are a result of the migration of Mexicans who are healthier than those who stay behind (Cho eta/. 2004; Abraido-Lanza eta/. 1999; Marmot eta/. 1984). At first blush this hypothesis should be easy to test: compare rates of any given health outcome among Mexican immigrants to the U.S. with rates for those who stay in Mexico. For example, UNICEF estimates that the rate of LBW in Mexico in 2000 was 9% (UNICEF 2004). The rate of LBW in Mexico is not that different from the overall rate of 8.1% for all Hispanics in Colorado in 2000 (see Table 1.2). However, some researchers argue that one cannot compare rates of an outcome at this level without taking into account the context of each population, such as comparisons of health behaviors, risk profiles, social support, and other psychosocial effects (Finch, B., personal interview February 27, 2007). Crude rates of outcomes do not account for individual and contextual effects and therefore one cannot immediately assume that any differences are or are not accounted for by the healthy migrant hypothesis. A direct test of this hypothesis, therefore, requires better birth outcome data on mothers living in Mexico than currently exist. Even when access to roughly comparable data is available the results may not support the logical conclusion of healthy migrant selection. Crimmins et a/. were able to compare data from the U.S. National Health and Nutrition Examination Survey and similar 20

PAGE 38

data from the Mexican Health and Aging Survey to compare the height of Mex ican immigrants to the U.S. with Mexicans who remained in their communities of origin. They found that immigrants are i ndeed selected for greater height after adjusting for socioeconomic status but that same group of immigrants was shorter than people of Mexican orig i n who were born in the U.S. (2005). In another study of households in Mexico, Frank and Hummer analyzed data from the 1997 Encuesta Nacional de Ia Dinam i ca Demografica [ENADID] (2002). They found that membership in a m i granf household (one in which at least one i ndiv i dual in the household took a m i gratory trip to the U S in the five years prior to the survey) i s protect i ve for LBW of an infant born in that household in Mexico after the migratory trip The authors suggest that the protective factor of migration so defined was probably largely based on rece i pt of remi ttances from the migratory household member This finding supports the general social grad i ent hypothesis-that access to more financ i al resources would be associated w ith better LBW rates However Frank and Hummer also split migrant households into two categor ies based on whether or not they rece i ved monetary remi ttances and found that i nfants born i nto non-remittance migrant households i n Mex i co had a s i gnificantly lower r i sk of LBW compared with those born into remittance migrant households The authors argue that some factor like social remittances" may be associated with the better health outcomes of migrant families but they have found no way to conceptualize or measure i t using the ENADID data. Even more interesting i s their f i nd ing of a m i n i -epidemiolog ical paradox w ithi n Mex i co. Women i n migrant households had riskier househo l d socioeconomic prof i les but their i nfants had more favorable birth outcomes compared with infants in non-migrant households (2002:755). A number of studies have shown statistically significant results that foreign-born Hispanic women have better low wei ght-related birth outcomes such as LBW and preterm b irth (Wingate eta/. 2006 ; Acevedo Garcia eta/. 2005 ; Gould et a/ 2003 ; Hummer eta/. 1999; Lansdale eta/. 1999). But Rosenberg eta/. found that after adj usting for r i sk factors the results did not support the healthy m i grant hypothes i s (2005) They concluded that the d i fference in outcomes reflected d i fferent i al distribut ion of risk not healthier' migrants. Healthy Immigrant Hypothesis The healthy i mmigrant exp l anation finds support in studies show ing that the soc i oeconomic r i sk profile of fore i gn-born i mmigrants i s poor but that their health behaviors are good immediately after they arr i ve in the United States (that i s they smoke and drink litt l e and eat a health i er diet). This hypothesis sounds in agency" because i t i s based on health 21

PAGE 39

behaviors, and suggests that immigrants follow their home cultural beliefs, which are thought to include healthier practices than those associated generally with life in the U.S It does appear that female immigrants from Mexico as a group engage in fewer deleterious health behaviors such as smoking and drinking of alcohol when compared with non-Hispanic Whites (Meneses-Gonzalez et a/. 2006; Abrafdo-Lanza et a/. 2005) Some studies have found that Mexican-born women have healthier dietary intakes (Guendelman & Abrams 1995). These findings might support the positive outcomes noted in the first generation, which seem to decay rapidly with time spent in the U.S. (Montez & Eschbach 2008; Kasirye et a/. 2005; Neuhauser eta/. 2004). Kaplan et a/. (2004) demonstrated a dose-response relationship between obesity and length of residence in the U .S. Adjusting for multiple factors, those who lived in the U.S for 15 or more years had an odds ratio of 4.0 for obesity compared to those who lived in the U.S for fewer than 5 years. Reduction in the odds of better birth outcomes is suggested to be a result of acculturation processes such as adoption of unhealthy diet and sedentary lifestyle (Mainous et a/. 2008; Abrafdo-Lanza et a/. 2005) Barcenas et a/ (2007), however, found that length of residence in the U S is associated with risk of obesity especially in Mexican-American women, but that degree of acculturation was not a predictor. Acculturation/Assimilation For several reasons, defining the concept of acculturation/assimilation" is complex and elusive. Acculturation and assimilation are sometimes conflated or parsed very finely to differentiate between the two concepts. Historically anthropologists and sociologists have used the concept of assimilation to address structural locations of immigrants as they move toward interacting with the more dominant social institutions of their new communities whereas acculturation addresses changes in beliefs, values and practices at the individual level (Wingate & Alexander 2006; Weigers & Sherraden 2001 ) Regardless of definition the idea that either process operates in a linear fashion does injustice to their complexity, especially with regard to Mexican immigrants. While in general Mexican-born immigrants sometimes experience better birth outcomes than later generations of Mexican descent born in the U.S., generational status alone is at best a gross measure of acculturation, which is more complex than just amount of time spent in the U .S. (Amaro & de Ia Torre 2002; Guendelman et a/. 1990). Moreover, acculturation itself is affected by exogenous factors such as the degree of, and need for, acculturation as well as the starting point of the immigrant when she arrives (Weigers & Sherraden 2001 ). Unlike immigration from many other countries, immigration from Mexico i s not necessarily an "unknown" because there are 22

PAGE 40

histor ical and geographic links between Mexico and U.S. along their long shared border Indeed land that is now part of the U S was once the northern part of Mexico. In addit ion, there i s substantial two-way immigration so that immigrat ion from Mexico cannot necessarily be seen as a permanent status (Weigers & Sherraden 2001 ) According to Portes et a/. (1985) education age poverty index and size of commun i ty have important effects on acculturation. Even when researchers clearly define assimilation or acculturation its operationalization is fraught with measurement d i fficulties. E i ther acculturation or assimilat ion as a process i s like a black box that masks transition from one status to another and i t is unclear which d i mens ions or factors should be included and which are more i mportant (Berry 1997) Acculturat ion may be a marker for values beli efs and lifestyle ( i nc l uding smoking and par i ty) (Scribner & Dwyer 1989) Some studies operationalize acculturation using the length of time immigrants have been in the country (Harley eta/. 2007) others look at the age at which immigrants came to the U.S. (Kasirye eta/. 2005) the language spoken (Ebin eta/. 2000) or other measures Cobas et at. (1996) crit i c ize "an" acculturat ion hypothes i s because whatever the concept i ncludes they say i t is multidimens i onal. They used structural equation modeling to address the conceptual and methodological weaknesses i n acculturation theory and found that not all components of acculturation had the same effect on LBW. They found that language spoken was a stronger predi ctor of smoking and LBW than ethn i c ident i ty and that while acculturation i nfluences LBW through d i et and smok ing, a significant direct effect of acculturation still persists. Therefore, more acculturated mothers were more likely to deliver LBW children partly because they smoked. They also found that d i etary i ntake was an i nterven ing variable between accu l turation and LBW status w ith accu l turat ion hav ing a negative effect on dietary i ntake and diet affect ing LBW. The mechanisms that account for relationsh ips between acculturation and health behav i ors remain elusive (Abra f do-Lanza eta/. 2005). In addition t o the d iffi culty of operationalizing acculturation however i t i s conceptualized there i s a dearth of population-w i de i ndividual-level data on accu l turat ion. B irth and death records do not record the length of t ime lived in the U .S., age at immi grat ion, date of immigration or primary language. Instead these data must be obtained from surveys that sample a population, usually one that i s quite small. 23

PAGE 41

Social Support Some researchers argue that studies of acculturation focus too much on negative health behaviors (e.g smoking and diet) and not enough on more general social support as a route to resources that improve outcomes (Weigers & Sherraden 2001 ). Guendelman eta/. agree and further suggest that social support variables may not correspond with degree of acculturation (1990) For example, protective factors for birth outcomes include strong cultural support for maternity, healthy diet, and marianismo {devotion to the maternal role) (McGlade eta/. 2004), which may exist for women regardless of their degree of acculturation. Social support may be facilitated by living in an area with a high concentration of culturally similar residents, which may provide better access to culturally appropriate prenatal care and other support during pregnancy that lead to better birth outcomes, either through social support pathways or through commonly-held cultural beliefs and practices. Finch eta/. (2007) found that foreign-born Hispanics living in Los Angeles had better birth outcomes than U.S. born-Hispanics, even when they lived in very deprived neighborhoods, so long as those neighborhoods had high concentrations of foreign-born residents. Another pathway to better birth outcomes may be the existence of protective cultural factors that may influence perceptions of resilience and risk and thus affect health outcomes. Bender and Castro conducted a qualitative study of Mexicans in North Carolina, testing the concept of resilience (universal capacity to prevent, minimize, or overcome damaging effects of adversity), as a "positive counterpart to 'risk factors"' (2000:156; see also Kawachi & Subramanian 2006). Bender and Castro found that protective factors may not directly promote positive health outcomes, but instead may buffer or increase resistance to negative events, through compensation (counteracting stressful/adverse with external characteristics or external sources of support); challenge (stressful events in prior life can enhance competence); and immunization (protective factors temper the impact of stressors, but may not be evident absent stress). Among other findings, they identified as a positive theme the strength of social support enjoyed by the immigrants, who had strong nuclear and extended families, especially during pregnancy These immigrants had larger numbers of friends and kin in North Carolina than the authors anticipated. Even small families had 8-10 members of extended family who would gather on weekends in the park for recreation and eating, as they did in their Mexican pueblos. Furthermore, pregnant women in the U .S. maintained ties to their mamas who were still in Mexico by relying on telephone calls. 24

PAGE 42

Cultural values may have negative effects as well. Culture shapes one's view of weight. Outside of the industrialized West, being large and round indicates positive health, and Hispanics are more likely than Whites to perceive overweight as closer to their ideal body size (Candib 2007). Ahluwalia et a/. find that Mexican American women are less likely to perceive that they are overweight (2007) and that less acculturation is associated with higher BMI among Mexican immigrants (Akresh 2007). Finally some researchers argue that the paradox is not real but is merely a statistical artifact (Palloni & Arias 2004). This particular view has gained adherents especially with respect to mortality and life expectancy studies. Palloni and Arias argue that mortality statistics of foreign-born immigrants are particularly suspect, because foreign-born individuals are especially likely to return home when they age or get sick, or to die. This "salmon effecf' is argued to account for any paradox in mortality statistics, especially among Mexican immigrants. Morales eta/. (2002) disagree, as do Hummer eta/. (2007) who report on a study that not only measures infant mortality among Mexican-born mothers, but also calculates how many infants (far too many to be plausible) would have to move back to Mexico to account for the paradox. The salmon effect cannot apply to weight-related birth outcomes, as the outcome is a measure of birth weight, which exists in the record An Alternative TheoryPolitical Economy Studies of health outcomes are situated in the broader context of theories of disease causation and health disparities. Disease is proximately caused by biomedical factors such as genetic inheritance and/or exposure to agents that affect the healthy working of the body. In regard to genetics, there is evidence that genetics plays a role in rates of diabetes among Mexicans. Those with more indigenous heritage have higher rates of obesity and diabetes (Martorell 2005). It is also interesting that immigrants from Mexico are not as exogamous as might be expected. In the 1990s Hispanic intermarriage rates with non-Hispanic Whites declined substantially (Lichter eta/. 2007) Based on these observations, one might expect that rates of diabetes and obesity and perhaps LGA would be comparable among Mexican born and U.S.-born mothers of Mexican origin. Intermediate and distal causes of disease include poverty, lack of education, poor nutrition exposure to environmental insults, or lack of access to health care, all of which increase susceptibility to disease (Link & Phelan 1995; Millard 1994) and which fit within the ideas of the social gradient of health and political economy. Therefore a number of theorists suggest that models of disease causation and their opposite, models of disease prevention 25

PAGE 43

must address macro-level processes (often economic) that affect indi viduals and their choices as well as individual health risks (Diez-Roux 1998) Studies that try to correlate a degree of assimilation or acculturation w ith diet other health behaviors and health outcomes necessarily focus on measures of assimilation and acculturation. While it is indeed likely that the social and cultural transitions after immigration to the U.S. affect d i et health behaviors and health outcomes stepping back to look at a larger frame of reference within which trans i tions occur may be more explanatory especially if the outcomes are not uniformly positive A succinct def i nition of political economy is elus ive, because i t is developed from a mix of i ntellectual and political movements (Roseberry 1988). It has i ts roots in Weber and Marx and at a m i n imum consists of a set of theories that center on Western economic and political domination of less developed societies and the large-scale effects of capitalism on specific countries regi ons, or communit ies, approached from an histor i cal perspect i ve (Roseberry 1988). A political economy approach i s fundamentally structural positing that indi vidual choices are framed and constrained by pervasive political and economic forces far from the individual. Candib (2007) diagrams pathways to obesity and diabetes, situated in a political economy perspective that appears applicable to Mexican m i grants to the U.S Figure 1 Pathways to obesity and diabetes G loba l wnion Enwlronment Povmy. Global ttadt policy High rost of produce Childhood obesity rnhhlnizDtion. poducn chtap bts low ll(aS.S lO good food mass media ond su911rs from ...... Sedenll!ry enwotalnment ...... Epidemic obesity Cl)rn ond K1'f No safe plaees 10 play -! @]-Sdlools without gyms Vascular @ physicai!IC!lvity Thlrd World Fostfood out Social and family of of ness 1: me n lngs of high or low Might low ilnd hig!Kalorl e fellll erwironmenu; physiolcglclll I Urb;,nization I and psycholcgklll effects ond immlgratlon Genetic Cl)fllt)Ct Figure 2 .1. PATHWAYS TO OBESITY AND DIABETES (Candib 2007 : 548) Mex i co i s a middle-income country that is affected by global i zat ion. It i s rapi d l y developing economically (with a capitalist system) and has been urbaniz ing s i nce the 1960s w ith people moving from villages to c i t ies (Martorell 2005) As a result Mexicans have access to processed foods and foods conta i ning refined sugars Refrescos (sodas) are ubi quitously 26

PAGE 44

consumed, even outside urban areas of Mexico (Jimenez-Cruz & Bacardi-Gascon 2004). According to Martorell, "Mexico is a country far along the nutrition transition" (2005:2). This nutritional transition within Mexico has led to higher rates of obesity and type II diabetes, with 58% of women of reproductive age either overweight or obese (Jimenez-Cruz & Bacardi Gascon 2004). In addition, there is evidence that malnutrition early in life followed by catch up growth among poor Mexicans (the same pattern seen among LBW and some SGA babies) is a high risk factor for obesity, diabetes, and cardiovascular disease in adulthood (Martorell 2005; Diamond 2003). Moreover, changes in Mexico s labor economy continue to create motivation to leave the country for the U.S. when people cannot find work in Mexico's cities or when they see better economic opportunities than exist in their villages (Partes & Bach 1985). Once in the U.S., immigrants recognize the inverse relationship between the cost of food and its energy density, with foods high in sugar and fats being the least expensive and the most energy dense (Carnethon 2008). Most acculturation studies try to define the construct and suggest that it operates over a generation or more to produce changes in diet and health behaviors such as smoking and weight gain. These changes are associated with poorer LBW outcomes in the second and third generations in the U.S. Yet it may be that an earlier, more personalized, nutritional transition occurs almost immediately upon immigration. This transition may explain lower rates of LBW and low weight outcomes and at the same time tend to increase LGA. When Mexican immigrants arrive in the U.S. they typically are poor. High density foods are cheaper and often more readily available; access to tropical fruits of the quality they are used to in Mexico is less (Yeh eta/. 2008; Drewnowski & Specter 2004). In addition ''fast food" is often viewed as high status in their native country its low cost and high availability in the U.S. is attractive to immigrants. Himmelgreen eta/. (2007) found that changes in the diet that result in weight gain, less exercise, and eating fast food are occurring as early as two and a half years after immigration to the U.S Moreover, they found that the proposed positive buffering of high immigrant density was not supported instead social isolation was the norm post immigration. Associating positive health outcomes with various facets of individual behavior and even neighborhood contextual factors likely represents oversimplification of a complex reality. All else being equal, regularly consuming a diet low in refined sugars and fats, living a lifestyle that includes physical exercise and less sedentary activity, refraining from smoking, 27

PAGE 45

and engaging in social and cultural activities that increase social support and cohesion result in better health outcomes generally, and better birth outcomes specifically. But centuries of social science research demonstrate that these behaviors are not completely voluntary. All of us live in larger contexts that affect both individual health behaviors and neighborhood contexts. Residents of Mexico and the U.S. live in a world economy that affects social conditions that in turn affect individual agency. The most common attempts at explaining any epidemiological paradox based on race, ethnicity, or immigrant status miss the mark because they are incomplete. More inclusive theories, such as political economy, need to be employed to place context in a much larger frame. These understandings may then allow more specific health behavior theories and cultural beliefs to inform interventions at multiple levels to improve birth outcomes. 28

PAGE 46

CHAPTER 3 RESEARCH DESIGN AND METHODS This study has two particular strengths compared with existing population studies of birth outcomes : the inclusion of a broader range of birth outcomes, with differentiation of high b irth weight from low birth weight outcomes; and the use of mixed methods. Prior studies tend to focus on low birth weight (sometimes with the addition of very low birth weight) and premature birth. Few include small for gestational age as an outcome. More importantly, population studies have focused on the low weight end of the spectrum of birth outcomes and ignored other weight-related outcomes. Broadening the scope of outcomes allows a stronger test of the hypotheses proposed to underlie population differences. In addition the use of mixed methods allows for the collection of qualitative data to explore the why" of differences that is not captured in the government-gathered official birth records. This study addresses four specific aims: to test the expected social gradient of health on birth outcomes by race!ethnicity and to identify any paradoxical outcomes (Aim 1 ); to test the expected social gradient of health on birth outcomes by nativity of Hispanic mothers of Mexican origin and to explore the healthy migrant and healthy immigrant hypotheses as explanations for any paradoxically better birth outcomes experienced by Mexican-born mothers (Aim 2); to examine the association of neighborhood deprivation and immigrant orientation on birth outcomes of mothers of Mexican origin in two large Colorado counties (Aim 3); and to explore potential reasons for differential outcomes among Mexican-born mothers and U.S.-born mothers of Mexican origin (Aim 4) In brief, risk profiles are compared for a cohort of Colorado mothers and odds ratios of birth outcomes are calculated for two statewide populations: all mothers based on their race/ethnicity and mothers of Mexican-origin based on the i r place of nativity (Mexico or U S.) using individual-level compositional data from the state s official birth records. Separately contextual neighborhood-level effects are examined to test whether neighborhood deprivation 29

PAGE 47

or immigrant orientation affects birth outcomes of mothers of Mexican origin in Adams and Denver Counties using data from the 2000 Decennial Census. All quantitative analysis is performed using SAS v. 9.1.3 (2003). Qualitative interviews with five key i nformants and ten recent mothers of Mexican origin provide insight into the quantitative results. Nested Mixed Method Design Mixed method designs include quantitative and qualitative approaches to a research question to minimize weaknesses and maximize strengths of each of the research paradigms (Medlinger & Cwikel 2008; Morse & Field 1995:164). In general, qualitative methods are used to explore problems about which relatively less is known, whereas quantitative methods are used to test theories and hypotheses (Medlinger & Cwikel) In reality, neither paradigm alone is usually sufficient to answer complex questions that seek to uncover both outcomes and reasons underlying them (Korbin 2008). A review of 57 mixed method designs by Greene at a/. (1989) identified five purposes for using this approach: triangulation (corroboration of findings using different methods), complementarity (enhancing or clarifying results), development (using results from one method sequentially to inform and develop the next stage), initiation (developing new perspectives by highlighting conflicting findings), and expansion. Mixed methods can be employed in at least three ways: sequentially (each is a distinct study), in a nested design (one paradigm is paramount and the other is used as a supporting method of analysis), or in a fully integrated design (both paradigms are used concurrently) (Medlinger & Cwikel 2008). This study originally was proposed as a mixed method design using a development approach because it was expected that qualitative data would be needed to identify appropriate variables for inclusion in the quantitative analysis. However, based on the availability of variables in the birth record and prior literature, it was decided to perform the quantitative analysis first and use qualitative methods to complement and corroborate the quantitative results. The choice of a mixed method design became particularly useful to explore differential odds of LGA among mothers of Mexican origin in Aim 2 As a result this study is a nested design, with the quantitative component dominant, using qualitative methods to complement the quantitative findings. Figure 3.1 depicts the nested design and shows that both methods inform each of the specific aims. 30

PAGE 48

Quantitative Methods Individual Level Aim 1 Race/Ethnicity Aim 2 Nativity Area Level Aim 3 Nativity Qualitative Figure 3.1. SCHEMATIC OF RESEARCH DESIGN Quant i tative Research Component The quantitative component of the study is a retrospective cohort design. For Aims 1 and 2 all singleton births recorded in Colorado s Vital Records for the years 2000 through 2005 are exam i ned. For the contextual analysis of Aim 3 data on singleton births to women of Mexican orig i n res i ding in Adams and Denver counties are linked with census tract data from the 2000 U.S Decennial Census The original study des ign proposed evaluating four birth outcomes : low birth weight preterm birth, small for gestational age, and i nfant mortality, all of which have several i nterrelating r i sk factors that focus on smaller size of the infant. Infant mortal i ty was dropped as an outcome based on its very low i ncidence in Colorado compared to the other study outcomes and the questionable quality of cases reported as l ive births at very early gestat ion i n the v i tal records dataset. In one of the earliest reviews of the apparent H i spanic paradox," Markides & Core i l (1986) noted that H i spanics had lower incidences of some specific health outcomes such as infant mortality and breast cancer but were not advantaged compared to the nonH i spanic Whi te population for other conditions such as stomach cancer and diabetes. Delivering an infant that is large for gestational age i s assoc i ated with maternal diabetes and a l so contributes to the infant s immediate and life course health risks such as diabetes chronic 31

PAGE 49

cardiovascular disease and stroke, in ways that are sim i lar to the life course effects of LBW preterm birth and SGA (Martorell 2005). LGA was therefore added to the study to broaden the scope of weight-related birth outcomes Addition of LGA adds depth to the explorat ion of adverse birth outcomes because, although LBW is an i nternational measure of population hea lth, birth weight of >2500 grams is a somewhat simplistic measure for a favorable birth outcome given the complexity of the relationship of birth weight and both acute and chron i c morbidities (Lubchenco 1971, 1972). In light of trends to overwe i ght obesity and type II d i abetes in the U.S. especially among Hispan i cs (Mainous et a/ 2008 ; Carnethon 2008 ; Martorell 2005), it i s appropr i ate to include a more precise classificat ion of weight-related birth outcomes that better captures health risk. Levels of Analysis Many stud i es examine the r elationsh i p between i ndividuall evel compositional characteristics and health outcomes or between area-level contextual characteristics and health outcomes. Results of studies at one level are sometimes misused to make statements about a different level creating fallac i es in logic The ecological fallacy occurs when one draws i nferences at the individual level based on area level data ; the atomist i c fallacy draws i nferences at the group level based on i ndividual level data. To avoid these valid criticisms researchers doing single-level studies use data from and make inferences at the same level. But single level studies alone even when they avoid fallac i es of cross-level interpretation may miss the effect of mechanisms operating at multiple levels (Green l and 2002 ; Diez Roux 1998). Contextual factors may constrain or enhance choices that individuals can make that affect the i r health (Diez-Rouz 1998) In a number of multilevel studies i nclud ing four stud i es on LBW ne i ghborhood socioeconom i c status and soc ial cli mate were shown to have a small ( 10%) effect on child health outcomes (Sellstrom & Bremberg 2006) (see also Krieger eta/. 2003). Contextual variables are usually classified in one of two ways. Derived var i ables sometimes referred to as aggregate variables summarize the character i stics of the indi v i duals in the group (Kawachi & Subramian 2006 ; D i ez-Roux 1998) Examples include the percentage of the population in an area/neighborhood that is below a defined poverty level or the percentage of adult males i n the area that i s unemployed This type of variable i s used to descr i be group properties that are more than the sum of the indi v i dual character i st i cs For example Kr i eger eta/. (2003) demonstrated that a higher risk of mu l t i ple adverse b irth outcomes i n poorer neighborhoods i s at least partly i ndependent of large differences in 32

PAGE 50

maternal education at the individual level in urban areas of Canada. The second type of contextual variable called an integral variable descr ibes a characteristic of an area that is not related to the aggregate characteristics of the individuals. Examples include the number of health clinics or the type of regulation such as zoning, in an area If variation i s sufficient among areas, multilevel analysis can i dentify contextual predictors of outcomes using both individual and neighborhood var i ables (Kawachi & Subramanian 2006) It is an appropriate methodology when observations are clustered, exposures operate simultaneously at more than one l evel or heterogeneity in exposure exists (Raudenbush & Bryk 2002). Various rather complex statistical methods permi t s i multaneous estimation of variability on the outcome at both levels and correction for any intraclass correlation at the area level (Sellstrom & Bremberg 2006). Some studies have linked i ndividual level birth outcomes with area effects (Finch et a/ 2007 ; Gorman 2005; Pickett eta/. 2005; O Campo et a/.1997). Most multilevel studies i n the U S examine particularly poor health outcomes of Blacks to try to i dentify how much place of residence may contribute to their poor outcomes (Pickett eta/. 2005 ; O'Campo eta/. 1997). Fewer studies have applied area level analys i s to study Hispan ics or why they might enjoy better outcomes than would be expected based on their socioeconomic position (but see F inch eta/. 2007). In lieu of undertaking the complexity of simultaneous estimation of the contribution of individual and contextual level factors to the odds of the four birth outcomes this study examines i ndividual level character i stics i n Aims 1 and 2 and area level characteristics in Aim 3 and interprets results at each l evel. Sample Study Data and Variables The CDPHE prov i ded an h i storical observational dataset cons isting of a census of de-identified records of all singleton live births occurring in Colorado for the years 2000 t hrough 2005 on a CD-ROM in SAS format pursuant to a written data agreement and after approval of the study by the Human Subjects Research Committee of the Univers i ty of Colorado Denver as exempt under 45 CFR 46.101 (b)(4) and later as nonhuman subj ect research (Appendix 8). Although it i s possible that some births are not recorded in the vital records system CDPHE states that its records account for over 99% of all b irths occurring in Colorado (Bol K. personal interview June 10, 2007) The sample then i s a census and need not be randomized (Stokes Davis & Koch 2000:7 -8). The dataset provides information on each b irth by year but omits the names of the i nfant and parents the specific day and month of b irth, and the date of mother s birth to 33

PAGE 51

preserve anonymity. The birth record includes demographic information and limited data on socioeconomic position of the parents information on the weight and gestation of the infant, medical conditions associated with the pregnancy and delivery, and data reflecting certain maternal health behaviors. The census tract of the mother's residence and the specific altitude in feet above sea level of her address i mmediately before the birth are also associated with the birth record. CDPHE calculates and reports the adequacy of prenatal care using the Adequacy of Prenatal Care Utilization Index (APNCU), also known as the Kotelchuck Index, and flags births of infants weighing less than 2500 grams. Dependent Variables Four birth outcomes are the dependent variables for this study: LBW preterm birth, SGA, and LGA. LBW is defined as an infant weighing less than 2500 grams at birth (approximately 5 5 pounds) (WHO 2006 : 36) Although CDPHE provided a flag for births <2500 grams, CDPHE's identification of LBW births was tested by writing SAS code to identify each birth <2500 grams and comparing those results to the flags provided by CDPHE. There were no differences between the two methods of identifying LBW infants. The Institute of Medicine s recent report on preterm birth considers a birth to be preterm if it occurs before 37 completed weeks of gestation (10M 2006) Each birth was coded as preterm or not based on the clinical estimate of gestation (in completed weeks) reported in the birth record. SGA is generally defined as birth weight of an infant below the 1oth percentile of weight for gestational age (Kiiegman & Das 2002; Alexander eta/. 1999). LGA is defined as birth weight of an infant above the 90th percentile of weight for gestational age (Alexander eta/. 1999). There is no single national standard to determ i ne percentiles of birth weight for gestation. This study uses the percentiles and associated birth weights, in grams, developed by Alexander et a/. (1999), which were based on a sample of 9 6 million births to mothers living in the U.S. contained in the 1994 1996 U.S. Natality Files from the National Center of Health Statistics The sample was racially/ethnically diverse, with approximately 61% non Hispanic Whites 18% Hispanics 15% Blacks, and the balance representing other racial/ethnic groups. Alexander's team calculated a percentile table for the total sample and separate tables for various racial/ethnic categories, including Hispanics. Using these tables, the 1oth and 90th percentiles were calculated for each gestational age in weeks. Births below the 1oth percentile by gestation were coded as SGA and those above the 90th percentile were coded as LGA. In addition, births that were neither SGA nor LGA were coded as appropriate for gestational age (AGA). For Aim 1 (births by race/ethnicity) percentile weights based on 34

PAGE 52

Alexander's entire, national sample population are used. For Aim 2 ( births to mothers o f Mexican orig i n categorized by mother's nativity) Alexander's percentile table for Hispanics is used. Independent Variables for Aims 1 and 2 The individual-level variables initially considered for inclusion in the analysis of Aims 1 and 2 are shown in Table 3.1. Also reported are the min imum and maximum values in the dataset and notes on their coding. CDPHE provided mother's age, previous pregnancies altitude of residence, years of education weight gain dur ing pregnancy, c i garettes smoked per day and alcoholic drinks consumed per week during pregnancy birth weight in grams and clinical estimate of gestation as continuous variables. As described below, continuous variables were coded into categorical levels. For any categorical variable with more than two values, multiple levels were created. Table 3.1. Individual-level variables for Aims 1 and 2 Variable Values Operationalization Year of infant birth 2000-2005 All 6 years = study cohort Age of mother 11-53 years S19 20-34 9000 ft Education of mother 0-28 years <9 9-11 90"' LGA or not (calculated) ; percentile Appropriate for gestat i onal age B i rth weight i n grams from 1 0 '" AGA or not ( AGA) (calculated) through 90th percentile 35

PAGE 53

Causes of all four adverse birth outcomes in this study are multi-factorial and include demographic, social, behavioral, and medical risk factors (Ahlsson et a/. 2007; March of Dimes 2007; 10M 2006; Ehrenberg eta/. 2004; Kramer 1987; 10M 1985). Demographic and Socioeconomic Risk Factors Age of mother at the time of the birth of her infant has been shown to affect the likelihood of all four birth outcomes (Gould eta/. 2003; Singh & Yu 1996; 10M 1985). Various age categories are used in studies of pregnancy outcomes; in this study mother's age is categorized into 'teen mothers" who are 19 years of age and younger mothers who are ages 20-34, and "older mothers who are age 35 and older. Teen mothers have higher risks of adverse reproductive outcomes (Fraser et a/ 1995), as do mothers over age 34 (Hansen 1986). The reference category for maternal age is 20 34 years old Parity of the mother at the time of birth is another demographic risk factor. In general, first and low parity births are associated with LBW and SGA; higher parity is associated with greater weight gain during pregnancy and LGA (Olsen eta/. 2007). "Parity" is a term that is defined inconsistently in obstetrics practice and studies of birth outcomes (Beebe 2005). Parity is properly defined as the number of pregnancies completed past 20 weeks of gestation (not the number of living infants delivered). "Gravidy" is the number of pregnancies conceived, regardless of outcome (Beebe 2005) The data on previous pregnancies in Colorado's birth record are categorized as previous pregnancies now living," previous pregnancies now dead,'' and other terminations. The intent of Colorado's previous pregnancy categories is to identify pregnancies completed past 20 weeks (whether issue are now living or dead) and to collect data on pregnancies ending before 20 weeks of gestation in "other terminations." However, the instructions for recording the information for these categories focus not on previous pregnancies but rather the number of infants. Therefore, it is possible that the categories used by CDPHE fail adequately to distinguish either viability or results of multiple gestations. For this study, parity is operationalized by summing the number of previous pregnancies now living and "previous pregnancies now dead" and using that number to form a base number for parity. Although this formulation is imperfect, it is a reasonable approach given the ambiguity in the collection of the data for the birth record in Colorado. From there, the variable for parity in this study is calculated using the Kleinman-Kessel index, which combines birth order and maternal age to account for the interaction effect of age and parity on outcomes, without producing collinearity between parity and age (Frisbie et a/. 1998; Kleinman & Kessel 1987) The Kleinman-Kessel Index 36

PAGE 54

categorizes parity into ''first births" at any age; "low parity," which is a second birth to mothers and up to a third birth to mothers (reference group); and "high parity," which is a second birth or higher to mothers <18 a third birth or higher to mothers <25, and a fourth or higher birth to a mother 25 years. The reference category for parity is "low'' parity. Colorado records the mother's address of residence before delivery. For the years 2000-2005, CDPHE geo-coded the altitude of the mother's residence address by latitude and longitude and elevation in feet above sea level and associated it with the birth record. Hypoxia, which can be caused by living at high altitude, is a demographic factor contributing to LBW and SGA (10M 1985). Although most births in Colorado are to mothers living at lower than 6000 feet above sea level, about 20% of births are to mothers living above 6000 feet and 5% of births are to mothers living above 7000 feet. Inclusion of residence altitude permits analysis of altitude effect on birth outcomes at a more discrete level than using the average elevation for a county, as is done in many studies where altitude is a variable of interest. Jensen and Moore have shown that an average increase of 1000 meters of altitude correlates with a drop of about 100 grams of birth weight {1997). In this study elevation of residence was coded as <5,000 feet, then in 1 ,000 foot increments between 5,000 and 8,999 feet, and> 9000 feet. Residence before delivery at <5,000 feet is the reference category. Social factors that are related to adverse birth outcomes and that are recorded in the birth register are education of the mother and father marital status, and adequacy of prenatal care. Education of the mother is categorized into three categories: less than high school education, some high school (grades 9-11 ), and greater than high school, because Mexican born immigrants often have only an elementary school education. Although there is often an association between more education and better health (Desai & Alva 1998; Link & Phelan 1995), some studies report contrary findings, where mothers of certain races/ethnicities with less education have better birth outcomes than mothers with more education (Acevedo-Garcia et a/. 2005; Gould et a/. 2003). More than 9.85% of cases are missing father's education in the dataset for Aim 1 and 13 03% of cases are missing father s education in the dataset for Aim 2. Accordingly, father's educational level is not included in the analysis. The reference category for mother's education is high school graduate and above. Marital status also contributes to birth outcomes. Being married or in a committed relationship is usually related to better birth outcomes (10M 1985), perhaps because it is a vehicle for access to more resources or social support {Williams et al. 2008; Schoenborn 37

PAGE 55

2004). Colorado records marital status as "legitimate" or "illegitimate" on the birth record, based on interview with the mother, and it includes a committed relationship in the category of "legitimate." Being married (legitimate) is the reference category for marital status. Although there is some suggestion that adequacy of prenatal care, as measured by various indices, may not correlate with birth outcomes (Alexander & Kotelchuck 2001; Fiscella 1995), U.S. and international policy encourages improvements in access to prenatal care (10M 2003). The Adequacy of Prenatal Care Utilization (APNCU), also known as the Kotelchuck Index, takes into account the length of gestation with the number of prenatal visits, thereby correcting for preterm births and distinguishing women who have more prenatal visits than the number recommended by the American College of Obstetricians and Gynecologists (and who likely are higher risk patients) (Kotelchuck 1994). Four categories are used to rate adequacy of care: inadequate, intermediate, adequate, and adequate plus. CDPHE calculates and provides data on the adequacy of prenatal care using the Kotelchuck Index. Adequate prenatal care is the reference group. Medical Risk Factors The birth record indicates whether the mother had certain medical risks during pregnancy and delivery The following risk factors are associated with one or more of the birth outcomes of the study: anemia (hematocrit <30%/hemoglobin <1 Og/dl); cardiac disease; acute or chronic lung disease; gestational diabetes; pre-existing diabetes; genital herpes; hydramnios/oligohydramnios; hemoglobinopathy; chronic hypertension; pregnancy associated hypertension; eclampsia; incompetent cervix; previous infant weighing 4000+ grams at birth; previous preterm or SGA infant; renal disease; Rh sensitization; uterine bleeding; other risk factors; premature rupture of the membranes (>12 hours); abruptio placenta; placenta previa; other excessive bleeding; and seizures during labor. Medical risk status is treated as dichotomous: any one or more of the medical risks or none. The reference group is no medical risks. Three medical risks are highly associated with LGA: pre-existing diabetes; gestational diabetes; and having given birth to a previous infant weighing more than 4000 grams. This set of variables is examined separately from the more general medical risks associated with pregnancy for LGA. 38

PAGE 56

Behavioral Risk Factors Inadequate weight gain of the mother during pregnancy is associated with LBW SGA and sometimes preterm birth (10M 2006, 1985). Higher than recommended weight gain is associated with LGA (Dyer eta/. 2007). The Institute of Medicine adopted weight gain guidelines in 1990 based on BMI but only weight gain in pounds is recorded in the Colorado birth record during the study period. For this study weight gain during pregnancy is coded into three categories: up to and including 15 pounds, 16 to 40 pounds, and more than 41 pounds, following stud ies by Frisbie and Song (2003) The Institute of Medicine is in the process of adopting new guidelines that will likely re-emphasize the importance of maternal BMI before pregnancy and recommend lower weight gain as a result of the trend to more overweight and obese mothers and concerns about the association between excessive weight gain and insulin overproduction in LGA infants (Barbour L., personal interview November 10, 2008). Weight gain between 16 and 40 pounds is the reference category. The effects of smoking and drinking on LBW and premature birth are well-recognized (10M 1985) The birth record includes the mother's report of the number of cigarettes smoked per day and number of alcoholic drinks taken per week during pregnancy. Smoking and drinking are each categorized using a dichotomous measure of smoking or not; drinking or not, with not smoking and not drinking as the reference categories. CDPHE collects data on smoking and drinking behaviors during pregnancy in two ways : from each mother's self report which is recorded in the off i cial birth register, and by survey using the Pregnancy Risk Assessment Moni toring System (PRAMS) a surveillance project of the CDC and state health departments (2006d) In 2004 and 2005 CDPHE identified discordance in reporting smoking and drinking between the two sources. For both years the birth record reported that about 1% of women drink alcoholic beverages and about 11% smoke during pregnancy. PRAMS survey reports for the same years estimated about 10% of women drank during pregnancy and 13% smoked Using multivariate analysis, CDPHE tested eight predictive factors of discordance and among other things determined that Hispanic women were less likely than White women to report discordantly for either smoking or drinking, and Hispanic women report less smoking and drinking than Whites. For these reasons and because Hispanic ethnicity is the most important racial/ethnic category of interest the data from the birth regi ster on smoking and drinking are used in this study (Drisko, J., personal interv i ew October 13 2007). In any event using the lower numbers from the birth record tend to bias 39

PAGE 57

results to the null. If smoking or drinking are influential factors in any of the outcomes their effects will be understated. Race/Ethnicity/Nativity Aim 1 compares Colorado's population of mothers using her identification of race/ethnicity" to capture the significant influence of Colorado's Hispanic population Mothers identify their race for the birth record. CDPHE uses the following categories to classify race: o White (includes Mexican Cajun Creole, Puerto Rican and all other Caucasian) o Black (Negro, Colored, Afro-American) o Indian (North American Central American, South American, Alaskan, Canadian) o Chinese o Japanese o Hawaiian (includes part-Hawaiian) o Filipino o Other Asian or Pacific Islanders (Korean Thai Amerasian, Vietnamese, Chamorro) o Other non-white o Unknown not stated, or not classifiable Mothers also identify their origin, which CDPHE records using the following categories: o Non-Hispanic o Mexican o Puerto Rican o Cuban o Central or South American o Other and unknown Hispanic Following Gonzalez-Quintero et a/. (2007), non-Hispanic women are classified by race and Hispanics are classified by origin. For this study, non-Hispanic White ('White ) mothers are those who identify for the birth record as 'White" and who do not also identify Mexican Puerto Rican, Cuban, Central or South American, or other and unknown Hispanic origin. Hispanic" mothers are those who identify White" race and who also identify one of the .previously mentioned Hispanic categories. Black mothers are those who identify Black" race and who also identify either Hispanic or non-Hispanic. Only 360 mothers out of the total study population of 356,389 identified both Black and "Hispanic;" accordingly 40

PAGE 58

most mothers in the Black category are Black non-Hispanic." Mothers who i dentify as Black and Hispanic could be analyzed in either the Black or Hispanic category. They are grouped w ith the Black" category in this study because of their small number and based on a study that found that Black/Hispanics first identified themselves as Black and identified H i spanic ethnicity only when the interviewer inqu ired directly about ethn i city perhaps suggesting that Hispanic heritage was not as important to thei r identity as race (Baker eta/. 2006). Mothers of any other race are categorized as Other. For Aim 1 White mothe r s constitute the reference group. The birth record also reports the mother's nativity. The choice for state of birth includes each of the fifty United States and the D i strict of Columbia, Puerto Rico Guam Canada, Mexico a number of additional countr ies, and remainder of the world. For Aim 2 mothers of Mexican origin are identified as Mexican-born or U.S.-born (born i n any of the United States or District of Columbia). No mothers of Mexican orig i n were born in either Puerto Rico or Guam. U S.-born mothers of Mexican origin constitute the reference group for Aim 2. Missing Data and Size of Study Population for Aim 1 The dataset of singleton births occurring in Colorado for the years 2000 2005 consists of 392 ,881 b irths. According to CDPHE the birth record represents only live b i rths. Yet within the dataset are records of births of very low birth wei ght and very short gestation CDPHE conducts quality control of the birth records i n two ways when data appear to be outs i de of normal value ranges It checks for data entry errors and i t queries the hosp i tal/location of b irth to inquire whether the data are correct as received (Sol K., personal i nterview June 10, 2007). After qual i ty control measures are exhausted CDPHE reta ins records that nevertheless appear to be outside the range of viability. One reason for these anomalous records may be some parents insistence on i ssuance of a live birth cert i f i cate (and subsequent death certificate) when a child is stillborn From this dataset of 392,881 records 254 presumed non-viable b irths that are recorded as weighing less than 400 grams and having completed fewer than 23 weeks of gestation were removed. These weight and gestation values follow the recommended cut points for consideration of non-in i tiation of resuscitation at birth and represent the most widely circulated resusc i tation guideline i n the U .S. (American Academy of Pediatrics 2005:e1-10}. The removal of these presumed non v i able births preserves births at very low birth weights that are extreme l y growth restricted but which have longer gestation and could have been born alive. The d i stribut ion of births by 41

PAGE 59

race/ethnicity of the resulting dataset of 392,627 is shown in Table 3.2. Thirty-three records were missing any racial/ethnic identifier Table 3.2. Original dataset by race/ethnicity Variable Race/Ethnicity Total 392,627 (100 00%) White Hispanic Black 239,921 118 130 17, 213 (61.11%) (30.09%) (4 38%) Other 17,330 (4.41%) Missing 33 (0 01%) Cases with any missing value for any of the variables of interest were successively deleted. Table 3 3 reports the number and percentage of cases missing any variable of interest. Table 3.3. Number and percentage of missing cases by variable for Aim 1 Variable Mother's altitude of residence Weight gain Adequacy of prenatal care Mother's education Smoking Drinking Parity Birth weight Estimated gestation Mother s age Mother's race Marital status Medical risk Total Number of Cases 13 414 11, 374 9 232 4 520 1,785 1,076 346 67 66 54 33 0 0 41, 940 % of Cases Missing 3.42% 2 90% 2.35% 1.15% 0.45% 0 27% 0 09% 0.02% 0 02% 0.01% 0.01% 0.00% 0 00% 10 68% After deleting cases with any missing variable of interest for Aim 1 the resulting dataset is 356,389 births (cases deleted equal 36,238). The number of cases deleted is lower than the total number of cases missing any variable shown in Table 3.3 because some cases were missing more than one variable. A total of 9.2% of cases were deleted from the original dataset. When availability of data is an exclusion criterion, the frequency of missing data for each variable is considered. If the amount of "overall missingness" is <10% then the dataset is generally considered sufficient, even when differences between independent variables are significant (Burton & Altman 2004). For Aim 1, no single variable has more than 3 42% of 42

PAGE 60

missing values and the entire dataset is missing fewer than 10% of cases after exclusion of cases missing any variable of interest. The distribution of missing data by race/ethnicity in Table 3.4 shows that there are differential percentages of missing data by race/ethnicity; in particular, there are higher percentages of missing data among Hispanics relative to the percentage of births of Hispanics statewide (30.09%) for weight gain, education, and parity To further address the question of whether and how much missing data might bias results regressions were run using the original dataset with "Missing" shown as a separate category and using the dataset after deleting cases with missing variables. Comparisons of the odds ratios for each outcome based on the fully adjusted model described at page 52 are shown in Figures 3.2 3.5. Odds ratios for each outcome are shown using the complete dataset followed by the odds using the dataset with missing cases removed for each outcome. For the low-weight associated outcomes, the largest drift occurs in the Others category; for Whites, Hispanics, and Blacks the drift is small, especially for Hispanics, the main race/ethnicity of interest. For LGA there is very little drift for any population Accordingly, the dataset with missing cases deleted is appropriate for analysis. Table 3.4. Number and percentage of missing variables by race/ethnicity Total Race Variable 41,940 White Hispanic Black Other Missing (%} (%} (%} (%} (%} Alt i tude 13,414 8 ,993 (67 04) 3 556 (26 51) 356 (2 65) 498 (3. 71) 11 (0 08) Weight gain 11,374 5,483 (48.21) 4,692 (41. 25} 578 (5.08) 603 (5.30) 18 (0 16) Prenatal care 9 232 5 526 (59 86) 2,469 (26.74) 668 (7.24) 549 (5.95) 20 (0 22) Education 4 520 1 ,809 (40.02) 2,085 (46 13} 315 (6 97) 283 (6 26) 28 (0 62) Smoking 1 785 1 085 (60 78) 517 {28.96) 90 (5 04) 77 (4.31) 16 (0 90) Alcohol 1 ,076 621 (57.19) 325 (30 20) 69 (6 41) 44 (4. 09) 17 (1. 58) Parity 346 137 (39 60) 128 (36 99} 39 (11. 27) 24 (6 94) 18 (5.20) Birth weight 67 33 (49.25) 18 (26.87) 6 (8.96) 4 (5 97) 6 (8.96) Gestation 66 55 (83.33) 4 (6.06) 0 (0.00) 1 (1. 52) 6 (9 09) Age 54 21 (38 89) 16 {29 63) 2 (3 70) 1 (1.85) 14 (25 93) Race 33 N/A N/A N/A N/A N / A 43

PAGE 61

2.5 2 .11 2 .16 2. 25 2 "' 1.64 0 1.75 __._. HispNiic 1 5 B la::k a: 1.18 1.18 Qhe< 1 25 "' '0 1 '0 0 0 .75 0 5 0 25 0 L BWA IIOata LBW Study Data Figure 3 2 AIM 1 LBW DRIFT OF ODDS 2.25 "' 1 .75 0 1 5 a: 1 25 "' '0 '0 0 .75 0 0 5 0 25 1.96 1 .98 -1.69 1.72 1.18 1.18 SGA A ll Dat a SGA S tudy Oat a Wile ---Hisparic I I Figure 3.4 AIM 1 SGA DRIFT OF ODDS 2.5 2.25 "' Q 1 75 1ii 1.4 1 4 2 1 5 -HispiV'Iic a: 1 25 Black "' .11 ..1.16 '0 -Qhe< '0 1.02 1.01 0 0 75 0 5 0 25 PretermA I IOata PretermStudyData F i gu r e 3 3 AIM 1 PRETERM DRIFT OF ODDS 2 5 2.25 2 -+--Wlite 1 5 a: 1 25 -------H ispanic "' B lack '0 1 0 95 0 .95 '0 0 .680 0 75 0 .68 0 5 0.59 OR 0 25 0 LGAAII Dat a LGAStudy Data Figure 3 5 AIM 1 LGA DRIFT OF ODDS Table 3 5 describes the f i na l study populat i on by race / ethnicity for A i m 1 Table 3.5. Final study population by race/ethnicity of mother-Aim 1 Study Population 356 389 Study Population by Race / Ethnic i ty White 219 029 ( 61.46 % ) Hispanic 106 ,291 ( 29 82 % ) Black 15 448 ( 4.33 %) M i ssing Data and S i ze of Study Population for Aim 2 Other 15 ,621 ( 4 38 % ) Aim 2 repl i cates Aim 1 except that it compares birth outcomes of mothers of Mexican or i g i n by nativ i ty. Table 3 6 displays the original statewide dataset by nat i v i ty Two hundred seven cases are miss ing mother s place of nativity. 44

PAGE 62

Table 3.6. Original dataset by place of nativity of mother Study Population 95,291 Study Population by Nativity U.S.-Born Mexican Origin 35, 357 (37 1 0%) Mexican-Born Missing 59, 727 (62.68%) 207 (0.22%) As with Aim 1, cases with any missing value for any of the variables of interest were successively deleted Table 3 7 reports the number and percentage of cases missing any variable of interest from the dataset for Aim 2. Table 3.7. Number and percentage of missing cases by variable for Aim 2 Variable Weight gain Mother's altitude of residence Adequacy of prenatal care Mother's education Smoking Alcohol Parity Mother's age Birth weight Estimated gestation Marital status Medical risk Total Number of Cases Missing Variable 3 716 2 760 1 982 1,779 370 249 105 12 7 1 0 0 10,981 % of Cases Missing Variable 3.90% 2 90% 2.08% 1.87% 0 39% 0 26% 0.11% 0.01% 0 01% 0 00% 0.00% 0 00% 11. 53% After deleting cases with any missing variable of interest for Aim 2, the resulting dataset consists of 85,755 births (cases deleted equal 9,536). The range of missing variables is 03.9%. As with Aim 1, the number of cases deleted is lower than shown in Table 3 7 because some cases had more than one missing variable. A total of 10.01% of cases were deleted from the original dataset of mothers of Mexican origin. To determine whether the missing data might affect the results of the analysis, the distribution of missing data by nativity is reported in Table 3.8. As can be seen, there are differential percentages of missing data by nativity; in particular there are higher percentages of missing data among Mexican-born Hispanics relative to their percentage of births statewide (62.68%) for weight gain altitude of residence, education, par i ty and gestation. 45

PAGE 63

Table 3.8. Number and percentage of missing variables by nativity U.S.-Bom Mexican Nativity Variable Total Mexican Origin Born Missing 10,981 {%} {%} {%} Weight gain 3716 826 (22 23) 2,869 (77.21) 21 (0.57) Altitude 2760 949 (34 38) 1794 (65.00) 17 (0.62) Prenatal care 1982 785 (39.61) 1181 (59.59) 16 (0.81) Education 1 779 421 (23.66) 1325 (74 48) 33 (1.85) Smoking 370 204 (55.14) 162 (43.78) 4 (1.08) Alcohol 249 104 (41. 77) 143 (57.43) 2 (0.80) Parity 105 21 (20 00) 81 (77.14) 3 (2.86) Age 12 2 (50 00) 6 (16.67) 4 (33.33) Birth weight 7 3 (43.86) 4 {57.14) 0 (0.00) Gestation 1 0 (0.00) 1 (1 00.00) 0 {0. 00) Nativity 207 N/A N/A N/A As with the dataset for Aim 1, to further address the question of whether and how much missing data may bias results, regressions of the fully adjusted models were run using the original dataset with "Missing" shown as a separate category and using the dataset w ith missing cases deleted. Figures 3.6 3.9 display the drift of odds ratios between the full dataset and the study dataset after removing cases with missing variables of interest. For LBW and preterm birth, odds for Mexican-born mothers are not significantly different from the odds for U.S.-born mothers. Given the small or non-existent drift in odds ratios for each outcome the dataset without cases having missing data i s used for analysis. 1.5 I I ---U S Bor 1.: .... Mexi can 0.11 us Born 1.5 1.: 1.01 1.118 ....... ---U S Bor I I MexicanBorn 0 0 PretermAII PretermStudy LBWAIIData LBWStudy Data Data Data Figure 3.6 AIM 2 LBW DRIFT OF ODDS Figure 3.7. AIM 2 PRETERM DRIFT OF ODDS 46

PAGE 64

1.5 1.5 -1.:.. ... 1.45 --U S B orn 1 : 1 t --U. S B orr ... \_ M exican o.75 74 Born 1.: 1 1 "-Mexican-Born 0 0 S GAAIIData SGAS t udy Data L GAA IIDat a LGAStudy D at a F i gure 3 .8. AIM 2 SGA DRIFT OF ODDS Figure 3.9 AIM 2 LGA DRIFT OF ODDS The resulting study population for Aim 2 is a total of 85,755 births: 32, 484 (37.88%) U.S.-born mothers of Mex i can origin (born in any of the 50 states or the District of Columbia) and 53,271 (62.12%) Mexican-born mothers. Table 3 9 reports births for Aim 2 by year The percentage of births by nativity is quite consistent throughout the study per i od. Table 3.9. Final study population by nativity of mothers of Mexican origin by year-Aim 2 Study Population All Years 85,755 2000-2005 2000 2001 2002 2003 2004 2005 U.S.-Born Mexican Origin 32 484 (37 88%) 4,606 (38 78 % ) 5 125 (37 57%) 5 553 (37 16 %) 5 612 (37.43 % ) 5 753 (37.64%) 5 835 (38 86 % ) Contextual Variables for Aim 3 Mexican-Born 53 ,271 (62 12 % ) 7 ,271 (61.22% ) 8 518 (62.43 % ) 9 389 (62.84 % ) 9 382 (62 57 % ) 9 530 (62 36 % ) 9 ,181 (61. 14 % ) To examine the effect of neighborhood composit ion on birth outcomes two scales used by Finch et a/. (2007} were calculated from i nformation collected from the 2000 Decennial Census (Summary File 3). The Scale of Immigrant Orientation consists of : o % Mexican-born individuals living i n the tract ; ca l culated by dividing the number of Mexican-born individuals from Census Table PCT20 divided by the population in the tract o % non-citizens born in Mexico living in the tract ; calculated by summing the numbe r s of individuals who are not citizens based on year of entry from Census Table PCT20 divided by the population in the tract 47

PAGE 65

o % linguist i cally isolated households speaking Spanish ; calculated by dividing the number of linguistically isolated Spanish households from Census Table P20 by the number of households in the tract. A linguistically isolated household is one in which all members of the household 14 years old and over speak Spanish and have at least some difficulty with Engl ish. The Scale of Neighborhood Deprivation consists of: o % individuals living in the tract in poverty calculated by dividing the number of indi viduals i n poverty in 1999 in the tract by the number of ind i v i duals from whom poverty status is determined in the tract from Census Table P87 o % of households receiving public assistance income, calculated by dividing by the number of households receiving public assistance income i n 1999 by the number of households in the tract from Census Table P64 o % female headed family households calculated by d i v i ding the number of female householders by the number of households i n the tract from Census Table P9 o %males unemployed in the civilian work force reported in Census Table QT P24 (no calculation necessary). Neither scale was normally distributed, so each was transformed. Neighborhood deprivation was transformed by taking the square root of the i ndex value following the practice of Finch et a/. (2007) and O'Campo et a/. (1997) which resulted i n a normal distribution Immigrant orientation was transformed by taking the natural log of the value, as that produced a better normal distr i bution than square root transformation The validity of each scale was tested using Cronbach s Alpha statistic (Cronbach 1951 ) The c l oser the Cronbach score i s to 1 0 the more reliable the generated scale i s Nunnaly (1978) i ndicates 0 7 to be an acceptable reli ability coefficient but l ower thresholds are sometimes reported in the literature. Cronbach s alpha ranges from 0 .725-0.970 for all scales except Denver s neighborhood deprivation scale which approaches 0.7 (value 0 674). Table 3.11 reports the characteristics of each index by county. Because the items i n each scale are correlated the items for each scale are summed and then averaged to create the scale values used i n linear modeling. 48

PAGE 66

Table 3.10. Values of contextual scales for Adams and Denver counties2000 Mean (SD) Minimum Maximum Cronbach's Alpha Adams Neighborhood Deprivation 7.99 62 2.41 17 .39 725 Denver Neighborhood Deprivation 10 78 .92 0.00 47.17 .674 Adams Immigrant Orientation 6.31 09 0.00 34.90 .970 Missing Data and Size of Population for Aim 3 Denver Immigrant Orientation 10.97 85 0 .00 40.63 .931 Aim 3 analyzes the effects of neighborhood deprivation and immigrant orientation on birth outcomes in two of Colorado's largest counties, Adams and Denver Counties. These two counties were chosen because they have both a large percentage and number of residents of Mexican origin. Adams County, with 86 census tracts, has a mixed urban/rural composition; Denver, with 136 tracts, is urban. Urban/rural characterization is based on Summary File 3, Table H5 of the 2000 Census, which counts housing units as "urban" if they are inside urbanized areas or urban clusters or "rural," whether farm or non-farm. In 2000, 31.7% of Denver County's population identified as Hispanic/Latina; 28.2% of Adams County identified as Hispanic/Latina (Census 2000b, Table GTC-P6). CDPHE reports that the proportion of linguistically isolated households speaking Spanish increased by 171% statewide between 1990 and 2000, based on census data. The proportion of linguistically isolated households in Adams County increased by 416% and by 158% in Denver County during this same time period (CDPHE 2005). Most critics of contextual studies lament the necessity of using administrative subdivisions, such as census definitions, for area-level descriptions of "neighborhood" (Diez Roux 1998, 2000). Unfortunately, the use of secondary data from the birth record, which provides census tract information as the smallest geographic area unit, makes this problem unavoidable. The Census Bureau creates census tracts that consist of 1 ,500 8,000 (optimum 4,000) residents. Tracts are intended to represent relatively homogenous areas that conform to local perceptions of neighborhood (Lee & Marlay 2007). It has been shown that measures at the census tract level perform about equally to those at the smaller census block group level (Krieger eta/. 2003). Therefore, consistent with other population studies that examine area factors (O'Campo eta/. 1997; Finch eta/. 2007), this study uses census tracts as a proxy for neighborhoods. 49

PAGE 67

Subramanian eta/. recommend that at least twenty observations per geographic area are needed for area-level analysis (2003:105). One tract in Adams County had no births during the study period and it was therefore deleted. Six tracts in Adams County had fewer than twenty births. These were combined with nearby tracts by matching the quartile of neighborhood deprivation and the quartile of immigrant orientation to an adjacent or close tract, resulting in 79 tracts for analysis. In Denver County, one tract also had no reported births and it was deleted. Thirty-six tracts in Denver County had fewer than 20 births. These tracts were deleted from the analysis because it was not possible to combine them by matching on both neighborhood deprivation and immigrant orientation with an adjacent or nearby tract. The analysis of Denver County therefore includes 100 tracts. Unlike the multiple logistic regression modeling used in Aims 1 and 2, general linear modeling requires only the mother's tract of residence, nativity and outcome. Cases missing census tract or nativity were deleted for Aim 3 For Adams County, 435 cases (2.63%) were removed from the dataset. For Denver County, a total of 325 cases in 36 tracts were removed because matching was not possible within the a priori rules established. An additional171 cases were deleted because they were missing tract or nativity (0. 7% of cases in the remaining 100 tracts). The final study populations for Adams and Denver Counties are described in Table 3 .11. Table 3.11. Population of mothers of Mexican origin by nativity in Adams and Denver Counties 2000 2005 U.S.-Born Primary County Mexican Origin Mexican-Born Character of Adams N=16,107 5 733 (35.59%) 10 374 (64 41%) Mixed Denver N = 23,332 4 914 (21.06%) 18 418 (78 94%) Urban Methods for Aims 1 and 2 The analytical tools for Aims 1 and 2 are univariate and bivariate descriptive statistics and multiple logistic regression Frequencies of risk factors and birth outcomes describe mothers by race/ethnicity and nativity Multiple logistic regression is used to predict each dependent birth outcome. As discussed in Chapter 2 according to the social gradient of health Hispan ics and Blacks are expected to have demographic/social, medical, and health behavior risk profiles that are less advantageous than those of Whites, and therefore to have worse birth 50

PAGE 68

outcomes. U.S.-born mothers of Mexican origin are expected to have better birth outcomes than Mexican-born mothers based on the same social gradient of health. The null and alternative hypotheses for each risk factor are: H10 : each risk factor is independent of race/ethnicity/nativity H1 A: each risk factor is related to race/ethnicity/nativity. The null and alternate hypotheses for incidence of birth outcomes without adjustment for risk factors are: H20 : each birth outcome is independent of race/ethnicity/nativity H2A: each birth outcome is related to race/ethnicity/nativity Pearson's chi-square is used to test the relationships between risk factors and race/ethnicity or nativity and each birth outcome and race/ethnicity or nativity. Unadjusted odds of any given birth outcome do not adequately account for individual level risk factors. Therefore, successive adjustments are made to examine the relationship between outcomes and race/ethnicity and nativity. H30 : race/ethnicity/nativity does not predict each birth outcome after adjusting for risk factors. H3A: race/ethnicity/nativity predicts each birth outcome after adjusting for risk factors. Model Building for Aims 1 and 2 Four nested logistic regression models are constructed for each outcome in Aims 1 and 2. Based on the existing literature and the availability of data from the birth record, the variables listed in Table 3.1 were selected as candidate variables. Model 1 tests the main effects of race/ethnicity or nativity in each aim. Each successive model adjusts for related variables as a block. Model 2 includes the main effects on each outcome from Model 1 and adjusts for demographic and social economic risk factors (age, parity of the birth, altitude of mother's residence immediately before the birth, marital status, level of mother's education, and adequacy of prenatal care). As described at pp. 36-39, reference categories are the lowest risk category of that variable. Model 2 tests the social gradient of health by examining whether there are differences in odds of a given outcome after adjustment for demographic and socioeconomic risk profiles. Model 3 includes the main effects and demographic and socioeconomic risk factors and adds the existence of one or more medical risks associated with pregnancy. For Aim 1 51

PAGE 69

this is a further test of the social gradient. For Aim 2, Model 3 tests the hypothesis that mothers born in Mexico are "healthier'' than mothers of Mexican origin born in the U.S. Although Model 3 cannot test the healthy migrant hypothesis directly, because it does not include data on birth outcomes for Mexican mothers who remain in Mexico, nevertheless it can assess the impact of pregnancy-associated medical risks on Mexican-born mothers. Model 4 includes the main effects, demographic and socioeconomic risk factors, medical risks, and adds health behaviors during pregnancy For Aim 2, Model 4 tests the healthy immigrant hypothesis (whether Mexican-born mothers engage in healthier behaviors than mothers of Mexican origin who are born in the U.S.). To screen the candidate variables, univariate testing of outcomes and potential explanatory variables was conducted on the statewide datasets by race/ethnicity and by nativity. Main effects were tested by examining the relationship of each outcome variable with each explanatory variable using the Pearson chi square test. According to Hosmer and Lemeshow, variables with a p-value of less than .25 may be deleted (2000:86). This method may result in over-identification of potential variables, but further model testing techniques can then be used to eliminate marginal variables. No variables were deleted in this step for any outcome or aim. Finally, the relationship of each explanatory variable against each other explanatory variable was tested using the chi square measure of association to determine whether there might be any interactions among the explanatory variables. For Aim 1, each variable showed a statistically significant interaction with each other variable with a p-value <0.05. For Aim 2, each variable showed a statistically significant interaction with every other variable, except drinking* age. This screening provided additional information for further model building. For each outcome in Aims 1 and 2, each candidate variable from Table 3.1 was entered into the logistic regression using backward and forward selection. Variables that were not significant to the model were discarded, unless the literature suggested that the variable should be retained. Then, using interactive modeling variables were entered into each model, one by one in order of significance, using forward selection modeling. Again variables that were not significant to the model were discarded, unless the literature suggested that the variable should be retained. The significance of each variable added to the previous model was tested using the -2 log likelihood ( -2LL) test. Any variable that changed the -2LL score by an amount that was significant using Pearson's chi square of the difference between the -2LL statistic of each successive model was retained. Finally, each 52

PAGE 70

model was run with all of the model variables retained after testing, as a block with forward selection of each interaction term The significance of each interaction was tested using the 2LL test, and significant interactions were retained. Figure 3.10 depicts the steps in model building Examination of Candidate Variables All variables significant at p=0 05 using chi square test ; Each outcome each independent variable Q all variables reta i ned Interactions All interactions significant at Each independent variable each other Q p=0.05 using chi square, except independent variable drinking*age, which is deleted Backward Selection of Candidate Variables Q Tested using chi square significance of difference of 2LL statistics Interactive Modeling Entry of Candidate Variables as block Q Tested using chi square forward selection of each interaction significance of difference of fi -2LL statistics Final Model for Each Aim and Outcome Figure 3 10 SCHEMATIC OF MODEL BUILDING FOR AIMS 1 AND 2 The results of model building are described below by Aim and outcome. Aim 1 LBW Drinking is not significant based on univariate tests, and is dropped. Five interactions are significant: medical risk*race, parity*marital status, prenatal care*weight gain, prenatal care*smoking, and smoking*age. o M1: Y (LBW) = a (intercept) + b,X, (race/ethnicity Hispanic) + b 2X2 (race/ethnicity Black)+ b3X3 (race/ethnicity Other)+ E (standard error) o M2: Y (LBW) =a+ b X (race/ethnicity Hispanic)+ (race/ethnicity Black)+ (race/ethnicity Other)+ (<20 years age)+ b 5 X 5 (>34 years age)+ bsXs (first birth) + b1X1 (high parity) + bsXs (elevation 5000-5999) + bg)(g (elevation 6000-6999) + b ,oX1o (elevation 7000-7999) + b,X, (elevation 8000-8999) + b 12X12 (elevation >9000) + b13X13 (school
PAGE 71

(unmarried)+ b1sX1s (inadequate pnc) + b11X11 (intermediate pnc) + b18X18 (adequate plus pnc) + b1!}-23X1!}-23 (parity*marital status-5 levels)+ E o M3: Y (LBW) =a+ b1X1 (race/ethnicity Hispanic)+ b2X2 (race/ethnicity Black)+ b 3Xa (race/ethnicity Other)+ (<20 years age)+ bsXs (>34 years age)+ bsXs (first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + bgXg (elevation 6000-6999) + b10X1o (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b1aX1a (school 34 years age)+ b6X6 (first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + b 9Xa (elevation 6000-6999) + b1oX1o (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b1aX1a (school 34 years age) + bsXa (first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + bgXa (elevation 60006999) + b1oX10 (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b1aX1a (school 34 years age) + bsXa 54

PAGE 72

(first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + b 9 X 9 (elevation 60006999) + b1oX1o (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school 34 years age) + bsXs (first birth) + b1X1 (high parity) + baXa (elevation 5000 5999) + bgXg (elevation 60006999) + b 1oX10 (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b1JX13 (school 34 years age)+ bsXs (first birth) + b?X1 (high parity) + baXs (elevation 5000-5999) + bgXg (elevation 6000-6999) + b10X10 (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school 34 years age)+ b6Xs (first birth) + b1X1 (high parity) + b8Xa (elevation 5000-5999) + bgXg (elevation 6000-6999) + b1oX1o (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b 12X12 (elevation >9000) + b13X13 (school
PAGE 73

(adequate plus pnc) + bHHsX19-2s (parity*marital status-7 levels) + b2eX2e (medical risk)+ b21-34X21-34 (race*medical risk-8 levels)+ E o M: Y (SGA) = a (intercept) + b,X, (race/ethnicity Hispanic) + b2X2 (race/ethnicity Black)+ b3Xa (race/ethnicity Other)+ (<20 years age)+ bsXs (>34 years age)+ bsXs (first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + b 9 X 9 (elevation 6000-6999) + b10X10 (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school 34 years age)+ beXa (first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + bgXg (elevation 6000-6999) + b1oX1o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school 34 years age)+ beXe (first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + bgXg (elevation 6000-6999) + b1oX1o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school 34 years age)+ beXe (first birth) + b 7 X 7 (high parity) + b8Xa (elevation 5000-5999) + b 9Xg (elevation 6000-6999) + b1oX10 (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school
PAGE 74

plus pnc) + b19X19 (medical risk) + b20X:!o (smoking) + b21X:!1 (low weight gain) + b22X22 (high weight gain) + E Appropriate for gestational age (babies born neither SGA nor LGA) was also tested for Aim 1 as a way to confirm the results of the more specific adverse outcomes. As with the other outcomes, drinking was not significant and was dropped from the models. There were two interactions: weight*parity and weight*marital status. Aim 2 LBW. Drinking is not significant based on univariate tests, and is dropped. Age and school do not add significantly to the model in backward selection, but as known contributors to LBW, they are retained. Only one interaction is significant: prenatal care*nativity. o M1: Y (LBW) =a (intercept)+ b1X1 (Mexican-born)+ E (standard error) o M2: Y (LBW) = a (intercept) + b1X1 (Mexican-born) + b2X2 (<20 years age) + b3Xa (>34 years age)+ (first birth)+ bsXs (high parity)+ bsXs (elevation 5000-5999) + b7X7 (elevation 6000-6999) + baXs (elevation 7000-7999) + bgXg (elevation 80008999) + b1oX1o (elevation >9000) + b11X11 (school 34 years age)+ (first birth)+ bsXs (high parity)+ bsXs (elevation 5000-5999) + b7X7 (elevation 6000-6999) + baXs (elevation 7000-7999) + bgXg (elevation 80008999) + b10X10 (elevation >9000) + b11X11 (school 34 years age)+ (first birth)+ bsXs (high parity)+ bsXs (elevation 5000-5999) + b7X7 (elevation 6000-6999) + baXa (elevation 7000-7999) + bgXg (elevation 80008999) + b1oX1o (elevation >9000) + b11X11 (school
PAGE 75

is included because it is a main effect and a parent term for one of the interactions. Two interact i ons are significant: prenatal care nat i vity and prenatal care weight. o M1: Y (Preterm birth)= a (intercept)+ b X (Mexican-born)+ E (standard error) o M2: Y (Preterm birth)= a (intercept)+ b,X, (Mexican-born)+ b 2 X 2 (<20 years age)+ b 3Xa (>34 years age) + b 4 Xt (first birth) + bsXs (hi gh parity) + bsXs (elevation 50005999) + b 1X1 (elevation 6000-6999) + beXs (elevation 7000-7999) + bgXg (elevation 8000-8999) + b10X10 (elevation >9000) + b11X 1 1 (school 34 years age) + b 4 Xt (first birth) + b5Xs (high parity) + b6Xs (elevation 5000 5999) + b 1X1 (elevation 6000-6999) + beXs (elevation 7000-7999) + bgXg (elevat i on 8000-8999) + b 1oX10 (elevation >9000) + b,,x,, (school 34 years age) + b 4Xt (first birth) + bsXs (high parity) + b sXs (elevation 50005999) + b1X1 (elevation 6000-6999) + beXs (elevat ion 7000-7999) + bgXg (elevat ion 8000-8999) + b,oX o (elevation >9000) + b,,x,, (school 34 years age) + b 4Xt (first birth) + b5Xs (high parity) + b sXs (elevation 5000-5999) + b7X7 (elevation 6000-6999) + b8Xs (elevation 7000-7999) + bgXg ( elevation 80008999) + b,0X10 (elevation >9000) + b,,x,, (school
PAGE 76

o M3 : Y ( SGA) = a ( i ntercept) + b1X1 (Mexican-born) + b 2X2 (<20 years age) + b3Xa (>34 years age)+ b4'4 (first birth) + bsXs (high parity) + bsXs (elevation 5000-5999) + b 1X1 (elevation 6000-6999) + bsXa (elevation 7000-7999) + b 9 X 9 (elevation 80008999} + b10X1o (elevation >9000) + b11X11 (school 34 years age)+ b 4'4 (first birth) + b5Xs (high parity) + bsXs (elevation 5000-5999) + b 1 X 1 (elevation 6000-6999) + bsXa (elevation 7000-7999) + b9)(g (elevation 80008999) + b 1oX10 (elevation >9000) + b11X11 (school 34 years age) + b4'4 (first birth) + b5X5 (high parity) + b 6 X 6 (elevation 5000-5999) + b 1 X 1 (elevation 6000-6999) + bsXa (elevation 7000-7999) + b9X9 (elevation 8000 8999} + b 1oX1 o (elevation >9000) + b 11X11 (school 34 years age) + b 4'4 (first birth) + bsXs (high parity) + beXs (elevation 5000-5999} + b 1 X 1 (elevat i on 6000 6999) + b8Xa (elevation 7000-7999} + b9X9 (elevation 8000-8999) + b1oX10 (elevation >9000) + b11X11 (school 34 years age) + b4'4 (first birth} + bsXs (high parity) + beXs (elevation 5000-5999) + b 1 X 1 (elevation 6000-6999) + bsXa (elevation 7000-7999) + b9)(g (elevation 8000-8999) + b 1oX10 (elevation >9000) + b11X11 (school
PAGE 77

As with Aim 1, AGA was also tested. Drinking was not significant and was dropped from the models Only one interaction is significant, weight *parity Significant standardized beta coefficients and odds ratios are reported for each outcome, model, and aim in Chapter 4. For models with interactions, the model was run substituting a newly created n-level interaction variable for the parent terms and the interaction to parse the effect of the interaction. The R2 statistic (for linear regressions) and the -2LL test (for logistic regressions) use differences in variance accounted for in successive models to test model fit. The c statistic tests how well a model predicts outcomes. The c statistic ranges from .51 to 1.0; with a c statistic of .51 representing random variation. The closer a c statistic is to 1 the better the fit. Each successive model should improve the fit. In addition, adequacy of fit is often tested using the Hosmer-Lemeshow statistic. The c statistic and Hosmer-Lemeshow statistic are reported for each model and outcome. For some models, where variables are retained notwithstanding their insignificant addition to the model, model fit is sacrificed in the name of previous research showing that the variables are important to the outcome. Model Building for Aim 3 Aim 3 examines the influence of neighborhood of the mother's residence immediately before birth to test the effect of neighborhood deprivation and immigrant orientation on birth outcomes by nativity in two Colorado counties. Neighborhood is operationalized by using census tract data from the 2000 Decennial Census. In Los Angeles County, California, Finch eta/. (2007) compared the probability of LBW by census tract for foreign-born and U.S.-born mothers of Mexican origin using hierarchical linear modeling They showed that as neighborhood deprivation increased, the probability of LBW increased. They also showed that immigrant orientation moderated the probability of LBW for foreign-born mothers and lowered the probability of LBW when compared with the probability of LBW by neighborhood deprivation alone. For this study, the effect of the two contextual variables on each outcome is tested separately and independent of individual level effects. The hypotheses for Aim 3 are: H40 : immigrant-oriented neighborhoods have no effect on the four birth outcomes H4A: immigrant-oriented neighborhoods affect the likelihood of birth outcomes H50 : neighborhood disadvantage has no effect on the four birth outcomes 60

PAGE 78

H5A: neighborhood disadvantage affects the likelihood of birth outcomes For each of the two counties the effects of neighborhood deprivation and immigrant orientation were modeled with the rate of each outcome in each tract using generalized linear modeling. F i rst the transformed values for the contextual variables (square root for neighborhood deprivation and natural log for neighborhood immigrant orientation) were used to model area effects The resulting models were then run using the actual values of the scales after centering at the mean. The number of births and the number of outcomes in combined tracts were summed and a new percentage of each outcome was calculated based on that new denominator to create the outcome rate for the combined tract. The scale values for combined tracts were also weighted based on births in the combined tracts. Y is the rate of each outcome by tract, modeled against the centered values for scale of immigrant orientation alone the scale of neighborhood deprivation alone, the combinat ion of the two scales, and the interaction of the two. The t statistic for each contextual variable are reported, as is the F statistic for each model. For each county, the general linear modeling formulae are: Y (rate of outcome by tract)= a+ b1X1 (neighborhood deprivat ion scale)+ E Y (rate of outcome by tract)= a+ b1X1 (immigrant orientation scale)+ E Y (rate of outcome by tract)= a+ b1X1 (neighborhood deprivation)+ b2X2 (immigrant orientation)+ E Y (rate of outcome by tract)= a+ b1X1 (neighborhood deprivat ion)+ b2X2 (immigrant orientation) + b3X3 (neighborhood deprivation immigrant orientation) + E Qualitative Research Component Sample and Study Data History" and much of social science is biased because it is constrained by the existence of what i s collected in government files Families and informal groups do not necessarily collect and maintain information that captures the lived experience (Boorstin 1987 :7). Aim 4 is designed to explore the context surrounding the quantitative results especially the finding of higher odds of LGA among Mexican-born mothers, by interviewing women of Mexican origin who had babies within the past year in Colorado This component of the study was approved by the Human Subjects Review Committee of the University of 61

PAGE 79

Colorado Denver on an expedited basis under 45 CFR 46.1 02(i) as presenting minimal risk to research subjects. Waiver of documentation of consent pursuant to 45 CFR 46.1179(c)(2) was also approved so as to avoid any link between the interviewee and identifying information, which could present concerns for undocumented immigrants. Approvals from the University's Human Subjects Review Committee are provided in Appendix B. In contrast to quantitative studies, where the sample is chosen to address representation, generalization, replication, and detection of bias, qualitative research is processual (Morse 2008a) and need not, therefore, follow quantitative sampling requirements. Purposive and snowball sampling, also known as chain referral sampling, were used to identify mothers for qualitative interviews (Schensul et a/ 1999:240-244). Purposive sampling was accomplished by interviewing mothers receiving medical care at Salud Family Health Centers (Salud). Salud operates a number of clinics designed to serve the poor and near-poor, and un-and under-insured, in various Colorado locations. Its client base includes many women of Mexican origin. Salud's medical director agreed to have its case managers ask recent mothers of Mexican origin attending the Brighton and Ft. Lupton clinics if they were interested in being interviewed for this study. If the woman said yes, clinic personnel provided the woman's first name and telephone number to the researcher to follow up and schedule an interview. Women who indicated interest in participating in the study were contacted and a face-to-face interview was scheduled. The consent was written in Spanish and English for comprehension at the 81 h grade level. A copy of the consent was provided to the interviewee in English or Spanish, at the preference of the interviewee, and read by or explained to each mother. Mothers were not asked to sign the consent, pursuant to the waiver of documentation authorized by the Human Subjects Review Committee. Each interviewee was paid $20 for a one hour interview. The interviews were held in a location chosen by the interviewee and were conducted in English or Spanish, at the interviewee's request. Snowball sampling was employed by asking each mother if she had a friend or relative who met the selection criteria who might be interested in participating. Two mothers were solicited this way. All interviews were audio-taped using a Sony digital recorder. Relevant portions of the interviews were transcribed and translated, as necessary. Ten women were interviewed; five U.S.-born mothers of Mexican origin, and five Mexican-born mothers. Interviews were conducted until it appeared that saturation on key themes of diet, exercise, health behaviors, and cultural beliefs, especially relating to LGA, was reached. 62

PAGE 80

The interviews were semi-structured and inquired into the mother's most recent b irth and any other births, infant outcomes her use of prenatal care, i nsurance status her neighborhood, sources of social support, and her health behaviors before and during pregnancy. Special emphasis was placed on eliciting differences between life in Mexico and in the U.S. for Mexican born mothers. In addition, five key informant interviews were conducted with professionals who had relevant insight into birth outcomes among Hispanics. One key informant has worked with poor Hispanic first time mothers (primarily immigrants from Mexico) i n the Nurse Family Partner Program. A second key informant is a certified nurse midwife pract i cing in Colorado for many years in rural locations and with Denver Health. The third key informant is an obstetrician working with a primarily poor Mexican population at Denver Health The fourth i s a physician-researcher who is an expert in diabetes during pregnancy at the Univers i ty of Colorado Denver who was interv i ewed by telephone. F i nally, a woman who was a practicing phys ician in Mexico who is now affiliated with La Clinica Campesina i n Colorado as a nurse, provided insights on the differences between mothers behaviors and b irth outcomes in Mex ico and in the U S All but the telephone interview were recorded using a Sony digital recorder and relevant portions transcribed. Field notes were also used to capture the researcher s observat ions and impressions of the interviewee as well as her surroundings The consent documents and the interview guides are at Appendix C. Methods for Aim 4 Content analys i s of interviews in qualitat ive research usually follows one of three approaches: conventional directed or summative (Hsieh & Shannon 2005). Through content analysis the researcher subjectively interprets text data using a systematic classification process of coding and i dentifying categories themes and patterns Categor ies help to identify whaf is in the data and develop a taxonomy that i dentifies relat i onsh ips between categories and subcategories. Themes are meanings that run through the data (Morse 2008b) Conventional content analysis is usually used to descr ibe a phenomenon ; codes a r e developed from the text without using preconceived categor ies or theoretical perspectives. Directed content analysis is used to validate or extend a theory; codes are developed using key concepts and operational definitions come from the theoretical framework applied in the study (Morse 2008b). F indings provide support ing and non support ing evidence for a theory. Summative analysis seeks to ident i fy and quantify certa i n words or content in the text so as to understand the contextual use of the words or content to 63

PAGE 81

explore usage, not meaning This study uses directed content analys i s because the purpose of the qualitative analysis is to explore certain contradictory quantitative data in the context of theories that are being tested. After transcription and translation, the data from the interviews were analyzed for a priori categories and themes that centered on maternal weight before pregnancy, weight gain, eating and exercise habits while pregnant, sources of social support, and neighborhood characteristics Each mother was also asked if she had insight into why Mexican-born mothers have higher risks of LGA. In addition to the a priori codes certain themes emerged that were identified and ultimately organized around cultural beliefs and political economic theory, including cultural beliefs concerning diet and exercise during pregnancy, practice of Ia cuarentena after the birth and male partners attitudes and behaviors in the form of machista -protection of women partners. Several Mexican-born mothers discussed the economic difficulties of life in Mexico and in the U.S. after immigration, and the differences in lifestyle, including levels of energy expenditure and nutrition in Mexico and the U .S. Results of the qualitative interviews are reported in Chapter 5. 64

PAGE 82

CHAPTER 4 QUANTITATIVE ANALYSIS Aim 1 compares the odds of each outcome by race/ethnicity using unadjusted and adjusted odds ratios Aim 2 compares odds of outcomes by nativity. For Aim 2 the successive models examine support for the healthy migrant or healthy immigrant hypotheses. A i m 3 examines the association of neighborhood deprivation and imm i grant orientation with birth outcomes of mothers of Mexican origin in two large counties, based on Finch et a/.'s use of immigrant orientation as a general proxy for ava i lability of social support. Aim 1 Comparison of Risk Factors by Race/Ethnicity Aim 1 tests whether racial/ethnic populations of mothers in Colorado differ in their risk profiles and whether there is congruence between risk factors and birth outcomes. Based on census data for the U .S. and Colorado, Hispanics and Blacks should have demographic social, medical and health behavior risk profiles that are less advantageous than Wh i tes The null and alternative hypotheses for risk profiles are: H10 : each r i sk factor is independent of race/ethnicity. H 1 A: each risk factor is related to race/ethn icity Table 4 1 shows the frequency distribution of risk factors by race/ethnicity and the significance of differences using Pearson s chi-square as the test of the relationship between each risk factor and race/ethnicity A risk factor is considered related to race/ethnicity if ps 0.05 (Gould eta/. 2003) In Colorado the frequency of each risk factor i s related to ethnicity ( p<0 .0001 ). In addition, Hispanics experience higher frequencies of some risk factors than Blacks experience, which are underlined in Table 4.1. For these risk factors, Hispanics have worse risk profiles than Blacks. 65

PAGE 83

Table 4.1. Percent frequency distribution of risk factors by racelethnicity 200G-2005 Characteristic Age of mother S19 20-34 Parity First Low High Altitude (ft above sea level) <5000 5000-5999 6000-6999 7000-7999 8000-8999 >9000 Education (mother) < 9 yrs 9-11 yrs 12+ yrs Prenatal care Inadequate Intermediate Adequate Adequate plus Marital status Married Unmarried Smoking No Yes Alcohol drinker No Yes Weight gained <151b 15-40 lb >401b Medical risk factors None One or more conditions LGA associated risk factors Gestational diabetes Preexisting diabetes Prev infant 4000+ grams Total 10 53% 75.37% 14.11% 42.14% 44.85% 13 01% 15.38% 64.40% 14.39% 3.51% 1.44% 0.88% 6 53% 15.60% 77.87% 13.46% 17 52% 42 93% 26 10% 73.93% 26.07% 91. 66% 8 34% 99 07% 0.93% 10 51% 73.12% 16.37% 70 05% 29 95% 2 52% 0.40% 0 78% White 6.60% 75 87% 17 52% 44. 39% 45 61% 10 01% 14 94% 60.54% 17.27% 4 10% 2 07% 1 07% 0.76% 7.57% 91.67% 8 51% 16 51% 46.06% 28.92% 82 18% 17 82% 89.98% 10.02% 98 91% 1 09% 7.91% 73.99% 18 09% 74.42% 25 58% 1.25% 0.22% 0.34% Hispanic 18 09% 74.40% 7.51% 37 23% 44.12% 18 65% 19.50% 68.14% 8 24% 2 97% 0.44% 0 71% 19.78% 32.62% 47 90% 22 67% 19 67% 37 14% 20 52% 59 81% 40. 19% 94.97% 5.03% 99.36% 0 64% 15 55% 71.92% 12 53% 61. 84% 38.16% 0.95% 0 14% 0 38% Black 17 47% 73 08% 9 .45% 40 24% 40.64% 19 12% 2.21% 83.60% 13.54% 0.40% 0.21% 0.04% 2 18% 17 78% 80.05% 19.37% 16.91% 37.69% 26 03% 47.48% 52 52% 89.68% 10 32% 99 07% 0 93% 13.79% 66.02% 14.56% 62 16% 37 84% 0 11% 0.02% 0.04% Other p-value 7.20% 77.16% 15 65% <0.0001 46.01% 43. 37% 13.01% <0.0001 6.45% 74.02% 16.65% 2 00% 0.61% 0.27% <0 0001 3.58% 10.20% 86.22% <0.0001 14.26% 17 62% 43.48% 26 10% <0.0001 80 58% 19.42% <0 0001 94.81% 5 19% <0.0001 99 37% 0 63% <0.0001 9 44% 76 00% 14 56% <0 .0001 72.56% 27.44% <0.0001 0.20% 0 02% 0 20% < 0001 In accord with expectations based on census data, Hispanics have higher frequencies of social and medical risk factors than Whites (and sometimes higher than Blacks noted by an *) for : o Teen births 66

PAGE 84

o High parity births o Lower education levels o Unmarried status o Inadequate*, i ntermediate*, and adequate levels of prenatal care o Low weight gain o One or more medical risk factors o Previous infant 4000+ grams Although Hispanics compare favorably with Whites for gestational and preex i sting diabetes Blacks and Others have much lower frequencies of the specific risk factors associated with LGA Hispan ics report the lowest frequency of smoking and the second lowest frequency of dr inking alcohol dur ing pregnancy ( second to Other''). Based on CDPHE s analysis of discordance in reporting of smoking and drinking behaviors during pregnancy (see page 39), smok ing and dr inking behav i ors reported in v i tal records may understate the true rate for some groups but the data are more likely to be reliable for Hispan ics. Overall, as shown i n F i gure 4 1 the data on risk factors support a finding that Hispanics have poore r social demographic, and medical risk factors compared to Whites Blacks and Others but better self-reported smok ing and drinking behav i ors compared with Whites and Blacks Risk factors spec i f i cally associated with LGA are not shown because thei r frequency i s so low that they do not register on the scale of the graph 100 90 80 70 -Ill 60 Cl I 50 40 30 20 10 0
PAGE 85

Frequency of Adverse Birth Outcomes by Race!Ethnicity The theory of the social gradient of health, supported by the poorer risk profiles of Hispanics, suggests that they should have higher frequencies of adverse birth outcomes than Whites and Blacks. However, prior research reports that Hispanics have paradoxically better low-weight associated birth outcomes than Blacks and outcomes that approach the White majority population for LBW and preterm birth (Rosenberg eta/. 2005; Frisbie & Song 2003; Gould et a/. 2003 ; Singh & Yu 1996). Few stud i es report on SGA although Gould et a/. (2003) report that foreign-born Mexican mothers have higher frequencies of SGA than U.S. born non-Hispanic Whites and much lower frequencies than U.S-born Black or foreign-born Indian mothers. While previous research shows higher rates of diabetes in Hispanics than other population groups in the U.S (Mainous et a/. 2008 ; Martorell 2005), no population studies of LGA were found. The null and alternate hypotheses for frequency of birth outcomes unadjusted for risk factors are: H20 : each birth outcome is independent of race/ethnicity. H2A: each birth outcome is related to race/ethnicity. Pearson s chi-square is used to test the relationship between each birth outcome and race/ethnicity. In accord with other research, the unadjusted frequency of each adverse birth outcome is related to race/ethnicity. However, Hispanics have disproportionately low frequencies of low-weight associated birth outcomes as shown in Table 4.2. Despite having poorer demographic social and medical risk profiles Hispanics are second to Whites in all categories of adverse b i rth outcomes except LGA where Hispanics are lower than the majority White population, but higher than Blacks and all Others. Figure 4.2 graphically demonstrates that based on unadjusted frequencies the epidemiological paradox exists for Hispanics in Colorado for LBW, preterm birth and SGA. Rates of LGA are about the same for Hispanics and Whites, but Blacks and Others enjoy much lower rates of LGA than either Hispanics or Whites Table 4.2. Percent frequency of LBW preterm birth, SGA, and LGA by race/ethnicity 200G-2005 Birth Outcome Total White Black Other p-value LBW 6 73% 6 18 % 6 83 % 12 14 % 8.48 % <0 0001 Preterm births 7 31% 7 00 % 7.40% 10 80% 7 66% <0 0001 SGA 12.13% 11. 16% 12 .51% 19 21% 16.12% <0 0001 LGA 5 07 % 5 34 % 5 00 % 3 .11% 3 67 % <0 0001 68

PAGE 86

25 20 -+-White ... 15 c: ----Hispanic Cll u ... Black Cll 10 D. ---*-Other 5 0 0'?0'?-v g v -:...0<; '0 Figure 4 .2. UNADJUSTED FREQUENCIES OF BIRTH OUTCOMES BY RACE / ETHNICITY Odds Ratios of Birth Outcomes by Race / Ethnicity Unadjusted frequencies of outcomes do not take into account various factors that are expected to contribute to any given outcome Multiple regression permits comparisons while adjusting for contributing factors To compare odds of each of the four birth outcomes multiple logistic regression for each dichotomous outcome was performed creating odds ratios based on race/ethnicity alone (Model 1 ) and then adjusting successively for demographic and socioeconomic characteristics (Model 2) medical conditions associated with pregnancy (Model 3), and health behaviors available in the birth record (Model 4). The reference group is White mothers based on their superior risk profiles and the theory of the social gradient of health. The risk profiles of Hispanics and Blacks in Colorado predict that they should have poorer birth outcomes than non-Hispanic Whites If Hispanics have odds of adverse outcomes similar to those of non-Hispanic Whites the data suggest that a paradox may exist at least for these outcomes at the individual/compositional level. H30 : race/ethnicity does not predict each birth outcome after adjusting for risk factors H3A: race/ethnicity predicts each birth outcome after adjusting for risk factors. The results of the multiple logistic regression are reported and discussed separately for each outcome. 69

PAGE 87

Low Birth Weight Table 4.3 reports the summary odds of LBW, including confidence intervals at 95%, adjusting for each successive model. Also reported are the c statistic and the Hosmer Lemeshow Goodness of Fit statistic for each model. Table 4.3. Unadjusted and adjusted odds ratios (95% Cl) of LBW by race/ethnicity Race White Hispanic Black Other c statistic Hosmer-Lemeshow Model 2 Model 3 Model 1 Adjusted for Additional Race/Ethnicity Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1 00 1 00 1 00 1 .11 (1.08-1. 15) 1.07 (1.03-1. 10) 1 15 (1.10-1.20) 2 10 (1. 99-2 21) 1 .91 (1. 81-2 02) 2 09 (1. 94-2 25) 1.41 (1.33-1 49) 1.46 (1. 38-1.55) 1.62 (1.50-1. 75) 0.535 0 687 0 704 0 9978 <.0001 < 0001 Model4 Additional Adjustment for Health Behaviors 1.00 1.18 (1.13-1. 23) 2 16 (2 01-2 33) 1.64 (1.52-1. 77) 0 733 0.0009 There is no monotonic increase or decrease in odds across the successive models although in the fully saturated model the odds of LBW are slightly higher than in Model 1 (race/ethnicity alone). Hispanics continue to have the closest odds of LBW to those of Whites notwithstanding their poorer risk profiles and almost 100% better odds than Blacks This particular model has a modest degree of discrimination (c statistic= 0 733) and the fit is not adequate suggesting that even with the high number of interactions (five) other factors account for much of the variation by race/ethnicity. 2 50 -+---Non-Hispanic White Ill 2 00 --0 ---Hispanic ; 1 50 Ill a: .. -Ill 1 00 Black "C "C 0 0 .50 0 00 2 3 4 Models Figure 4 3 LBW ODDS RATIOS BY RACEIETHNICITY AND MODEL 70

PAGE 88

Table 4.4 reports the significant standardized beta coefficients and odds ratios for each variable in the models for LBW by race/ethnicity. LBW by race/ethnicity has five interactions: parity*marital status, medical risk*race/ethnicity, smoking*age, prenatal care*weight gain and prenatal care*smoking. Table4.4. Estimated coefficients and odds ratios for LBW and racelethnlclty Variable Model1 Model2 Model3 Model4 OR OR OR OR Race White(r) 1 00 1 00 1.00 1 00 His12anic 0 1076. 1.11(1.08-1.15) 0 06421 1.07(1 03-1.10) 0 1383. 1.15(1 1 0-1.20) 0 1650. 1.18(1 13 1.23) Black 0 7414" 2.1 0(1.99 2 21) 0 6489. 1 91(1.81 2 02) 0 7365 2.09(1 94-2 25) 0 7706. 2 16(2 .012 33) Other o.34o8 1 41! 1.33 1.49) 0 3799 1.46! 1 .38-1 55) 0 4802 1 .62(1. 50 1 75) 0 4944. 1 .64(1.52-1. 77) A e 20 -34(r) 1.00 1 00 1.00 S 19 0 08281 1 .09(1. 04 1 14) 0 0981. 1.1 0(1.05-1.16) 0 1794. 1 .20(1.14 1 26) > 35 0 1704. 1.19(1.14 1.23) 0 1223. 1.1 0.0874. 1 09(1.05-1.14) Pari Low(r) 1.00 1.00 1.00 F i rst 0.4326. 1 .54(1.491 60) 0 4248. 0 4473. High 0 2572 1 .29(1.23-1. 36) 0 2369. 1 .27(1.20-1. 34) 0.1673. 1.16(1.1 0-1.22) E l evat i on < 5000 (r) 1 .00 1.00 1.00 5000 5999 0 0371 1.00 0 0240 1.00 0 0796 1 .08(1. 04 1 13) 6000-6999 0 .0793! 1.08(1.03-1.14) 0 1080 1.11 (1. 06-1.17) 0 1619. 1.18(1.12-1. 24) 7000 7999 0 2435. 1 .28(1. 08 1 38) 0 1895. 1.21(1.12 1 31) 0 2228. 1 .26(1. 16 1 36) 8000 8999 0 3250. 1 .38(1. 24-1.55) 0 3284. 1.39(1 241. 55) 0.4166. 1 .52(1. 38-1. 70) > 9000 0 5897" 1.80(1 .602.04) 0 5977" 1.82(1 60 2 06) 0 6775. 1.97(1. 73 2 24) Education HS grad (r) 1.00 1 .00 1.00 No h i gh schoo l 0 0537 1.00 0 0173 1 .00 -0 0323 1.00 Some high school 0 2284. 1.26(1 .211.31) 0.2097 0 1395. 1.15(1.1 0-1.20) Mari tal status Married (r) 1 .00 1.00 1.00 Not marr i ed 0.4385 1 55(1.47-1 63 0 4143. 1.51 (1.44-1. 59) 0 3094. 1.36(1 29 1.44) Prenatal care Adeguate (r) 1.00 1 00 1 00 ln adeguate 0 5303 1 .70(1.63-1.78 0 4526 1.57(1.50 1 64) 0 3874. 1.4 7! 1.39 1 56) Intermediate .0 1007" 0 90(0 .860 95) 0 1082 0 90(0 86 0 94) .0 13539 0 86(0 8 10 91) Adeguate 121us 1 2532 3.50(3 39 3 61) 1.2167. 3 38(3.27 3 49) 1 2609 Pari mar i tal Low-married (r) 1.00 1.00 1.00 111 married 0.4326 1.54(1.49-1.60) 0 4282. 0 5225. 1.69(1 .63-1. 75) H i ghmarried 0 2572. 1 .29(1.23-1. 36) 0 2369. 1 27(1.20-1.34) 0 1472. 1.16(1 1 0 -1. 22) 111 unmarr i ed 0 5694. 1.77(1.69 -1. 85) 0 5419. 1 72(1. 641 80) 0 5769. 1.78(1 70-1.87) LoW" unmarried 0 4383. 1.55(1 .471 63) 0.4143. 1.51 (1.441. 59) 0 3094. 1.36(1 29 1.44) H ighunmarr i ed 0 6550. 0 6268. 1 .87(1. 75 2 00) 0.4601. 1 58(1.48-1.70) Medical r i sk N one r 1 .00 1.00 One or more 0 7668. 2 15(2 8 2 23) 0.7208. 2 06( 1. 98 2 13) Med i cal riskrace Wh i te no risk (r) 1.00 1 00 His12anic no r i sk 0 1383 1 .15(1.10-1. 20) 0 16so 1 .18(1.13-1. 23) Black no risk 0 7365. 2 .09(1. 94 2 25) 0 7706. 2.16(2 01-2.33) Other no risk 0 4802. 1 .62(1. 50 1 75) 0.4944. 1 .64(1. 52 1 77) Wh ite w / r i sk 0 7668 2 15(2 08-2 23) 0 7208. 2 .06(1. 98 2 13) His12ani c w / r i sk 0 6241. 1 .88(1.78-1. 96) 0 6002. 1.82(1.741 9 1 ) B lack w /risk 1.1829. 3 2 6!3.02 3.53) 1 1213 3 07(2 83-3.32) Other w / risk 0.9791. 2 66(2.42-2 93) 0 9711. 2 64(2.40 2 9 1 ) Smok i n None r 1 .00 Smok i ng 0 7584. 2 .14(1. 97 2 31) 71

PAGE 89

Table 4.4. Estimated coefficients and odds ratios for LBW and racelethnicity (continued) Variable Wei ht aln 16-40 r <16 >40 Smokin a e No sm"20-34 (r) No smoke"<19 No smoke">35 Smoke"<19 Smoke"20-34 Smoke>35 Prenatal"wei ht Adegmed wt (r) lnad"lowwt Inter" low wt Adeg"lowwt Adeg+"low wt lnad"med wt lntermed wt Adeg+"med wt lnad"high wt h i gh wt Adeg"high wt Adeg+"high wt Prenatal" smoke Adegno sm (r) lnad"no smoke Inter no smoke Adeg+"nosm lnad"smoke Inter smoke Adegsmoke Adeg+"smoke (r) Indicates reference category p <.0001. t p <.001 Model1 Model2 Model3 OR OR I! OR The contribution of some risk factors is not unexpected based on prior studies : Model4 I! OR 1 00 0 8510" 2 34(2 18 51) 0 7491" 0 47(0 43-0 52) 1 .00 0.1794" 1.20(1.14-1.26) 0.0874" 1.09(1 05 1 14) 0 4761" 1 61(1.55-1 67) 0.0221 1.00 0 4802" 1.62(1 40-1 86) 1.00 1 2008" 3 32(2 .90-3. 81) 1.7243" 5 61(4 71-6 68) 0 8510" 2 34(2 18-2 51) 1 5836" 4 87(4 50-5 28) 0 7208" 2.06(1 81-2 34) 0 8407" 2.32(1 98-2 72) 0 7894" 2 20(2 06-2 35) 1 7366" 0 18(0 12-Q.26) -3 7446" 0 02(0 02-Q 02) -Q7491" 0 47(0 43-Q 52) -Q.9561 0 38(0 37-Q.40) 1 .00 -Q. 3334" 0.72(0 63-0 81) -Q.9946" 0 37(0 .32-Q. 43) 0 4715" 1 60(1 49-1 72) 0 5184" 1 68(1.43-1 97) -Q. 1850 1 .00 0 7364" 2 09(1 86-2 34) 0 9454" 2 57(2.28-2 91) o Odds of delivering an LBW inf ant increase for teen mothers {1.20) and older mothers (1.09) o First babies have higher odds (1.69) of being LBW o Odds of LBW show a monotonic increase as residence at elevation above 5000 feet increases {1.08-1.97) o Being unmarried increases the risk of LBW {1.36) o Mothers with inadequate (1.47) and adequate plus prenatal care (3.53) have increased odds of LBW o The presence of one or more medical risks associated with pregnancy is a marker tor increased odds of LBW (2.06) o Smoking increases the odds of LBW by over 100% (2.14) 72

PAGE 90

o The odds of LBW with weight gain of less than 16 pounds increases by 134% (2.34) o Weight gain of >40 pounds is very protective against LBW gaining more than 40 pounds reduces the odds of LBW by 53% (0.47) Some results are unanticipated: o Low education (less than high school} is not significant o Intermediate prenatal care (one level below adequate) does not increase the odds of LBW (0.86) In the fully saturated model the factors having the greatest predictive power for increased odds of LBW are adequate plus prenatal care (3.53), presence of one or more medical risks (2.06), being Black (2.16} smoking (2.14), and first parity (1.69). Interactions highlight the effect of some of these risks for LBW. The interaction of parity and marital status has a small impact on odds of LBW : first parity married = 1.69, first parity*unmarried = 1. 78. Whereas intermediate prenatal care seems protective (0. 86), the interaction of intermediate prenatal care and low weight gain increases odds to 5.61. The interaction of race and medical risk is most interesting. In Model 4, White mothers with at least one medical risk have 2.06 higher odds of LBW compared to White mothers with no medical risks (1.00 reference) Even though Hispanics have poorer medical risk profiles than Whites Hispanics with at least one medical risk have an elevated odds of LBW compared to White mothers with no risk (1. 82), but lower odds than those of Whites with medical risks (2.06) Hispanics with no medical risk compare favorably with Whites with no medical risk with odds of an LBW birth of 1.18. The ranking of odds of LBW by race/ethnicity for mothers with one or more risks places Hispanics first (1. 82), then Whites (2. 06), followed by Others (2.64) and Blacks (3.07). Preterm Birth Table 4.5 reports unadjusted and adjusted odds ratios of preterm birth by race/ethnicity Each successive model improves the fit, with Model 4 having adequate fit and modest discrimination Although Hispanics have higher frequencies of social and medical risk factors than Whites, and sometimes higher than Blacks Hispan i cs odds of having a preterm birth are close to those of Whites in Models 1, 2, and 3, and are no different than the odds of Whites i n Model 4. 73

PAGE 91

Table 4.5. Unadjusted and adjusted odds ratios of preterm birth by race/ethnicity Race White Hispanic Black Other c statistic Hosmer-Lemeshow Model 2 Model 3 Model 1 Adjusted for Additional Race/Ethnicity Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1 00 1.00 1 00 1.06 (1.03-1. 09) 1 08 (1.05-1. 12) 1.03 (1.00-1.07) 1.61 (1.53-1. 70) 1 54 (1.45-1 63} 1.45 (1. 37-1.54) 1 10 (1.04-1. 17) 1 .17 (1.10-1. 25} 1 16 (1.09-1. 24) 0.518 0 735 0 753 0 9999 < .0001 0 0006 Odds not significantly different from 1 00 Model4 Additional Adjustment for Health Behaviors 1 00 *1.01 (0. 98-1.05) 1.42 (1. 34 1 .51) 1 16 (1. 09-1.23 ) 0 765 0 1892 Figure 4.4 shows the change in odds for preterm birth with the addition of risk factors in the successive models For Hispanics and Blacks the movement of odds trends generally down with successive models ; with Others the odds increase as risk factors are accounted for 2.00 Ill 1 50 0 ;; Ill a: 1 .00 Ill "C "C 0 .50 0 0 .00 2 3 Models 4 -+--Non-Hispanic ------Hispanic Black Figure 4.4 PRETERM BIRTH ODDS RATIOS BY RACE / ETHNICITY AND MODEL Table 4.6 reports the significant standardized beta coefficients and odds ratios for each variable in the models for preterm birth by race / ethnicity Preterm birth by race/ethnicity has four interactions: prenatal care medical risk, prenatal care*weight gain parity*weight gain and prenatal care*smoking. 74

PAGE 92

T a b l e 4 6 Estimated coeff i cients and odds r at i os for preterm birth and racelethn i c i ty Variable Model 1 Model2 Model3 Model4 13 OR 13 OR I! OR 13 OR Race White(r) 1 00 1 00 1.00 1 00 H i span i c 0 0607 1.0 6(1. 03-1.09) 0.0791" 1 .08(1. 05-1 12) 0 0317 1 .00 0 0130 1 00 Black 0 4761" 1.61(1 .53-1. 70) 0 4291" 1 .55(1. 45 1.63) 0 3710" 1.45(1 .37-1. 54) 0 3507" 1.42(1 34 1.51) Other 0 0971" 1 .10(1.04-1. 17) 0 1582" 1 .17(1. 10 1 25) 0 1515 1.16(1.09-1.24) 0 1457" 1 16(1.09 1.23) A e 20-34(r) 1 00 1 .00 1.00 S19 0 0264 1 00 0 0441 1.00 0 .0757! 1.08(1 .031 13) > 35 0 1319" 1 14(1. 10 1.19) 0 0835" 1 .09(1. 05 1 13) 0 0749" 1 .08(1. 04 1 12) Pari Low(r) 1 .00 1 .00 1 00 F ir st 0 1823" 1.20( 1 .17 1 24) 0 1768" 1 .19(1.16-1. 23) 0 2163" 1 24( 1 99 1 29) H i gh 0 268 6 1 .31(1.26-1. 36) 0 2542" 1 .29(1.241 34) 0 2260 1 .25(1.19-1. 32) E l evat i on < 5000 (r) 1 00 1 .00 1 .00 5000 5999 -0. 0083 1 00 -0 0246 1 00 0 0083 1 00 6000 6999 -0 0085 1 00 0 0176 1 00 0 0483 1 .00 7000 7999 0 0701 1 00 0.0112 1.00 0 0261 1 .00 8000-8999 0 1752! 1 19(1.07 1 33) o 1n9t 1 20(1.07-1. 34) 0 2287" 1. 26(1.12-1. 41) > 9000 0 1563 1 .17(1. 02 1 35) 0 1606 1 .17(1. 02 1 35) 0 .2040! 1.23( 1 061 41) Education HS grad (r) 1 .00 1 .00 1 00 No high school 0 0587 1.00 0 0122 1 .00 -0 0250 1 00 Some h i gh school 0 1504" 1 .16(1. 12 1 21) o .12n 1 .14(1. 09 1 18) 0 0974" 1 10( 1 06 -1.1 5) Marital status Married (r) 1 .00 1 .00 1 00 Not manried 0 1650" 1 .18(1. 14 -1. 22) 0 1454" 1 16( 1 .12-1. 19) 0 1225" 1.13( 1.09-1.17) Prenata l care Adequate (r) 1 00 1 .00 1 00 Inadequate 0.8838" 2.42(2 .312 53) 0 9714" 2 64(2 48-2 82) 0 938 6 2.56(2 38-2 75) Intermediate -0.0131 1 00 0.0211 1 00 -0. 0193 1.00 Adequate plus 1 8939" 6 65(6 42-6 88) 1 9620" 7 11(6 80-7 44) 1 9959" 7 36(7 00-7 74) Medical risk None r 1 .00 1.00 One or more 0 8538" 2 35(2 .212.49) 0 8181" 2 27(2 13-2 41) Prenatal"medrisk Adeq no risk (r) 1 .00 1 00 lnadeq no risk 0 9714" 2 64(2.48-2 82) 1 2501" 3 50(2 94-4 1 4) Inter no risk 0 .0211 1 .00 -0 2997 0 74(0 6 1-0. 91) Adeq+ no risk 1 9620" 7 11(6 80-7 44) 1.8346 6 23(5 50-7 14) l nadeq w /ri sk 1 4576" 4 30(4 04-4 57) 1.6737" 5 33(4 50-6 32) I nter w / risk 0 7669" 2 .15(1. 99-2 33) 0 38671 1.4 7( 1 .2Q-1 8 1 ) Adeq w /ri sk 0 8538" 2 35(2 21-2 49) 0 8181" 2 27 ( 2 1 3-2.4 1 ) 2 5864" 1 3 28(12 661 3 93) 2 4390 1 1 46( 1 0 05 13 07) 1 00 0 3360" 1 .40(1. 27 1 54) 1 00 0.7471" 2 .11(1.932 31) -0. 5325" 0 59(0 52-0 66) Prenatal"wei ht Adeq"med wt (r) 1 00 lnad"lowwt 0 7151" 2 .04(1. 73 2.42) Inter low wt 1 2367" 3 44(2 76-4.29) Adeq"lowwt 0 9215" 2 51(2 27 2 78) Adeq+ "low wt 0 8793" 2 .41 (2. 09 2 78) lnad"med wt 0 2007 1 .22(1. 05 -1. 43) l ntermed wt 0 .2804! 1 .32(1. 091 62) Adeq + "med wt 0 1613 1 1 8(1.04-1. 33) 2 0080" 7 45(7 .337 57 1 4301" 0 24(0 24-0 24) lnad h i gh wt lnter"high wt 75

PAGE 93

Table 4.6. Estimated coefficients and odds ratios for preterm birth and racelethnicity (continued) Variable Model1 13 OR Adeg'high wt Adeg+'high wt Pari 'wei ht Low"med wt (r) 111'1owwt Low"lowwt High'lowwt 151 'medwt High'med wt 1". high wt Low" high wt High'high wt Prenatal" smoke Adegno sm (r) lnad"no smoke Inter" no smoke Adeg+'no sm lnad'smoke Inter' smoke Adeg'smoke Adeq+'smoke (r) in d i cates reference category p<. 0001 tp < .001. :j:p <.01 p < 05 Model2 Model3 13 OR 13 OR As with LBW, some results are not surprising based on prior studies: Model4 13 OR -Q.2771t 0.76(0.65-0 88 3 .2330' 25 36(25 .2 4 -25.4 8) 1 .00 0 .3279' 1.39(1 27-1 51) 0 1744' 0 84(0 77-0 91) 2 1689' 8 75(7 .719 92) 0 .2163' 1 .24(1.20-1. 29) 0 .2260' 1 .25(1.19-1. 32) -0 1573 0 85(0 75-0 97) 0.25541 0 78(0 68-0 89) 3 9465" 51. 75(46 19-58 00) 1 .00 -0. 5122" 0 60(0 55-0 65) -0 3250" 0.72(0 56-0 94) 4.1705" 64 75(53.00-79 10) 3 .4360' 31.06(21 1 0-40 03) 0.4120" 1.51(1.30-1 75) 0 3360" 1.40(1.27-1 54) 0 .1030:1: 1 .11(1. 04-1 18) o The presence of medical risks is a marker for increased odds of preterm birth, raising odds by 127% (2.27) o Smoking increases the odds of preterm birth by only 40% compared with an increase of 114% for LBW, but smoking with inadequate prenatal care raise odds to 31.06 o Inadequate and adequate plus prenatal care are markers for increased odds of preterm birth (inadequate 2.56; adequate plus 7.36)-much higher increases than for LBW. Intermediate prenatal care alone is not significantly different from adequate care. o Being unmarried increases the odds of preterm birth slightly (1.13) o Both first birth (1.24) and high parity (1.25) increase the odds of preterm birth o The odds of preterm birth increase over 1 00% with weight gain of less than 16 pounds (2.11 ). Weight gain >40 pounds is protective against preterm birth -gaining more than 40 pounds reduces the odds of preterm birth by 41% 76

PAGE 94

Some results are unanticipated: o Age has little effect on odds of preterm birth. Odds increase only 8% for both teen births and births to older mothers. o Altitude affects preterm birth only if mother's residence is above 8000 feet o No high school is the same as being a high school graduate in all models o Intermediate prenatal care (one level below adequate) is the same as adequate prenatal care o The interaction of high parity and high weight gain raises the odds to 51.75 o The interaction of adequate plus prenatal care and no smoking raises odds to 64.75, which is highly perplexing In the fully saturated model, the factors having the greatest predictive power for increased odds of preterm birth are being Black (1.42) having inadequate (2.56} or adequate plus (7 .36) prenatal care, gaining less than 16 pounds during pregnancy (2.11 ), the interaction of high parity*high weight gain (51.75), and the interaction of smoking*inadequate prenatal care (31.06). Although intermediate prenatal care alone is the same as adequate care, when intermediate care is combined with low weight gain, the odds of preterm birth rise to 3.44, and when it is combined with smoking, the odds rise to 1.51. The interaction of prenatal care and medical risk is striking. Needing adequate plus prenatal care combined with one or more medical risks is associated with 11.46 higher odds of preterm birth. Indeed, with the exception of intermediate care without any medical risks, having anything other than adequate care with no medical risks raises the odds of preterm birth substantially (from 1.47-11.46). Some interactions raise the odds even higher. Small for Gestational Age SGA has some overlap with LBW and preterm birth, but not all SGA babies are low birth weight or preterm. Indeed, the frequency of SGA babies is almost double the frequency of LBW in Colorado during the study period as reported in Table 4.2 above. Table 4.7 reports unadjusted and adjusted odds ratios of SGA by race/ethnicity As with LBW and preterm birth, Hispanics have odds ratios closer to Whites (but in no model are the odds of SGA the same as Whites), while Blacks and Others have much higher odds of SGA. Hispanics continue to have the closest odds of SGA to those of Whites, notwithstanding their poorer risk profiles. This particular model has a poor degree of discrimination (c statistic = 0.634) and the fit is not adequate, suggesting that, even with the 77

PAGE 95

Table 4.7. Unadjusted and adjusted odds ratios (95% Cl) of SGA by race/ethnicity Race White Hispanic Black Other c statistic Hosmer-Lemeshow Model 2 Model 3 Model1 Adjusted for Additional Race/Ethnicity Only Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1 00 1 00 1 00 1 14(1 .11-1. 16) 1.05(1 .02-1. 08) 1 .11 (1.07-1.14) 1 89 (1. 82-1.98) 1 73 (1.66-1. 81) 1.80 (1.79-1. 90) 1 .53 (1.46-1 60) 1 .53 (1.47-1 61) 1.60 (1.51-1 69) 0 535 0 593 0 597 0.9998 < .0001 <.0001 Model4 Additional Adjustment for Health Behaviors 1.00 1 .18 (1. 15-1.22) 1 98 (1. 86-2.09) 1 72 (1.63-1 82 ) 0 634 0 0005 As shown in Figure 4.5, the change in odds for SGA moves in the same pattern for Hispanics and Blacks: odds decrease in Model 2 and increase in Models 3 and 4 to levels higher than all previous models. There is no change in odds between Models 1 and 2 for Others; odds increase in Model 3 and 4 to levels higher than previous models 2 50 --+--Non-Hispanic 2 00 -----, White Ill ---Hispanic 0 ; 1 .50 Ill a: ... Ill ... Black "C 1 00 "C 0 0 50 0 00 2 3 4 Models Figure 4 5 SGA ODDS RATIOS BY RACE / ETHNICITY AND MODEL Table 4.8 reports significant standardized beta coefficients and odds ratios for each variable in the models for SGA by race/ethnicity 78

PAGE 96

Table4.8. Estimated coefficients and odds ratios for S GA and racalethnlclty Variable Model1 Model2 Model3 Model4 OR OR OR OR Race White(r) 1 00 1 00 1 00 1 00 His11anl c 0 1298. 1 .14(1.11-1. 18) 0 .0468! 1 .05(1. 02 1 08) o 1oo5 1 .11(1. 07 1 14) 0 1683. 1 .18(1.15-1. 22) B l ack 0 6384 1 89(1 82-1.98) 0 5478. o 5855. 1 80(1 79 1 90) 0 6804. 1 98( 1 .862 09) Other 0 4253. 0 4277. 0 4682. 1 6Qi1.51-1 69) 0 5431. 1 72(1.631 82) A e 20 -34(r) 1 00 1.00 1 00 S 1 9 0 0232 1 00 0 0262 1 00 0 0938. 1 10(1.06-1. 14) >35 0 0228 1 00 0 0097 1.00 -0. 0053 1 00 Pari Low(r) 1 00 1 00 1 00 First 0 3972. 1.49(1 45 1 53) 0 3960. 1.49(1 45 1 53) 0.4591. 1 .58(1.54-1. 63) H i gh 0 0442 1.05(1 .00-1. 09) 0 0379 1 00 -0. 0307 1 00 Elevat i on <5000 ( r ) 1 00 1 00 1.00 5000 5999 0 0174 1 00 0 0157 1 00 0 .0553! 1 .06(1.03-1. 09) 6000 6999 0 1235. 0 1316 1 .14(1. 10 1.19) 0 1721. 1 .19(1. 141 23) 7000 7999 0 2630. 1 .30(1. 23 1 38) 0 2495. 1 .28(1.21-1. 36) 0 2861. 1 .33(1. 26-1.41) 8000 8999 0 3643. 1 .44(1.33-1. 56) 0 3650. 1.44(1.33-1. 56) 0.4248. >9000 0 7740. 2 .17(1. 98-2 37) 0 7751. 2 .17(1 98-2 38) 0 8334. 2 30(2 1 0-2 52) Educat i on HS grad (r) 1 00 1 00 1 00 No h igh school 0 0115 1 00 0 0047 1 00 -0. 0313 1.00 Some h i gh school 0 2123. 1 .24(1. 20 1 28) 0 2076 1 00 0 1369. 1 15(1.111.18) Marita l status Married (r) 1 00 1 00 1 00 Not married 0 3940. 1 .48(1. 43 1 54) 0 3856. 1 .47(1.411 53) 0 2773. 1 .32(1.27-1. 37) P r enata l care Adeguate ( r ) 1 00 1 00 1 00 lnadeguate 0 1214" 1 .1;M1. 09 1 17) 0 1036. 1 11(1.07 1 1 5) 0 .0301 1 00 Intermed i ate -0. 0216 1 00 -0. 0225 1 00 -0. 0 23 0 97(0 941 00) Adeguate 1:11us 0 1507" 1 .16(1. 1 31 19) 0 1 363. 1.15(1. 1 2 1 1 8) 0 1 279. 1 14(1. 11-1.17) Pari marital Low'"married (r) 1 00 1.00 1 00 111 married 0 3972. 1.49(1 45-1. 53) 0 3960. 1 .49(1. 45 1 53) 0 4591. 1 .58(1.54-1. 63) H i gh married 0 0442 1 .05(1. 00 1 09) 0 0379 1.00 -0. 0307 1 00 11 1 u nmarried 0 5821. 1.79(1.73-1 85) 0 5728. 1 77(1.71 1 84) 0 5697" 1 .77(1.711 83) Low'" unmarried 0 3940. 1 .48(1. 43 1 52) 0 3856. 1 .47(1.411 53) 0 2773. 1 32(1.27-1 37) High unmarried 0 4048. 1 .50(1.42-1. 58) 0 3951. 1 49(1.41-1 57) 0 2427" 1 .28(1.21-1. 35) Medi cal risk None r 1.00 1 00 One or more 0 2686. 1 31( 1 27 -1. 35) 0 2309. 1 26( 1 .22-1. 30) Medi cal riskrace Whiteno risk (r) 1.00 1.00 H is11anl c no risk o 1oo5 1 .11(1. 07 1 14) 1.1 683. 1 .18(1. 15-1. 22) Black no risk 0 5855. 1 .80(1. 79 1.90) 0.6804. 1 .98(1. 86-2 09) Other no risk 0 4682. 1 .60(1.51-1. 69) 0 5431. 1 .72(1.63-1. 82) Whitew/risk 0 2686. 1 3 1(1. 27 -1. 35) 0 2309. 1 26( 1 22 1 30) 0 1 687" 1 1 8(1.14 1 23) 0 2079. 0 6922. 2 00(1.87 2 1 4) 0 7364. 2 09( 1 95-2 24) 0 5912. 1 .81(1. 67 1 96) 0 6712. 1 96( 1 80 2 .13) 1 00 0 8383. 2 .31 (2.22 2 4 1 ) 1 00 0 3045. 1 36( 1 32-1. 40) -0. 6227" 0 54(0 52-0 55) Smok l n a e No s m o-34 (rl 1 00 No smoke.<19 0 0938. 1 10(1.061 .14) No smoke >35 0 0053 1 00 79

PAGE 97

Table 4.8. Estimated Coefficients and Odds Ratios for SGA and RaceiEthnicity (continued) Variable Model1 Model2 Model3 Model4 13 OR 13 OR 13 OR OR Smoke.<19 -<>. 0032 1 00 Smoke-34 0 1505 1 00 Smoke >35 0.4433. 1.56(1 .27-1. 91) Smokin race Whiteno smoke (r) 1 00 Whitesmoke 0 6878. 1 .99(1. 65-2.40) Hispsmoke B l acksmoke Other smoke H i spno smoke Black no smoke Other no smoke (r) indicates reference category p < 0001 t p <.001 t:ps01 p < 05 0 6258. 1 .87(1. 54-2 28) 0 38871 1.48(1. 1 8-1.84) -1. 1117" 0.33(0 32-{) 34) -<>.2 196. 0 78(0 69-{) 88) 1 .0281. 2 8(2 55-3 06) 2 3260. 10 24(8 75-11.97) Risks for SGA are similar to those of LBW, but their effects are generally more attenuated. The contribution of some risk factors is not unexpected based on prior studies: o First babies have higher odds (1.58} of being SGA o As with LBW, SGA is sensitive to elevation, showing a monotonic increase with residence at elevation above 5000 feet (1.06-2.30) o Being unmarried increases the risk of SGA (1.32) o Adequacy of prenatal care is not strongly associated with SGA. Inadequate care in Model 4 is not significantly different than adequate care and intermediate care has slightly lower odds than adequate care (0.97). Even adequate plus prenatal care is not a strong marker for SGA (1.14} o The presence of medical risks associated with pregnancy has a relatively weak association with SGA (1.26) o Weight gain of >40 pounds is very protective against SGA ga i ning more than 40 pounds reduces the odds of SGA by 46% o Age was not significant in building the model for SGA by race/ethnicity, but it was retained based on the literature. Age is significant only in Model 4 and raises the odds of SGA minimally (1 0%} and only for teen mothers Nevertheless, some results are unanticipated: o Smoking increases the odds of SGA by a greater margin than i t does for LBW (2.31 com pared with 2.14) 80

PAGE 98

o The odds of SGA with weight gain of less than 16 pounds increases by only 36% compared with 134% for LBW o Low education (less than high school) is not significant in any model; some high school is associated with slightly elevated odds of SGA (15%) o Intermediate prenatal care (one level below adequate) does not increase the odds of LBW (0.86) In the fully saturated model, the factors having the greatest predictive power for increased odds of LBW are race (Black 1.98, Other 1.72), residence above 8000 feet (1.53-2.30), and smoking (2.31 ). Interactions highlight the effect of some of these risks for SGA. While smoking is a risk factor for SGA in general, Whites who smoke have higher odds of delivering an SGA baby (1.99) than Hispanics who smoke (1. 87) or Blacks (1.48) The results for the interaction of smoking and Other are perplexing. Other*non-smoking has an odds of 1 0.24 while Other* smoking has an odds of 0 33 These odds are a result of linear combinations of other variables. Repeated checks disclose no coding or computational error. The interaction of race and medical risk is similar to that for LBW. In Model 4, White mothers with at least one medical risk have slightly higher odds of SGA (1.26) compared to White mothers with no medical risks. Even though more Hispanics have a poorer medical risk profile than Whites Hispanics with at least one medical risk have an elevated odds of SGA (1. 23) compared to White mothers with no risk, but slightly lower odds than those of Whites with medical risks (1.26) The ranking of odds of SGA by race/ethnicity for mothers with one or more risks places Hispanics first (1.23), then Whites (1. 26), Others (1.96), and Blacks (2.09), thus supporting an epidemiological paradox in favor or Hispanics for SGA. Large for Gestational Age Table 4.9 reports unadjusted and adjusted odds ratios of LGA by race/ethnicity For LGA Hispanics have equal odds ratios to those of Whites in Model 2, and have statistically significant lower odds ratios than Whites for Models 1, 3 and 4, although the differences are small, ranging from 5-7%. Blacks and Others have much lower odds ratios for LGA than either Whites or Hispanics ranging from 32% to 43% decreased odds of LGA, making LGA a White and Hispanic phenomenon Although the fit is adequate, is discriminating power (0.650) is low. 81

PAGE 99

Table 4.9. Unadjusted and adjusted odds ratios (95% Cl) of LGA by race/ethnicity Race Non-Hispanic White Hispanic Black Other c statistic Hosmer Lemeshow Model 2 Model 3 Model 1 Adjusted for Additional Race/Ethnicity Only Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1 00 1.00 1 00 0 .93 (0. 90-0.97) 0.97 (0.93-1. 01) 0.95 (0.92-0 .99) 0 .57 (0. 52-0.62) 0 .63 (0. 57-0 69) 0 .61 (0. 56-0 68) 0 .68 (0. 62-0 74) 0 .68 (0. 62-0 74) 0 .68 (0. 62-0 74) 0 522 0 602 0 605 0 9999 0 0403 0 0084 Odds not significantly different from 1 .00. Model4 Additional Adjustment for Health Behaviors 1 .00 0 .95 (0.91-0 .99) 0 .59 (0. 54-0 .65) 0.68 (0. 62-0 74) 0.650 0 1912 Figure 4.6 reports the odds ratios of LGA by race/ethnicity and model. Odds of LGA behave differently than odds of LBW SGA or preterm birth. As noted above Whites have the highest odds of LGA, followed closely by Hispanics Blacks and Others have much lower odds of LGA For the category of Others the odds remains the same across all models 1 20 .-)E -Ill 1 00 0 0 .80 i a: 0 60 Ill "C 0.40 "C 0 0 20 0 00 2 3 Models -4 -+-LGA Non Hispanic White --LGA Hispanic LGA Black --*-LGA Other Figure 4 6 LGA ODDS RATIOS BY RACE/ETHNICITY AND MODEL Table 4.10 reports the significant estimated coefficients and odds ratios for each var i able. Unlike the situation with LBW preterm birth and SGA there are no significant interactions among any of the variables so the odds remain relatively constant across all models. In addition adequacy of prenatal care was not significant in the model building for LGA but that variable is retained nonetheless based on the literature 82

PAGE 100

Table 4.1 0. Estimated coefficients and odds ratios for LGA and racelethnlclty Variable Model1 I! Race White(r) Hispanic -0.0690' Black -0.5645' Other -0 .3910' A e 20-34(r) S19 >35 Pari Low(r) First High Elevat i on <5000 (r) 5000 5999 6000 6999 7000 7999 8000 8999 >9000 Education HS grad (r) No high school Some high school Marital status Married r Not married Prenatal care Adequate (r) Inade quate Intermediat e (r) i nd i cates reference category p < 0001 t p < .001. tps01 p < 05 OR 1 00 0 93(0 90-0 97) 0 57(0 52-0 62) 0 68(0 62-0 74) Model2 ModEII3 I! OR I! OR 1.00 1 00 -0. 0353 1 00 -0.0493 0 95(0 92-0 99) -0 4697" 0 63(0 57-0.69) 0.4871' 0 .61 (0 56-0 68) 0 .3889' 0 68(0 62-0 74) -0.3904' 0 68(0 62-0 74) 1 00 1.00 -0 4007" 0 67(0 62-0 72) -0 3965' 0 67(0 63-0 72) 0 1844' 1 .20(1. 06 1 25) 0 1744' 1 .19(1.14-1. 24) 1 00 1 .00 0 .4076' 0 67(0 64-0 69) -0.4089' 0 66(0 64-0 69) o onot 1 08(1.03-1 13) 0 .0723! 1 .08(1. 03-1 12) 1 00 1 00 -0 .0874' 0 92(0 88-0 96) -0.0930' 0 91(0 87-0 95) -0 2357" 0 79(0 75-0 84) -0.2316' 0 79(0 75-0 84) 0 .4849' 0 62(0 56-0 68) -0.5012' 0 .61 (0. 55-0 67) -0 .7565' 0 47(0 40-0 56) -0.7586' 0.47(0 40-0 55) -1.0688' 0 34(0 27-0 44) 1 .0693' 0 34(0 27.0.44) 1 .00 1 00 0 1440' 1 .16(1.08-1. 23) 0 12861 1.14(1 .07-1. 21) -0 0649 0 94(0 89-0 99) -Q.0745t 0 93(0 88-0 98) -0.2589' 0 77(0 74-0 81) -0.2650' 0 77(0 74-0 80) 1 .00 1 00 -0. 1158' 0 89(0 85-0 94) -0 .1371' 0 87(0 83-0 92) 0.0016 1 00 0 00027 1 00 -0 0419 0.96(0 92-1.00) -Q.0635t 0 95(0 91-0 98) 1 .00 0 1723' 1 .19(1. 15 1 23) As is expected from the medical literature discussed in Chapter 1 at pages 6-8, Model4 I! OR 1 00 -Q.0556t 0 95(0 91-0 99) 0 .5268' 0 59(0 54-0 65) -0.3861' 0 68(0 62-0.74) 1 00 -0 .4396' 0 64(0.60-0 69) 0 .1812' 1 20(1.51-1 25) 1.00 -0.4958' 0 .61 (0. 59-0 63) 0 1189' 1 .13(1.08-1. 18) 1 .00 -0 1253' 0.88(0 85-0 92) -0 .2659' 0 77(0 73-0.81) -0.5254' 0 59(0 54-0 65) -0.7972' 0.45(0 38-0 53) -1.0971' 0 33(0 26-0 43) 1 .00 0 .1869' 1 .21(1. 13-1.29) -0.0114 1 .00 -0.2251' 0 80(0 77-0 83) 1 00 -0.0736! 0 93(0 88-0 98) 0.0095 1 .00 -0 0449 0 96(0.92-0.99) 1.00 0 1865' 1 .21(1.17-1. 25) 1 00 -0.8560' 0.43(0 39-0.46) 1 .00 -0 .2649' 0.77(0 72.0.81) 0 .8281' 2 29(2 21-2 37) o Higher age and high parity increase odds of LGA (1.20 for higher age; 1 13 for high parity) o Like LBW and SGA, elevation has a dose-response relationship with LGA except the effect is inverse: as altitude increases, odds of LGA decrease (0. 88 0.33) o The presence of medical risks increases odds of LGA by 21% 83

PAGE 101

o Smoking is not associated with higher odds of LGA: mothers who smoke during pregnancy have 57% lower odds of LGA o High weight gain raises odds of LGA by 129% (2.29) ; low weight gain is protective, lowering odds of LGA by 23% There are some surprising results. o In Model 4 some high school is the same as high school graduate o In contrast to LBW, preterm birth, and SGA, being unmarried is protective (odds of LGA decline to 0.80) o Neither inadequate nor adequate plus prenatal care is a marker for LGA odds of LGA are lower by 4-7%; intermediate prenatal care is not statistically different than adequate care In the fully saturated model, the factor having the greatest predictive power for increased odds of LGA is high weight gain (OR 2.29). The most predictive protective factors are first birth (OR 0.61 ), increasing elevation of residence before birth (OR 0.88 0.33), low weight gain (OR 0.77), and being unmarried (OR 0.80). Because gestational diabetes and delivering a previous infant of 4000 or more grams are specific predictors of LGA, it is instructive to examine the distribution of these two factors by race/ethnicity and to run the fully saturated model replacing the more generic medical risk category with gestational diabetes and previous large infant. Hispanics have the highest frequency of previous infants weighing 4000+ grams, and the second highest frequency of gestational diabetes (second to Whites) Interestingly, this model is slightly more discriminating with a c statistic of 0 .661 (compared with 0.650) and the predictive power of both specific risk factors far exceeds the predictive power of medical risks generally: gestational diabetes 2.23 (2.07-2 39) ; previous infant 4000+ grams 5.42 (4.96-5.93) The odds of LGA by race/ethnicity, however, change very little. Odds for Hispanics fall slightly (0.95 to 0.93) as do odds for Others (0. 68 to 0.66) Odds for Blacks remain the same (0. 59) Discussion of Aim 1 Hispanics have worse risk profiles than Whites or Blacks for the frequency of teen births, low educational attainment inadequate and intermediate prenatal care, low weight gain the presence of one or more medical risk factors and the specific risk of having a previous infant who weighed more than 4000 grams at birth. In addition Hispanics have a risk profile that is inferior to Whites for low and high parity and recommended weight gain In contrast, Hispanics report the lowest frequency of smoking. Overall, the data on risk factors 84

PAGE 102

suggest that Hispanics have poorer social, demographic and medical risk factors compared to other races but better self-reported smoking behavior compared with non-Hispanic Whites and Blacks. Yet after controlling for all risk factors Hispanics have odds of LBW and SGA that are more similar to Whites (and much lower odds than Blacks). In light of the disparity of socioeconomic position between Hispanics and Whites and the somewhat worse social profile of Hispanics compared with Blacks, an odds ratio of 1.18 is considered to reflect a relatively positive outcome. Hispanics also have the same odds of preterm birth (and much lower odds than Blacks), and slightly lower odds of LGA than Whites have although odds of LGA for Whites and Hispanics are much higher than for Blacks and Others Table 4.11. Comparison of fully adjusted odds ratios of birth outcomes by race/ethnicity LBW Preterm Birth SGA LGA White 1.00 1.00 1 00 1 00 Hispanic 1 18 (1.13-1 23) 1 .01 (0.98-1.05) 1 18 ( 1 15 1 12) 0 95 (0.91-0 99) Odds not significantly different from 1 00 2 5 2 VI 0 1 5 += IV a: VI "0 1 "0 0 0.5 0 LBW Preterm Birth Outcome SGA Black 2.16 (2. 01-2 33) 1.42 (1.34-1. 51) 1 98 (1. 86-2 09) 0.59 (0.54-0 65) LGA Other 1.64 (1.52-1. 77) 1 16 (1.09-1.23) 1.72 (1.63-1.82) 0 68 (0 62-0 74 ) -+-White H i spanic Black Figure 4 .7. FULLY ADJUSTED ODDS RATIOS BY RACE / ETHNICITY In addition to suggesting that the epidemiological paradox exists for Hispanics, Aim 1 confirms the positive linear association of elevation with LBW and SGA and the negative linear association of elevation with LGA. Education has little effect on any of the outcomes a result that contradicts the usual association of higher education and better health but which 85

PAGE 103

is consistent with studies showing that education is not necessarily related to better outcomes among Hispanics (Acevedo-Garcia et a/. 2005; Gould et a/. 2003) Prenatal care is also less clearly associated with outcomes. Adequate plus care is a marker for all poor birth outcomes although most markedly for preterm birth Hispanics enter prenatal care later than other groups, but inadequate and intermediate levels of care are less clear l y negatively associated with heightened adverse outcomes The interaction of race/ethnicity and medical risk for both LBW and SGA show Hispanics with lower odds of the respective outcomes followed in ascending order by Whites, Others, and Blacks. Although the paradox can be said to exist for LGA because Hispanics and Whites have about the same odds both are well above the odds for Blacks and Others. Hispan i cs have slightly lower frequency of gestat i onal diabetes than Whites, but that may be a resu l t of the fact that more Hispanics have inadequate and intermediate prenatal care, which may reflect under-diagnos i s of gestational diabetes in that population. More surprising i s the protective effect of being unmarried on LGA. Resource-based theories of social determinants of health suggest that being married improves outcomes but this is not the case for LGA. Examining only those non-interacting factors that increase odds of the outcome by 50.0% or more, Table 4.12 indicates which factors influence each outcome at that level. Table 4.12. Comparison of contributing risk factors by birth outcome and race/ethnicity LBW Preterm Birth SGA LGA Race/Ethnicity Black Race/Ethnicity Other Adequate + PNC Inadequate PNC Low weight gain High we i ght gain Smoking Medical risks First b irth Elevation As a check on the effect of SGA and LGA on birth outcomes appropriate for gestat i onal age (AGA) was also tested (babies born ne i ther SGA nor LGA). Because AGA i s a desired outcome, odds ratios lower than 1 00 indicate less favorable outcomes As can be seen i n Table 4 13 Hispanics are second to Whites in their odds of having an AGA baby while Blacks and Others are much more likely to have a baby that i s not appropriate for gestat i onal age These results confirm, again that the epidemiological paradox ex i sts for Hispanics in Colorado. 86

PAGE 104

Table 4.13. Comparison of fully adjusted odds ratios of AGA by race/ethnicity AGA White 1 00 Hispanic 0 96 (0 93-0.98) Black 0 72 (0 69-0.75) Other 0.78 (0.75-0 82) This model has adequate fit (Hosmer Lemeshow 0.3804), but little discriminating value Aim 2 Aim 2 examines birth outcomes of mothers of Mexican origin by her place of birth (nativity), using the same candidate variables as are used in Aim 1 Comparison of Risk Factors of Mothers of Mexican Origin by Nativity Aim 2 tests whether the place of birth of mothers of Mexican origin in Colorado affects their risk profiles, and whether there is congruence between risk factors and birth outcomes. Based on the 2000 Census, Mexican-born mothers are poorer and less well educated than U.S.-born mothers of Mexican origin. Prior population studies report that foreign-born Mexican immigrants to the U.S. enjoy unexpectedly better outcomes for low weight-associated birth outcomes. No national studies of LGA as an outcome for mothers of Mexican or igin were found. As with Aim 1, the null and alternative hypotheses for each risk factor are: H40 : each risk factor is independent of race/ethnicity. H4A: each risk factor is related to race/ethnicity Table 4.14 reports frequencies of each risk factor by nativity and the significance of any differences using Pearson's chi-square. A risk factor is considered related to nativity if p:S 0 .05 (Gould eta/. 2003). Mexican-born mothers have higher frequencies of social risk factors for low weight-associated outcomes and two LGA-specifi c medical risk factors than U.S.-born Hispanics of Mexican origin, shown by underlining in Table 4.14, in the following categories: o Older mothers o Much lower education levels o Inadequate and intermediate levels of prenatal care o Presence of one or more medical risk factors o Low weight gain o Gestational diabetes o Previous 4000+ gram infant 87

PAGE 105

In contrast, Mexican-born mothers report lower frequency of smoking, drinking alcohol during pregnancy, high weight gain and preexisting diabetes They are also more likely to be married, and they have lower frequencies of teen births and first births. As shown in Table 4.14, the frequency of each risk factor is significantly different by nativity (p<.0001 for all factors except for preexisting diabetes p=0.0198). Figure 4.8 displays the same information using a bar graph. Table 4.14. Percent frequency distribution of risk factors of mothers of Mexican origin by nativity 200G-2005 Characteristic Total U.S.-Bom Mexican-Born p-value Age of mother S19 17 85% 24 37% 13 88% 20 34 75 06% 69 99% 78.15% 9000 0 .67% 0.29% 0.91% <0 0001 Education (mother) < 9 yrs 22 57% 4.01% 33.89% 9-11 yrs 33 72% 31. 23% 35 24% 12+ yrs 43.71% 64 .76% 30 88% <0 0001 Prenatal care Inadequate 23 59% 18 62% 26 62% Intermediate 19 56% 16 .21% 21. 60% Adequate 36 90% 39.24% 35.47% Adequate Plus 19 95% 25 93% 16.31% <0.0001 Marital status Married 61. 27 % 52 66% 66 53% Unmarried 38 73% 47.34% 33 47% <0 0001 Smoking No 96.08% 92 66% 98 77% Yes 3 .92% 8.34% 1 23% <0 0001 Alcohol drinker No 99.50% 99 13% 99 72 % Yes 0 50% 0 87% 0 .28% <0 0001 Weight gained <151b 15 .61% 12 98% 17 22% 15-40 lb 72 .45% 71. 39% 73 09% >401b 11.94% 15 63% 9 69% <0 0001 88

PAGE 106

Table 4.14. Percent frequency distribution of risk factors of mothers of Mexican origin by nativity 2000-2005 (continued) Characteristic Total U.S.-Born Mexican-Born Medical risk factors None 62.84 % 29 37 % 41.91% One or more 37 16 % 70 63 % 58 09 % Gestational diabetes 3 25 % 2 57 % 3 67 % Preexisting diabetes 0.45 % 0 .52% 0.41% Prev infant 4000+ gr 1 28 % 0.40 % 1 .81% A"ev 4000+ g A"eexist [)abetes Gest Diabetes 1-igh Wei ght Aec Weight ( r ) i Low Weight ,...._ Dri nker Smoker I l'v'edical Risk Adequate + PNC 1.....1... I 0 Adequate ( r ) PNC lnterrred i ate PNC u.. Inadequat e PNC a: I -l Mexican -Born U S -Born U1rrarried HS Grad ( r ) Sorre HS I I 1 NoHS 1-igh Parity r-l i I Low Par i ty ( r ) Rrst Birth >34 -! I 2034 ( r ) Teen Birth I 0 1 0 20 30 40 50 60 70 80 90 Percentage Figure 4 .8. DISTRIBUTION OF RISKS BY NATIVITY (r) designates reference category 89 p-value <0.0001 <0.0001 0 0198 <0 .0001

PAGE 107

Frequency of Adverse Birth Outcomes by Nativity Although the risk profile of Mexican-born mothers is something of a mosaic, the social gradient of health and their generally poor risk profiles suggest that they should have higher rates of the adverse birth outcomes under study than U.S.-born Hispanics of Mexican origin. However, prior research shows that Hispanics, particularly foreign-born Hispanics have paradoxically better LBW and SGA outcomes than U.S.-born women Hispanics (Rosenberg eta/. 2005; Frisbie & Song 2003; Gould eta/. 2003; Singh & Yu 1996). The null and alternate hypotheses for incidence of birth outcomes unadjusted for risk factors are : H50 : each birth outcome (LBW, preterm birth, SGA, and LGA) is independent of nativity. HSA: each birth outcome (LBW preterm birth, SGA, and LGA) is related to nativity. Pearson's chi-square is used to test the relationship between each birth outcome and nativity On the basis of frequencies of risk factors by nativity, particularly education, less than adequate prenatal care, and low weight gain it is surprising that Mexican-born mothers fare so well when examining the frequencies of LBW, preterm births, and SGA. However, the paradox does not hold for LGA, for Mexican-born mothers have a higher frequency of LGA, even while they also have higher frequencies of low weight gain and lower frequencies of high weight gain. These unadjusted frequencies indicate the importance of including LGA in the analysis. While Mexican-born mothers demonstrate the "paradox'' with respect to LBW preterm birth, and SGA, they do not demonstrate the paradox with respect to LGA. Table 4.15. Percent frequency of LBW, preterm birth, SGA, and LGA by nativity 2000.2005 Birth Outcome All Mexican U.S.-Born Mexican-Born p-value Origin Mexican Origin LBW 6 44% 7.67 % 5 69% <0.0001 Preterm births 7 23% 8 19 % 6.64% <0 0001 SGA 12 94% 15 32/o 11.49% <0.0001 LGA 6.14% 4.69 % 7 02% <0.0001 90

PAGE 108

18 16 A 14 12 / \ --+--U S.-Born .. / ..... J 10 8 .... \ Mexican-6 Born 4 2 0 v :...0
PAGE 109

Table 4.16. Unadjusted and adjusted odds ratios (95% Cl) of LBW by nativity Nativity U.S. Born Mexican Mexican-Born c statistic Hosmer-Lemeshow Model1 Nativity Only Model2 Adjusted for Demographic & Socioeconomic Position Model3 Additional Adjustment for Medical Conditions Odds Ratios (Confidence Interval 95%) 1 00 1.00 1.00 0.73 (0.69-0.77) 0 95(0 .85-1. 07) 0 .91 (0 81-1.02} 0.539 0.669 0.681 0.1166 0 .1331 Odds not significantly different from 1.00. Model4 Additional Adjustment for Health Behaviors 1.00 *0. 93 (0.83-1.05} 0.710 0 5107 Mexican-born mothers have lower odds of LBW than mothers of Mexican origin who are born in the U.S. before any adjustments In Models 2 3, and 4 after adjusting for all factors, Mexican-born mothers have the same odds as U S.-born mothers of Mexican orig i n Despite their less advantageous risk profiles, Mexican-born mothers have the same odds of LBW U S.-born mothers of Mexican origin demonstrating the existence of the paradox based on these i ndividual level characteristics. This model has a modest degree of discriminat ion (c statist ic= 0 .7 10) and adequate fit. 1 20 1 .00 ---. II) . --+-U.S.-Born 0 0 .80 a: 0 .60 II) Mexican"tl t .! "tl 0.40 Born 0 0 .20 0 00 1 2 3 4 Models Figure 4.1 0. LBW ODDS RATIOS BY NATIVITY AND MODEL Table 4 17 reports the significant standardized beta coefficients and odds ratios for each variable in the models of LBW by nat i vity Only one i nteraction is significant for LBW by nativityprenatal care*nativity. 92

PAGE 110

Table 4.17. Estimated coefficients and odds ratios for LBW and nativity Variable Nativi U S -bom (rl Mexican-born -0. 3205 A e 20-34(r) :S19 >35 Pari Low(r) First High Elevation <5000 (r) 5000-5999 6000-6999 7000-7999 8000 8999 > 9000 Education HS grad(r) No high school Some high school Marital status Married (r) Not married P r enatal care Adeguate (r) lnadeguate Intermediate Adeguate f21Us Prenatalnativit Adegu S.-bom (r) lnadegu S.-bom lntermedU.S -bom Adeg+U.S -bom lnadeg Mexican-born lntermMexican-born AdegMexican bom Adeg+Mexican-bom Medical risk (r) indicates reference category p<. 0001 tp<.001. 1:p:S01 p<. 05 Model1 OR 1.00 Model2 Model3 OR OR 1 00 1 00 -0.0518 1 00 -0. 0999 1 00 1 00 1 00 0 0588 1 00 0 0675 1 00 0 2535. 1.29(1. 16-1. 43) 0 20341 1 00 1 00 0 3342. 1.40(1 .31-1 49) 0 3317" 1 .39(1. 30-1.49) 0 1966. 1.22(1 .13-1. 32) 0 1869. 1 .21(1.11-1. 31) 1.00 1 00 0 0483 1.00 0 0216 1 00 -0. 0521 1.00 -0. 0278 1.00 0 4551. 1 58(1.32 1 88) 0.4394 1 .55(1.30-1. 85) 0 .5072! 1 .66(1. 14-2 42) 0 .5249! 1 70(1.16-2.47) 0 6758. 1 .97(1. 14 -1. 29) 0 7166 2 .05(1. 55-2 70) 1 00 1.00 0 1062 1 .11(1.02-1. 21) 0 0826 1.00 0.11641 1.12(1 .14-1. 29) 0.0971! 1 .10(1.03-1. 18) 1.00 1 00 0 1923. 1 .21(1.14-1. 29) 0 1819. 1 .20(1.13-1. 27) 1.00 1 00 0 6775. 1 .97(1.74-2 23) 0 6138. 1 .85(1. 63 2 09) -0.0010 1.00 -0. 0014 1 00 1.2616. 3 53(3 1 8-3 92) 1.2438. 3.46(3 12-3 85) 1.00 1 00 1 00 0 6775. 1 .97(1. 74-2 23) 0 6138. 1 85(1.63-2 09) -0. 0010 1.00 -0. 0014 1 00 1.2616. 8-3 92) 1.2438. 3 47(3 12-3 85) 0 1202 -0. 0032 1.00 -0. 2728. 0 76(0 66-0 87) -0. 3380. 0 71(0 62-0 82) -0. 0518 1.00 0 0999 1.00 1 1398. 3 13(2 80-3 50) 1.1043. 3 02(2 70 3 37) 1 00 0 4998. 1.65(1 .56-1. 75) Model4 OR 1 00 -0. 0688 1 00 1 00 0.1056! 1.11(1.03-1 20) 0 .1723! 1 .19(1.07-1. 32) 1 00 0.4259. 0 .1340! 1.14(1.06-1.24) 1.00 0 0498 1.00 0 0117 1 00 0.4576. 1 .58(1. 32 -1. 89) 0 .5711! 1.77(1.21-2 60) 0 8076. 2 .24(1. 70-2 96) 1.00 0 0433 1.00 0.0769 1 08(1.01-1. 16) 1 00 0.1540. 1.17(1. 10-1.24) 1.00 0 5294. 1 .7Q!1.50-1. 93) -0. 0248 1.00 1 2393. 3 45(3 11-3 84) 1 00 0 5294. 1.70(1.50 1 93)) -0. 0248 1.00 1 2393. 3 45(3 11-3.84) 0 1239 0 88(0 .78-1. 00) 0 3592. 0 70(0 61-0 80) -0. 0688 1.00 1 1263. 3 08(2 76-3 45) 1.00 0 4592. 1 .58(1.49-1. 68) 1.00 0.6708. 1 96(1.75 2 19) 1.00 0 7027" 2 .02(1. 89-2 16) -0.8424. 0 43(0 38-0 48) The contributors and markers for LBW by nativity are the same as those for LBW by race/ethnicity, although the order of their importance and magnitude vary somewhat. 93

PAGE 111

o Odds of delivering an LBW infant increase for teen mothers (1.11) and older mothers (1.19) o First babies have higher odds (1.53) of being LBW o Odds of LBW show a monotonic increase as residence at elevation above 7000 (1.58 2.24) o Being unmarried increases the risk of LBW (1.17) o Mothers with inadequate (1. 70) and adequate plus prenatal care (3.45) have increased odds of LBW o The presence of medical risks associated with pregnancy is a marker for increased odds of LBW (1.58) o Smoking increases the odds of LBW (1.96) o The odds of LBW with weight gain of less than 16 pounds increases by 1 02% (2.02) o Weight gain of >40 pounds is very protective against LBW -gaining more than 40 pounds reduces the odds of LBW by 57% (0.43) As with LBW by race/ethnicity some results are unanticipated: o Low education (less than high school) is not significant o Intermediate prenatal care (one level below adequate) does not increase the odds of LBW In the fully saturated model, the factors having the greatest predictive power for increased odds of LBW are adequate plus prenatal care (3.45), presence of one or more medical risks (1.58) smoking (1.96), and first parity (1.53). The one interaction, adequacy of prenatal care by nativity supports the paradox in favor of Mexican-born mothers. Mothers reporting adequate prenatal care have the same odds regardless of nativity (1. 00) For those reporting inadequate care, U.S.-born mothers have more than 80% greater odds of LBW than Mexi can-born mothers (1. 70 compared with 0.88) Mexican-born mothers with intermediate prenatal care are 30% less likely (0.70) to have an LBW baby than U S.-born mothers with adequate or intermediate care (1.00 for both). And even with adequate plus prenatal care, a marker for a more complicated pregnancy Mexican-born mothers have lower odds compared with U.S.-born mothers of Mexican origin (3.08 compared with 3.45). Preterm Birth Table 4 18 reports unadjusted and adjusted odds ratios of preterm birth by nativity. Interestingly, Model 2 has adequate fit and modest discrimination (c statistic = 0.711), but although Models 3 and 4 increase discrimination somewhat, the fit drops to inadequate, 94

PAGE 112

suggesting that medical risk and health behaviors are not particularly relevant to the overall fit of the models. Table 4.18. Unadjusted and adjusted odds ratios (95% Cl) of preterm birth by nativity Nativity U S -Born Mexican Mex i can-Born c statistic Hosmer Lemeshow Model 2 Model 3 Model1 Adjusted for Additional Nativity Only Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1 00 1.00 1 00 0.80 (0 .760 84) 1 .14 (1.011 30) 1 08 (0 .96-1. 23 ) 0 527 0 .711 0 726 0.2214 0 0016 Odds not significantly different from 1 .00. Model4 Additional Adjustment for Health Behaviors 1 .00 1 08 (0.951 23) 0 743 0 0082 F i gure 4 .11 shows the change i n odds rat i ons for preterm birth across models. After i ncreas ing from 0 80 in Model 1 to 1.14 in Model 2 i n Models 3 and 4 Mex i can born mothers have the same odds of preterm b i rth as those mothers of Mex i can or i gin born i n the U .S. 1 .20 ......... 1 00 . II) _, -+-U.S Born 0 0 .80 ;:: ""I ftl a: 0 .60 II) .. Mexican-"tl "tl 0 40 Born 0 0 .20 0 .00 1 2 3 4 Models F i gure 4 .11. PRETERM B IRTH ODDS RATIOS BY NATIVITY AND MODEL Table 4 19 r eports the s i gnif i cant standardized beta coeffic i ents and odds rat i os f or each variable i n the models for preterm birth by nativ i ty Preterm b irth by nat i vity has two i nteract i ons : prenatal care nativity and prenatal care weight. 95

PAGE 113

Table 4.19. Est i m ated coefficients and odds ratios for preterm birth and nativity Variable Model1 Model2 Model3 Model4 OR OR OR OR Nat i v i U S born (r) 1.00 1 00 1 00 1 00 Mexican born 0 2272 0 80(0 76.() 84) 0 1349 1.14(1.01 1 30) 0 0793 1 00 0 0772 1 00 A e 20 34(r) 1 00 1.00 1.00 S 19 0.0329 1 00 0 0436 1.00 0 0682 1 00 > 35 0 2684 1 3 1(1. 19 44) 0 2096' 0 1873! 1 21(1.09 33) Pari low(r) 1 00 1.00 1.00 F i rs t 0.1705' 1 .19(1. 1 H .2Z) 0.1678' 1.18(1 1 H 26) 0 2444' 1 28(1 2 0. 36 ) H i gh 0 2274' 1 .26(1. 17. 10) 0 2179' 1 .24(1. 16 1 34) 0 1806 1.20( 1 1 1 1.29) Elevation <5000 (r) 1 00 1.00 1 00 5 000 5999 0 0239 1 00 .0 0043 1.00 0 0189 1 00 6000 6999 0 .1790! 0 84(0 75 0 94) .0.1505! 0 86(0 80 0 96) .0. 1159 0 89(0 80 1 00) 7 000-7999 0 1788 1.00 0 1590 1.00 0 1693 1 00 8000-8999 0.3875 1.47(1.01 2 15) 0.4046 1 .50(1. 02 1 19) 0 4335 1 .54(1. 05 2 27) > 9000 0 1291 1 00 0 1678 1 00 0 2320 1 00 Educat i on HSgrad (r) 1 00 1.00 1.00 N o high s chool 0 0565 1 00 0 0905 1.00 .0. 0029 1.00 S o m e hig h school 0 11141 1 .12(1. 05. 19) 0 0905 1.1 Q!1 03-1.17) 0 0784 1.08(1.01-1 06) M arit a l status Married (r) 1 00 1 00 1 00 Not married 0 1220' 1 13 07 1 20) 0 .1111! 1.12(1 05. 18) 0 1000! 1.11(1 04 1 17) Prenatal care Adeguate (r) 1 00 1.00 1 00 lna deg u ate 1 0212' 2 78(2.44 3 17) 0 9482 2 58(2 26 2 95) 0 9978' 2 27(2 35 3.13) i n t er medi ate .0. 0394 1.00 0 .0401 1.00 0 0887 1 00 A de g uat e 21us 1.8356' 6 27(5 61-7 01) P r en ata i' nativit 1.8187" 6 1 6(5 51-6 89) 1 8560' 6 46(5 66 7 23) Adeg'U. S bom ( r ) 1.00 1.00 1 00 1.00 1.00 lna deg'U S bom 1.0212' 2 78(2 44 3 17) 0 .9482' 2.58(2 26 2 95) 0 3734 1.45(1.13-1. 86) lnterrned'U. S born .0. 0394 1.00 .0 .0401 1.00 .0042' 0 37(0 24.0 57) A deg+'U S bom 1 .8356' 6 27(5 .61 7 01) 1.8187" 6.16(5 51-6 89) 1.1826' 3 26(2 7 4 89) In ad' Mexican-born 0 6111' 1 .84(1. 63-2 08) 0 .4715' 1 .6Q!1 421.81) .0. 1379 1.00 l n ter Mexican bom .0 0344 1 00 .0. 1103 1 00 1 1245' 0 33(0 2 1.0. 51) A deg'Mexicanborn 0 1349 1.14(1.01. 30) 0 0793 1.00 0 0772 1 00 Ad+'Mexi canbom 1 8 152' 6 14(5 47 6 89) 1.7794' 1 1438' 3 14(2 62 3 77) 1 00 1.00 0 5680' 1.77(1.67 1 86) 0 5396' 1.72(1. 62. 81) 1 00 0 2799' 1.32(1.17. 49) 1 00 0 8067" 2 24( 1 94 2 59) > 40 .0 6399' 0 9658' l n t er'low wt 1 8662' 6 46(4 13 10 12) Adeg 'l owwt 0 8067" Adeg+'low wt 1 2797" 3 60(3 04-4 25) lnad'med wt 0 6244' 1 87 ( 1 48 2 35) inter med wt 0 9156 2 .50(1 62 3 86) Adeg+ med wt 0 6734 1 .96(1. 69 2 27) 96

PAGE 114

Table 4.19. Estimated coefficients and odds ratios for preterm birth and nativity (continued) Variable lnad.high wt lnterhigh wt Adeghigh wt Adeq+high wt (r) i ndicates reference category p <.0001 t p < 001 f p S01 p < 05. Model1 Model2 Model3 OR OR OR Model4 ll OR -1. 3547" 0 26(0 .25-Q. 27) -4.9105. 0 01(0.01 0.01) -o .6399. 0 53(0 .42-Q. 67) 0 .3733. 1 45(1 39-1 52) Some of the results are consistent with those of preterm birth by race/ethnicity. For example smoking, marital status, education, altitude, and parity have similar effects. However, there are some differences. o Teen births do not increase the odds of having a preterm birth among the population studied in Aim 2 o The presence of medical risks associated with pregnancy increases the odds of preterm birth by 72%, but this increase is quite a bit lower than the increase for race/ethnicity (127%) As might be expected, interactions provide the most interesting information. Inadequate and adequate plus prenatal care alone are markers for increased odds of preterm birth (inadequate by 127%; adequate plus by 546%), these are somewhat smaller increases in odds than by race/ethnicity. Intermediate prenatal care alone is not significantly different from adequate care. However the interaction of prenatal care and weight gain shows that intermediate prenatal care combined with low weight gain creates an odds ratio of 6.46. Low weight gain in combination with any level of prenatal care, including adequate increases odds of preterm birth by 2.24 to 6.46. The interaction of prenatal care and nativity shows that Mexican-born women with either adequate or inadequate prenatal care have the same odds as the reference category-U.S.-born mothers with adequate care; Mexican-born mothers with inadequate care (1. 00) compare favorably with U.S.-born mothers with inadequate prenatal care (1.45). Although both groups have roughly the same odds if they have adequate plus prenatal care (U.S.-born 3.26 compared with Mexican-born 3.14), Mexicanborn mothers' odds are lower. 97

PAGE 115

Small for Gestational Age Table 4 .20 reports unadjusted and adjusted odds ratios of SGA by nativity. Table 4.20. Unadjusted and adjusted odds ratios (95% Cl) of SGA by nativity Nativity U.S.-Born Mexican Mexican-Born c stat i stic Hosmer Lemeshow Model 2 Model 3 Model 1 Adjusted for Additional Nativity Only Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1 00 1 00 1.00 0 72 (0 69-0 75) 0 73 (0 70-0 77) 0 73 (0)00 76) 0 540 0 586 0 587 0 0007 0.0025 Model4 Additional Adjustment for Health Behaviors 1 00 0 74 (0.71-0.78) 0 614 0.4307 As shown in Figure 4.12, Mexican-born mothers have much lower odds of SGA than their U.S -born counterparts There are no significant i nteract i ons which i s one reason the odds ratios rema i n so stable across the models. The fully saturated model has low discrimination ( c statistic = 0 614) but adequate fit. Although Mexican-born mothers have higher frequencies of social r i sk factors than U.S.-born mothers of Mexican orig i n i n all cases Mexican-born mothers have 26-28% lower odds of having an SGA baby than Mexican mothers born in the U S., supporting the exi s t ence of the epidemiological paradox in favor of Mexican-born mothers in Colorado. 1 20 1 00 Ill 0 .80 0 ................... ; ca a: 0 .60 Ill rv1exica n -'tl 'tl 0 .40 Born 0 0 20 0 .00 1 2 3 4 Models F i gure 4 12. SGA ODDS RATIOS BY NATIVITY AND MODEL 98

PAGE 116

Table 4.21 reports significant standardized beta coefficients and odds ratios for each variable in the models for SGA by nativity. Table 4.21. Estimated coefficients and odds ratios for SGA and nativity Variable Model1 Model2 OR OR Nalivi U S -born (r) 1.00 1.00 Mexican-born 0 3323. 0 72(0 69-0 75) -0 31oo 0 73(0 70-Q.n} A e 20-34(r) S19 >35 Pari Low(r) First High Elevation <5000 (r) 5000-5999 6000-6999 7000-7999 8000-8999 >9000 Education HS grad (r) No high school Some high school Marital status Married r Not married Prenatal care Adequate (r) Inadequate Intermediate Adequate plus (r) in dicates reference category p <.0001 1 p < .001. tps01 p < 05 1 .00 0 .0771t 1.08(1 .02-1 14) 0 0192 1.00 1.00 o 34oo 1.41 (1.34-1. 47) 0 0013 1 00 1 .00 0 0275 1.00 0 .1283t 1 14(1.05 1 23) 0 3876. 1.4 7(1.29-1. 69) 0 .4428t 1 .56(1. 17 2.08) 0 8238. 2 .28(1. 87 2 78) 1.00 0 0450 1 00 0 .0730t 1 .08(1.02-1. 13) 0 1529. 1 17(1.12-1.22) 1 .00 0 09501 1 .10(1.04-1. 16) 0.0173 1 .00 0 1405. 1 .15(1.09-1. 22) Model3 OR 1 00 -0 3174. 0 73(0 70-0 76) 1.00 0 .0784t 1.08(1 .02-1. 15) -0.0266 1.00 1 00 0 3392. 1 .40(1.34-1. 47) -0.0034 1 00 1.00 0 0207 1.00 0 .1313t 1.14(1 05-1 24) 0 3841. 1 .47(1.28-1. 68) 0.4461t 1.56(1 17-2.08) 0 8308. 2 .30(1. 88-2 80) 1 00 0 0406 1 00 0.0690t 1 07(1.02-1 13) 0 1509. 1.16(1 .11-1. 22) 1 00 0 .0828t 1 .09(1. 03-1.15) 0.0156 1 .00 0 1389. 1 .15(1. 09 1 21) 1.00 0 0873. 1 .09(1.05-1. 14) Model4 OR 1.00 -0 2979. 0.74(0.71-0.78) 1.00 0.10181 1 .11(1.05-1. 17) -0 0493 1 00 1 00 0.4005 1.49(1.42-1 57) 0 0393 1.00 1 .00 0.0361 1 00 0 15381 1 .17(1. 08-1.27) 0 3968. 1.49(1.30-1 70) 0 .4744t 1.61 (1.20-2 15) 0 8704. 2 .39(1. 95-2 92) 1 00 0 0178 1 00 0.0554 1 .06(1.01-1. 11) 0 1276. 1.14(1 .09-1. 19) 1 00 0 0176 1 00 -0 0025 1 00 0 1297" 1 .14(1.08-1. 20) 1 00 0.0640t 1.07(1 .02-1. 11) 1 00 0 6501. 1 .92(1. 76-2 09) 1 00 0 3371. 1.40(1 .33-1. 48) 0 6325. 0 53(0 49-0 57) The model for SGA is uncomplicated by interactions. Similarities and differences with respect to prior analyses include: o In contrast to the results for SGA by race/ethnicity, teen births are significant across all models, although the increase in odds is small (1.11) 99

PAGE 117

o There is a dose-response relationship between increasing altitude of residence and increased odds of SGA in all models starting at 6000 feet (1.17 2.39) compared with race/ethnicity where the increase in odds began above 5000 feet. o Smoking increases the odds of SGA by 92%, lower than for race/ethnicity (131%) o First birth increases the odds of SGA by 49% o Low weight gain is associated with SGA increasing odds by 40%; high weight gain reduces odds by 47%, the same pattern that is observed by race/ethnicity o The presence of medical risks increases the odds of SGA by only 7% o Education does not have the expected effect ; having less education does not increase odds of SGA In Model 4 having inadequate or intermediate prenatal care is not statistically different from adequate prenatal care; adequate plus prenatal care is associated with SGA but it raises the odds by only 14%. These results suggest that SGA may not have as clear predictor conditions as LBW and preterm birth may have and that it may not be diagnosed during prenatal visits and then followed more closely resulting in adequate plus prenatal care The factors having the greatest predictive power for increased odds of SGA are birth above 7000 feet (1.49 increasing to 2.39 at 9000 feet and above) smoking during pregnancy (1.92), and first birth (1.49). Large for Gestational Age Table 4 22 reports unadjusted and adjusted odds ratios of LGA by nativity. Here the paradox does not hold for Mexican-born mothers In marked contrast to odds rat i os for LBW, preterm birth and SGA Mexican-born mothers have 45% higher odds of delivering an LGA baby than U.S.-born mothers of Mexican origin in the fully saturated model. Table 4.22. Unadjusted and adjusted odds ratios (95% Cl) of LGA by nativity Nativity U S.-Born Mexican Mexican-Born c statistic Hosmer-Lemeshow Model 2 Model 3 Model 1 Adjusted for Additional Nativity Only Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1 00 1 00 1.00 1 53 (1. 44-1.63) 1.43 (1.34-1 54) 1 .41 (1.32-1 52) 0 548 0.616 0.619 0.0111 0 0325 100 Model4 Additional Adjustment for Health Behaviors 1.00 1.45 (1.35-1. 55) 0 649 0.8893

PAGE 118

Figure 4.13 reports the odds ratios of LGA by nativ i ty Aim 1 showed that Hispanics as a group have slightly lower odds of LGA compared with Whites. But when the Hispanic population is limited to Hispanics of Mexican origin and then compared by nativity, i t can be seen that LGA i s a phenomenon of recent immigration. 1.80 1 .60 1 .40 ......... -... In 1 .20 U.S.-Born 0 ; "' 1.00 a: In 0 80 l'v1exican-, 0 60 Born 0 0.40 0 20 0 00 1 2 3 4 Models F i gure 4 13. LGA ODDS RATIOS BY NATIVITY AND MODEL Tab l e 4.23 reports the s i gnificant standardized beta coefficients and odds ratios for var i able in the models for LGA by nativity. Table 4.23. Estimated coefficients and odds ratios for LGA and nativity Variable Model1 Modei2 Modei3 Modei4 OR OR OR OR Nativi U S -bom (r) 1 00 1 00 1 00 1.00 Mexican-born 0 .4278' 0 .3593' 0 .3453' 1 .41(1.32-1. 52) 0.3692' 1.45(1 .35-1. 55) A e 20{r) 1 .00 1.00 1 .00 S19 -0.4332' 0 65(0.58-0 72) .4302' 0 65(0 59-0 72) -0 .4625' >35 0 3009' 1 35(1.23-1 48) 0 2851' 0 .3121' 1 .37(1.24-1. 50) Pari Low{r) 1 .00 1 .00 1 .00 First -0.3 9 18' 0 68(0 63-0 73) -0 3933' 0.68(0 63-0.73) 0 .4770' 0 62(0 58-0 67) High 0 0461 1.0 0 0 0415 1 .00 0 0759 1 .08{1 00 1 16) Elevat i on <5000 {r) 1 .00 1 00 1 00 5000-5999 -0. 1439' 0 8 7(0 8 1-0 93) -0 1627' 0 85(0 79-0 91) -0 1784' 0 84(0 78-0 90) 6000-6999 -0.2965' 0 74(0 66-0 84) -0.2919' 0 75!0 66-0 85) -0 .3264' 0 75!0 64-0 82) 7000-7999 -0 .6162' -0.6250' 0 54(0 42-0 69) -0 .6424' 8000 8999 1 .0765! 0 34!0 17-0 69) 1 .0706! 0 34(0 17-0 69) -1.0813! 0 34!0 17-0 69) >9000 -1.2345' 0 29(0.17 0 51) -1.2219' 0 30!0 1 7-0 51) -1.264 8 0 28(0 16-0 49) Education HSgrad {r) 1 .00 1 .00 1.00 No high school 0 0720 1 00 0 0612 1 .00 0.0926 1 1Q!1.02 1 19) Some high school 0 0 624 1 00 0.0523 1 00 0.0642 1 .00 Marital status Married {r) 1 .00 1 00 1.00 Not married -0 1455' 0 87(0 81-0 92) -0 1500' 0 86(0.81-0 92) -0 1433' 0 87(0 81-0 92) 101

PAGE 119

Table 4.23. Estimated coefficients and odds ratios for LGA and nativity (continued) Variable Model1 Model2 Model3 Model4 13 OR 13 OR 13 OR 13 OR Prenatal care Adequate (r) 1.0 0 1 00 1.0 0 Inadequate 0 1494' 0.86(0.80..() 93) ..(). 1762' 0 84(0 .78..(). 91) -0.1 034t 0 90(0 .83..(). 98) Intermediate 0.0039 1 00 ..().0011 1 00 0.0268 1 00 Adequate p lus ..().0364 1.00 ..
PAGE 120

o As with race/ethnicity being unmarried is protective against LGA, reducing odds by about 13% (for race/ethnicity, being unmarried is slightly more protective reducing odds by 20%) o Inadequate prenatal care is associated with 1 0% lower odds of LGA; intermediate and adequate plus prenatal care are not significantly different than adequate prenatal care o Smoking is associated with lower odds of LGA by 33% (57% by race/ethnicity) In Model 4, the factors having the greatest predictive power for increased odds of LGA are high weight gain (2. 21) and nativity in Mexico (1.45). Replacing the category of general pregnancy associated medical risks with the specific risk factors of gestational diabetes preexisting diabetes, and having a previous infant weighing more than 4000 grams lowers the odds of LGA for Mexican-born mothers slightly compared to the model with the general medical risks (1.42 compared with 1.45), but the point estimates for the specific medical risks then become the highest contributors to odds of LGA. o Gestational diabetes increases the odds of LGA by 163% o Preexisting diabetes increases the odds of LGA by 357% o Previous infant weighing more than 4000 grams at birth i ncreases the odds of LGA by 293% Table 4.24 presents a summary comparison of the fully saturated models using presence of one or more general medical risks and LGA specific risks Using specific risks improves the discrimination slightly (although the effect is still low) and retains adequate fit. Table 4.24. Model 4 adjusted odds ratios (95% Cl) of LGA by nativity Nativity U S.-Born Mexican Mexican Born c statistic Hosmer-Lemeshow Model 4 Model 4 General Medical Risk Specific Medical Risks for LGA Odds Ratios (Confidence Interval 95%) 1.00 1.45 (1.35 1 55) 0 649 0 8893 1 00 1 42 ( 1.32-1 53) 0.668 0.3409 Given the higher risk of Mexican-born mothers of having an LGA baby i t i s appropriate to explore factors that m i ght explain these unexpected results. As can be seen from Figure 4 14 the frequency of LGA by nativity by year within the study period shows a 103 j

PAGE 121

pattern whereby Mex i can-born mothers consistently have about 2% greater frequency of LGA babies than U.S. born mothers, except for 2005 when the gap narrows to about 1.7%. 8 7 ......... .. ... -. 6 5 ---... -Mexican-c Born Ill 4 ... --4-U.S. born Q. 3 2 1 0 # a, r., # b 'V 'V ,.,_,r:SJ Year of Infant Birth Figure 4 .14. FREQUENCY OF LGA BY NATIVITY BY YEAR F i gures 4 15 4.17 compare demographic risk factors thought to affect the rate of LGA: weight gain age, and parity. As each increases, LGA increases, and the frequency of LGA for Mexican born mothers increases more dramatically than for U S.-born mothers. 1 4 12 10 - Mexicanc 8 --Born 6 fl / ------U.S. Born t. 4 _,. ...2 0 Low Medium Hgh ( ref) Weight Gain Figure 4.15. FREQUENCY OF LGA BY WEIGHT GAIN 12 10 8 .. -+-MexicanJ .. ... Born 6 ..,.-----U S Born 4 v 2 .. 0 <20 20-34 35+ (ref) Age Categories Figure 4.16. FREQUENCY OF LGA BY AGE 104

PAGE 122

10 8 f'. ,. . +-1\texican. -6 Born c 4 __.__ U S Born Gl .. 0.. 2 0 Rrst Low High (ref) Parity Figure 4 17 FREQUENCY OF LGA BY PARITY As shown in Figure 4.18, although Mexican-born mothers have higher frequencies of gestational and preexisting diabetes and a previous 4000+ gram infant, the differences between the two population groups are quite constant. . s i l +-----------1 Gest Pre-exist Prev Diabetes Diabetes 4000+ gram Fisk Factors 1 +-I Born --u.s. Born Figure 4.18. FREQUENCY OF LGA-SPECIFIC RISKS Discussion of Aim 2 In accord with previous studies the population of Hispan i c mothers in Colorado is not homogeneous. Looking at race/ethnicity a l one misses the disproport i onate impact of LGA on Mexican-born mothers Stratifying by Mexican origin and then by nativity shows that Mexican-born mothers contribute to the lower odds for SGA but have higher odds LGA which tends to be masked if only race/ethnicity is considered. 105

PAGE 123

!CI a: 14 "0 8 Table 4.25. Fully adjusted odds ratios of birth outcomes by nativity LBW Preterm Birth SGA LGA U.S.-Born Mexican Origin 1 00 1 00 1.00 1 00 Mexican-Born *0 93 *1. 08 0 .74 1.45 Odds not significantly different from 1.00 1 6 1.4 1 2 ... '"" 0 8 .... . . .... .... I I -+-U. S Born Mexican-Born 0.6 0.4 0.2 0 LBW Pre term SGA LGA Birth Outcome Figure 4.19. FULLY ADJUSTED ODDS RATIOS BY NATIVITY Aim 2 demonstrates that the epidemiological paradox exists for Mexican-born mothers at the low weight end of the birth outcome spectrum, but not for LGA. It also confirms the positive l i near association of elevation with LBW and SGA and the negative linear association with LGA As with Aim 1 education has little effect on any of the outcomes, a result that conflicts with studies that assoc i ates higher levels of education w ith better birth outcomes as part of the social gradient. This is especially evident for mothers of Mexican origin who have even lower educational levels than Hispanics generally. Also consistent with Aim 1, prenatal care is less obviously associated with outcomes as adequate and i ntermediate levels of care are less clearly associated w ith heightened adverse outcomes. Mexican-born mothers have higher frequencies of LGA with higher weight gain, higher parity and higher age of pregnancy than U S -born mothers of Mexican origin. It is worth noting that the reports of lower frequencies of preexisting diabetes in the birth record for Mexican-born mothers (as compared to h i gher reported frequencies of gestational diabetes and previous infants weighing more than 4000 grams) suggest that preexisting 106

PAGE 124

diabetes may not always be diagnosed in Mexican-born mothers In light of Mexican-born mothers higher frequencies of inadequate and intermediate prenatal care, it i s likely that they have limited access to health care outside of pregnancy ; hence d i abetes that is i dentified during prenatal care and reported as gestational may in some instances be undiagnosed preexisting diabetes. Complementing the discussion of adverse outcomes is consideration of presumed healthy birth weight babies those who are born appropr i ate for gestational age. For the population of all mothers of Mexican origin 80.92% have AGA babies Mexican -born mothers have a slightly higher frequency of AGA bab ies (81.49%} compared w ith U.S.-born mothers (79 98%}. Table 4.26 reports the odds of hav ing an AGA baby using the fully saturated Model 4 w ith the one interact ion for AGA i n thi s population -wei ght parity. The results of logi stic regression for the fully-saturated model show that Mexican-born mothers are 8% more l i kely to have an AGA baby than U.S.-born mothers of Mex ican origin Thus a l though Mexican-born mothers have lower odds of having SGA babies and higher odds of hav ing LGA babies they also have higher odds of having AGA babies compared to U.S.born mothers in Colorado Table 4.26. Comparison of fully adjusted odds ratios of AGA by nativity AGA U.S.-Born Mexican Origin 1 00 Mexican-Born 1 08 (1.04-1. 12) Model has adequate fit (Hosmer-Lemeshow 0 1458) but l i ttle discr i m i nating v alue c=0 540 This study was designed with nested models to test whether any inferences m i ght be drawn about the h ea l thy m i grant and healthy immi g r ant hypotheses. Thi s study does not test the healthy m i g r ant hypothesis d i rectly, because adequate data are not ava i lab l e on rates of the b irth outcomes under study among Mex ican mothe r s who rema i n i n Mex ico. Nevertheless those who suggest that Mexican-born mothers who migrate to the U.S. are health i er than those Mex i can-born mothers who do not m i grate m iss the point. It i s i ndeed possi ble that Mexican-born mothers who migrate to the U .S. may be healthier than those who rema i n i n Mexico but one part i cularly appropr i ate comparison for purposes of the epi demiological paradox i s the population of U.S. resident mothers of Mexican origin not those who rema i n i n Mexico. If the healthy m i grant hypothesis exp l a ins d i fferences i n outcomes odds ratios should change between Models 2 and 3 when medical risk factors are taken i nto account. In the same manner i f the healthy i mmigrant hypothesis exp l ains 107

PAGE 125

differences in outcomes, the odds should change between Models 3 and 4 after health behaviors relating to smoking and weight gain are taken into account. Mexican-born mothers have a worse demographic and socioeconomic profile than U.S-born mothers as measured by the frequency of adequate prenatal care, level of education, and higher age at delivery (Table 4 14) Table 4 27 reports the odds of Mexican born mothers compared with the reference group of U.S.-born mothers of Mexican origin by outcome and by model using general medical risks associated with pregnancy. Table 4.28 reports odds of LGA using adjustments for LGA-specific medical risks in Models 3a and 4a Differences based on the nature of medical risks are small. Table 4.27. Fully adjusted odds ratios (95% Cl) by nativity and model Nativity LBW Preterrn B irth SGA LGA Model1 Nativity Only Model2 Adjust for Demographic & Socioeconomic Position Model3 Adjust for General Medical Conditions Odds Ratios (Confidence Interval 95%) 0 73 (0. 69-0 77) 0 95(0 .85-1. 07) 0 .91 (0.81-1. 02) 0.80 (0. 76-0 84) 1 14 (1.01-1. 30) 1.08 (0.96-1. 23) 0 72 (0. 69-0 75) 0 73 (0. 70-0 77) 0 73 (0. 70-0 76) 1 53 (1.44-1 63) 1 43 (1. 34-1.54) 1 .41 (1.32-1. 52) Odds not significantly different from 1 00 Model4 Adjust for Health Behaviors 0 93 (0.83-1.05) 1 08 (0.95-1.23) 0 74 (0. 71-0 .78) 1.45 (1.35-1.55) Table 4.28. Fully adjusted odds ratios (95% Cl) by nativity for LGA using LGA-specific medical risks Nativity LGA Model1 Nativity Only Model2 Adjust for Demographic & Socioeconomic Position Model3a Adjust for Medical Conditions Specific to LGA Odds Ratios (Confidence Interval 95%) 1 53 (1.44-1.63) 1 43 (1.34-1.54) 1.38 (1. 29 1 48) Model4a Adjust for Health Behaviors 1.42 (1.32-1. 53 ) Model 2 adjusts for demographic characterist ics and socioeconomic position, and thus puts Mexican-born women and U S.-born women of Mexican origin on the same level for the further tests of the healthy migrant and healthy i mmigrant hypotheses Model 3 adjusts for general medical risks associated with pregnancy and delivery One might argue that Mexican-born mothers are somewhat healthier'' than U.S.-born mothers as measured by the frequency of one or more general medical risks associated with pregnancy because 58.1% of Mexican-born mothers have one or more medical risks 108

PAGE 126

compared with 70.6% borne by U.S.-born mothers. The healthy migrant hypothesis would suggest that odds would change after adjustment for medical risks if Mexican-born mothers are healthier. However, the odds of LBW and SGA remain the same for Mexican-born mothers between Models 2 and 3 when medical risks are adjusted for. They decrease 4% for preterm birth to the same level as U.S.-born mothers and decrease 2% for LGA using the general medical risk model. However, Mexican-born mothers have higher frequencies of medical risks specific to LGA-gestational diabetes and previous infants weighing 4000 or more grams. Odds for Mexican-born mothers drop 5% between Models 2 and 3a. The difference in odds after adjusting for these risks remains smallbetween 2-4% depending on outcome and medical risk model. Finally, in the fully-saturated model, after the addition of health behaviors associated with pregnancy weight gain and smoking the odds stay the same for preterm birth and increase only 1% for SGA. The odds increase 2% for LBW but the increase is not statistically significant. The odds increase slightly (4% in both Models 4 and 4a) for LGA, suggesting that the healthy immigrant hypothesis does not overly affect the odds of these outcomes, at least as measured by these two behavioral factors. In short, neither the healthy migrant nor the healthy immigrant hypotheses is supported using this design for measuring the influence of pregnancy related health factors (healthy migrant) or health behaviors (healthy immigrant) for the population of mothers of Mexican origin in Colorado. Before discussing the effect of the contextual variables on outcomes, it is worth testing whether the paradox persists for mothers of Mexican origin at geographic levels below the state. Table 4.29 and Figure 4.20 compare the odds ratios of each of the four outcomes for Mexican-born mothers in Adams County, Denver County, and statewide, with U.S.-born women of Mexican origin as the reference group. All models have adequate fit. For LBW, preterm birth, SGA, and LGA the results for the two counties and the state are consistent, although mothers of Mexican origin in Denver County have higher odds of LGA than mothers in either Adams County or statewide. In addition, the confidence intervals for LBW, preterm birth, and LGA increase at the county level. For AGA, whereas statewide Mexican-born mothers have 8% higher odds of having an AGA baby, in the two counties their odds are the same as U.S.-born mothers. The paradox continues to exist in Adams and Denver counties for low weight related birth outcomes, but not for LGA. 109

PAGE 127

Table 4 .29. Comparison of fully adjusted odds ratios of birth outcomes of Mexican-born mothers In Adams and Denver Counties and statewide Adams County Denver County LBW *0.89 (0.67 -1.17} 1 .19 (0.90-1.57) Preterm Birth *0.96 (0. 71-1.30) 1 .26 (0.95-1.67) SGA 0.70 (0.63-0.79) 0 .75 (0.67-0.84) LGA 1.54 (1.30-1.82) 1 .86 (1.55-2.23) AGA 1 .08 (0.99-1.18) *1.02 (0.94-1.12} Odds ratio not significantly different from 1 .00. 2.00 1 80 1 60 1 40 1/) 2 1 20 1ii a: 1/) 1 00 0 80 0 0 60 0.40 0 20 0 00 LBW A'eterm Birth SGA Outcomes LGA AGA Statewide 0 .93 (0.83-1.05) *1.08 (0.95-1.23) 0 .74 (0.71-0.78) 1.45 (1.35-1.55) 1 .08 (1.04-1.12) -+-Adarrs -Denver -.-state Figure 4 .20. ODDS OF OUTCOMES FOR MEXICAN -BORN MOTHERS IN ADAMS AND DENVER COUNTIES AND STATEWIDE Aim 3 Aim 3 is designed to examine the role of neighborhood on birth outcomes of mothers of Mexican origin, specifically the effect of neighborhood deprivation and immigrant orientation. Neighborhood is defined as census tract based on the 2000 Decennial Census As described more fully in Chapter 3 pages 47-48 neighborhood deprivation is operation alized for each tract using data from the 2000 Census by summing and averaging the percent of : individuals living in poverty households receiving public assistance income female headed family households males unemployed in the civilian work force 110

PAGE 128

Immigrant orientation is operationalized by summing and averaging the percent of: Mexican-born individuals non-citizens born in Mex i co linguistically isolated households These values were centered at the mean for the linear modeling. are : Following Finch eta/. (2007), the hypotheses for contextual influence on birth outcomes H70 : immigrant-oriented neighborhoods have no effect on the four birth outcomes H7 A: immigrant-oriented neighborhoods affect the likel i hood of birth outcomes H80 : neighborhood disadvantage has no effect on the four birth outcomes H8A: neighborhood disadvantage affects the likelihood of birth outcomes To test the effect of contextual conditions on outcomes the rate of each birth outcome in each tract is modeled against each contextual var i able separately both variables and their i nteraction using generalized linear modeling. In Adams County, the index of immigrant or i entation is correlated with neighborhood deprivation with Pearson s coefficient of 75375 In Denver County the correlation is less pronounced with a Pearson s coefficient of 51979 Influence of Neighborhood Deprivation and Immigrant Orientat i on on Outcomes As discussed in more depth i n Chapter 3 in Adams County there were no births in one tract for the entire study period and fewer than twenty births in five tracts The five tracts with fewer than twenty b i rths were matched on both the quartile of neighborhoods depr i vation and immigrant orientation and combined with a matching tract leav i ng 79 tracts and 16, 107 births for analysis The number of births per tract afte r combination ranged from 20936 Table 4.30 summar i zes the range of frequenc i es of each outcome in Adams County Table 4.30. Outcomes in Adams County Outcome Frequency Range Frequency In Tracts Countvwide LBW N=1071 1 .82-13 33 % 6 65 % Preterm Birth N=1201 0-13. 00 % 7.46 % SGA N = 2044 5.11-35.00% 12.69 % LGA N=1040 0-13. 33 % 6.46 % 111

PAGE 129

Table 4 .31 summar izes the results of the general l i near model ing on outcomes i n Adams County In Adams County neither contextual variable influences rates of LBW, preterm b irth, or SGA. For LGA however neighborhood deprivat ion and the interact ion of neighborhood deprivation and immigrant orientation are significant at the 0 .10 level which some researchers suggest i s appropriate because contextual variables are generally considered to contribute less to outcomes than individual level factors (Sellstrom & Bremberg 2006; P i ckett & Pearl 2001 ). Nei ghborhood depr i vation i s marginally s igni ficant w ith respect to LGA i n Adams County (p=0 .09). As neighborhood deprivat ion i ncreases 10% from the average measure of deprivation the rate of LGA i ncreases 1 4 % Immigrant orientat ion alone i s not s i gnificant. However, the interaction of deprivation and i mmigrant or i entat ion on LGA i n Adams County i s marg i nally s igni f i cant (p= 0.10). P l ott ing hypothet ical values of the i nteract ion shows that when immigrant ori entation is low, as deprivation i ncreases so does LGA; when i mmigrat ion orientat ion i s h i gher i ncreas ing depr i vation decreases LGA. These results although weak suggest that immigrant orientation may moderate the effect of neighborhood depr i vation in Adams County when there are h igh l evels of immi grant orientation in the tract. Table 4.31. Estimated coefficients for neighborhood deprivat i on and Immigrant ori entation In Adams County Outcome Model1 Model2 Model3 Model4 Neighborhood Immigrant Deprivation & Interaction of Deprivation Deprivation Orientation Imm i grant Orient & Immigrant Or i ent F statistic F statistic F statistic F statistic t statist i c Estimate t statistic Estimate t statistic Estimate t statistic Estimate p-value p-value p-value p-value LBW F=0 03 F=0 .87 F = 0 77 0 1065 F = 1 1 6 ..().0143 ..(). 0348 depriv t = 0 .82 0.0140 t=..Q. 17 t =..Q. 93 immi a rant t =1 2 3 ..().0706 t= 1 .39 ns ns ns ns P r e t erm b irth F = 1 .11 F=O.OO F-1.3 9 0 2148 F-0.9 9 t=1.05 0 0 888 t=..Q.04 ..(). 0015 d e priv t= 1 12 ..().0737 t=0 .48 0 0049 i mm i grant t = 1 6 6 ns ns n s ns F =0 .01 F = 0 22 F = 0 .33 0 1 406 F = 0 25 SGA t=0 0 8 0 .0108 t =..().47 ..().0286 de p riv t=0 .67 0 .076 0 t=0 .30 0 004 9 immi g r a nt -0 8 1 ns ns ns ns F = 2 .86 F = 1 8 7 F = 1 4 2 F = 1 9 2 L G A t=1. 6 9 0 1413 t = 1 3 7 0 0 509 d epriv t = 0 .99 0 1275 t=-1.68 ..(). 0 1 68 immigrant t = 0 1 4 0 .008 1 p = 0 .09 n s ns P 0 1 0 In Denver County, 36 tracts had fewer than twenty births As discussed in Chapter 3 page 50, these tracts were removed because it was not poss i ble to match them on both nei ghborhood deprivation and immigrant orientation withi n the a pri ori match ing rules One 112

PAGE 130

hundred tracts with 23, 332 births are analyzed. The number of births per tract ranges from 21 902. Table 4 .32 summarizes the range of frequencies of each outcome in Denver County Table 4.32. Outcomes in Denver County Outcome Frequency Range LBW Preterm B i rth SGA LGA N=1594 N=1928 N=2802 N=1483 In Tracts 0-13 13 % 0-19. 05 % 2 27 25. 00 % 0-11.11% Frequency Countywide 6 83 % 8 26% 12.01% 6 36% In Denver County LBW and SGA are influenced by neighborhood depr i vat ion. For each 10% increase from average deprivation in Denver County, the rate of LBW i ncreases 0.9 % from the average in the county and the rate of SGA i ncreases 1.3% Contex tual varia bles have no influence on preterm birth or LGA. Table 4.33. Estimated coefficients for neighborhood deprivation and Immigrant orientation In Denver County Outcome Model1 Model2 Model3 Model4 Neighborhood Immigrant Deprivation & Interaction of Deprivation Deprivation Orientation Immigrant Orient & Immigrant Orient F statistic F statistic F statistic F statistic t statistic Estimate t statistic Estimate t statistic Estimate t statistic Estimate p-value p-vaiue _I)-Value _p-value LBW F=3 .4 9 F=1. 12 F=1.74 0 0868 F=1.36 1=. 67 0 0910 1=-1.06 0 0311 depriv 1=1.53 0 0049 I=-Q 78 -o .0057 immigrant I=-Q 15 P=0 07 ns ns ns Pre ter m birth F=1.62 F=0 .09 F=0.88 0 0850 F 0 .64 1=1.27 0 .07207 I=-Q.31 0 0104 depriv I= 1 .29 -o 0152 I=-Q .41 -o 0034 i mmigrant I=-Q .39 ns ns ns ns F=3 .61 F=0. 13 F=2.04 0.1616 F=1. 34 SGA 1=1.90 0 1331 I=-Q .36 0 .0152 depriv 1=1.99 -o o336 I=-Q.04 -o 0004 i mmiarant--Q .69 P=0 06 ns ns ns F=0 16 F=1. 90 F=1.00 F=0 .73 LGA 1=0.40 O.Q188 1=1.38 0 .0383 depriv I=-Q.34 -o.0183 1=--0.45 -o 0031 im m i grant 1=1.35 0 0438 ns ns ns ns Contextual variables affect birth outcomes Adams and Denver County differently. Differences in the character of the counties may i nfluence these results Adams County is a mixed urban and rural county with a population density per square mile of 305; Denver is urban, with a population density of 3 ,617. Only LGA i s affected by contextual variables in 113

PAGE 131

Adams County where it is weakly influenced by neighborhood deprivation and by the interaction of deprivation and immigrant orientation where high levels of immigrant orientation moderate the effect of high neighborhood deprivation. In Denver County, which is more densely populated, LBW and SGA are weakly affected by neighborhood deprivation. Finch eta/. (2007) reported a monotonic moderation of immigrant orientation on rates of LBW as neighborhood deprivation and immigrant orientation increased, using multi-level hierarchical modeling It is not clear why results for LBW in Colorado are not more fully consistent with those in Los Angeles. Adams County is much less densely populated than either Los Angeles County (2,344 per square mile according to the 2000 Census) or Denver and that may suggest why these contextual variables have no effect on LBW in Adams County Just as different regions of the U.S. were settled by different religious groups in the 1600s and 1700s (Phillips 1999; 80-122), residents of various Mexican states settle in specific parts of the U.S. Los Angeles may have denser networks of immigrants from these Mexican states and cities of origin compared with Colorado. It may also be that the multilevel hierarchical study by Finch eta/. was able to capture more variation than this study's design. Discussion of Quantitative Analysis The design of this study sought to expand the spectrum of previous weight-related birth outcome studies with the addition of LGA. The results of Aim 1 show that the epidemiological paradox exists with respect to LBW, preterm birth, and SGA and race/ethnicity. It also exists with respect to LGA, but both Whites and Hispanics have higher odds than less advantaged Blacks or Others. Technically this is a paradox based on the reference group of White mothers, but LGA behaves differently than LBW, preterm birth and SGA by race/ethnicity. In addition, the study design sought to test the existence of the social gradient. Both Hispanic and Mexican-born mothers have much worse socioeconomic profiles than Whites. Adding medical risks in Model 3 sought to elucidate the influence of pregnancy-associated medical risks. In Aim 1 Hispanics have worse medical profiles that those of Whites In Aim 2, the object was to test whether Mexican-born immigrants are healthier'' than their U.S.-born counterparts. Although Mexican-born mothers have a better general medical risk profile (one or more of general medical risks associated with pregnancy), their LGA-specific risk profiles are worse, suggesting that they are not necessarily healthy'' migrants Finally, adding smoking and weight gain into Model 4 tested the healthy immigrant hypothesis whereby Mexican-born mothers are postulated to have healthier' behaviors which explain their 114

PAGE 132

favorable outcomes results. None of the fully saturated models had superior discriminat i ng power Nevertheless the lack of change i n odds from Models 2 to 3 and from Models 3 to 4 suggests that the proposed exp l anations in the literature do not prov i de powerful rationales for differences in weight-related birth outcomes of Hispanics generally or mothers of Mexican or i g i n specifically who live in Colorado The results of the analysis also raise questions about the influence of certain var i ables on these outcomes Educat ion was uniformly not explanatory even though i t is a key factor in socioeconomic position and the theory of the social gradient. The information in the birth record on education levels i s probably accurate as it is cons i stent with census data showing low levels of education among Mexican immigrants. Smoking i s a strong pred i ctor of LBW and SGA. Colorado PRAMS data suggest self-report of smoking by Hispanics is fa i rly accurate, but under reporting may be greater for other population groups. The low prevalence of smoking among Hispan i c mothers likely accounts for an even greater share of LBW and SGA than birth registry data suggest. In addition the low rate of smoking among Hispanics removes the antagonistic effect of smok ing on LGA. Alt i tude operates on LBW SGA and LGA just as predicted. There i s a monotonic increase in low we i ght outcomes w i th i ncreasing altitude and decrease for LGA. The low rate of drink i ng reported in the b i rth registry probably explains why it i s insign i f i cant in all model building. The influence of measures of adequacy of prenatal care i s a puzz l e. Intermed i ate care is almost always not s i gnificant or the same as adequate care. The bOUr)daries between i nadequate intermediate and adequate care are necessarily arbitrary in the Kotelchuck i ndex Perhaps as F i scella (1995) and Alexander & Kotelchuck (2001) discuss the value of prenatal care remains poorly measured Given the late entry i nto prenatal care by H i span i cs and Mexican-born mothers however it may be that medical risks are not adequately d i agnosed. Their prenatal care may also be interrupted by i nability to pay. The interact i on of we i ght gain and prenatal care show that low weight ga i n with any level of prenatal care raises the odds of preterm b irth by 2 24 to 6.46 among women of Mexican orig i n so prenatal care should not be written off as irrelevant. Area measures of neighborhood deprivat ion and i mmigrant or i entation had weak association with outcomes The difference in population density between Adams and Denver Count i es probab l y expla i ns why LBW and SGA are not affected by contextual var i ab l es i n Adams but are in Denver Most i nterest ing are the results for LGA. Just as incidence of LGA does not behave i n the same way as LBW preterm birth and SGA for mothers of Mex i can 115

PAGE 133

ong1n based on compositional characterist ics, neither does the influence of contextual variables on LGA behave as they appear to on LBW and preterm birth. Contextual variables have only weak influence in both counties, but even there the results are not consistent. Only in Adams County is there is a weak moderating effect of immigrant orientation on the negative effect of neighborhood deprivation on LGA The qualitative portion of the study sheds some light on the importance of neighborhood on mothers of Mexican origin in Colorado. 116

PAGE 134

CHAPTER 5 QUALITATIVE RESULTS The qualitative portion of this study complements the quantitative results and is designed to be directive: based on the statistical results certain matters were explored with key informants and recent mothers of Mexican origin. In light of quantitative findings suggesting that mothers of Mexican origin have odds of LBW, preterm birth, and SGA that are similar, regardless of their place of nativity, but that Mexican-born mothers have h i gher odds of LGA than their U.S -born counterparts the interviews focused on contributors to LGA -diet, exercise, weight gain, diabetes sources of support including the neighborhood cultural beliefs concerning exercise and diet while pregnant, and with Mexican-born women differences between life i n Mexico and the U S Recent Mothers Ten mothers of Mexican origin were interviewed; five were born in the U.S. and five were born in Mexico. All women were either married to or in a committed relationship with a man of Mexican origin ; several U S.-born mothers were married to men born in Mexico. Mothers ranged in age from 18 to their early 30s. Three were first-time mothers; the others had between two and four children. Five of the mothers had full time jobs, one worked part time in a medical clinic, and four were not employed in the formal economy. One woman had worked for a nursing home but lost her job after her baby was born prematurely and she spent a lot of time with him in the hospital (about f i ve weeks) She wants to return to school to earn an LPN degree. The women entered prenatal care between eight weeks (two were planning on getting pregnant and started care as soon as they had a positive test) and four months into the pregnancy. Some women recalled being tested for diabetes and all sa i d the test was normal. One mother, though, probably had preexisting diabetes. She spent her seventh month in the hospital after premature rupture of the membranes, and was put on a str i ct diet and given i nsulin shots when her blood sugar rose Three of the i nterviews w ith mothers were conducted in Spanish ; the balance were conducted in Engl ish. Three mothers were interviewed at their place of work during a lunch break ; six were interviewed in their 117

PAGE 135

homes; and one was interviewed outside one of the Salud clinics. Pseudonyms are assigned to the recent mothers. Anna, Isabel, Sara, Linda, and Marta were born in Mexico. Juanita, Graciela, Carmen, Yolanda, and Alicia were born in the U.S. Table 5.1 describes certain characteristics of the women interviewed. Table 5 .1. Selected characteristics of women interviewed Characteristic Mexican-Born CN=5) U.S.-Born (N=Sl Age 18 -low 30s 19-low 30s Employment full-time outside home 1 4 First time mother 2 1 2-4 children at time of interview 3 4 Weioht oain >40 or hioh BMI 2 2 Medical risk/problem pregnancy 2 0 Key Informants Five key informant interviews were conducted with professionals who work in various capacities with Hispanic mothers. One key informant works with poor, mostly Hispanic, first time mothers in the Nurse Family Partnership Program in Colorado. A second key informant is a certified nurse midwife, practicing in Colorado for many years in rural locations and with Denver Health, the public health hospital in the city and county of Denver. The third key informant is an obstetrician working with a primarily poor, Mexican population at Denver Health The fourth is a woman who was a practicing physician in Mexico, who now works as a nurse with La Clinica Campesina a public health clinic for the poor, many of whom are of Mexican origin, with clinics in Adams, Boulder and Denver counties. Her experience in both countries provided insights on the differences between mothers behaviors in Mexico and in the U .S. Each of these women is immersed in providing health care and counseling to women of Mexican origin in Colorado. The final interview was with a physician-researcher who is an expert in diabetes in pregnancy at the University of Colorado Denver. All key informant interviews were conducted in English; four were conducted at various locations convenient to the interviewee (office, coffee shop, home); one person was interviewed by telephone. Diet and Exercise During Pregnancy Diet All mothers except one described their diet as primarily "Mexican" eggs, beans, tortillas, ham, enchiladas, rice, and chile Several mentioned cooking the way their mothers 118

PAGE 136

did or cooking to please their husbands. Their diets were influenced by cultural preference for Mex ican foods and some mentioned a belief that it was not only permissible, but necessary, to eat a lot during pregnancy, a view that was reinforced by the i r mothers or mothers-in-law. [We eat] green chile with potatoes and meat ; beans rice the basic Mexican meal. Juan i ta The number one thing we have on our plate all the t i me i s beans Beans have to be there ... it's weird but food doesn t taste the same if we don t have beans. That s our ma i n d ish. Some days i t s just flour t ortillas with beans and cheese and ch ile. Grac i e Ia The way my mom cooks i s the way I cook Mostly everyth i ng I make i s Mexican foods like enchiladas . Like yesterday was my husband s birthday so I made mole which i s a real trad i tion of Mexico green chi l e red chile rice and beans sopa. Alic i a All but one mother descr ibed a relatively healthy diet for the i r most recent pregnancy citing cereal fruits vegetables salad milk and avoiding (for the most part) fast foods although there were times when lunch consisted of pizza burritos from a local restaurant or hamburgers. Their staples however leaned heavily to carbohydrates i n the form of beans tort i llas and rice and their diet typically included various fr ied foods such as ench i ladas and tostadas. Their sopas (soups) contain pasta like bits of fr ied flour w ith onion garlic and tomatoes. Sara s d i et was part icul arly calor i c ; she consumed four to s i x meals a day ea ting t acos and tostadas at home with her chi ldren and inst i tutiona l meals (meat and potatoes) with t he res i dents at the nur s ing home where she worked. Two U.S.-born mothers ment i oned receiv ing WIC ass i stance dur i ng their pregnancy and said they ate foods from WIC t hat they probably wouldn t have bought or eaten on the i r own such as eggs peanut butter milk and cheese. With my second child I had a lot of problems and I didn t really eat at all. So he was only 6 pounds when he was born-he was underweight. And I m sur e that was the reason why. And when I got pregnant with thi s one I was also on WIC. So they introduced me to the nutrition and what s healthy and what s not healthy like the sugars and that I needed to dr i nk more m i lk and cheese and stuff like that. So I k ind of balanced myself out. ... So . i t was a big change i n diet between my second child and the third. Carmen 119

PAGE 137

Key informants reported that a diet high in carbohydrates and fats and fast foods characterized the Mexican families they worked with. The 08 from Denver Health talked about the Diabetes Clinic where she emphasizes diet and logging dietary intake. She sees a direct correlation of the number of tortillas mothers have eaten and their blood sugar levels Staples, they have the same kinds of staples [here as in Mexico]. They have a hard time finding fresh tortillas, but they eat home-made beans, rice, and some are conscious of eating the way they did in Mexico. And others they're conscious of trying to make as much money as they can, and so they are working before the baby comes and they're driving around and they stop for fast food Public health nurse The normal diet in Mexico is tortillas, chile and a big soda. Soda is available even in rural areas. And also we are pretty fond of dessert ... In the old towns they would slaughter pigs, and then butcher the meat and make lard and use that for cooking. My father-in-law would do that. He was diabetic and continued to do it all the same .... [And we keep it up in the U.S.] Former Mexican physician Both mothers and key informants discussed the cultural belief that pregnancy is a time to eat for two. Mothers also indicated that their mothers urged them to "eat for the baby." Carmen's mother would tell her "you have to eat; because it's not for you, it's for your baby. Marta suggested that women in Mexico eat what they want when they are pregnant without regard to the health benefits of the food, eating foods cooked with lard. Key informants reported similar experiences with pregnant moms. They think they have to eat for two. So people start increasing the calories, both in Mexico and in the U.S. . They say I have to eat; I have to eat for two," but they eat this type of carbs, this much protein [and] they just increase the amounts .... Portion sizes are so different here compared to Mexico. Former Mexican physic ian This idea that when you're pregnant you have to eat a lot probably comes from the sense of having had food insecurity in the past and you want to be sure that this baby comes out all right, so you eat a lot during your pregnancy. Public health nurse 120

PAGE 138

Exercise/Energy Expenditure Among all the mothers interviewed, exercise levels during pregnancy were relatively low. Two mothers responded positively when asked about exercise during pregnancy; Alicia would walk after dinner and Yolanda walked two miles each day until her seventh month. Then she said she slept until noon and sat around the last three months of her pregnancy doing nothing. Others described walking a few blocks once or twice a month (Juanita) or not working out or doing anything other than what was required at their jobs (Carmen and Juanita). Anna said she would walk a little but that her neighborhood (a trailer park) was not that pretty or friendly. Graciela, born in the U.S., explained that she stopped exercising during her most recent pregnancy because she had a previous miscarriage. I was active before I was pregnant I was going to Curves every day it wasn't something like really major but I had been going and once we started planning the baby-because before this baby, I had a miscarriage . . and I got this gut feeling that I was pregnant and then I stopped going . . My doctor said wait 'til you 're three months and you can start up again. But I honestly didn't want to because I had that miscarriage already and I didn't want anything to happen to it. Graciela Key informants affirmed that, in their experience, exercise is not as much a part of the lifestyle of Mexican-born mothers after they first come to the U.S. Whereas in Mexico people eat to work, here women in particular are less likely to have jobs and they feel physically isolated or afraid of being out and about as they were in Mexico. In addition, cultural norms work to dampen exercise when a woman is pregnant. When they move here from Mexico they may be isolated, and they aren't working in the fields or walking everywhere, as they did in Mexico, and so their base rate of exercise is much less. In Mexico people eat to have enough energy to work. 08 A lot of the moms lived in Mexico in areas where they lived in the cal/e, in the streets, and they would go outside all the time ... and they would go to the market on foot and if they were in a rural area they were outside walking around. When they come to the U.S. they're stuck, because they don't necessarily drive or have cars. Public health nurse When I take care of women here, that's a huge problem -that they don't exercise ... the exercise they might get is walking their kids to and from school. Midwife 121

PAGE 139

In Mexico, as soon as they are pregnant, they don't work as hard. People say you don t have to run. Former Mexican physician And yet no group is homogeneous. Marta, born in Mexico refused to give in to morning sickness or tiredness while pregnant. She read several books about pregnancy and decided that she would be just as active as she was before becoming pregnant. This did not extend to working out, but she convinced herself that she would not allow her pregnancy to slow her down Maternal Weight and Weight Gain Although weight gain is the most commonly used predictor of weight-related birth outcomes, more recent research suggests that maternal BMI immediately before pregnancy rather than weight gain is a stronger marker for poor birth outcomes (Casey eta/. 1997). The quantitative results show that Hispanics generally, and mothers born in Mexico, have higher rates of low weight gain during pregnancy. Observations by key informants corroborate the literature with respect to LGA. The most important predictor of LGA is maternal weight, not we i ght gain during pregnancy. You could have high BMI before pregnancy gain just a little weight, and smoke, and end up having an AGA baby even though all signs point to macrosomia Diabetes researcher It may be that Mexican-born moms gain less weight during pregnancy, but that they enter pregnancy at heavier weights. BMI is high among Mexican women in the Diabetes Clinicsome have a BMI between 40 49 08 Gainalthough it' s not very predictive because a lot of these women don t gain much weight, but especially if they are overweight to start it' s key. Midwife Mothers interviewed gained varying amounts of weight. Table 5.2. Weight gain of mothers interviewed and birth weight of infants Mexican-born Parity Age Maternal Weight Birth Weight Complications Weight Gain (Pound-Ounce) Anna 1 18 98 52 6 8 None Isabel 3 30s 198 31 7-7 Induced at 34 weeks Marta 1 20s 110 20 6-12 None Sara 4 30s n/a n/a 3-0 C-section at 32 weeks Linda 2 Early 20s 123 39 7-8 Born at 38 weeks 122

PAGE 140

Table 5.2. Weight gain of mothers interviewed and birth weight of infants (continued) U.S.-born Parity Age Maternal Weight Birth Weight Complications Weight Gain (Pound-Ounce) Graciela 3 25 140 35 7-1 None Carmen 3 Late 20s 148 42 8-5 Induced at 39 weeks Alicia 2 Late 20s n/a 22 6-1 None Juanita 2 Mid 20s 150 29 7 2 None Yolanda 1 24 140 45 7-10 None Two mothers, both born in Mexico had complications. At the time of her interview, Isabel appeared to have high BMI. She said she had been tested for diabetes during pregnancy and the results were negative. Nevertheless, when she delivered her third child at 34 weeks of gestation, he already weighed 7 pounds 7 ounces. At this weight, 3 373.7 grams, he was almost LGA for his gestation (3,595 grams) according to Alexander et at. (1999). If she had carried him to term, he would probably have been LGA, because the fetus typically gains several pounds during the last trimester. Sara has four children. Both of her daughters were large 9 pounds 14 ounces and 10 pounds. She needed a c-section to deliver her 1 0-pound daughter. During her most recent pregnancy she ate four to six meals a day. She was hospitalized at six months with premature rupture of her membranes. Her blood sugar was very high, and she was put on a strict diet and given insulin shots. The baby was born e i ght weeks early and weighed only 3 pounds (1360 grams definitely an LBW baby and also SGA). Although she said her glucose levels were fine after the birth mothers with pre existing diabetes often have small babies, while mothers with gestational diabetes tend to have LGA babies Body Image Some studies have noted differing views of ideal body i mage across cultures (Candib 2007 Ahluwalia eta/. 2007). This was also noted by two key informants And some of them who have high BMis ... And also it's amazing to me ... they see it as a positive thing. A big body for a woman is what is desired. It's desirable that looks good I worked in a clinic -it was 95% Latina and most of them were immigrants. The staff was also Mexican, bilingual. .. We would talk and they were ''That's how my man wants me to be; this is what looks good; and you sk i nny ladies look terrible It' s a different, completely different culture. They're in health care, they know it's not healthy to be overweight but it's difficult for them to believe that or overcome those beliefs values that they have Midwife 123

PAGE 141

There is this switch from being somehow a young woman to being a mother, and the switch in body image -it' s OK to be overweight, to care a little less about how you keep yourself and that you are the cook. You make all the food you feed everybody, and food becomes the center of your life in a way. Public health nurse Whereas being married is generally protective for birth outcomes at the low-weight end of the spectrum, the quantitative results showed an unexpected protective effect on LGA of being unmarried. This finding is likely consistent with the idea of the marr ied mother as the "cook" as noted above and perhaps also consistent with some Hispanic men s preference for larger women. Sara linked this preference with machista." Mexican men are very machista. They're always like really possessive, they want to have their woman stay home ; some won' t let you work. Some they just want like if you are out at the street, they don t let you look any where else or they get mad. When asked if she agrees that Mexican men like larger women Sara said it has to do with control over the woman. I think it's they think, because they don t want nobody else to look at you and they like them to get like that I think ... I was telling my baby's dad I was telling him that I want to start doing exercise because I was getting chunky, and he said "No, you're looking better like that. You look good like that. You don t need to do anything." Smoking and Drinking In accord with the literature and results of the quantitative results the mothers interviewed said they did not smoke during pregnancy. A couple said they smoked occasionally or maybe had three cigarettes a day'' before getting pregnant but once they got pregnant they stopped. A few said members of their households smoked but they always smoked outside. When asked why they didn t smoke during pregnancy, they all said i t wouldn't be good for the baby, and a couple said it wasn t good for them e i ther. Mothers didn t drink alcoholic drinks during pregnancy either, although a few said they had before they were pregnant or afterwards, if they weren t nursing. The public health nurse agreed that smoking was minimal. My impression after working with these families is that women are careful; they really don't smoke. I m in their homes and I don t smell it; they don t smoke. Public health nurse 124

PAGE 142

Other Cultural Beliefs About Pregnancy I asked each woman if she practiced Ia cuarentena after birth. All said they d i d to one degree or another. I then asked them to describe the practices they engaged in. The answers ranged from no sex for 40 days to relatively elaborate rituals for the first forty days after giving birth usually encouraged strongly by her mother and/or mother-in-law. The most elaborate descriptions came from U S -born mothers Oh yeah i t was like a big deal for my mom and my mother-inlaw. More on my mother-in-law; since she s a bit older than my mother. I mean she told me stories about -she had six kids -and i n all her s i x k ids and /a cuarentena she never showered and to be honest, I only did it for e i ght days And just so I wouldn t like, I didn't want to get her upset. My mom said she bathed after she had the baby . You have to cover your back and if you don t take care of yourself in those 40 days your m i lk production won' t be good if you don t cover your back. You're not supposed to be out in the cold cause you're i n the healing process They were really str ict. With my f i rst one it was even worse these other ones, they would let me at least go to the kitchen. My first one I was just in the bedroom for 40 days and I was ''Thi s is just ridiculous Mom, I think I could go to the living room!" She d say "No, just stay i n your room because it s warm. .. The second one was a little bit easier but I do believe i n that Ia cuarentena. Gracie Ia La cuarentena I take that seriously in a woman. And my mom, being there tell ing me a woman should always take care of herself before and after birth. And she would tell me "Don' t go nowhere you're not supposed to be out and about you know, because how would she say it hemorrag i a "-kind of l ike bleed ing to death-and so she would tell me. So I really took care of myself. I really highly beli eve in taking care of your body after -you and the baby as well. I would stay home And my mom would say "Don' t go up and down the sta irs, don' t be outside when there i s wind If you have sit there all day then sit there all day and t ake care of your baby. So I so when a baby is born, you have to see the doctor withi n two days so that was the only time I went out. I went out for h i s check ups and came straight home. I was at home the whole time . She would tell me to well i n the tub you have to s tep over and she would always tell me like when you l ift up your leg and all that, you need to be very careful [not to let air go up the birth canal]. But I would t ake a shower every day for that for the same reason that I was breastfeed ing at that time and to be clean because you are breastfeeding the baby. I d idn' t eat different. I ate the same I just didn t do any house chores. Carmen 125

PAGE 143

Sources of Social Support All mothers were asked what they would do or whom they would call if they were six months pregnant and they hadn t felt the baby move for a while and were getting scared. None mentioned getting support from their neighbors all said they would seek support from members of their familywhether or not that person was close by or from their medical provider. I d call the doctor. Right away. And if the clinic wasn t available, I would call the hospital, like the nurse line and get information. And I would also turn to my mom. So I would call several people, all on the same day, so I would know what is going on. I would also call my sisters Carmen My mother my sister-in-law, a friend at work. Juanita Mi suegra (mother-in-law) before even talking to my mother. I might go to the clinic or hospital if there was a problem. Anna My mom .... And then my mother-in-law lives with me so she s an older woman and she was raised in Mexico so she knows a lot of home remedies, and my mom's like that too. Gracie Ia When asked generally where they got support generally, mothers mentioned parents, in-laws, siblings, and husbands. One mother mentioned that since her miscarriage she had become more spiritual and went to church more I kind of started going a lot closer to church after I had my miscarriage and then after I had this baby and I kind of felt I needed more spiritual so I would say I pray a lot more. Graciela The striking cultural zero" in all the interviews was the unimportance of the neighborhood. No one spontaneously mentioned their neighborhood or neighbors as sources of support or even much in the way of socializing (Kutsche 1998:1 0). Mothers occasionally mentioned that they knew a neighbor or two, but not well enough to call on for help My neighborhood is a trailer park. I don t like it. Anna [My neighborhood] it's not very big and everybody works, so by the time I get home or they get home before me, and I don't really see my neighbors ... I think I could [get support] if I needed it. I know that my 126

PAGE 144

neighbors would be there, but I m never home. The hours I m never home they're not home when I m home. Carmen I can probably say I don t have a real neighborhood. I live way out in the country My nearest neighbor i s probably maybe a mile away. My only neighbor he s not very nice. Gracie Ia I talk a lot to my neighbor over here She s an old lady, she s really nice but friends? No Sara Political Economy and Birth Outcomes All mothers were asked if they had any ideas why women of Mexican origin tended to have healthy birth outcomes at the low-we i ght end of the spectrum but why recent immigrants had worse LGA outcomes than mothers of Mexican or i gin who were born in the U S Answers ranged from differences in cultural beliefs to pregnancy diet to economic diff icul ties in Mexico and the U.S to big changes in d i et and exercise after they moved to the U .S. Although not always stated explicitly lack of resources i n general framed the i r experiences Four mothers had lived with in-laws or parents during one or more of their pregnancies and spoke of the difficulty generally of making it in the U .S. Others talked about how hard life in Mexico i s compared to the U .S. I walked i n the park at the rec center whi le I was pregnant, but didn t use the rec center for any type of exercise. [although not stated, it seemed that the reason was lack of money to use the rec center] Anna I know that being in the U S tor a Mexican and try to make it here i n the Uni ted States is a lot harder . .. to buy anything in general i t s harder than over there [in Mexico] I have heard that from my motheri n-law that from her point, her boys they were all large babies you know, bigger than normal. And she had them over there And when I had my k i ds except for the most recent the other two were smaller and she would tell me how come I wasn t eating enough because my baby wasn t growing Carmen The adjustment to pract i cal exigencies of life in the U.S creates changes i n diet and energy expenditure almost immediately. Key i nformants spontaneously talked about economic pressures to m i grate and the abrupt changes in diet activity patterns and isolation of recent immigrants all of which are likely to contribute to LGA bab ies. 127

PAGE 145

I'm not so convinced that it's the healthier women that come, because a lot of the time it's impoverished women that come; they are the ones that are willing to take the risk because the poverty levels in their families in Mexico are so desperate that they have to leave. They leave behind their kids .... It used to be the male from Mexico who comes to support the family, but now it's more, the whole character is changing face. A lot of women who're coming over to work and getting jobs related to childcare and cleaning ... the women are often sending money home and a lot of them still have their kids back in Mexico with their parents .... A lot of it is just to seek employment and help improve their family's situation in Mexico Midwife I think most come because the economy is really bad; so mostly poor people immigrate. Former Mexican physician The more physically active ones have hotel jobs, they make beds and then clean houses, and they work in fast food joints. It' s becoming harder and harder for them to find work because people are asking for social security numbers. And so informal labor, like caring for each other s children while one of them goes out for work. Public health nurse Sara described her life i n Mexico and after immigration to the U.S. She and her sister stayed in Mexico with her grandparents for three years while her parents and two younger siblings worked to make enough money to bring Sara and her sister to the U.S. I lived in a little village [in Mexico). I used to go to school. After school I would cuz I was with my grandma my parents came over here first. Life over there is so different. They don't pay good on the job or anything like that. So I used to help my grandma after school. To clean, to cook, to feed the chickens, and cows. I used to help milk the cows. I was sometimes hungry. It was hard. When my parents came over here it was a little bit better, but when we were all there [in Mexico] it was hard sometimes and we were hungry ... My parents used to buy shoes that we both could use, that way sometimes I used his shoes and sometimes he would. It was hard It was really hard. So that's why they came over here so they could give us a better life. Sara described the difference in diet between Mexico and the U.S Her mother worked two jobs and her father worked late hours. As the oldest child, she was in charge of the children and helped with cleaning and cooking after they arrived in the U.S. In Mexico we used to eat beans, fruit, potatoes Sometimes, not a lot of times, we used to eat meat not so much. Fruit, a lot of fruit ; vegetables not that much. My grandma had a lot of like peaches apples, in trees, like that, and we used to eat that. When we came here, we used to eat more 128

PAGE 146

like junk food. But not there-more like homemade food. Especially when you are at school you go "Oh, let's go get a hamburger, nachos, pizza." Or like on the weekends, we were all together junk food. "Oh I don't feel like cooking, let s go get some junk food." We would be tired during the week, cooking and stuff like that Friday, Saturday and you don t want to do anything, you buy junk Key informants noticed these same changes the abrupt nutritional transition coupled with changes in energy expenditure and increased social isolation observations also made by some researchers such as Himmelgreen eta/. (2007) These elements added to a traditional Mexican diet, provide the predicate for higher maternal weight at pregnancy and higher odds of LGA among Mexican-born women Food insecurity leads to energy dense food diets because such foods are cheaper in the U S It has an i mpact in terms of food security ... a lot of times when they f i rst come there is food insecurity because they just don t have money to buy food. But then when they start working and get a little more settled then I think you see the changes in their diet. And I think there is less food security i n Mexico than they end up having once they get settled here. . It' s partly a lack of understanding of nutr i tion and a change of d i et when they come here and they eat very differently than they did at home because they have all these cheap, cheap fast foods that they can get and so i t s partly educational deficits and partly economic reasons why they eat high fat h igh calorie foods. Public health nurse When they come they start adopting the American lifestyle and start eating carbs because i n Mexico poor families mostly have carbs i n the form of tortillas Here they start eating doughnuts and pizza Former Mexican physician Changes i n energy expenditure and social isolation also contribute to higher maternal wei ght. And they're in the trailer park and the trailer park is far from the supermarket and so their husbands drive them where they need to go and other than that they are at home And they're also nervous about walking around because it' s an unfamiliar area . They're socially i solated they don t know many people and they are physically isolated in their homes and of course culturally isolated because they have left a lot of fami liarity behind. [and] that leads to all sort of lifestyle changes around physical activity and diet. Public health nurse Some people come from small towns ; they don t have washing machines so they do their wash by hand [on a washboard] [When they are i n the U .S.] they are staying at home waiting for the husband to come home at 129

PAGE 147

night, and when he comes, they start eating and eating. And when she stays at home, taking care of the kids, and in the evenings they are afraid to go out. Former Mexican physician So you look at health disparities in some populations, you see what foods are available to them in their neighborhoods, for example, where do they go they go to fast food places because food is cheap and they don't go to large supermarkets and buy fresh fruits and vegetables because they don't see them as being cheap, so they go to the 7-11 and you can buy a lot of food cheaply but it's soda and junk food. . that's one of the shifts access to more processed foods. .. A lot of them tell me they don t feel safe exercising, going out and exercising in the areas they live in-or they don't have access they can't even go to the neighborhood rec center because they don't have $10. Because they have small children, they don't have anyone to leave their children with. I think there s this idea that Hispanic families are these tight-knit, extended, large families that really support one another, and I think that can be true but many of these new immigrant women are here alone, and they are very isolated. Midwife Discussion of Qualitative Interviews The observations of key informants represent a convergence of opinion that is consistent with the medical literature, social and political aspects of immigration and the effects of immigration on lifestyle Maternal BMI before pregnancy and the presence of gestational diabetes are predictors of LGA Abrupt change in diet combined with access to more and cheaper energy-dense foods changes in daily energy expenditure, cultural beliefs about eating for two and not running" when pregnant, and isolation characterize many of the Mexican immigrants the key informants see in the course of their work. They also suggest that escape from poverty in Mexico is the primary motivating factor for immigration. Based on the key informant interviews, I anticipated sharper differences between the mothers born in Mexico and those born here. To a surprising degree, though, there was little discrepancy overall between the two groups of women All mothers considered themselves rooted in Mexican culture, at least with respect to diet and certain beliefs of post-pregnancy practices. Every woman was married to or in a committed relationship with a man of Mexican origin, and several of the U.S.-born women were married to men who had been born in Mexico, consistent with a strong degree of endogamy All mothers also generally eschewed exercise during pregnancy because they felt they got enough exercise in their daily routines, because they were concerned about exercise being harmful to their pregnancy, or because they believed it was permissible to be less active or more lazy" (Yolanda). 130

PAGE 148

Neighborhood was decidedly not a source of support or even a meaningful construct from which to seek support for these women, in line with the quantitative results showing little association of immigrant orientation on birth outcomes However social isolation seemed greater for women born in Mexico. It is likely that three of the Mexican-born mothers didn t drive I met them in their homes or they arranged for someone to drive them to meet me. These three mothers also asked to conduct the interview in Spanish (although each had some facility in English}. In these senses, they may have been more soc i ally and linguistically isolated although only the youngest mother seemed so in the interview and that may have been because she was only 18. She referred often to her mother-in-law as the person to whom she looked for advice, usually with clear subservience in her tone. All mothers had some family within the Front Range area-either in the same town or within fifty miles of where they lived, with whom they described friendly relations In-laws were cited more often than even parents as sources of support of all types (close ties, sources of advice and encouragement for Ia cuarentena). Isabel summed it up in a torrent of Spanish without interruption, about why she thought Mexican-born women have more LGA babies. Although there are some cultural beliefs at work (women wanting to be rounder, eating carbohydrates to provide nutrition to the baby), her focus is on the structural constraints within the larger political system that affect Mexican immigrants (poverty, isolation, depression the inability to get around without a car) Isabel is one of eleven children and migrated to the U.S. when she was 24. It is possible she was describing herself to some extent, because she alternated between "we" and 'they." Perhaps because we like food too much. Mexican women in the U.S., we are inclined to eat more at meals and to get fat. And to be good for the baby you should be more round more fat. Women in Mexico eat more foods like lentils, beans, favas all to provide nutrition to the baby. The majority of [Mexican] women in the U S have problems with money; economically and there are many women who are not economically well off. Therefore it is more difficult to feed ourselves and to live well and to have a good environment for the baby. And sometimes we can t go to the doctor because we don t have money for the appointment. And I think all of that affects many things and emotionally too. You feel alone here. You don t feel you have support. Even if you have family in the U.S., sometimes you don t feel much support There are women who are here alone Alone! Sometimes they don t have help getting to the clinic or they don t understand so they don t go. And many people do not know that you can get food stamps to eat. When you are pregnant and alone your are emotional. You feel sad. Melancholy And your self esteem drops. And sometimes you don t have anything to eat and you don t take care of yourself The difference in Mexico, is, for example most people have their 131

PAGE 149

house And they don' t have payments hanging over them. So I believe these things contribute to a person's emotional state. Also in Mexico if you don' t have money you don' t lose your house. And to get around you don' t need a car or a truck. It' s not difficult there you can walk, or take the bus. 132

PAGE 150

CHAPTER 6 DISCUSSION This study sought to determine whether an epidemiological paradox exists for four birth outcomes related to the weight of the infant (LBW preterm birth SGA and LGA) among Hispanic mothers giving birth in Colorado during the period 2000 2005. The present study compared Hispanic mothers with mothers of other races/ethnicities and Mexican born mothers with U S.-born mothers of Mexican origin to identify individual-level contributors to those outcomes. This study also sought to determine whether contextual/area level factors contribute to birth outcomes among mothers of Mexican origin in Adams and Denver Counties and used qualitative methods to enrich interpretation of quantitative findings. Previous studies of the epidemiological paradox focused on low weight related birth outcomes only. Favorable odds of low birth weight preterm birth and infant mortality led to considerable controversy over the existence of an epidemiological paradox and various proposed hypotheses to explain these results. Many explanations seem unsatisfying, either because they are too simplistic or too narrow to provide insight into the complex factors that affect birth outcomes. Four key findings emerged from this study. First an epidemiological paradox exists tor Hispanics in Colorado with respect to low birth weight, preterm birth, small for gestational age status, and large for gestational age status. Despite having worse social and medical profiles than non-Hispanic White mothers, Hispanic mothers have similar odds of each of the tour birth outcomes. Second the epidemiological paradox also exists for Mexican-born mothers compared with U.S -born mothers of Mexican origin for LBW preterm birth and SGA. The paradox does not however exist for LGA, where Mexican-born mothers have much higher odds of LGA than U S.-born mothers of Mexican origin Third the hypotheses offered to explain the paradox, are, indeed, unsatisfying The results of the study do not support the healthy migrant or healthy immigrant hypotheses. Finally, and unexpectedly neighborhood measures of immigrant orientation and neighborhood deprivation do not influence the likelihood of outcomes as strongly as certain other studies have shown. 133

PAGE 151

This study has several limitations. First, it is cross sectional and therefore cannot measure outcomes across generations or determine causation. Second, important data that are lacking in the birth record would strengthen the study. Age at migration, length of residence of foreign-born mothers and primary language spoken would be informative for categorizing mothers as well as for creating predictors for behaviors and outcomes. Maternal wei ght and BMI before pregnancy would be very useful for understanding LGA. Finally if i t i s believed that soc ial neighborhoods are i mportant i nfluences on health outcomes reli ance on census tracts must cede to defin i tion of neighborhoods by the people who live there or perhaps it is time to move away from neighborhood analysis and focus instead on soc ial networks of the type mothers in this study suggest are meaningful to them. Is There an Epidem i o logi cal Paradox i n Weight Related B irth Outcomes? The quant i tative results of this study demonstrate that the epi demiological paradox ex i sts i n Colorado for Hispanics. Hispanics have the same odds of preterm birth and lower odds of LGA compared w ith non-Hispanic White mothers. Hispanics have 18% higher odds of LBW and SGA than White mothers but these odds are much lower than those of Black mothers whose SES is similar to that of Hispanics For LGA while H i spanics have odds 5% l ower than those of Wh i tes the i r odds are 36% h i gher than those of Blacks. H i spanic B l ack and Other mothers have lower odds of having an AGA baby -t hat i s t he i r l ikel i hood of hav ing an appropriate weight for gestat i onal age i s l ess than the ma j or it y Wh i te populat ion, although the odds for Hispanic mothers are only 4% lower than those of Whites. Table 6.1. Comparison of fully adjusted odds ratios by race/ethnicity White 1 00 1 00 1 00 1 00 1 00 Hispanic 1 18 1 .01 1.18 t 0 95 t0. 96 Black LBW Preterm Birth SGA LGA AGA No statistical difference from odds for White mothers. t Statistically significant at p so.os. 2 16 1 .42 1 98 t0.59 t0. 72 Other 1 64 1 1 6 1 72 t0. 68 t0. 78 The paradox also exists for low wei ght related outcomes for Mexican-born mothers compared w ith U.S -born mothers of Mexican or i gin even though Mex i can-born mothers have lower SES than U S.-born mothers In contrast to results based on race/ethnc i ty 134

PAGE 152

however, the paradox does not hold for LGA, because Mexican-born mothers have 45% higher odds of LGA than U.S.-born mothers of Mexican origin. Mexican-born mothers also have higher (better) odds of having an AGA baby (1.08). Table 6.2. Comparison of fully adjusted odds ratios by nativity LBW Preterm Birth SGA LGA AGA U.S.-Born Mexican Origin 1.00 1.00 1.00 1.00 1 00 No statistical difference from odds for U S.-bom mothers t Results statistically significant p :>0.05. Mexican-Born *0.93 *1.08 t0.74 t1.45 t1. 08 This study confirms the results of many previous population studies with respect to low weight outcomes. The intense and continuing public health emphasis on LBW and preterm birth is understandable, given the large immediate costs both personal and to the healthcare system, of caring for LBW and preterm babies, and the longer-term health consequences, such as cardiovascular disease and diabetes, that are associated with being born early or small. It is somewhat surprising that less public health emphasis has been placed on SGA babies, who share the increase in risk for negative long-term consequences with LBW babies, in light of the high frequency of SGA births in Colorado (12.13%), which is almost double the rate of LBW. SGA has multifactorial causal contours, and is likely undiagnosed during prenatal care. Given the life course complications from SGA, it deserves more study as an adverse outcome. More important, however studying outcomes only at the low weight end of the spectrum misses an important public health risk -that of LGA. Overall, the public health focus on LBW has created a false sense of well-be ing for babies born wei ghing more than five and a half pounds. All four outcomes in this study are associated with higher risks for developing obesity and metabolic disorders, especially diabetes later in life. Certainly the focus of population studies on low wei ght birth outcomes has left LGA out of the public eye and policy formulation, and even suggested that various populations may be healthier than they really are (Amaro & de Ia Torre 2002, Borrell 2005). In a recent meeting of agencies interested in health baby outcomes in Colorado a funder mentioned that her agency was considering allocating fewer resources to programs focusing on Hispanics because of their 135

PAGE 153

" better birth outcomes. Informat ion on the results of this study for LGA was a surprise to the audience. The Hispanic Paradox is mistakenly reported even in the mass media together with cultural stereotypes explaining it. Women who have recently arrived from Mexico have bigger healthier babies than more affluent non-Hispanic white natives. That s because strong family and social networks support these pregnant women reminding them what to eat and do But the longer they stay and the more assim i lated they become the more bad habits they acquire and the more problems the i r subsequent babies have. (Brooks 2006}. Perhaps most distressing i s Colorado s recent campaign to reduce low birth weight with a social marketing message A Healthy Baby is Worth the Weight," which is not appropr i ate messaging to mothers of Mexican origin. A l though the state i s now rethinking i ts campaign to better align with recommendat i ons to eat a healthy diet and to continue to exercise during pregnancy the ex i stence of a three year state wide campaign that focused on ga i ning weight during pregnancy shows the policy emphasis on low weight outcomes to the exclusion of others. It is tempting to suggest that the better outcomes at the low weight end of the spectrum for Hispanics and mothers of Mexican origin merely represent a shift to LGA outcomes But the analysis of odds of AGA babies at least for mothers of Mex i can or i g i n bel i es that not ion because they h ave better odds of an AGA baby. Although i t i s possible that a mother w i th h i gh maternal weight entering pregnancy who smokes w ill del i ver an AGA baby when all signs point to LGA (Barbour, L., personal i nterv i ew November 1 2008) the i ncrease in LGA babies among Mex i can-born mothers i s not expla i ned by their odds of AGA. I t i s poss i ble however that these mothers have lower odds of LBW and SGA because they have more AGA and LGA babies. Figure 6 1 shows the h i stograms of b irth weight i n grams for Mexican born mothers (above) and U.S.-born mothers of Mexican origin {below). The r i ght shi ft of the populat ion curve for Mex i can-born mothers i s cons i stent w ith the i r h i gher odds of LGA bab i es and sli ghtly larger bab i es overall but it is unlikely to explain the much h i gher r i sk of LGA among Mexican-born mothers. Neverthe l ess shi ft ing the curve even slightly to the left to would improve outcomes for Mexican-born mothers (Rose 1992) 136

PAGE 154

Figure 6.1. BIRTH WEIGHT IN GRAMS OF MEXICAN-BORN MOTHERS (ABOVE) WITH U.S.-BORN MOTHERS Do the Hypotheses in the Literature Explain the Paradox? How Should Health Be Measured? Since two of the hypotheses that try to explain paradoxically positive are based on the relative health of a population, it is useful to ask how one should measure health" in this context. The healthy migrant and the healthy immigrant hypotheses suggest that foreign-born women are either healthier overall (healthy migrant) or engage in healthier behaviors notably with respect to diet smoking, and weight gain during pregnancy and are therefore healthier during their pregnancy (healthy immigrant). The epidemiological paradox literature equates health" with better specific outcomes, while recognizing that immigrant populations have better outcomes for some health conditions, but not for others Existing literature about birth outcomes has focused on low weight birth outcomes such as LBW and preterm birth with a few studies including SGA as an outcome. At a minimum, it is reasonable to suggest that healthy birth outcomes based on weight should include outcomes along the full range of weight-related outcomes including LGA and perhaps AGA. Another way to measure health is to examine the risk factors associated with various outcomes and determine whether different populations have different health risk profiles. 137

PAGE 155

Hispanic mothers have worse medical risk profiles than Whites (and to a large degree worse than those of Blacks). Mexican-born mothers have higher odds of specific medical risks associated with LGA babies (gestational diabetes, pre-existing diabetes and previous 4000+ gram infant) than their U.S.-born counterparts. By these measures, neither Hispanics nor Mexican-born mothers can be said to be healthier than their comparison groups in the U .S. Healthy Migrant Hypothesis The healthy migrant hypothesis posits that healthier people migrate to the U.S and that thi s selection bias explains the paradoxically better outcomes of Mexican-born mothers. Although this study does not test the healthy migrant hypothesis directly it is possible to critique this explanation. While i t is undoubtedly accurate to suggest that those who make the migration to the U S from Mexico are sufficiently able to make the trip, else they would not be here that fact alone does not mean they are health i er' than those who do not migrate nor does i t mean that they make the choice to m i grate based on their health. Moreover the explanation fails on logi c alone. To explain birth outcomes in the U.S. the appropriate comparison i s between Mexican-born mothers in the U.S. and other U.S. mothers, not a comparison with mothers in Mexico who are irrelevant t o rates of birth outcomes i n the U.S. Thus although they do not speak to the relative health of Mexican women who immigrate and those who do not the data show that Colorado s Mexican-born mothers are more l i kely than U.S -born mothers of Mexican orig i n to have LGA related medical risks so they cannot be pai nted broadly w ith the hea lthi er' brush. In add i t ion, the very small d i fferences on odds for each outcome between Models 2 and 3 (where medical risks are entered into the predict i ve model) suggest that underlying health related to pregnancy does not exp lain differences i n outcomes. For these reasons alone the healthy migrant explanation falls short. Moreover stud ies of immi gration from Mex i co to the U.S i dentify var i ous aspects of politi cal economy wr i t large to expla i n the motivation to m i grate (Portes & Bach 1985). The i r indi vidual decis i ons to migrate (agency) are spurred almost exclusively by structural economic factors. Key informants descr i bed the changing face of m i grants from Mex i co and the poor economic conditions that cause them to come here. I'm not so conv i nced that i t s the healthier women that come because a lot of time i t s i mpover i shed women that come ; they are the ones that are willing to take the risk because the poverty levels i n their families i n Mexico are so desperate that they have to l eave Midwife 138

PAGE 156

Another Colorado mother born in Mexico described the cultural beliefs about food, the harsher economic conditions, and despair of Mexican-born mothers in the U.S. The majority of women in the U.S. have problems with money; economically and there are many women who are not economically well. Therefore it is more difficult to feed ourselves and to live well and to have a good environment for the baby. And sometimes we can't go to the doctor because we don't have money for the appointment. And I think all of that affects many things and emotionally too. You feel alone here. You don't feel you have support. Even if you have family in the U.S., sometimes you don't feel much support. There are women who are here alone. Alone! Healthy Immigrant Hypothesis The healthy immigrant hypothesis suggests that immigrants engage in healthier behaviors, at least for some period of time, and then the positive outcomes decay over time spent living in the U.S. The data on Colorado mothers support the observation that Mexican born mothers smoke less than U.S.-born mothers of Mexican origin, by a factor of 6. But, Mexican-barn mothers are more likely to gain less weight and less likely to gain excessive weight during pregnancy than Whites or U.S.-born mothers of Mexican origin. Yet Mexican-born mothers have much better outcomes than U S.-born mothers of Mexican origin at the low-weight end of the spectrum (which is consistent with not smoking but not consistent with low weight gain) and they have higher odds of LGA (even though they gain less weight). Weight gain during pregnancy does not provide a complete picture of the effect of weight on the birth outcomes in this study. If a mother begins her pregnancy overweight or obese, even if she gains a normal amount of weight during the pregnancy, she may be at risk for LGA. At least one of the key informants noted that her Mexican-born population of mothers was generally overweight and often obese key risk factors for LGA and gestational diabetes. These observations call into question the validity of "healthy immigranf' behaviors. The slight differences in odds between Models 3 and 4 (where smoking and weight gain are added into the predictive models), suggest that the healthy immigrant hypothesis does not explain differences in outcomes. Smoking is a significant factor in increasing odds of LBW and SGA, and should be avoided. Low weight gain is associated with LBW, preterm birth, and SGA and high weight gain is associated with LGA. However, focusing just on weight gain, rather than maternal weight before pregnancy and the quality of the weight gain is simplistic and does not speak to Mexican-born mothers, for whom pre-pregnancy weight and LGA are significant concerns. Discussions with key informants and mothers describe cultural 139

PAGE 157

beliefs of eating for two during pregnancy and cultural values of their male partners favoring rounder and heavier women. Perhaps most important, differences in availability and quality of food in Mexico and the U .S., abrupt changes in energy expenditure and social and linguistic isolation likely contribute to a mini-nutritional transition upon immigration. These factors speak more to choices constrained by structure than to immigrants arriving with individually-based and realized healthy behaviors. Neighborhood Effects Neighborhood effects, especially immigrant orientation, might be a gross measure of availability of social support As Finch et a/. (2007) demonstrated in Los Angeles neighborhoods consisting of immigrant ethnic enclaves might moderate the effect of neighborhood deprivation on LBW. In contrast, contextual effects are weak in the two counties measured i n this study. Neighborhood deprivation and the interaction of deprivation and immigrant orientation have a marginally significant influence on LGA in Adams County Immigrant orientat ion may slightly moderate the negative effect of deprivation on LGA there. In Denver County, only LBW and SGA are weakly influenced by neighborhood deprivation which acts to i ncrease slightly the rate of these two outcomes Immigrant orientation has no effect. Interviews with mothers in Colorado strikingly confirmed that neighborhood is not a meaningful construct for them nor is it a proxy for social support. It is possible that the slight influence of immigrant orientation in Adams County has little to do with social support but instead merely reflects residential housing patterns among immigrants Political EconomyA Broader Perspective Even with a broader view of health one that includes individual and area factors ascr i b ing the reasons for an epidemiological paradox to selective migration of healthier'' migrants cultural influences that (may) support healthier lifestyles or the effect of neighborhood enclaves that (may) provide social support does not explain the complexity of human behavior and health outcomes not even for the few birth outcomes of this study. Instead it may be useful to step back and examine the structural effects of political economy. Behavior is complex and informed by multiple, often contradictory, cultural and health beliefs Political economy can be used to explain those global economic conditions that underlie the genetic condition of humans in the current era, the reasons for migration from one economic environment to another, the dietary and cultural changes associated with immigration to a different culture and the structural constraints on individual behavior in any political 140

PAGE 158

environment in which one lives. While such sweeping statements sound distal to birth outcomes that also have some genetic and biological influences, it is this broader view that provides a framework for the epidemiological findings of this study. Only then might it be helpful to suggest interventions to modify health behaviors that pertain to the specific outcomes of specific populations at risk. Significance The public health importance of this study centers on its revelation of important current health realities and insight into the structure of health disparities. By including LGA, it broadens the range of weight-related outcomes so that any paradox does not mask other negative outcomes In the same vein, it shows how important it is not to make policy decisions based on the traditional view of the paradox that misses both heterogeneity of the population and a broader view of outcomes It is hoped that Colorado will adopt culturally appropriate messaging about maternal weight and weight gain during pregnancy that addresses the specific risks and needs of Hispanic women, who represent 30% of the singleton births in Colorado. The broader political economic perspective on LGA suggests that reliance on individual interventions or social marketing alone will be insufficient. To the extent that Mexican-born immigrants are constrained by structural barriers to better health outcomes such as poverty lack of access to food stamps and healthy foods, real social support, especially for Mexican women living alone in the U.S., and social and linguistic isolation, broader approaches to pre-pregnancy and pregnancy assistance will be needed. And not to be lost in a study of the whys," it is i mportant to remember that as the odds of positive low weight outcomes among Mexican-born women decay over time the health of the largest subpopulation in the U.S. will deteriorate rapidly in coming years and set the stage for trans-generational perpetuation of ill health related to adverse health outcomes. Indeed, it may be time to eschew focus on paradoxical outcomes and instead work to improve birth outcomes for all populations. Even those who may temporarily enjoy paradoxically better outcomes today for LBW, preterm birth, and SGA also represent its disappearance later their own childbearing years or in the next generation. And the focus on low weight-associated outcomes obscures LGA where there i s no immigrant paradox. Accordingly, several next steps to expand on this study might include policy recommendations for Improved data collection relating to causes of LGA in PRAMS (Pregnancy Risk Assessment Monitoring System) Although questions about pre-existing and 141

PAGE 159

gestational diabetes are included in PRAMS, there are no detailed questions about nutrition (nothing about fruits, vegetables, methods of cooking [fried, baked, steamed, etc.]}, or questions about depression generally or during pregnancy (as compared to post-partum) Greater availability of fruits and vegetables on the WIG list of approved foods Improved prenatal and inter-conceptual counseling, especially for recent immigrants from Mexico. In addition, LGA may be an outcome that affects immigrants from countries other than Mexico. The political economy approach suggests that the conditions that may lead to increased risk of LGA in Mexican immigrants may apply equally to other immigrants. A national population study of LGA among foreign-born immigrants is warranted. 142

PAGE 160

APPENDIX A SELECTED STUDIES 143

PAGE 161

Appendix A. Selected Individual-level studies Author/Year Cobas eta/. (1996) Singh & Yu (1996) Frisbie eta/. (1998) Data HHANES 1980s 1985-1987 NCHS linked birtMnfant death files 1989-1991 NCHS linked birth/infant death files Hummer eta/. 1989-1991 NCHS (1999) linked birth/infant death files Fuentes-Afflick 1992 CA birth et a/ certificates (1999) Lansdale eta/. Puerto Rican (1999) Maternal and Infant Health Study 19941995 Hessel & 1990-1993 CA Fuentes-Afflick linked birtMnfant (2000) death files Buekens eta/. 1994 US birth (2000) Certificates Chung JH eta/. 1997-2002 (2003) Retrospective Cohort Memorial Health Care S stem Frisbie & Song 1995-1997 NCHS (2003) Linked birtMnfant death files compared with 1989-1991 files Gould eta/. (2003) 1995-1997 CA linked birth/infant death records Outcomes LBW LBW Preterm birth Infant mortality LBW Preterm birth Infant mortality Infant mortality VLBW1 MLBW2 LBW Infant mortality LBW Infant mortality Birth weight LBWby gestation LBW Preterm birth Infant mortality VLBW MLBW Preterm birth SGA <3% IM Results 1 Acculturation factors differentially affect LBW 2 Acculturation affects LBW through diet and smoking 3 Language is more important than ethnic ID on acculturation 4 Independent of dietary intake acculturated women are more likely to have LBW babies 1 Foreign-born status associated with reduced risk of IM and LBW for, among others, Mexicans 1 Paradox applies to Hispanic subpopulations for IM adjusted for LBW and preterm birth 1 Nativity affects pregnancy outcome by race/ethnicity, when large % is foreign-born 2 Mexican-born have less risky health profiles than US-born (especially re smoking) 3. BW and preterm birth are intervening variables in IM 1 VLBW OR 0 93 Latina/White 2 MLBW OR 1 0 Latina/White 3 VLBW OR 1 0 foreign/US-born (ns) 4 MLBW OR 0 93 foreign/US-born 1 Recent immigrants experience fewer stressful life events ; less likely to engage in negative health behaviors during pregnancy 2. Recent immigrants have better outcomes than earlier immigrants or US-born of Puerto Rican descent 1 LBW OR 0 98 Latina/White 2. IM OR 0 88 Latina/White 3 Some risk factors for LBW vary from risk factors for IM, e.g., age <18 = less LBW but higher IM 1 Lower LBW of Mexican American b/c fewer small, preterm babies 2 But mean BW lower for Mex i can American than White and overall preterm BW higher than for Whites which may represent m isclassification 1 Descending ranking of infants by BW for gestation =Whites, Hispanics Blacks 2.Differences in LBW due to differences in size at birth because of gestation 1. All groups showed increased rates of adverse birth outcomes but decreased rate of IM 2 Mexican Americans had higher preterm birth rate than Whites in both time periods but lower rates of IM than Whites 1 Despite high risk demographic profile, Mexican-born did not have elevated levels of LBW or neonatal mortality 2 But Asian Indians with lower risk demographic profile had high levels of LBW SGA, fetal mortality 3 Higher education better prenatal care and private insurance was protective for White and African American but not Mexican or Asian Indian mothers dual paradox" 144

PAGE 162

Appendix A. Selected Individual-level studies (continued) Author/Year Cho eta/. (2004) Data 1989-1994 National Health Interview Survey Rosenberg et 1996-1997 NYC a/ (2005) birth files Outcomes Self-reported overall health Daily activity limitations # da sin bed LBW Acevedo Garcia eta/. (2005) 1998 Detail Natality LBW dataset Appendix A. Selected multi-level studies Author/Year O'Campo eta/. (1997) Data 180 Census tracts in Baltimore linked to birth certificate data 1985-1989 Outcomes LBW Johnson et a/ CO 1992-1994 ( 1999) Birth certificate data BWas continuous variable Preterm birth Pearl eta/. (2001) 1994-1995 birth records 18 CA hospitals BWas continuous variable Gorman (1999) 1990 linked birth LBW Reagan and Salsberry (2005) Sellstrom and Bremberg (2006) Finch eta/. (2007 and death files 1979-1998 National Preterm birth Longitudinal Survey (very and of Youth 1979 moderate) cohort native bom only Review of multiple level studies LBW Child behavior Child injuries Child maltreatment 2000 birth records LBW for Los Angeles Coun CA Results 1 Hispanic subgroup differentials wide 2 Foreign nativity = more favorable outcomes supporting healthy migrant hypothesis 1 Positive outcomes of foreign bom largely due to more favorable distribution of behavioral risk factors 2 Nativity is significant predictor of LBW only among Mexicans (OR 0 .6) and other Hispanics (OR 0.7) 1 Interaction between race/ethnicity and nativity (foreign-bom) significant 2. Protective effect of foreign-bom for Blacks and Hispanics 3 Inverse relationship between low education and LBW for foreign-bom Hispanics Results 1 Some neighborhood characteristics directly associated with and had interactions with higher odds of LBW among Blacks 2. Individual risk factors for LBW behaved differently depending on characteristics of neighborhood 1 Male unemployment has larger effect than crime rates 2 No particular effects for Latino population 1 Less favorable neighborhood = lower BW for Black and Asian 2. No consistent relationship of neighborhood for foreign-bom or US-bom Latinas or Whites 3 BW increased with less-favorable neighborhoods among foreign-bom Latinas in high poverty or high unemployment neighborhoods 1 Differences by race/ethnicity are function of individual and area characteristics 2 Negative relationship between % foreign-bom in county and LBW 3. Specific contextual variables vary by race/ethnicity 1. Neighborhood disadvantage highly sensitive across race/ethnicity depending on measure 2 Direct effect of cumulative income inequality only for Hispanics; Female head of household fl Hisp. very preterm fl 1 Risk of LBW increased 1 0%+ for mothers in disadvantaged neighborhoods 1 Protective effect of Hispanic immigrant co-residence at neighborhood level especially for foreign-bom 145

PAGE 163

APPENDIX B HUMAN SUBJECTS APPROVALS 146

PAGE 164

University of Colorado at Denver and Health Sciences; Center Human Subjects ResearCh Committee Institutional Review Board Downtown Denver Campus Box 120 P O Box 173384 Denver Col orado 00217-3384 Phone : 303-556-4060 Fax : 303-556 -33n DATE: October 5, 2006 TO: Sharon Devine FROM: Debbie Kellogg, HSRC Chair SUBJECT: Human Subjects Research Protocol #2007-038The Latina Epidemiological Paradox in Colorado Your protocol has been approved as exempt under CFR Title 45 Part 46.10l.b. This approval is good for up to one year from this date Your responsibilities as a researcher include : If you make changes to your research prolocoi-
PAGE 165

r1a:n OfHV\:1< CAMPU 0912112007 To : Sharon Devine From : UCD Human Subjects Research Committee Subject HSRC Protocol 2007 Initial Review (APP001) 2nd Review/Panel : Exempt I Panel S Review Date : 19 September 2007 Not Human Research Tille: Hum a n Subjects Research Convnlltee Lawr!'n<:e Str ee t Center SUite 300 Your research project submitted to HSRC under protocol number has been reviewed and our determination Is that it is not human research as defined by our policies and wrrent regulations and i n accordance with OHRP (http : /twww hhs gov/ohrp) and FDA (http : / /www fda gov/oclohrtlirbslde _faulthtmLgyldeli Therefore, you may proceed wlth the project, stricUy following the protocol as submitted and reviewed by HSRC No continuing review of the project will be required However, you must resubmit the protocol to HSRC for approval if My substantive changes are made to the protocol in question. Comments: Re-reviewed lhls project {Investigator requested eX18nllon) and have determined It is Not Human Subject Research Thent Is no interaction or with subjects and there Is no obtaining of Identifiable private information as well identillera such as names, addresses, birth day and mont\ have been stripped by COPHE Mery Geda MSN Tony Robinson 148

PAGE 166

(fili, University of Colorado Denver Downtown Campus 03/1212008 Certificate of Approval Investigator : Sharon Devine Sponsor(s): Uod H8lllll1 And Behavioral Sciences Dept Subject: nu.: 1380 L a w r ence S1reet, S uite 300 Campus Box 120 P .O Box 1 7 3 364 Denver c o 80217 3364 HSRC Protocol2008-110 lniUal Review (APP001) 2nd WHY DO WOM&N IIOfiH t1 MeXICO DelNER HEAL TH I ER B.AI!IES THAN 01' MEJm the laat apprc!Oial d otharwtsa In the ......;.w cycle liMed below If )'OU '-a rw1ricledlhlgh rlak protocd. spadflo detallt wt1 be oulllned In lhla ....... Nan-compllanoa with Conllnulng Ravtaw will result In the ..,l nallon of this ltudy. TNa projac:t hn been anlgnad the folowlnG r...taw C)'dl: HSRC Continuing Review Cycle: 12monlhs We wll send you a Continuing Review Form to be oompleced prior to lhe due date Mary Geda MSN Tony Robinson Ravised 03105 200&-110 Penal: S 149

PAGE 167

@....n Univendty of Colorado Denver Do\"utown C'anqms 138 0 Lawrence Street. Suite 300 C ampus Box 120 P O.Box 173364 Denver. CO 80217-3364 08/1912008 Certificate of Approval 1nvestlga1or. Sponsor : Subject: Approval Date : Approval Includes : Sharon Devine lJcd Health And Behavi
PAGE 168

of Colm-a do Dt-nvtr D0\\'1\tO\Ql (' IURp1U 08/1912008 Mary Gada, MSN Revised 03105 Protocol with highlighted revisions Consent for Key Informants Discussion Guide for Key Informant Interviews Tony Robinson 151 1 380 Lawrence Street Suite 300 Campus Box 120 P O.Box 173364 Denver CO 80217 3364 20011-110 Panel; S

PAGE 169

APPENDIXC SOLICITATION GUIDE INTERVIEW GUIDES CONSENTS 152

PAGE 170

Solicitation Script Approved by Humans Subjects Review Committee March 11, 2008 Would you like to participate in a study about differences in the size of babies at birth? Salud is helping Sharon Devine fmd women of Mexican origin for a research study about their pregnancy experiences. This study is being conducted by Sharon through the University of Colorado Denver. If you o are at least 18 years old o were born in Mexico or born in the U.S and of Mexican origin o have had a baby in the past year o are willing to be interviewed for an hour I will give Sharon your telephone number and she will call you to see if you want to be in the study. If you agree to be in the study you will be paid for your time Is it OK for me to give Sharon your first name and phone number? She will call you if you are interested. You can say no and it will not affect your access to Salud's services. 153

PAGE 171

(,Le gustarfa participar en un estudio sobre las diferencias de tarnafi.o de los bebes? Salud esta ayudando a Sharon Devine a encontrar mujeres mexicanas para un estudio sobre sus experiencias durante el embarazo. Sharon conduce este estudio con la Universidad de Colorado Denver. Si usted tiene por lo menos 18 afi.os es Mexicana o de origen mexicano (Sharon quiere hablar con madres que nacieron en Mexico o en los Estados Unidos) tuvo un bebe el afi.o pasado esta de acuerdo en hablar con Sharon por no mas de una hora le dare su nllinero de telephono a Sharon y ella le llarnara para preguntarle si quiere participar en la entrevista. Siesta de acuerdo en participar, Sharon leva a pagar por su tiempo. Por favor dfgarne si le puedo dar a Sharon su nombre y numero de telefono. Ella le llamara por telefono y preguntara cuando y d6nde la puede ver. Usted decide si participar o no. Sino quiere participar en el estudio, va a seguir recibiendo atenci6n medica en la clfnica Salud como de costumbre. 154

PAGE 172

Topics for Semi-Structured Interview After Consent Process These are not the exact questions, but the topics and questioning approaches. Approved by Human Subjects Review Committee March 11, 2008 Introduction Going to ask questions, maybe share stories, want to know about your pregnancy, neighborhood, where/from whom you got support while pregnant. No right answers, just want you to tell me what is comfortable about your pregnancy. Audio tape so I don't miss anything. You can stop at any time; you can decide not to answer any question that you would rather not answer. Confirm whether born in Mexico or born in U.S. and of Mexican origin. If born in Mexico at what age did you come to U.S.Ihow long have you been in U.S.? Repeat that do not need to know full name, address, or immigration status About the baby How long ago was your baby born? Boy? Girl? Other children? First pregnancy? All babies living? Remember if baby born early (premature)? How much baby weighed when born (approximate is OK). Any complications with birth? Probe: c section? gestational diabetes, hemorrhage, other? Things going well now? Health behaviors What kinds of foods did you eat while pregnant? Probe for vegetables, home cooked, prepared (like from grocery store), fast foods What did you eat on a typical day? Breakfast/lunch/dinner. Did you smoke while you were pregnant? Before pregnant? If yes, probe how far along when found out pregnant, continue to smoke? Less? More? If no, why didn't you smoke? 155

PAGE 173

Did others in household smoke while you were pregnant, when in same room? How much? Did you drink alcohol (including beer) while you were pregnant? Before pregnant? If yes, probe how far along when found out pregnant, continue to drink? Less? More? If not, why didn't you drink? Did you gain a lot of weight while pregnant? Any idea how much? Exercise/activity about same before pregnant as when pregnant? Do you remember when you first went to the doctor/clinic after you became pregnant for prenatal care? Probe: how far along in the pregnancy? Diabetes? Check in I'm going to ask questions next about your neighborhood and the types of support you received during your pregnancy. Are you comfortable continuing? Need a break? Neighborhood How would you describe your neighborhood? Probe: where generally-city/more rural? Types of people who live there? Mostly people of Mexican origin? More recent immigrants than not? Probe for how much/%. Feel safe in neighborhood? Friends or family in neighborhood? Would you describe your neighborhood as poor? While pregnant did you spend most of your time in your neighborhood? How long have you lived in your neighborhood? Support Tell me/list the people who gave you support while you were pregnant not their actual names, but, for example, my sister, who lives next door, or my mother, who lives in Mexico -I would call her each week, or my husband, or whatever. Elicit list. Ask which ones provide which types of support probe: information? rides? maternity clothes? baby clothes/equipment? financial support? emotional support? Groups of people, church? Which ones most important to you? Why? 156

PAGE 174

An alternative if the listing is not eliciting responses would be to tell a story and ask if the woman were in this situation, where would she find solace/support. Example: You are 6 months pregnant. You can't remember when you felt the baby last move and you are getting worried. Would you talk to someone about your fears? Whom? Example: You are pregnant and your family lives far away, except your husband. This is your first child and you don't know what to expect. You are feeling overwhelmed. What would you do? Probe for people, groups, church, as sources of support. Familiar with concept of machismo/marianismo? Was marianismo a strong value during your pregnancy? How would you say it was expressed? By whom, examples? Did you practice La cuarentena after pregnancy? Probe-completely, partially? Did you know you would do La cuarentena before the baby was born? Did La cuarentena make you less anxious about your pregnancy? Can you think of anything else you think would explain how you coped with being pregnant and who/what was helpful? Sisters Do you have any sisters who live in Mexico? Have they had babies in Mexico? If so, tell me about their babies size, premature? How old were they when they had the/each baby? Finding that mothers born in Mexican mothers are very "healthy" when look at low weight gain outcomes that is, much less low birth weight babies, babies that are premature, or that are small given their gestational age. BUT they have more large for gestational age babies. And with women of Mexican origin who were born in the US, the pattern is the opposite. What do you think is happening? notice this based on your family or acquaintances? Significant changes in behavior between US and Mexico that are important and significant to them? What do you think is happening? Thank you 157

PAGE 175

Thank you very much for your help. Make payment. This research is for my dissertation, which will result in a long written report of my study. When it is finished (in about a year!), I will give a copy if it to Salud, in case you would like to see it. 158

PAGE 176

Consent Form Approval Date: Approved by HSRC March 11, 2008 Valid for Use Through: Study Title: Why do Women Born in Mexico Deliver Healthier Babies in the U.S. than Women of Mexican Origin Born in the U.S.? Principal Investigator: Sharon Devine HSRC No: 2008-110 Version Date: Version No: 1 You are being asked to be in a research study This form provides you with information about the study. A member of the research team will describe this study to you and answer all of your questions. Please read the information below and ask questions about anything you don t understand before deciding whether or not to take part. Why is this study being done? The goal of this study is to learn about the difference in size of babies. You are being asked to be in this study because you did all the following. You had a baby within the last year. You are of Mexican origin. You got health services from Salud. Up to 50 people will be in the study. What happens if I join this study? If you join the study, you will answer some questions. This should take less than one hour. You will be asked how you took care of yourself while you were pregnant. You will be asked whether you smoked or drank alcohol while you were pregnant. You will also be asked to tell me about your neighborhood. I will ask you questions about the types of people who live there, and the support the area provided during your pregnancy. In the interview I will share with you some stories about pregnancy. I will ask you whether your experience was the same or not. 159

PAGE 177

Consent Form Approval If you have sisters living in Mexico who have had children, you will be asked to tell me a little about what happened. Were their babies born early or small? I will also ask how old you sister was when the baby was born. What are the possible discomforts or risks? Answering my questions may make you uncomfortable. You may be upset or sad if your pregnancy was difficult or you felt that you did not have much support, embarrassment. You may also feel guilty you smoke or drank during your pregnancy. Finally, you may be embarrassed if you felt you did not have enough money to get care while you were pregnant. What are the possible benefits of the study? There is no direct benefit to you for being in this study. We hope to learn how to increase the number of healthy babies born to women of Mexican origin. Who is paying for this study? This research is being paid for by the University of Colorado Denver. Willi be paid for being in the study? Will I have to pay for anything? You will be paid $20 for participating in the interview. It will not cost you anything to be in the study. Is my participation voluntary? Taking part in this study is voluntary. You have the right to choose not to be in this study. If you choose to be in the study, you have the right to stop at any time, or to refuse to answer specific questions. If you refuse or decide to withdraw later, you will not lose any benefits or rights to which you are entitled. If you choose not to be in the study, you will still be able to get health services from Salud. 160

PAGE 178

Consent Form Approval Who do I call if I have questions? The researcher carrying out this study is Sharon Devine You may ask any questions you have now. If you have questions later, you may call Sharon at 303-556-6797. You may have questions about your rights as someone in this study. You can call Sharon Devine with questions. You can also call the Human Subject Research Committee (HSRC). You can call them at 303-315-2732. Who will see my research information? Information from the interview may be looked at by others. They are: o Federal agencies that monitor human subject research o Human Subject Research Committee o The researchers o The funder (the Health and Behavioral Sciences program at the University of Colorado Denver). People who work at Salud will not know if you agree to be in the study. People who work at Salud will not see information from the interview that identifies you. You will not be asked to give us your name, specific address, or your immigration status. The results from the research may be shared at a meeting or in published articles. Your name will be kept private when information is presented. Audio recording You may choose to speak either English or Spanish during your interview. The interview will be audio recorded and then written in a paper document. The audio file will be deleted after it is transferred to paper and all copies of the interview will be kept in a locked office file or on a secure computer that can be accessed only by the researchers. Three years after the study is concluded, the paper and computer copies of the interviews will be destroyed. 161

PAGE 179

Consent Form Approval Agreement to be in this study I have read this paper about the study or it was read to me. I understand the possible risks and benefits of this study. I know that being in this study is voluntary. I choose to be in this study: I will get a copy of this consent form. Date: __ Consent form explained by: ________ Date: ___ Investigator: ______________ Date: ___ 162

PAGE 180

Consent Form Approval Fecha: Approved by HSRC March 11, 2008 Esta forma es valida hasta: Titulo del estudio: <,Por que las mujeres mexicanas tienen bebes mas saludables que las mujeres de origen mexicano que nacen en los Estados Unidos? lnvestigadora Principal: Sharon Devine Numero HSRC: 2008-110 Fecha de esta versi6n: Versi6n Numero: 1 Se le esta invitando a participar en un estudio de investigacion. Esta forma le da informacion sobre este estudio. Una persona del equipo de investigacion le va a describir de que se trata este estudio y respondera a sus preguntas. Por favor lea Ia informacion que se presenta a continuacion y pregunte cualquier cosa que no entienda antes de decidir si quiere participar o no. (.Por que estamos haciendo este estudio? Porque queremos aprender sobre las diferencias de tamano de los bebes. La estamos invitando a que participe en el estudio porque: Tuvo un bebe en este ano que paso Es mexicana Se atendio en una clfnica Salud Hasta 50 mujeres van a ser parte de este estudio. (.Que pasa si participo en este estudio? Si participa en el estudio, va a contestar algunas preguntas. La entrevista no tomara mas de una hora. Se le va a preguntar sobre como se cuido mientras estaba embarazada. Se le va a preguntar si fumo o tomo alcohol cuando estaba embarazada. Algunas preguntas son sobre ellugar donde vive. Le voy a hacer preguntas sobre las diferentes personas que viven en Ia misma area, y el apoyo que recibi6 durante su embarazo. En Ia entrevista, voy a contarle historias de algunos embarazos y le voy a preguntar si su experiencia fue Ia misma o no. 163

PAGE 181

Consent Form Approval Si usted tiene hermanas en Mexico que tienen hijos, le voy a pedir que me cuente un poco sobre sus bebes, si los bebes de sus hermanas nacieron antes de tiempo o muy pequeiios. Tambien le voy a preguntar que edad tenlan sus hermanas cuando sus hijos nacieron. t.Cuales son los posibles riesgos del estudio o Ia posibilidad de que me sienta inc6moda? Responder estas preguntas podrla hacerle sentir inc6moda. Puede ser que las preguntas le causen tristeza si usted tuvo un embarazo diflcil o si no recibi6 mucho apoyo o si su embarazo fue causa de vergOenza. Puede ser que sienta culpa si fumo o tom6 alcohol durante su embarazo. Tal vez sienta vergOenza si no tuvo suficiente dinero para ir al medico a checarse cuando estaba embarazada. t.Cuales son los posibles beneficios de este estudio? Usted no obtendra ningun beneficia por participar en este estudio. Nosotros esperamos aprender como aumentar el numero de bebes saludables de mamas de origen mexicano. (.Quien esta pagando por este estudio? Esta investigaci6n es pagada por Ia Universidad de Colorado en Denver. t.Me pagaran por participar en este estudio? l. Tengo que pagar algo por participar? Le pagaremos $20 por participar en Ia entrevista. No le costara nada ser parte de este estudio. (.Que pasa si no quiero partlcipar? t.Es mi participaci6n voluntarla? Participar en este estudio es completamente voluntario. Usted decide si quiere o no formar parte. Si decide ser parte del estudio, puede parar Ia entrevista en cualquier momenta, o puede no contestar algunas preguntas si no quiere. Si no quiere participar o si una vez que empecemos decide parar 164

PAGE 182

Consent Form Approval Ia entrevista, usted no perdera ningun beneficia o derecho que le corresponda. Si no quiere participar en el estudio, puede seguir recibiendo atencion medica en una cHnica Salud. <,A quien le llamo si tengo preguntas? La investigadora a cargo de este estudio es Sharon Devine Puede hacer cualquier pregunta que usted tenga ahora. Si tiene preguntas despues, puede llamar a Sharon al 303-556-6797. Puede ser que tenga preguntas sobre sus derechos al participar en este estudio. Puede llamar a Sharon Devine con preguntas. Tambien puede llamar al Comite de Sujetos Humanos que participan en investigacion (Human Subject Research Committee) al303-315-2732. <,Quien podra ver Ia informacion de esta investigaci6n? Las instituciones o personas que podrfan ver Ia informacion de las entrevistas son: o Agencias federales que se encargar de vigilar Ia investigacion con sujetos humanos o El Comite de lnvestigacion con Sujetos Humanos o Los investigadores o El donador (el programa de Salud y Ciencias de Ia Conducta de Ia Universidad de Colorado en Denver). El personal de Salud no va a saber si usted aprueba ser parte del estudio. El personal de Salud no podra ver informacion de las entrevistas que Ia pueda identificar. No le vamos a pedir su nombre, direccion o su estado migratorio. Los resultados de esta investigacion podran ser difundidos en conferencias o publicados en artfculos de revistas. No se usara su nombre cuando se presente informacion en publico. 165

PAGE 183

Consent Form Approval Grabaci6n de audio de las entrevistas Puede elegir si prefiere hablar en ingles o en espaiiol durante Ia entrevista La entrevista sera grabada para poder ponerla por escrito mas tarde. La grabaci6n sera borrada en cuanto Ia entrevista se encuentre por escrito y las copias de las entrevistas se guardaran en un archivero bajo llave o en una computadora con medidas de seguridad de manera que solo los investigadores las puedan ver. Tres aiios despues de que se termine el estudio tanto las copias en papel como los archivos de Ia computadora seran destruidos. Acuerdo para participar en este estudio He lefdo este documento sobre el estudio o alguien me lo ha lefdo. Entiendo los posibles riesgos y beneficios del estudio. Se que participar en el estudio es voluntario. Deseo participar en el estudio : voy a obtener una copia de esta forma de consentimiento. Fecha: ____ Forma de consentimiento explicada por: _______ Fecha: __ lnvestigador(a): _______________ Fecha: __ 166

PAGE 184

Topics for Discussion in Key Informant Interv i ews 2008-110 Approved by HSRC August 18, 2008 1 Description of quantitative findings In accord with expectations and the Hispanic paradox," Mexican born mothers in Colorado have much better (lower) odds than U S. born mothers of Mexican origin for low birth wei ght small for gestational age, and preterm birth. Mexican-born mothers have much higher odds of LGA than U.S. born mothers of Mexican origin. 2. The "healthy immigranf' hypothesis suggests that women who migrate to the US are "healthier' than those who are born in the US. What does healthier mean in this context? 3 What factors might account for the higher rate of LGA among Mexican born mothers? Diet Exercise Cultural values (and how expressed) 167

PAGE 185

Consent Form Approval Date: Approved by HSRC August 18, 2008 Valid for Use Through: Study Title: Why do Women Born in Mexico Deliver Healthier Babies in the U.S. than Women of Mexican Origin Born in the U.S.? Principal Investigator: Sharon Devine HSRC No: 2008-110 Version Date: Version No: 2 (Key Informants) You are being asked to be in a research study. Why is this study being done? The goal of this study is to learn about four birth outcomes low birth weight, preterm birth, small for gestational age, and large for gestational age among women who have given birth in Colorado during the years 2000-2005. You are being asked to be in this study because you have experience working with mothers who have recently given birth in Colorado. Up to 50 people will be in the study. What happens if I join this study? If you join the study, you will answer some questions. This should take less than one hour. You will be asked if you have insights, based on your experience, into findings that suggest that: In accord with expectations and the "Hispanic paradox," Mexican-born mothers in Colorado have much better (lower) odds than U.S.-born mothers of Mexican origin for low birth weight, small for gestational age, and preterm birth. Mexican-born mothers have much higher odds of LGA than U.S.-born mothers of Mexican origin. What are the possible discomforts or risks? Answering my questions may make you uncomfortable if you are unable to shed light on the findings. You will not be asked to divulge or discuss 168

PAGE 186

Consent Form Approval personally identifiable information about any of your present or past patients or clients. What are the possible benefits of the study? There is no direct benefit to you for being in this study. We hope to learn how to increase the number of healthy babies born to women of Mexican origin. Who is paying for this study? This research is being paid for by the University of Colorado Denver. Willi be paid for being in the study? Willi have to pay for anything? You will not be paid for participating in the study. It will not cost you anything to be in the study. Is my participation voluntary? Taking part in this study is voluntary. You have the right to choose not to be in this study. If you choose to be in the study, you have the right to stop at any time, or to refuse to answer specific questions. If you refuse or decide to withdraw later you will not lose any benefits or rights to which you are entitled. Who do I call if I have questions? The researcher carrying out this study is Sharon Devine. You may ask any questions you have now If you have questions later, you may call Sharon at 303-556-6797. You may have questions about your rights as someone in this study. You can call Sharon Devine with questions. You can also call the Human Subject Research Committee (HSRC). You can call them at 303-315-2732. Who will see my research information? Information from the interview may be looked at by others. They are: o Federal agencies that monitor human subject research o Human Subject Research Committee o The researchers o The funder (the Health and Behavioral Sciences program at the University of Colorado Denver). The results from the research may be shared at a meeting or in published articles. 169

PAGE 187

Consent Form Approval Audio recording The interview will be audio recorded and transcribed. The audio file will be deleted after it is transferred to paper and all copies of the interview will be kept in a locked office file or on a secure computer that can be accessed only by the researchers. Three years after the study is concluded, the paper and computer copies of the interviews will be destroyed. Agreement to be in this study I have read this paper about the study and I understand its possible risks and benefits. I know that being in this study is voluntary I choose to be in this study: I will get a copy of this consent form. Date: __ Printed Name : ________ Signature: __________ 170

PAGE 188

BIBLIOGRAPHY Abrafdo, A Dohrenwend, B., Ng-Mak D., & Turner J. (1999) The Lat i no mortality paradox: a test of the salmon bias and healthy migrant hypotheses. American Journal of Public Health 89 (1 0) 1543-1548. Abrafdo-Lanza A., Chao, M & Florez K. (2005). Do healthy behaviors decline with greater acculturat i on? ; Implications for the Latino mortality paradox Social Science & Medicine 61, 1243-1255 Acevedo-Garc i a D., Soobader M., & Berkman L. (2005). The different i al effect of fore i gn born status on low birth weight by race/ethnicity and education. Pediatrics, 115 (1 ) e29-e30 Ahlsson F., Gustafsson J., Tuvemo T., & Lundgren M (2007). Females born large for gestational age have a doubled risk of giving birth to large for gestational age infants. Acta Paed iatri ca 96, 358 362. Ahluwalia, I.B Ford E.S. Link M., & Bolen J .C. (2007). Acculturation we i ght and we i ght related behaviors among Mexican Americans in the United States. Ethnicity & Disease 17 643 649) Akresh I.A. (2007) D i etary assimilation and health among Hispanic i mm i grants to the Un i ted States. Journal of Health and Social Behavior 48 (Dec), 404-417. Alexander, G.R. & Kotelchuck, M (2001 ) Assessing the role and effectiveness of prenatal care : history challenges and d i rections for future research Publ i c Health Report 116 (4), 306-316 Alexander G .R., Kogan M.D. & Himes J.H. (1999). 1994 1996 U .S. singleton birth we i ght percent i les for gestational age by race Hispan i c or i gin and gender Maternal and Child Health Journal 3 (4) 225-231 Amaro, H & de Ia Torre A. (2002) Public health needs and scientific opportunities in research on Latinas. American Journal of Public Health 92 (4), 525-529 Amer i can Heart Assoc i ation and Amer i can Academy of Ped i atrics (2005) 2005 Amer i can Heart Assoc i ation guidelines for cardiopulmonary resuscitat i on (CPR) and emergency cardiovascular care (ECC) of pediatrics and neonatal patients: neonatal resuscitation guidelines Pediatrics, 117 (5) e1-1 0. Baker, D.W. Cameron, K.A., Fe i nglass J Thompson J.A. Georgas P Foster S., P i erce D., & Hasnain-Wyn i a R. (2006) A system for rap i dly and accurately collect i ng patients race and ethnicity American Journal of Public Health 96 (3) 532 537. 171

PAGE 189

Balcazar, H., Krull, J., & Peterson, G. (2001 ). Acculturation and family functioning are related to health risks among pregnant Mexican American women. Behavioral Medicine, 27, 62-60. Barcenas, C .H., Wilkinson, A.V., Strom, S.S., Gao, Y., Saunders, K.S., Mahabir, S., Hernandez-Valera, M.A., Forman, M.A., Spitz, K.R., & Bondy, M.L. (2007). Birthplace, years of residence in the United States, and obesity among Mexican American adults. Obesity, 15 (4), 1043-1052. Barker, D. (1998). In utero programming of chronic disease. Clinical Science, 95 (2), 115128. Barker, D (2001 ). A new model for the origins of chronic disease. Medical Health Care Philosophy, 4 (1 ), 31-35. Barker D. (2002). Fetal programming of coronary heart disease. Trends in Endocrinology Metabolics, 13 (9), 364 368 Basch, P. (1999). Textbook of International Health. New York: Oxford University Press Baumeister, L., Marchi, K., Pearl, M., Williams, R., & Braveman, P. (2000). The validity of information on "race" and "Hispanic ethnicity" in California birth certificate data. Health Services Research, 35 (4), 869-883. Beebe, K .R. (2005). The perplexing parity puzzle. Nursing for Women s Health 9 (5), 394399. Bender D., & Castro, D. (2000) Explaining the birth weight paradox: Latina immigrants perceptions of resilience and risk. Journal of Immigrant Health 2 (3), 115-173 Berkman, L., & Clark C. (2003). Neighborhoods and networks: the construction of safe places and bridges. In I. Kawachi & L. Berkman (Eds.), Neighborhoods and Health (pp 288-302). Oxford, UK: Oxford University Press. Berkman L., & Glass, T. (2000). Social integration, social networks, social support, and health. In L. Berkman&. I. Kawachi (Eds.), Social Epidemiology(pp. 137-173) Oxford UK: Oxford University Press. Berry J.W (1997). Immigration acculturation, and adaptation. Applied Psychology: An International Review 46 (1 ) 5-68. Boorstin, D.J. (1987). Hidden History: Exploring our Secret Past. New York, NY: Random House, Inc. Breier B .H., Vickers, M.H., lkenasio, B.A. Chan, K.Y & Wong, W.P.S. (2001). Fetal programming of appetite and obesity. Molecular and Cellular Endocrinology, 185, 7379. 172

PAGE 190

Buekens P., Notzon F., Kotelchuck M. & W i lcox A. (2000). Why do Mexican Americans give b i rth to few low-birth-weight i nfants? American Journal of Epidemiology 152 (4) 347-351 Borrell, L.N. (2005). Racial identity among Hispanics: implications for health and well-being American Journal of Public Health, 95 (3) 379 381. Brooks D (2006 March 30). Immigrants to be proud of The New York Times. Retr i eved November 15 2008 from The New York Times website: h ttp :// select. nvtimes.com / 2006 / 03 / 30 / opin ion / 30br ooks htm I ? r= 1 &scp= 1 &sq = Dav i d % 20Brooks % 201mmiqrat i on % 20March % 2030 % 202006&st = cse. Burton A. & Altman D.C. (2004) proposed guidelines for reporting m i ssing covariate data. British Journal of Cancer 91 (1 ) 4 8. Candib, L. M (2007). Obesity and diabetes in vulnerable populations : reflections on prox i mal and distal causes Annals of Family Medicine 5 (6), 547-556. Carnethon M .R. (2008). Diabetes prevention in US ethnic minorities: Role of the env i ronment. Journal of American Dietetic Association 108 (6) 942-944 Casey B.M., Lucas M.J Mcintire D .O. & Leveno K.J. (1997). Pregnancy outcomes i n women with gestational diabetes compared w ith the general obstetr i c population Obstetrics & Gynecology 90 (6) 869-873 Centers for Disease Control & Prevention (2002a). Births : f inal data for 2000. National V i tal Statistics Reports 50 (5): Tables 43 44, 46 Retrieved January 3 2007, from the CDC' s website : http :// www cdc gov / nchs / data/nvs r/ nvsr50 / nvs r 50 OS. pdf). Centers for Disease Control & Prevention (2006a). America's children in brief : key national indicators of well being 2006 Table HEALTH 7 : infant mortality: death rates among i nfants by detailed race and Hispan i c ori gin of mother selected years 1983-2003. Retr i eved January 4 2007 from the CDC' s website : http : //www .ch i ldstats.gov/americaschildren / tables/health 7 .asp. Centers for Disease Control & Prevention. (2006b). Births: final data for 2004. National V ital Stat i st i cs Reports 55 (1) : Table 32. Retrieved January 3 2007 from the CDC' s website : h ttp :// www ded.gov / nchs / data/nvsr / nvsr55 / nvsr55 01. pd f. Centers for Disease Control & Prevention. (2006c). Eliminating disparit i es in infant health. Fact Sheet. Retrieved January 9 2006 from the CDC' s webs ite: h ttp :// www cdc gov / om h / AMH / factsheets /i nfant. h t m Centers for Disease Control & Prevention (2006d). Pregnancy Risk Assessment Mon i toring System (PRAMS) Methodology Retrieved April 15 2007 from the CDC' s webs i te: h ttp :// www cdc.gov / prams / m e t hodo logy. htm. Chadwick E. (1842) Report of an enquiry into the san i tary conditions of the labouring population of Great Britain London UK: Poor Law Commiss i on. 173

PAGE 191

Cho, Y., Frisbie W Hummer R., & Rogers R. (2004). Nativity duration of residence and the health of Hispanic adults in the United States. International Migration Review 38 (1), 184 212. Chung J., Boscardin W., Garite T., Lagrew, D., & Porto M. (2003). Ethnic differences in birth weight by gestational age : At least a part i al explanation for the Hispanic epidemiolog i cal paradox? American Journal of Obstetrics & Gynecology 189 (4) 1058-1062 Clark, L. (2002) Mexican-orig i n mothers experiences using children s health care serv i ces. Western Journal of Nursing Research 24 (2) 159-179. Cobas J. Balcazar H., Benin, M., Ke i th V., & Chong Y. (1996). Acculturation and low birthweight infants among Lat i no women: A reanalysis of HHANES data with structural equation models American Journal of Public Health 86 (3) 394-396. Collins J. & David, R. (1990) The d i fferential effect of traditional risk factors on i nfant birthweight among blacks and Whites in Chicago. American Journal of Public Health 80 679-681 Coll i ns J., David A., Symons R., Handler A., Wall S., & Andes S (1998) Afr i can American mothers percept i ons of their residential environment stressful life events and very low b i rthweight. Epidemiology 9 286 289. Collins J. David R., Symons, A., Handler A Wall S., & Dwyer, L. (2000). Low income African American mothers perception of exposure to racial d i scriminat ion and infant birthweight. Epidemiology 11(3) 337 339. Collins J., & Schulte N. (2003) Infant Health : Race R i sk and Res i dence. In I. Kawach i & L. Berkman (Eds.), Neighborhoods and Health (pp. 223-232). Oxford UK : Oxford University Press Colorado Department of Public Health and Environment (2000a) Tipp i ng the scales: weighing i n on solutions to the low birth weight problem in Colorado. Denver CO: Colorado Department of Publ i c Health & Environment. Colorado Department of Public Health and Environment (2000b). Vital statistics. Retrieved January 4 2007 from the CDPHE website: www cdphe state co / hs / statebirthtab l es2000f2 pdf Colorado Department of Public Health & Environment. (2000c) Vital statistics. Retr i eved January 4 2007 from CDPHE s website: www. cdphe state .c o /h s / cou nt y 2000 / Co l orado OOb. pdf. Colorado Department of Public Health & Environment. (2000d). Technical notes. Retr i eved March 10 2007 from CDPHE s website : h ttp :// www cdphe s t a t e co u s /h s / 2000 append i x 1 pdf Colorado Department of Public Health & Env i ronment. (2004a) V i tal statistics Retrieved January 4 2007 from CDPHE s website: www cdphe.state.co/hs / vs/2004/b15.pdf 174

PAGE 192

Colorado Department of Publ i c Health & Environment (2004b). Vital statistics. Retrieved January 4 2007 from CDPHE s website : h ttp :// www cdphe state co us /h s / vs / 2004 /b11. pdf Colorado Department of Public Health and Environment. (2004c) Vital statistics. Retrieved January 4, 2007, from CDPHE s website: http://www. cdphe state co.us /h s/ vs / 2004 / Colorado 2004 pdf. Colorado Department of Public Health and Environment (2005a) Racial and ethnic disparities in Colorado. Denver CO: Colorado Department of Public Health & Environment. Colorado Department of Public Health and Environment (2005b) The health status of Colorado s maternal and child health population. Retrieved March 7 2007 from CDPH E s website : www cdp h e s t ate co u s/ ps / mch / heal t h status2005 pdf. Crimm i ns E.M., Soldo B.J. Kin J.K., & Alley, D. E. ((2005). Using anthropometric ind i cators for Mexicans in the United States and Mexico to understand the selection of migrants and the Hispanic paradox. Social Biology 52(3-4), 164-177. Cronbach L.J (1951 ) Coeffic i ent alpha and the internal structure of tests Psychometr i ka 16 297-334. Day J (2001). National population projections. Retrieved March 30, 2007 from the U .S. Census website http :// www ce n sus gov / popu lat i on/www/ pop p r o fil e / natproj. ht m I. Desai, S. & A lva S. (1998) Maternal educat i on and chi l d health : i s the r e a strong causal relationsh i p? Demography 35, 71-81. Diamond J.(2003). The double puzzle of diabetes Nature 423(5) 599-602 Diez-Roux, A. (2000). Multilevel analysis in public health research. Annual Review of Public Health 21 171-192. D i ez-Roux A (1998) Br i nging context back i nto epi demiology: var i ables and fallac i es i n multilevel analysis. American Journal of Public Health 88 (2) 216 -222. Di Leonardo M. (1984) The varieties of ethnic experience Ithaca NY: Cornell University Press. Dollberg S., Marom R., Mimouni F B & Yeruchimovich M (2000). Normoblasts i n large for gestat i onal age infants Archives of Disease in Childhood Fetal Neonatal Edi t i on 83, F148-F149. Donato K., Kana 'i aupuni S., & Sta i nback M. (2003) Sex differences in child health : effects of Mexico-US migration. Journal of Comparative Family Studies 34 (3) 455-477 Drake A.J. & Walker, B.A. (2004) The i ntergenerational effects of feta l programming: non genomic mechanisms for the inheritance of low b i rth weight and card i ovascular r isk. Journal of Endocrinology 180 1-16 175

PAGE 193

Drewnowski, A. & Specter, S.E. (2004). Poverty and obesity: the role of energy density and energy costs. American Journal of Clinical Nutrition, 79, 6-16. Dubowitz, T., Acevedo-Garcia, D., Salkeld, J., Lindsay A.C., Subramanian, S.V., & Peterson, K.E. (2007). Lifecourse, immigrant status and scculturation in food purchasing and preparation among low-income mothers Public Health Nutrition, 104 (4), 396-4-4. Dyer, J.S., Rosenfeld, C.R., Rice, J., Rice, M., & Hardin, D.S. (2007) Insulin resistance in Hispanic large-for-gestational neonates at birth. Journal of Clinical Endocrinology & Metabolism, 92 (1 0), 3836-3843. Ebin V., Sneed, C., Morisky, D., Rotheram-Borus, M., Mangusson, A., & Malotte C. (2000}. Acculturation and interrelationships between problem and health-promoting behaviors among Latino adolescents. Journal of Adolescent Health 28, 62-72 Ehrenberg, H.M., Mercer B.M., & Catalano P.M (2004). The influence of obesity and diabetes on the prevalence of macrosomia American Journal of Obstetrics and Gynecology, 191, 964-968 Eriksson, J. Forsen, T., Tuuomilehto, J., Osmond C., & Barker, D. (2001}. Early growth and coronary heart disease in later life: longitudinal study. British Medical Journal, 322, 949-953 Fang, J., Madhavan, S., & Alderman M (1999) Low birth weight: race and maternal nativityimpact of community income Pediatrics, 103 (1 ) e5-e1 0. Finch, B., Lim, N., Perez, W., & Do D (2007}. Towards a population health model of segmented assimilation : the case of low birth-weight in Los Angeles. Sociological Perspectives, 50 (3), 445-468. Fiscella K (1995). Does prenatal care improve birth outcomes? A critical review. Obstetrics & Gynecology 85 (3} 468 479 Flores E., & Armijo C. (2001}. Colorado's Latino population grows. Research Brief 1. Denver, CO: Latina/a Research & Policy Center. Flores G., & Brotanek, J. (2005). The healthy immigrant effect: a greater understanding might help us improve the health of all children. Archives of Pediatric and Adolescent Medicine 159, 295-297. Frank, R., & Hummer, R. (2002). The other side of the paradox: the risk of low birth weight among infants of migrant and nonmigrant households within Mexico. International Migration Review 36 (3) 746-765. Fraser, A., Brockert, J., & Ward, R. (1995). Association of young maternal age with adverse reproductive outcomes. New England Journal of Medicine 332 1113-1117. Frisbie, W., Forbes, D., & Hummer R. (1998) Hispanic pregnancy outcomes : additional evidence. Social Science Quarterly, 79 (1 ), 149-169 176

PAGE 194

Frisbie W., & Song S. (2003) Hispanic pregnancy outcomes: Differentials over time and current risk factor effects The Policy Studies Journal 31 (2), 237-252 Fuentes-Afflick, E., Hessol N. & Perez-Stable E. (1999). Testing the epidemiological paradox of low birth weight in Latinos Archives of Pediatric and Adolescent Medicine 153, 14 7-153 Gorman B (1999). Rac i al and ethnic variation in low birthwe i ght in the United States: individual and contextual determ i nants Health & Place 5, 195 207. Gould E. (2006) Health i nsurance erod i ng for working families : employer-prov i ded coverage declines for fifth consecutive year. EPI Brief i ng Paper 175. Retrieved May 14 2007 from the Economic Policy Institute website: h tt p ://w ww epinet.o r g / con t ent.cfm /bp17 5 Gou l d J Madan A., Qin C & Chavez G. (2003). Perinatal outcomes i n two d i ssim i lar immigrant populations in the Un i ted States : a dual ep i dem i olog i cal paradox Pediatr i cs 111 (6) e676-e682 Greenland S. (2002). A rev i ew of multilevel theory for ecolog i c analyses Stat i stics in Medicine 21 389-395 Guendelman, S. & Abrams B. (1995). Dietary intake among Mexican-Amer i can women : generat i onal differences and a comparison with white non Hispan i c women. American Journal of Public Health 85 20-25. Guendelman S., Buekens, P Blondel, B., Kaminski, M., Notzon, F & Masuy-Strubant G (1999). Birth outcomes of i mm i grant women i n the United States France and Belgium Maternal and Ch ild Health Journal 3 (4) 177-189 Guendelman S Gould J. Hudes M., & Eskenazi, B. (1990) Generational d i fferences i n perinata l hea lt h among the Mex i can American populat i on : f i nd i ngs from HHANES 1982-1984. Ameri can Journal of Public Health 80, 61-65 (Supplement). Hansen J.P. (1986). Older materna l age and pregnancy outcome : a r eview of the literature Obstetr i cs and Gynecology Survey 41 (11) 726-742 Hales C N & Barker D.J.P (2001). The thrifty phenotype hypothes is. Bri tish Medical Bulletin 60 5-20. Hales, C.N. & Ozanne S .E. (2003) For debate : fetal and early postnatal growth restr i ction lead to d i abetes the metabolic syndrome and renal failure. D i abetologia, 46 (7) 1013-1019 Haml i n C. & Sheard S. (1998). Revolutions in health: 1848 and 1998? Bri t i sh Medical Journal 317, 587-591 Hediger M L., Overpeck M .D., Kuzmarski, R.J. McGlynn A., Maurer K .R. & Davis W W (1998) Muscularity and fatness of infants and young ch i ldren born small or large for-gestational-age Pediatr i cs 102 e60. 177

PAGE 195

Hessol, N., & Fuentes-Afflick, E. (2000). The perinatal advantage of Mexican-origin Latina women Annals of Epidemiology 10 516-523. Hill J O Wyatt, H.r., & Melanson, E.L. (2000}. Genetic and environmental contributions to obesity. Medical Clinics of North America, 84 (2), 333-346. Himmelgreen D Daza N .R., Cooper E., & Martinez D. (2007) I don t make soups anymore : preto post-migration dietary and lifestyle changes among Latinos living in west-central Florida. Ecology of Food and Nutrition 46, 427-444. Hosmer, D.W. & Lemeshow S. (2000) Applied Logistic Regression (2 d ed.). New York NY:John W i ley & Sons. Hsieh, H-F. & Shannon S .E. (2005) Three approaches to qualitative content analysis Qualitative Health Research 15 1277-1288 Hummer R., Biegler, M., deTurk P Forbes, D., Frisbie, W Hong, Y et al. (1999). Race/ethnicity nativity and infant mortality in the Unites States. Social Forces, 77 (3} 1083 1118 Hummer, R., Powers D., Pullum S Gossman G., & Fr i sbie W. (2007). Paradox found (again): i nfant mortality among the Mexican origin population in the United States Demography, 44 (3), 441-451. Institute of Medicine. (1985). Preventing low birthweight. Washington, D.C.: National Academies Press. Institute of Med i cine. (2006 pre-publication). Preterm birth: causes, consequences and prevention. Washington, D C.: National Academies Press Institute of Medicine (2003). Improving birth outcomes: meeting the challenge in the develop ing world Washington D.C.: Nationa l Academies Press Jensen, G. & Moore L. (1997) The effect of h i gh altitude and other risk factors on birthweight: independent or i nteractive effects? American Journal of Public Health 87, 1 003-1 007. Johnson T., Drisko, J Gallagher K & Barela C (1999). Low birth we i ght: a women s health issue Women s Health Issues 9 (5), 223-230. Jimenez-Cruz, A. & Bacardi-Gascon M. (2004). The fattening burden of type 2 diabetes on Mexicans. Diabetes Care 27 (5), 1213-1215. Jung C G. (1980). Psychology and Alchemy: Collected Works of C.J Jung (Vol. 12) (R.F C Hull, Trans.) Princeton NJ: Princeton University Press Kana iaupuni S., Donato K., Thompson-Colon T & Stainback M. (2005). Counting on k in: Soc i al networks soc i al support and child health status. Soc ial Forces, 83 (3) 11371164. 178

PAGE 196

Kaplan M., Huguet N., Newsom J., & McFarland B. (2004). The association between length of r esidence and obesity among Hispanic i mmigrants. American Journal of Preventive Medicine 27 (4) 323-326. Kas i rye 0., Walsh J. Romano P., Beckett, L., Gardia, J., Elvine-Kreis B., et al. (2005) Acculturation and i ts associat i on with health-r i sk behaviors in a rural Latina population. Ethnicity & Disease 15, 733-739. Kassel, J. (1964}. Social science theory as a source of hypotheses in ep i dem i ological research. American Journal of Public Health 54 1482 1488 Kawachi, 1., & Berkman L. (Eds ) (2003) Neighborhoods and Health. Oxford UK : Oxford University Press Kawachi 1., & Subramanian S (2006) Measuring and modeling the social and geograph i c context of trauma: a multilevel modeling approach. Journal of Traumat i c Stress 19 (2}, 195-203 Kelaher, M. & Jessop D. (2002) D i fferences in low-birthweight among documented and undocumented foreign born and US born Latinas Social Science & Medicine 55, 2171-2175. Kliegman R., & Das U (2002). Intrauterine growth retardat i on. In A. Fanaroff & R. Martin (Eds.) Neonatal Perinatal Medicine : Diseases of the Newborn (pp 228 262). St. Louis MO : Mosby Kleinman, K., & Kessel S. (1987). Racial differences i n low birth we ight-trends and r i sk factors. New England Journal of Medicine 317 (12} 749 753. Kotelchuck M. (1994}. The adequacy of prenatal care index : i ts US distr i bution and assoc i ation with low birthweight. American Journal of Public Health 84 1486-1489. Kramer M (1987). Determinants of low birth weight: methodological assessment and meta analys is. Bullet i n of the World Health Organ izati on 65 (5) 663 737. Kramer M (2003) The epidemiology of adverse pregnancy outcomes : an overview Journal of Nutrition 5 (Suppl 2) 1592S-1596S Kr i eger N., Chen J., Waterman P., Rehkopf, D & Subramanian, S. (2003). Race/ethn i c ity, gender and monitoring soc i oeconomic grad i ents in health: a comparison of area based socioeconomic measures-the publ i c health disparities geocoding project. Ameri can Journal of Public Health 93 (1 0) 1655-1671 Krieger N., & Gordon D. (1999) Letter to the editor re: use of census-based aggregate variables to proxy for socioeconomic group: evidence from national samples Ameri can Journal of Ep i dem iology 150 (8} 892 894. Kutsche P (1998) Field ethnography : a manual for doing cultural anthropology Upper Saddle River NJ: Prentice Hall. 179

PAGE 197

Lansdale, N Oropesa R., Llanes D & Gorman, B. (1999) Does Americanization have adverse effects on health?: stress health habits and infant health outcomes among Puerto Ricans Social Forces 78 (2) 613-641. LaVeist T. (1989). Linking residential segregation to the i nfant mortality race disparity in U.S. cities. Sociological Research 73 90-94 Lee B.A. & Mar l ay M (2007). The right side of the tracks : affluent neighborhoods in the metropolitan United States Social Science Quarterly 88 (3) 766-789. Lee P.A. Chernausek, S.D. Hokken-Koelega, A.C S., & Czernichow P (2003). International small for gestational age advisory board consensus development conference statement: management of short chi ldren born small for gestational age Apr i l 24-0ctober 1 2002 Ped iatri cs 111, 1253-1261 Lichter, D.T., Brown J.B., Qian Z., & Carmalt J H (2007). Marital ass i milation among Hispanics : evidence of declining cultural and economic i ncorporation. Social Science Quarterly 88 (3) 745-765. Link B., & Phelan J (1995). Social conditions as fundamental causes of d i sease Journal of Health and Social Behavior 35 (Extra Issue) 80 94. Lubchenco L.O., Searls D.T. & Brazie J .V. (1972). Neonatality nortality rate: relationship t o birth we i ght and gestational age. Pediatrics, 81 (4) 814 822. Lubchenco L.O. & Bard, H. (1971). Inc i dence of hypoglycemia in newborn i nfants classif i ed by birth we i ght and gestational age. Pediatrics 47, 831-838. Lynch, J., & Kaplan G. (2000). Socioeconomic position. In L. Berkman & I. Kawachi (Eds.) Social Epidem i ology (pp. 13 35). Oxford UK : Oxford Univers i ty Press Mac i ntyre S., & Ellaway A (2000). Ecolog i cal approaches: rediscover i ng the role of the physical and social env i ronment. In L. Berkman & I. Kawach i (Eds ) Social Ep i dem i ology (pp. 332-348). Oxford UK: Oxford Uni vers i ty P r ess Mac i ntyre S., & Ellaway A. (2003). Neighborhoods and health : an overv i ew. In I. Kawach i & L. Berkman (Eds ) Neighborhoods and Health (pp 20-44). Oxford UK: Oxford Uni vers i ty Press. Madan A., Pa l aniappan L., Urizar G., Wang Y ., Fortmann S & Gou l d J (2006). Sociocultural factors that affect pregnancy outcomes in two dissimilar immigrant groups i n the United States. Journal of Pediatrics March 341-346. Mainous A.G. Diaz V.A. & Geesey M.E. (2008) Acculturation and healthy lifestyle among Latinos with d i abetes. Annals of Family Med i cine 6 (2) 131137). March of Dimes. (2007). Professionals & researchers. Low birth weight. Retr i eved March 9 2007 from March of D i mes webs i te: h ttp: // www marchofd i mes com / profess i ona l s /681 1153 .a sp. 180

PAGE 198

Markides, K., & Coreil, J. (1986). The health of Hispanics in the southwestern United States: an epidemiological paradox. Public Health Reports, 101, 253-265 Marmot, M., Adelstein A., & Bulusu, L. (1984). Lessons from the study of immigrant mortality. Lancet, 112, 1455-1457. Marmot, M., Kogevinas, M., & Elston, M. (1987). Social/economic status and disease. Annual Review of Public Health, 8, 111-137. Martorell A. (2005). Diabetes and Mexicans: Why the two are linked. Preventing Chronic Disease. Retrieved November 20 2008, from CDC website: http://www cdc.gov/pcd/issues/2005/jan04 01 OO.htm. McGlade, M., Saha, S., & Dahlstrom, M. (2004). The Latina paradox: an opportunity for restructuring prenatal care delivery. American Journal of Public Health, 94 (12), 2062-2065. Medlinger, S. & Cwikei,J Spiraling between qualitative and quantitative data on women s health behaviors : a double helix model for mixed methods Qualitative Health Research 18, 280-293 2008. Meneses-Ganzalez, F, Romieu 1., Salgado de Snyder N., Camargo-Bohorquez C., Hennessy T & Schenker M. (2006). Socioeconomic status, workforce and determinants of health among Mexican immigrant women in the U.S Journal of Epidemiology, 17 (6 Supp), S385-386) Merson, M., Black, A., & Mills, A. (Eds .). (2001 ) International Public Health: Diseases Programs Systems, and Policies Gaithersburg, MD : Aspen Publishers, Inc Miech A., Kumanyika, S., Stettler, N., Link, B., Phelan J., & Chang, V (2006). Trends in the association of poverty with overweight among US adolescents 1971 2004. Journal of the American Medical Association, 295 (20) 2385-2393. Millard, A. (1994) A causal model of high rates of child mortality. Social Science & Medicine, 38, 253-268. Modanlou, H.D., Komatsu, G., Dorchester W. Freeman, R.K. & Bosu S. (1982). Large-tor gestational-age neonates: anthropometric reasons for shoulder distocia. Obstetrics & Gynecology, 60, 417-423. Montez J.K. & Eschbach K. (2008). Country of birth and language are uni quely associated with intakes of tat, fiber and fruits and vegetables among Mexican-American women in the United States. Journal of the American Dietetic Association 108 (3), 473-480). Morales, L., Lara M., Kington, A., Valdez, A., & Escarce J. (2002). Socioeconomic cultural and behavioral factors affecting Hispanic health outcomes. Journal of Health Care for the Poor and Underserved 13(4), 447-503. 181

PAGE 199

Morse, J.M. (2008a). 'What's your favorite color? Reporting irrelevant demographics in qualitative research. Qualitative Health Research, 18, 299-300. Morse, J.M. (2008b) Confusing categories and themes. Qualitative Health Research, 18, 727-728 Morse, J .. M. & Field, P.A. Qualitative Research Methods for Health Professionals (2d ed.) Thousand Oaks, CA: Sage, 1995. Murray, C., & Lopez, A. (Eds.) (1996) The Global Burden of Disease. Cambridge, MA: Harvard University Press. National Institute of Nurs i ng Research, National Institute of Child Health & Human Development, & National Institute of Dental & Craniofacial Research (2003). Reducing preterm and low birth weight in minority families. PA-04-027. Washington, D.C. Neuhauser, M., Thompson, B Coronado, G., & Solomon C. (2004). Higher fat intake and lower fruit and vegetables intakes are associated with greater acculturation among Mexicans living in Washington state. Journal of the American Dietetic Association 104 (1) 51-57 Niermeyer, S., Wells C., Williford, D., Honigman, B., Moore, L., Asmus, 1., Shupe, A., Jacobellis J. Lezotte, D., Egbert, M. (2006). Analysis of low birth weight, high altitude and smoking using geographic information systems. Platform paper presented at the Pediatric Academic Societies Meeting: May 2006, San Francisco, CA. Nunnaly, J. (1978). Psychometric Theory. New York, NY:McGraw Hill. O Campo, P., Xue, X., Wang, M., & Caughy, M. (1997). Neighborhood risk factors for low birthweight in Baltimore: a multilevel analysis. American Journal of Public Health, 87(7), 1113-1118. Oken, N., & Gillman G (2003). Fetal origins of obesity Obesity Research 11 (4), 496-506 Olsen S.F. Halldorsson T l., Willett, W C Knudsen V.K. Gillman M W Mikkelsen T .. Olsen, J., & the NUTRIX Consortium (2007). Milk consumption during pregnancy is associated with increased infant size at birth: Prospective cohort study American Journal of Clinical Nutrition, 86, 11 04-1110. Ostir G., Eschbach, K., Markides K., & Goodwin J (2003). Neighbourhood composition and depressive symptoms among older Mexican Americans. Journal of Epidemiology of Community Health, 57, 987-992. Palloni, A., & Arias, E. (2004). Paradox lost: explaining the Hispanic adult mortality advantage. Demography, 41(3) 385-415. Palloni, A., & Morenoff, J. (2001) Interpreting the paradoxical in the Hispanic paradox. Annals of the New York Academy of Sciences 954, 140-174. 182

PAGE 200

Pearl, M. Braveman, P., & Abrams, B. (2001). The relationship of neighborhood socioeconomic characterists to birthweight among 5 ethnic groups i n Californ ia. American Journal of Public Health 91 (11 ) 1808-1814. Philli ps K (1999) The Cousins Wars : Religion Politics & the Triumph of Anglo-America New York NY : Basic Books P i ckett K Coll i ns J., Masi C. & RG, W. (2005) The effects of racial density and income incongruity on pregnancy outcomes Social Science & Medic ine, 60 2229-2238. P i ckett K., & Pearl, M. (2001). Multilevel analyses of neighbourhood socioeconomic context and health outcomes: a critical review Journal of Epidemiology and Commun i ty Health 55, 111-122. Portes A., & Bach R. (1985). Latin Journey: Cuban and Mexican Immigrants i n the Un i ted States. Berkeley CA: University of Californ i a Press Gonzalez-Quintero V.H Tolaymat L., Luke B., Gonzalez-Garcia, A., Duthely L., O Sullivan, M J. & Martin, D (2007). Outcome of pregnanc i es among H i spanics. Journal of Reproductive Medicine 51 (1) 10 14 Raudenbush S & Bryk A. (2002) Hierarchical Linear Models: Applications and Data Analysis Methods (2nd ed.). Thousand Oaks CA: Sage Publications Reagan P., & Salsberry, P (2005). Race and ethnic differences in determinants of preterm birth in the USA: broadening the social context Soc ial Science & Medic ine, 60 1952-1957 Re i chman N.E. & Kenney G M (1998). Prenatal care birth outcomes and newborn hospital i zation costs: patterns among Hispanics in New Jersey. Fam ily Planning Perspectives 30 (4) 182-187 & 200. Ricketts S Murray E., & Schwalberg R. (2005). Reducing low birthwe i ght by resolving risks : results from Colorado s Prenatal Plus Prog r am. American Journal of Public Health 95 (11) 1952-1957. Robert S. (1999). Socioeconomic position and health : the independent contribution of community socioeconomic context Annual Review of Sociology 25, 489 516. Rose G. (1992) The strategy of preventive medic i ne. New York NY : Oxford University Press. Rose G., & Marmot M. (1981). Social class and coronary heart disease. British Heart Journal 45, 13-19. Roseberry W (1988). Political economy. Annual Review of Anthropology 17 161-185 183

PAGE 201

Rosenberg, T. Raggio T., & Chiasson, M. (2005). A further examination of the epidemiologic paradox'': birth outcomes among Latinas. Journal of the National Medical Association, 97 (4) 550-556 Rumbaut, R., & Weeks, J. (1996). Unraveling a public health enigma: why do immigrants experience superior perinatal health outcomes? Research in the Sociology of Health Care, 13 (B), 337-391. SAS. (2003). Version 9.1. 3. Cary, NC: SAS Institute Inc Schensul, S., Schensul J.J., & LeCompte, M.D. (1999). Essential Ethnographic Methods: Observations Interviews and Questionnaires Walnut Creek CA: Altamira Press Schoenborn, C.A. (2004). Marital status and health : United States 1999-2002. Advance Data from Vital & Health Statistics 351, 1-33. Retrieved September 24 2008 from the website of NCHS : www.cdc gov / nchs/data/ad / ad351 pdf. Scr i bner R. (1996) Editorial: paradox as paradigm -the health outcomes of Mexican Americans. American Journal of Public Health 86 (3), 303-305 Scribner R., & Dwyer J (1989) Acculturation and low b i rthweight among Latinos in the Hispanic HANES American Journal of Public Health 79 1263--1267. Sellstrom, E., & Bremberg, S (2006) The significance of neighbourhood context to child and adolescent health and well-being: a systematic review of multilevel studies Scandinavian Journal of Public Health 34 544-554. S i ngh G., & Yu S (1996) Adverse pregnancy outcomes: differences between USand foreign born women in major US racial and ethnic groups American Journal of Public Health 86 (6) 837-843 Sm i th J., & Edmondston B (1997). The New Americans. Washington D.C.: National Academy Press. Sonfield A. (2007) The impact of ant i -immigrant policy on publicly subsidized reproduct i ve health care. Guttmacher Pol icy Review 10 (1) 7-11. Stokes, M.E. Davis C S., & Kock, G .G. (2000). Categorical Data Analysis Us ing the SAS System (2nd ed.) Ca r y NC : SAS Institute Inc Subramanian S., Jones, K., & C, D. (2003) Multilevel methods for public health research I n I. Kawachi & L. Berkman (Eds.) Neighborhoods and Health (pp 65-111 ) Oxford UK: Oxford Uni versity Press. Sull i van L., Dukes K., & Los i na E. (1999). Tutor i al i n b i ostatistics an introduction to hierarch i cal linear model i ng. Statistics in Medic i ne 18, 855 888 Surkan P.J. Hsieh C C Johansson A.V.L. Dickman P W. & Cnatt i ngu i s S (2004). Reasons for increasing trends i n large for gestational age births. Obstetrics & Gynecology 104, 720-726 184

PAGE 202

Syme, S., & Berkman, L. {1876) Social class, susceptibility and sickness. American Journal of Epidemiology, 104, 1-8 Teller C., & Clyburn S. (1974). Trends in infant mortality. Texas Business Review 48, 240-246. United Nations (2008) The millennium development goals report. New York NY: United Nations Retrieved October 8 2008 from the United Nations webs i te : h ttp :/ / mdgs un .orq / unsd / mdq / Resources / Stat i c / Products/Proqress2008 / MDG Report 2008 En. pdf. U.S. Census. (2000a). PCT-1. 100 percent data ; corrected counts Retr i eved March 10, 2007, from the U .S. Census website : http : //www census.gov/Press Release /www/ 2001 /sumfile2.htm I. U S. Census (2000b). Census 2000 Summary File 1 (SF1) 100 Percent Data for Colorado. QTP9 Hispano or Latino by Type DP-1 and GTC-P6 Retrieved November 13, 2006 from the U.S. Census website : http :// factf i nde r .census.gov. U.S. Census. (2000c). American FactFinder Colorado by selected county Retr i eved November 15, 2006 from the U.S Census website: http : //factf i nder census.gov/home/saff/main.html? lang=en. U .S. Census (2000d) Hispanic population in the U.S Current population reports. Retrieved March 30, 2007 from the U.S. Census website: http ://www. census gov / prod / 2001 pubs/p20 535 pdf U .S. Census (2000e). Population Housing Units Area and Density:2000. SF1 GCT-PH1. Retrieved Nobember 11, 2008 from the U S Census website : h ttp ://factfin de r census gov /. U .S. Census. (2002b). Hispanic population in the U .S. Current population reports Retrieved March 30 2007 from the U.S. Census website: http:/ /www census gov/prod / 2003pubs/p20-545. pdf U S Census. (2002c). Census 2000 gazeteer. Retr i eved June 10 2007 f rom thhe U.S. Census website: http://www census gov / qeo /www/qaz etteer / places2k .h tm I. U. S Census (2007) State and county quickfacts Retr i eved March 10 2007 from the U S Census website: http://qu i ckfac t s.census.gov / qfd / states / 08000.h t ml. U .S. Department of Health & Human Services (2003) Healthy people 2010. Vol. II. Retrieved March 10 2007 from the Department of Health and Human Serv i ces webs i te : htt p ://w ww .hea l thypeop l e gov / De f au lt. htm U.S. Department of Health and Human Services {2000) Healthy People 2010 (2"d ed .). Understand ing and i mprov i ng health and objectives for improving health (2 vols.). Washington DC: U.S. Government Print i ng Office 185

PAGE 203

UNICEF. (2004). Low birthweight: country regional and global estimates Retrieved AprilS 2007 from the World Health Organization website: h ttp :/ /www .ch ildinfo org/areas/ birthweiqht/LBW WHO UN ICEF % 202000 pdf. Weigers M., & Sherraden, M (2001) A critical examination of acculturation: the impact of health behaviors social support and economic resources on birth weight among women of Mexican descent. International Migration Review 35 (3) 803-839. Wingate, M S & Alexander G .R. (2006). The healthy m i grant theory : variations in pregnancy outcomes among USborn immigrants Social Science and Medicine 62 491498 W i nkleby, M.A. Jatu lis, D.E., Frank E., & Fortmann S.P. (1992). Socioeconomic status and health: how education income and occupation contribute to risk factors for cardiovascular disease. American Journal of Public Health 82, 816-820. World Health Organization (2006) Reproductive health indicators-gui delines for the i r generation i nterpretation and analysis for global monitoring. Retrieved March 1 0 2007 from the World Health Organization website: http :/ /www.who.int/reproduc t ive h ea lth/ publicat i ons /rh i n d i cato r s/quidel i nes pdf. Yang Q ., Greenland S., & Flanders, W. (2006) Associations of maternal ageand par i ty related factors with trends i n low-birthweight rates: United States 1989 through 2000 American Journal of Public Health 96 (5) 856-861 Yeh M -C., Ickes S .B., Lowenstein L.M., Shuval K., Ammerman A.S., Farris R.. & Ka t z D .L. (2008). Understanding barr i ers and facilitators of fruit and vegetable consumption among a d i verse multi-ethnic popu l ation in the USA. Health Promotion International 23 (1 ) 42-51. Zambrana R., Scrimshaw, S., Collins N., & Dunkel-Schetter C (1997) Prenatal hea l th behaviors and psychosocial r i sk factors in pregnant women of Mexican origin : The role of acculturation. Ameri can Journal of Public Health 87 (6) 1022-1026. Zapata, 8. Rebolledo A., Atalah, E., Newman B., & K i ng M. (1992) The influence of soc ial and political violence on the risk of pregnancy complications American Journal of Public Health 82 (5) 685-690 186



PAGE 1

_]_ r IS THERE AN EPIDEMIOLOGICAL PARADOX FOR BIRTH OUTCOMES AMONG COLORADO WOMEN OF MEXICAN ORIGIN? Sl Y NO: IT DEPENDS ON THE OUTCOME by Sharon Jean Devine B.S. Linguistics, Georgetown University, 1970 J.D., Boston University School of Law, 1975 M.A. Anthropology, University of Colorado Denver, 2005 A thesis submitted to the University of Colorado Denver in partial fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Science 2009

PAGE 2

2009 by Sharon Jean Devine All rights reserved.

PAGE 3

This thesis for the Doctor of Philosophy degree by Sharon Jean Devine has been approved Date

PAGE 4

Devine, Sharon Jean (Ph.D., Health and Behavioral Science) Is There An Epidemiological Paradox For Birth Outcomes Among Colorado Women Of Mexican Origin? Si Y No: It Depends On The Outcome Thesis directed by Professor Susan Niermeyer ABSTRACT This study examines whether an epidemiological paradox exists for low birth weight, preterm birth, small for gestational age, and large for gestational age among Hispanic mothers in Colorado. It compares birth outcomes by race/ethnicity and place of birth (Mexico or U.S.) to identify individualand neighborhood-level contributors and to contextualize quantitative findings. The study analyses two retrospective cohorts all mothers (N=356,389) and mothers of Mexican origin (N=85,755) delivering singletons in Colorado during 2000 2005, using multiple logistic regression to test the social gradient of health by race/ethnicity and by nativity, to identify any paradoxical outcomes, and to explore the healthy migrant and healthy immigrant explanations for better outcomes among Mexican-born mothers. General lil"lear regression analyzes the association of neighborhood deprivation and immigrant orientation for mothers of Mexican origin in Adams (N=16, 1 07) and Denver (N=23,332) Counties. Five interviews with key informants and ten interviews with mothers of Mexican origin, half of whom were born in Mexico, are analyzed using directed content analysis. Four key findings emerge. First, an epidemiological paradox exists for Hispanics for all four birth outcomes, despite having worse social and medical profiles than non-Hispanic White mothers. Second, the paradox exists for Mexican-born mothers for low birth weight, preterm birth, and small for gestational age. No paradox exists for large for gestational age. Third, neither the healthy migrant nor healthy immigrant explanation is supported. Finally, neighborhood measures of immigrant orientation and neighborhood deprivation do not influence the likelihood of outcomes in Adams and Denver Counties. The public health importance centers on the identification of a hidden epidemic of large for gestational age among Mexican-born mothers and insight into the structure of health

PAGE 5

disparities. Any paradox at the low-weight end of the spectrum of birth outcomes no longer obscures the existence of negative high-weight outcomes, an important finding in Colorado where Hispanics represent 30% of singleton births. The broader political economic perspective suggests that reliance on individual-level interventions alone is insufficient to reduce LGA disparities because Mexican-born immigrants are constrained by structural barriers to better health outcomes, including poverty, lack of access to healthy foods, and social and linguistic isolation. This abstract accurately represents the content of the candidate's thesis. I recommend its publication. p Signed Su.n Niermeyer

PAGE 6

DEDICATION I dedicate this dissertation with gratitude to my parents, Ethel and George, both scholars, and especially to Mom, who showed me how to be a life-long learner; to my children, Devin and Katharine, may you be life-long learners too; to my sister, Josie, with thanks for all the flowers that brightened my journey.

PAGE 7

ACKNOWLEDGMENTS This research was supported in part through a grant from the Craig R. Janes Fund for Graduate Research at the University of Colorado Denver. I wish to acknowledge Dr. Susan Niermeyer, who took a chance on mentoring me; Dr. John Brett, who convinced Susan to take that chance; Dr. Jean Scandlyn whose every conversation opened up a new thought; Dr. Richard Miech, who encouraged me ''to fly the plane;" Dr. Allison Sabei-Soteres for her statistics advice, and Dr. Lorna Moore, who encouraged me to pursue a doctorate before it was even on my radar screen. To my cohort go many thanks for their support and friendship, especially Maria de Jesus Diaz-Perez who made all things possible at Salud Family Health Centers. iMil gracias to Salud and the women who told me their stories! And finally, heartfelt thanks to Dr. Susan Dreisbach who employed me during this journey and to Brenda Beaty, who helped fill in the holes in my knowledge of SAS.

PAGE 8

TABLE OF CONTENTS FIGURES ................................................................................................ xiii TABLES ................................................................................................... xv CHAPTER 1. INTRODUCTION .................................................................................... 1 Epidemiological Paradoxes ................................................................. 1 Birth Outcomes ................................................................................ 2 Low Birth Weight. .................................................................. 3 Preterm Birth ........................................................................ 5 Small for Gestational Age ........................................................ 6 Large for Gestational Age ........................................................ 6 Fetal Programming ................................................................ ? Population Characteristics of Colorado .................................................. 8 Why Study Paradoxical Health Outcomes? ..................................................... 9 Research Design and Specific Aims ................................................... 11 Overview of Research Methods ......................................................... 11 Description of the Study Population .................................................... 13 Organization of the Dissertation ......................................................... 13 2. EPIDEMIOLOGICAL PARADOXES ........................................................... 14 The Social Gradient of Health ............................................................ 14 The Social Gradient of Health at the Individual Level. .................. 15 The Social Gradient of Health at the Area Level. ........................ 15 Unexpected Deviances from the Social Gradient of HealthThe Paradox .... 17 viii

PAGE 9

Traditional Explanations of the Hispanic Immigrant Paradox .................... 19 Healthy Migrant Hypothesis .................................................... 20 Healthy Immigrant Hypothesis ................................................ 21 Acculturation/Assimilation ...................................................... 22 Social Support ..................................................................... 24 An Alternative TheoryPolitical Economy ........................................... 25 3. RESEARCH DESIGN AND METHODS ...................................................... 29 Nested Mixed Method Design ............................................................ 30 Quantitative Research Component. .................................................... 31 Levels of Analysis ................................................................. 32 Sample, Study Data, and Variables .......................................... 33 Dependent Variables ............................................................. 34 Independent Variables for Aims 1 and 2 .................................... 35 Demographic and Socioeconomic Risk Factors ................ 36 Medical Risk Factors ................................................... 38 Behavioral Risk Factors ............................................... 39 Race/Ethn icity/Nativity ................................................ .40 Missing Data and Size of Study Population for Aim 1 ................... 41 Missing Data and Size of Study Population for Aim 2 ................... 44 Contextual Variables for Aim 3 ................................................ 47 Missing Data and Size of Population for Aim 3 .......................... .49 Methods for Aims 1 and 2 ................................................................. 50 Model Building for Aims 1 and 2 ............................................ 51 Model Building for Aim 3 ...................................................... 60 Qualitative Research Component.. ..................................................... 61 ix

PAGE 10

Sample and Study Data ......................................................... 61 Methods for Aim 4 ........................................................................... 63 4. QUANTITATIVE ANALYSIS .................................................................... 65 Aim 1 ........................................................................................... 65 Comparison of Risk Factors by Race/Ethnicity ............................. 65 Frequency of Adverse Birth Outcomes by Race/Ethnicity .............. 68 Odds Ratios of Birth Outcomes by Race/Ethnicity ....................... 69 Low Birth Weight. ...................................................... 70 Preterm Birth ........................................................... 73 Small for Gestational Age ........................................... 77 Large for Gestational Age ............................................ 81 Discussion of Aim 1 ............................................................ 84 Aim 2 ........................................................................................... 87 Comparison of Risk Factors of Mothers of Mexican Origin by Nativity ........ 87 Frequency of Adverse Birth outcomes by Nativity ........................ 90 Odds Ratios of Birth Outcomes by Nativity ................................ 91 Low Birth Weight. ...................................................... 91 Preterm Birth ............................................................ 94 Small for Gestational Age ............................................ 98 Large for Gestational Age .......................................... 1 00 Discussion of Aim 2 ............................................................ 1 05 Aim 3 .......................................................................................... 110 Influence of Neighborhood Deprivation and Immigrant Orientation on Outcomes .................................................................................... 111 Discussion of Quantitative Analysis ................................................... 114 X

PAGE 11

5. QUALITATIVE RESULTS ....................................................................... 117 Recent Mothers ............................................................................. 117 Key Inform ants ............................................................................. 11 8 Diet and Exercise During Pregnancy ...................................... 118 Diet. ..................................................................... 118 Exercise/Energy Expenditure ...................................... 121 Maternal Weight and Weight Gain .......................................... 122 Body lmage ............................................................ 123 Smoking and Drinking .......................................................... 124 Other Cultural Beliefs about Pregnancy ................................... 125 Sources of Social Support .................................................... 126 Political Economy and Birth Outcomes .............................................. 127 Discussion of Qualitative interviews ....................................... 130 6. DISCUSSION .................................................................................... 133 Limitations ................................................................................... 134 Is There an Epidemiological Paradox in Weight-Related Birth Outcomes? ...................................................................................... 134 Do the Hypotheses in the Literature Explain the Paradox? ......................... 137 How Should Health be Measured? ................................................. 137 Healthy Migrant Hypothesis .................................................. 138 Healthy Immigrant Hypothesis ............................................... 139 Neighborhood Effects .......................................................... 140 Political Economy-A Broader Perspective ......................................... 140 Significance .................................................................................. 141 xi

PAGE 12

APPENDIX ....................................................................................................... 143 A. Summary of Selected Population Studies ................................................ 144 B. Human Subjects Approvals ................................................................... 146 C. Solicitation Guide, Interview Guides, Consents ......................................... 152 BIBLIOGRAPHY ................................................................................................. 171 xii

PAGE 13

LIST OF FIGURES Figure 2.1 PATHWAYS TO OBESITY AND DIABETES ....................................................... 26 3.1 SCHEMATIC OF RESEARCH DESIGN ............................................................. 31 3.2 AIM 1 LBW DRIFT OF ODDS ........................................................................... 44 3.3 AIM 1 PRETERM DRIFT OF ODDS ................................................................. .44 3.4 AIM 1 SGA DRIFT OF ODDS ........................................................................... 44 3.5 AIM 1 LGA DRIFT OF ODDS ........................................................................... 44 3.6 AIM 2 LBW DRIFT OF ODDS ........................................................................... 46 3.7 AIM 2 PRETERM DRIFT OF ODDS ................................................................... 46 3.8 AIM 2 SGA DRIFT OF ODDS ........................................................................... 47 3.9 AIM 2 LGA DRIFT OF ODDS ........................................................................... 47 3.10 SCHEMATIC OF MODEL BUILDING FOR AIMS 1 AND 2 ..................................... 53 4.1 DISTRIBUTION OF RISK FACTORS BY RACEIETHNICITY .................................. 67 4.2 UNADJUSTED FREQUENCIES OF BIRTH OUTCOMES BY RACEIETHNICITY ....... 69 4.3 LBW ODDS RATIOS BY RACEIETHNICITY AND MODEL. ................................... 70 4.4 PRETERM BIRTH ODDS RATIOS BY RACEIETHNICITY AND MODEL. ................. 74 4.5 SGA ODDS RATIOS BY RACE/ETHNIC lTV AND MODEL. .................................... 78 4.6 LGA ODDS RATIOS BY RACEIETHNICITY AND MODEL. .................................... 82 4.7 FULLY ADJUSTED ODDS RATIOS BY RACE/ETHNICITY ................................... 85 4.8 DISTRIBUTION OF RISKS BY NATIVITY .......................................................... 89 4.9 UNADJUSTED FREQUENCIES BY NATIVITY ................................................... 91 4.10 LBW ODDS RATIOS BY NATIVITY AND MODEL.. .............................................. 92 xiii

PAGE 14

4.11 PRETERM BIRTH ODDS RATIOS BY NATIVITY AND MODEL. ............................. 95 4.12 SGA ODDS RATIOS BY NATIVITY AND MODEL. ............................................... 98 4.13 LGA ODDS RATIOS BY NATIVITY AND MODEL. .............................................. 101 4.14 FREQUENCY OF LGA BY NATIVITY BY YEAR ................................................. 104 4.15 FREQUENCY OF LGA BY WEIGHT GAIN ....................................................... 104 4.16 FREQUENCY OF LGA BY AGE ..................................................................... 104 4.17 FREQUENCY OF LGA BY PARITY ................................................................. 105 4.18 FREQUENCY OF LGA-SPECIFIC RISKS ........................................................ 1 05 4.19 FULLY ADJUSTED ODDS RATIOS BY NATIVITY ............................................. 1 06 4.20 ODDS OF OUTCOMES FOR MEXICAN -BORN MOTHERS IN ADAMS AND DENVER COUNTIES AND STATEWIDE ........................................................... 110 6.1 BIRTH WEIGHT IN GRAMS OF MEXICAN-BORN MOTHERS (ABOVE) WITH U.S.-BORN MOTHERS ........................................................................ 137 xiv

PAGE 15

LIST OF TABLES Table 1.1 Principal individual risk factors for LBW ................................................................ 4 1.2 LBW and preterm birth rates by race/ethnicity in U.S. and Colorado ............................ 5 1.3 Population of Mexican origin as% of Hispanic population in Colorado-2000 .............. 8 1.4 Comparison of socioeconomic position of Hispanic, Mexican, and non-Hispanic White populationsU.S. Census 2000 and 2002 ................................ 9 1.5 Study population by race/ethnicity of mother-Aim 1 ............................................ 13 1.6 Study population by nativity of mothers of Mexican origin-Aim 2 ............................ 13 1.7 Study population of mothers of Mexican origin in Denver and Adams Counties-Aim 3 ............................................................................................................................... 13 2.1 Measures of socioeconomic position .................................................................. 17 3.1 Individual-level variables for Aims 1 and 2 ........................................................... 35 3.2 Original dataset by race/ethnicity ....................................................................... 42 3.3 Number and percentage of missing cases by variable for Aim 1 ............................... 42 3.4 Number and percentage of missing variables by race/ethnicity ............................... .43 3.5 Final study population by race/ethnicity of mother-Aim 1 ...................................... 44 3.6 Original dataset by place of nativity of mother ...................................................... 45 3.7 Number and percentage of missing cases by variable for Aim 2 ............................... 45 3.8 Number and percentage of missing variables by nativity ......................................... 46 3.9 Final study population by nativity of mothers of Mexican origin by year-Aim 2 .......... 47 3.10 Values of contextual scales for Adams and Denver Counties2000 ......................... 49 3.11 Population of mothers of Mexican origin by nativity 2000-2005 in Adams and Denver Counties ..................................................................................... 50 4.1 Percent frequency distribution of risk factors by race/ethnicity 2000 2005 ................. 66 XV

PAGE 16

4.2 Percent frequency of LBW, preterm birth, SGA, and LGA by race/ethnicity 2000-2005 ................................................................................................. 68 4.3 Unadjusted and adjusted odds ratios (95% Cl) of LBW and race/ethnicity .................. 70 4.4 Estimated coefficients and odds ratios for LBW by race/ethnicity ............................. 71 4.5 Unadjusted and adjusted odds ratios of preterm birth by race/ethnicity .................... 74 4.6 Estimated coefficients and odds ratios for preterm birth and race/ethnicity .................. 75 4.7 Unadjusted and adjusted odds ratios (95% Cl) of SGA by race/ethnicity .................... 78 4.8 Estimated coefficients and odds ratios for SGA and race/ethnicity ............................ 79 4.9 Unadjusted and adjusted odds ratios (95% Cl) of LGA by race/ethnicity .................... 82 4.10 Estimated coefficients and odds ratios for LGA and race/ethnicity ............................ 83 4.11 Comparison of fully adjusted odds ratios of birth outcomes by race/ethnicity .............. 85 4.12 Comparison of contributing risk factors by birth outcome and race/ethnicity ................ 86 4.13 Comparison of fully adjusted odds ratios of AGA by race/ethnicity ............................ 87 4.14 Percent frequency distribution of risk factors of mothers of Mexican origin by nativity 2000 2005 ................................................................................... 88 4.15 Percent frequency of LBW, preterm birth, SGA, and LGA by nativity 2000-2005 .................................................................................................. 90 4.16 Unadjusted and adjusted odds ratios (95% Cl) of LBW by nativity ............................ 92 4.17 Estimated coefficients and odds ratios for LBW and nativity .................................... 93 4.18 Unadjusted and adjusted odds ratios (95% Cl) of preterm birth by nativity ................. 95 4.19 Estimated coefficients and odds ratios for preterm birth and nativity ......................... 96 4.20 Unadjusted and adjusted odds ratios (95% Cl) of SGA by nativity ............................ 98 4.21 Estimated coefficients and odds ratios for SGA and nativity .................................... 99 4.22 Unadjusted and adjusted odds ratios (95% Cl) of LGA by nativity ........................... 1 00 4.23 Estimated coefficients and odds ratios for LGA and nativity ................................... 1 01 4.24 Model 4 adjusted odds ratios (95% Cl) of LGA by nativity ..................................... 1 03 xvi

PAGE 17

4.25 Fully adjusted odds ratios of birth outcomes by nativity ........................................ 1 06 4.26 Comparison of fully adjusted odds ratios of AGA by nativity .................................. 1 07 4.27 Fully adjusted odds ratios (95% Cl) by nativity and model ..................................... 108 4.28 Fully adjusted odds ratios (95% Cl) by nativity for LGA using LGA-specific medical risks ............................................................................................... 108 4.29 Comparison of fully adjusted odds ratios of birth outcomes of Mexican-born mothers in Adams and Denver Counties and statewide ........................................ 110 4.30 Outcomes in Adams County ............................................................................ 111 4.31 Estimated coefficients for neighborhood deprivation and immigrant orientation in Adams County ......................................................................................... 112 4.32 Outcomes in Denver County ........................................................................... 113 4.33 Estimated coefficients for neighborhood deprivation and immigrant orientation in Denver County ......................................................................................... 113 5.1 Selected characteristics of women interviewed ................................................... 118 5.2 Weight gain of mothers interviewed and birth weight of infants .............................. 122 6.1 Comparison of fully adjusted odds ratios by race/ethnicity ..................................... 134 6.2 Comparison of fully adjusted odds ratios by nativity ............................................. 135 xvii

PAGE 18

CHAPTER 1 INTRODUCTION "Only the paradox comes anywhere near to comprehending the fullness of life."1 Epidemiological Paradoxes Since at least the Middle Ages, observers have noted a general social gradient of health that associates health outcomes with certain compositional aspects of populations. Better outcomes are associated with an individual's increasing income/wealth, social status, and education; worse health outcomes are associated with lower income, status, and education (Lynch & Kaplan 2000:13-35). In public health terms, these differences are usually called health disparities, and much effort is expended to understand the causes of poorer health and to try to reduce those disparities (Berkman & Kawachi 2000). The expected social gradient is not invariant, however. For example, it does not appear to hold for all diseases or conditions, for all subgroups within a racial or ethnic population, or for some foreign-born individuals who immigrate to the Unites States (Gorman 1999; Singh & Yu 1996). In addition to a social gradient related to the compositional aspects of populations, public health researchers have identified a similar social gradient of health at the area, or contextual level, where better health is often associated with wealthier, better-endowed neighborhoods (Diez-Roux 1998). As is the case with compositional characteristics of a population, research on neighborhood contextual factors affecting health is usually focused on poorer neighborhoods and their correlation with poorer health outcomes (Reagan & Salsberry 2005; Pearl eta/. 2001; Gorman 1999; O'Campo eta/. 1997). One exception is the recent work by Finch et a/. (2007), who associated the predicted probability of low birth weight among the offspring of native and foreign-born Hispanics in Los Angeles County, California with measures of neighborhood disadvantage and "immigrant-orientation" and found a protective effect of immigrant orientation at the neighborhood level. 1 Although this quotation from Carl Jung refers to the richness of paradoxes in religion, this and his further observation that non-ambiguity and lack of contradiction are one-sided, can apply equally, if more prosaically, to understanding unexpected occurrences in population health (Jung 1980: 18). 1

PAGE 19

Studies from the 1970s and 1980s identified a "paradox" from the general social gradient of health among Hispanics as a population; that is, despite socioeconomic and demographic risk profiles that are less advantageous than those of the majority reference population, Hispanics statistically showed a number of unexpectedly better health outcomes (Markides & Coreil1986; Teller & Clyburn 1974). Thereafter, researchers noted that the lack of health disparity appeared even more pronounced for certain birth outcomes, usually low birth weight and premature birth, among foreign-born immigrants, whether Hispanic or other raciaVethnic groups such as immigrants from Africa and India. These immigrant populations typically come to the U.S. with little wealth, low education, and low social status, and they usually live in poorer neighborhoods (Hummer eta/. 1999; Singh & Yu 1996), thus putting them at the low end of the social gradient. Birth Outcomes Countries adopt various measures of population health, including life expectancy, measures of the burden of disease on the functioning of the population, and birth outcomes. Birth outcomes are considered signal measures of health internationally (United Nations 2008; UNICEF 2004) and in the U.S. (USDHHS 2000) because they represent the future health of the population. Although low birth weight (LBW) and preterm birth have long been identified as birth outcomes of interest, small for gestational age status (SGA) is increasingly recognized as an adverse birth outcome (Lee eta/. 2003). In addition, large for gestational age status (LGA) is gaining recognition as an adverse birth outcome that should be better studied (Dyer eta/. 2007; Dollberg eta/. 2000). Weight-related birth outcomes, specifically low birth weight, preterm birth, small for gestational age, and large for gestational age, are important indicators of well-being, both with respect to the infant's immediate health and over the life course (Oken & Gillman 2003; Barker 2002; Eriksson et a/. 2001; Basch 1999:79-81 ). All four measures are key indicators of health; understanding their patterns in different populations can help guide public health policies and allocation of resources. Rates of LBW, preterm birth, SGA, and LGA vary among countries and among population groups within countries (Frisbie & Song 2003; Gould eta/. 2003; Hummer eta/. 1999; Singh & Yu 1996). In its initiative Healthy People 2010, the U.S. Department of Health and Human Services has adopted aggressive objectives to reduce rates of LBW and preterm birth, both in the aggregate and with regard to disparities in their rates across race and ethnicity by 201 0 (2000). Although the highest risk of morbidity and 2

PAGE 20

mortality is associated with low-weight births, higher morbidity and mortality occur among LGA infants compared with those who are appropriate for gestational age (AGA) (Lubchenco et a/. 1972; Lubchenco & Bard 1971 ). Moreover, considering AGA and LGA births together may obscure significant outcomes, especially when the rates of LGA births are rising (Barbour, L., personal interview, November 10, 2008). Low Birth Weight Low birth weight is defined as an infant weighing less than 2500 grams2 at birth (approximately 5.5 pounds) (WHO 2006:36). LBW is an important measure of public health because LBW infants are almost 40 times more likely to die during their first month of life than infants weighing more than 2500 grams at birth (Collins & Schulte 2003:223; 10M 1985). In addition, LBW infants comprise two-thirds of deaths in the first 28 days of life and 20% of deaths from 28 days until one year of age (10M 1985). Moreover, LBW infants who survive have elevated risks of morbidity from conditions such as cardiovascular disease, diabetes, and obesity during their life course (Barker 2002; Merson eta/. 2001: 122). LBW results proximately from premature birth (<37 completed weeks of gestation), poor weight gain of the fetus, or both. Although poor weight gain and preterm birth often occur together, they have independent and multiple causes (March of Dimes 2007; 10M 1985). Because fundamental (or more distal) causes of LBW are multiple and poorly understood, most medical research into LBW has focused on identifying and addressing risk factors that are associated with the increased probability of delivering an LBW infant. Table 1.1, adapted from the Institute of Medicine (1985:7), lists the principal individual level risk factors for LBW. These factors sort into demographic and socioeconomic factors, medical risks, and environmental/behavioral factors. 2 Interestingly, some experts define low birth weight as 2500 grams or less (Merson et a/. 2001 :213; 10M 1985), while others use less than 2500 grams (March of Dimes 2007). This study follows the definition used by the WHO (2006) and the Colorado Department of Public Health and Environment (2000a), each of which defines LBW as less than 2500 grams at birth. 3

PAGE 21

Table 1.1. Principal Individual risk factors for LBW Demographic/Socioeconomic Age (<17; >34) Race (Black) Low socioeconomic status Medical Risks Genitourinary anomalies/surgery Diabetes Parity (0 or >4) Low weight for height Chronic hypertension Non-immune for selected infections Previous LBW infant Low birth weight of mother Multiple parity Poor weight gain of mother Short inter-pregnancy interval Behavioral/Environmental Maternal smoking Poor nutritional status of mother Alcohol or other substance abuse Unmarried Low level of education Hypotension Hypertension/preeclampsia/eclampsia Infections Placenta previa/abruptio placenta Hyperemesis gravidarum Oligohydramnios/polyhydramnios Anemia Isoimmunization (Rh incompatibility) Fetal anomalies Incompetent cervix Iatrogenic preterm birth Toxic exposures High altitude/hypoxia In 2000, the rate3 of LBW in the United States overall was 7.6%, and the disparity among races and ethnicities was large; the rate among non-Hispanic Blacks was 13.1 %, among non-Hispanic Whites it was 6.6%, and among Hispanics of all races it was 6.4% (CDC 2002: Table 44). All of Colorado's rates exceeded the national rates by race/ethnicity. Colorado's statewide LBW rate was 8.4% in 2000, while the rate among non-Hispanic Blacks was 15.0%, among non-Hispanic Whites it was 8.0%, and among Hispanics of all races the rate was 8.1% (CDC 2002: Table 46). According to the Colorado Department of Public Health & Environment (CDPHE), the major contributing factors to LBW in the state during 1995-1997 were multiple births (1 in 5 LBW in Colorado was a multiple birth during that period); inadequate maternal weight gain; smoking; and premature rupture of membranes (CDPHE 2000a). The state identified inadequate maternal weight gain as the largest contributor to Colorado's LBW rate. In response to this finding, the state adopted a marketing campaign entitled "A Healthy Baby is Worth the Weighf' (CDPHE 2005b). One objective of Healthy People 2010 is to reduce LBW in the U.S. to 5.0% and to eliminate disparities by race and ethnicity (USDHHS 2000). In this regard, the National 3 The use of "rate" refers to the frequency or percentage of all live births rather than a rate per 1000 births, the usual epidemiological use of "rate." 4

PAGE 22

Institutes of Health has called for research to better understand mechanisms that underlie racial and ethnic disparities in LBW. Preterm Birth The Institute of Medicine's recent report on preterm birth considers a birth to be premature if it occurs before 37 completed weeks of gestation (10M 2006). Colorado uses the same definition (CDPHE 2000d). The Institute of Medicine describes preterm birth as a "cluster of problems with a set of overlapping factors of influence" associated with risks that include individual-level factors, neighborhood characteristics, medical conditions, genetics and environmental exposures (2006:1). As with LBW, the causes of preterm birth are complex, and rates vary by race and ethnicity. The costs of preterm birth include increased infant mortality and morbidity, such as ophthalmic, neurologic, and psychomotor development difficulties, as well as deficits in pulmonary health over the life course (Kramer 2003). In 2000, the overall U.S. preterm birth rate was 11.6%; the rate for non-Hispanic Blacks was 17.4%, for non-Hispanic Whites it was 1 0.4%, and for Hispanics of all races the rate was 11.2% (CDC 2002a:Table 44). Rates of preterm birth in the U.S. increased in 2004 by 0.5% to 1.1 %, depending on the race/ethnicity category. In contrast to Colorado's relatively high LBW rate, its rate of preterm birth is lower than the national average. In 2000, the rate of preterm births for all races in Colorado was 9.0%, while the U.S. average was 11.6% (CDPHE 2000b). The rate of preterm births is rising, however, with the U.S. average for all races/ethnicities reaching 12.5% in 2004 (10M 2006), and the Colorado average reaching 9.8% (CDC 2006b; CDPHE 2004). Healthy People 2010 has set a target to reduce preterm birth to 7.6% for all races and ethnicities by 2010 (USDHHS 2000). Table 1.2. LBW and preterm birth rates by racafethnlclty In U.S. and Colorado RacaiEthnlclty Low Birth Weight % Preterm Birth % 2000 2004 2000 2004 us CO' co us co CO" All races/ethnicities 7.6 8.4 8.1 9.0 11.6 9.0 12.5 9.8 White (non Hispanic) 6.6 8.0 7.2 8.7 10.4 8.8 11.5 9.9 Black (non-Hispanic) 13.1 15.0 13.7 14.5 17.4 13.5 17.9 13.6 Hispanic (all races) 6.4 8.1 6.8 8.6 11.2 8.7 12.0 9.2 1. CDC 2002a: Table 44. 2. CDC 2002a: Table 46. 3. CDC 2006b. 4. CDPHE 2004b. 5. CDPHE 2000b. 6. CDPHE 2004a. 5

PAGE 23

Small for Gestational Age Small for gestational age is generally defined as birth weight of an infant below the 1 Olh percentile of weight for gestational age (Kiiegman & Das 2002), although some researchers use the 151, 3rd, or 51 h percentile (Gould eta/. 2003). SGA measures size in relation to gestational age to adjust for lower weight due to preterm birth and is therefore a more targeted measure of intrauterine growth. Its accuracy depends on the clinical estimate of gestation or gestation reported by the mother, neither of which is as accurate as birth weight. Some studies calculate SGA from data on birth weight and gestation in the birth record; more studies use LBW because it requires only one measure. Chung et a/. (2003) argue that there are baseline differences in birth weight by gestational age among different races/ethnicities. Their work suggests that ethnic differences in weight by gestation may contribute to the differences in SGA rates. There are no nationally-adopted tables of the distribution of birth weight by gestation. For the reasons described in Chapter 3 at pages 3435, this study uses the tables created by Alexander et a/., which report weight by weeks of gestation by race and ethnicity (1999). Factors associated with SGA include underlying medical risks of the type reported in the birth record (hypertension, renal disease, diabetes, chronic pulmonary disease, poor placentation), low pregnancy weight, smoking, very young maternal age, older maternal age, and first births (Lee et a/. 2003). SGA has been associated with the development of non insulin-dependent diabetes mellitus (Barker 1998), hypertension (Barker 2002), and coronary heart disease (Barker 2001) later in life. Large for Gestational Age LGA is defined as birth weight of an infant above the 901 h percentile of weight for gestational age (Alexander et a/. 1999). LGA results in babies who are not only large, but whose excess weight is found in adipose issue around the trunk of the baby. Some babies are too large to fit through the birth canal, leading to shoulder dystocia and increased incidence of delivery by c-section (Casey et a/. 1997; Modanlou et a/. 1982). An LGA infant often has reduced ability to regulate glucose immediately after birth and higher risk of developing metabolic syndrome and type II diabetes during the life course (Dyer et a/. 2007; Hediger eta/. 1998). Maternal diabetes, gestational diabetes, high weight gain during pregnancy, and high maternal BMI before pregnancy are associated with increased odds of delivering an LGA 6

PAGE 24

baby. In addition, mothers who were themselves born LGA are more likely to give birth to an LGA infant (Ahlsson eta/. 2007; Ehrenberg eta/. 2004). It is estimated that 7.7% of Mexican Americans had diagnosed diabetes and 5.4% had undiagnosed diabetes in 2000 (Martorell 2005; see also Mainous et a/. 2008). The rate of overweight and obesity is extremely high in Mexico 62% of adults in 2000 and the rate of diabetes among the Mexican population continues to increase (Jimenez-Cruz & Barcardi-Gascon 2004). While genetics most likely plays a part in obesity, environmental influences also play a role, such as lower levels of physical activity, lower total daily energy expenditure, and increasing availability of high-fat, energy-dense foods and larger portion sizes (Hill eta/. 2000). Moreover, people are products of their life histories, which in turn are products of the regional and global economic environments in which they live (di Leonardo 1984:12). Fetal Programming Studies of health over the life course of babies born small (LBW, some preterm births, and SGA) have suggested a theory of ''thrifty phenotype" or ''fetal programming" to explain why infants born small have higher odds of obesity and chronic conditions associated with obesity, such as type II diabetes and cardiovascular disease, than babies born appropriate for gestational age (Barker 1998, 2001, 2002). In brief, the health of an infant is partly programmed by the environment in utero. Some LBW, preterm, and SGA infants experience a fetal environment in which the available nutrition is limited perhaps by poor perfusion of the placenta, undernutrition in the mother, maternal smoking, or other causes of inadequate nutrition reaching the growing fetus. Scarcity causes the fetus to take in as much nutrition as it can obtain and programs its phenotype to be ''thrifty" and "save" it (Breier eta/. 2001; Hales & Barker 2001 ). When the infant is born, if the environment is one of abundance, the programming nonetheless remains in place and raises the odds that the infant, and later adult, will continue to gather nutrients as if still in a nutritionally-deprived environment. Catch-up growth often results in overweight and obesity in adolescence and adulthood. Fetal programming may also account for adverse health conditions over the life course for LGA babies. Mothers of LGA babies often supply excess glucose to their fetuses, either because the mothers have habitually high and continuous carbohydrate intake or because they have metabolic syndrome or gestational diabetes. At birth an LGA infant has higher odds of reduced ability to regulate glucose within the first few days after birth (Dyer et a/. 2007). In response to glucose challenge, even asymptomatic infants display early 7

PAGE 25

markers of insulin resistance. It is believed by a number of researchers that a fetal environment that includes excess glucose programs the fetus with a predisposition for insulin resistance, which later in life may manifest in obesity, type II diabetes, and cardiovascular disease. Women who are born LGA have higher odds of delivering LGA infants of their own, suggesting that fetal programming can persist from one generation to the next (Drake & Walker 2004; Barbour, L., personal interview, November 10, 2008). Population Characteristics of Colorado Based on the U.S. Census, the percentage of "Hispanics or Latinos'.4 in Colorado has risen from 17.1% in 2000 to 19.5% in 2005 (2007a). Hispanics represent one of the fastest growing ethnic groups in the U.S.; by 2050, Hispanics are expected to account for 22.5% of the U.S. population (U.S. Census 2001 ). Notwithstanding current U.S. policies intended to reduce immigration from Mexico, the Hispanic population would double by 2050 even if immigration stopped (Smith & Edmonston 1997). Table 1.3 presents selected information about the Hispanic population in Colorado (U.S. Census 2000d). Individuals of Mexican origin are the second largest raciaVethnic group in Colorado, making up at least 1 0.5% of the state's population (Census 2000c). Persons of Mexican origin are more likely to live in poverty, not speak English at home, and be foreign-born than the Hispanic/Latina population as a whole. Similarly, Coloradans of Mexican origin are less likely to have completed high school or college than the Hispanic/Latina population in the state. Table 1.3. Population of Mexican origin as % of Hispanic population In Colorado 2000 Characteristic Population Total Female Population Average Family Size High School Graduate Bachelor Degree + Foreign Born Non-English at home Families < poverty Individuals< poverty Hispanic 735,601 350,795 4 218,902 39,335 201,072 366,528 25,903 135,421 Mexican 450,760 205,156 4 113,129 19,478 170,356 259,294 16,613 90,575 61.28% 58.48% 51.68% 49.52% 84.72% 70.74% 64.89% 66.68% 4 The federal government classifies those of Spanish, Central American, and Latin American origin as "Hispanic or Latino." Many in this group prefer one ethnic label over another. For this study, persons in this category as referred to generally as "Hispanic." Hispanics of Mexican origin are referred to as Mexican-born or U.S.-born of Mexican origin. 8

PAGE 26

National estimates by the U.S. Bureau of the Census, as reported in Table 1.4, confirm that Hispanics, and Mexicans as a particular subpopulation, have worse economic and socioeconomic position profiles than non-Hispanic Whites in 2000. Based on Colorado's Pregnancy Risk Assessment Monitoring System (PRAMS) data from 1998 2002, Hispanic women in Colorado were less likely to receive adequate prenatal care than their counterparts, a result that is influenced primarily by the inability of Hispanics who are not born in the U.S. and are undocumented to obtain prenatal care on the recommended schedule (CDPHE 2005b). In most categories, Mexicans fare worse than Hispanics as a whole (U.S. Census 2000d). In 2000, 31.7% of Denver County's population identified as Hispanic/Latina and 28.2% of Adams County identified as Hispanic/Latina (U.S. Census 2000b). CDPHE reports that the proportion of linguistically isolated Spanish-speaking households increased by 171% statewide between 1990 and 2000, based on census data. The proportion of linguistically isolated households in Adams County increased by 416% and by 158% in Denver County during this same time period (CDPHE 2005a). Table 1.4. Comparison of socioeconomic position of Hispanic, Mexican, and non Hispanic White populationsU.S. Census 2000 Measure Mexican HI 1 Non-Hispanic span c White 2000 2000 2000 High school graduate 51.0% 57.0% 88.4% Unemployment 7.0% 6.8% 3.4% Employment in service occupations NIA 19.4% 11.8% ProfessionaVmanagement employment 11.9% 14.0% 33.2% Income of $35,000 or more 20.6% 23.3% 49.3% Living below poverty level 24.1% 22.8% 7.7% By most measures, people of Mexican origin in the U.S. and Colorado are at the low end of the spectrum for education, employment, and income, and at the high end for being classified as poor. Why Study Paradoxical Health Outcomes? It is important to study paradoxical birth outcomes, first, to learn if they exist in Colorado. Previous studies have found some paradoxical results in California, New York, and national samples (Rosenberg et a/. 2005; Zambrana et a/. 1997; Singh & Yu 1996). Second, it is important to "unpack" the apparent epidemiological paradox within the general grouping of "Hispanics" because of the heterogeneity of subpopulations (Hummer eta/. 1999; Frisbie et a/. 1998; Singh & Yu 1996) and to disentangle the effect of nativity (Gould et a/. 9

PAGE 27

2003; Singh & Yu 1996). Evidence that the paradox is short-lived points to a need to improve outcomes for all populations. Third, it is beneficial to examine "non-small" infants for evidence of previously ignored adverse outcomes. Finally population studies alone are insufficient to answer complex questions about outcomes and their underlying causes. Several hypotheses have been proposed to explain the unexpected deviance from the general social gradient of health among Hispanic infants, all of which suggest that the incidence of paradoxically good birth outcomes among infants of foreign-born immigrants improves the rates for Hispanics as a whole. Three of the most common are {1) the healthy migrant hypothesis [people who choose to migrate to the U.S are healthier than those who stay in their native countries] (Cho eta/. 2004; Marmot eta/. 1984), (2) the healthy immigrant hypothesis [immigrants engage in healthier behaviors, such as better diet, less smoking, less alcohol consumption, at least immediately after they immigrate to the U.S.] (Flores & Brotanek 2005; Kasirye et a/. 2005; Abrafdo et a/. 1999; Guendelman et a/. 1999), and (3) a social support hypothesis [immigrants enjoy increased social support, social networking, and/or social capital, probably founded in retained cultural beliefs, which compensate for less education, income, and access to health care] (Landsale et a/. 1999; Rumbaut & Weeks 1996). Other researchers argue that examining the degree of assimilation or acculturation explains differing health outcomes among immigrants. Finch eta/. suggest that segmented assimilation theory explains divergent health trajectories among immigrant women and that immigrants who live in ethnic enclaves may engage in healthy behaviors brought from their native countries or continue to live in accordance with cultural practices or values, which may be protective (2007). Finally, some researchers argue that there is no paradox; rather it is merely a statistical artifact (Palloni & Arias 2004; but see Hummer eta/. 2007). Amaro and de Ia Torre (2002) warn that the assumption of some public health agencies that Hispanic mothers are not at risk for poor birth outcomes because of the "paradox" may not in fact be accurate for all subgroups or all conditions. In addition, if Mexican-born immigrants in Colorado are shown to enjoy unexpectedly better birth outcomes than the social gradient would predict, and if their advantage decays over time as shown by some, it would be beneficial to generate initiatives to promote conditions that might extend the longevity of those better outcomes (Ricketts eta/. 2005; McGlade eta/. 2004). If there is no paradox, then it will be important to direct attention to the poor outcomes. 10

PAGE 28

Research Design and Specific Aims Originally this study sought to determine whether an epidemiological paradox exists for each of four birth outcomes: LBW, preterm birth, SGA, and infant mortality, among Hispanics, primarily Mexican-born, in Colorado during the years 2000 2005 at both the individual and neighborhood level. During the analysis of data, when it became clear that the paradox does exist for LBW, preterm birth, and SGA among Colorado's Hispanic population and even more clearly among Mexican-born mothers, curiosity about large for gestational age infants led to the addition of LGA as a birth outcome. The data on infant mortality proved less reliable than was hoped and the rates of infant mortality were very low resulting in small measurable variances among the population groups of interest. Accordingly, infant mortality was dropped. This multi-level, nested, mixed method study has four aims: to test the expected social gradient of health on birth outcomes by racelethnicity and to identify any paradoxical outcomes (Aim 1 ); to test the expected social gradient of health on birth outcomes by nativity of Hispanic mothers of Mexican origin and to explore the healthy migrant and healthy immigrant hypotheses as explanations for any paradoxically better birth outcomes experienced by Mexican-born mothers (Aim 2); to examine the association of neighborhood deprivation and immigrant orientation on birth outcomes of mothers of Mexican origin in two large Colorado counties (Aim 3); and to explore potential reasons for differential outcomes among Mexican-born mothers and U.S.-born mothers of Mexican origin (Aim 4). Overview of Research Methods Most literature on the subject of the paradox is either quantitative or qualitative. The quantitative studies that have identified a paradox in health outcomes tend to explain their results by referencing suggested explanations from the qualitative literature with little contextual analysis. The qualitative literature tends to report small contextual studies that do not include quantitative observations of outcomes. Use of both approaches, in the form of a mixed-method study, provides a fuller interpretation and understanding of the social epidemiology of birth outcomes through corroboration of findings using multiple methods and enhancement and clarification of results (Greene et a/. 1989). In addition, existing birth 11

PAGE 29

outcome studies focus on the "lower weighf' end of the spectrum of outcomes. No study was found that included birth outcomes at both the lowand the high-weight ends of the birth weight spectrum. Thus, the use of mixed methods and the selection of birth outcomes relevant to the current clinical reality among the largest ethnic population in Colorado permit more nuanced conclusions and reveal potential connections that have been unintentionally masked by less comprehensive designs. The quantitative component of the study is a retrospective cohort design, intended to determine the existence of any epidemiological paradoxes and to explore the validity of the healthy migrant and healthy immigrant hypotheses and the contribution of neighborhood deprivation and immigrant orientation to outcomes. The qualitative portion consists of semi structured interviews with key informants and recent Mexican-born mothers and U.S.-born mothers of Mexican origin to explore the results of the quantitative analysis and to identify potential explanations for the outcomes. Based on the expected social gradient of health and census data, the ranking of social and health profiles should be: Mexican-born mothers poorest, U.S.-born mothers of Mexican descent-poor but not as poor as Mexican-born mothers, and White non-Hispanic mothers (wherever born)best. However, both the healthy migrant and the healthy immigrant hypotheses suggest that the health profiles of Mexican-born immigrants should be better than U.S.-born women of Mexican descent. Multiple logistic regression is used to test the likelihood of each outcome by race/ethnicity among all Colorado mothers and by nativity for Mexican-born and U.S.-born mothers of Mexican origin (Aims 1 and 2). By testing successive regression models adjusting for the main effect alone (race/ethnicity or nativity) (Model 1 ), adding demographic and social economic factors (Model 2), adding medical risk factors from the birth record (Model 3), and finally adding health behavior factors during pregnancy (smoking, consumption of alcohol, and weight gain) (Model 4), it is possible to compare births by race/ethnicity and then to explore the healthy migrant and immigrant hypotheses. The influence of neighborhood effects at the census tract level in Adams and Denver Counties is tested using general linear modeling. The final stage of this mixed method study includes qualitative interviews of five key informants and ten recent mothers of Mexican origin recruited from Salud Family Health Clinics, a provider of health services to the poor and underserved in Colorado. Five mothers were born in Mexico and five were born in the U.S. 12

PAGE 30

Description of the Study Population The Division of Vital Records in the Colorado Department of Public Health & Environment provided a CD-ROM containing a de-identified state census of all recorded singleton births, with flags for infant mortality from its linked death records, for the years 2000 2005. This dataset consists of 392,881 recorded births. After removal of cases for which any study variable was missing, the dataset for Aim 1 consists of 356,389 births and the dataset for Aim 2 consists of 85,755 births .. Table 1.5. Study population by racelethnlclty of mother-Aim 1 Study Population 356,389 White Hispanic Black Other Study Population by Race/Ethnicity 219,029 (61.46%) 106,291 (29.82%) 15,448 (4.33%) 15,621 (4.38%) Table 1.6. Study population by nativity of mothers of Mexican origin Aim 2 Study Population 85,755 Study Population by Nativity U.S.-Bom Mexican Origin 32,484 (37.88%) Mexican-Born 53,271 (62.12%) The study population for the contextual analysis in Aim 3 is reported in Table 1.3. Table 1. 7. Study population of mothers of Mexican origin In Denver and Adams Counties Alm3 County Adams Denver Study Population 16,107 23,332 U.S.-Bom Mothers Mexican Origin 5,733 (35.59%) 4,914 (21.06%) Mexican-Born Mothers 10,374 (64.41%) 18,418 (78.94%) Organization of the Dissertation Number of Census Tracts 79 100 This chapter introduces the issues, briefly reviews the rationale for this type of study, states its specific aims, describes the overall research methods, and enumerates the study population. Chapter 2 reviews the literature of the epidemiological paradox, especially as applied to Hispanics and immigrants in the U.S., and the various hypotheses and related theories that have been suggested to explain it. Chapter 3 describes the research design and methods for the quantitative and qualitative analyses. Chapter 4 discusses the quantitative analysis and results. Chapter 5 discusses the qualitative analysis and results. Chapter 6 synthesizes the results and discusses the importance of the findings. 13

PAGE 31

CHAPTER 2 EPIDEMIOLOGICAL PARADOXES The Social Gradient of Health The biomedical tradition has historically approached health and disease from an individual perspective; it is the individual, after all, who gets sick. In this paradigm, treatment focuses on the individual and often emphasizes perceived unifying, biological causes of disease. The biomedical approach emphasizes personal agency in the form of theories of individual behavior change that assign the individual control of health promotion, such as smoking cessation, eating a healthier diet, and avoidance of unhealthy exposures. In contrast, public health approaches rely more heavily on structure, seeking causes of health and disease in conditions that affect a population, such as access to clean water, and pursuing health promotion in the form of enhancing social conditions that facilitate access to health benefits and eliminating health disparities (Hamlin & Sheard 1998). These different perspectives are important, because they inform the questions one asks about health outcomes and the potential approaches to improve them. One of the foundations of public health is belief in a general social gradient of health, in which there is a positive relationship of better health with increasing income/wealth, social status, and education (Lynch & Kaplan 2000:13-35; Cassel 1964). Epidemiological studies have identified the social gradient in various historical time periods, among different population groups, for multiple health conditions, and for both morbidity and mortality (Lynch & Kaplan 2000:13). Two well-known demonstrations of the social gradient are the Chadwick Study from the mid-1800s in the U.K. (Chadwick 1842) and the more recent Whitehall Studies showing a social gradient of coronary heart disease by U.K. occupational classification (Rose & Marmot 2001 ). Differences in health corresponding to socioeconomic position are thought to result from structural societal factors that lead to differential access to health resources and damaging exposures (Miech et a/. 2006; Link & Phelan 1995; Marmot et a/. 1987; Syme & Berkman 1976). Socioeconomic position, also variously referred to asSES, social class, and social status, is defined by Lynch & Kaplan (2000:14) as [t]he social and economic factors that influence what position(s) individuals and groups hold within the structure of society, i.e., what social and economic factors 14

PAGE 32

are the best indicators of location in the social structure that may have influences on health. The concept of socioeconomic position is influenced by three schools of thought: Marx's analysis of "class" based on relations of production; Weber's stratification of society by class, status, and power, leading to unequal distribution of resources; and the Functionalist stratification of society as a necessary outgrowth of the complex workings of modern society (Lynch & Kaplan 2000). Each of these formulations is based in the broader theoretical perspective of political economy, which suggests that much of what happens in the world is driven by access to resources or lack thereto. The Social Gradient of Health at the Individual Level Belief in the social gradient of health has led to identification of epidemiologic "exposures" that translate into measures of position and health risk factors at the level of the individual. Some social epidemiologists look at socioeconomic position as reflective of access to resources (Link & Phelan 1995). Thus, income and wealth are each resources in themselves, as well as instrumental vehicles for access to resources of other kinds, including healthier food, safer neighborhoods, and health care. Education is viewed as instrumental for access to information and knowledge (such as the value of good nutrition, exercise, or appropriate prenatal care) (Winkleby eta/. 1992) as well as correlated with higher-paying and safer jobs and perhaps with employer assistance in obtaining and paying for health insurance (although employer-provided health insurance is diminishing in frequency in the U.S.) (Economic Policy Institute 2006). Marriage is generally seen also as a vehicle for access to more resources, with the potential for a married pregnant woman to share more resources than she might have on her own. In general, having low income, being unmarried, being unemployed, and reaching lower educational levels are associated with higher rates of the adverse birth outcomes of interest in this study. The Social Gradient of Health at the Area Level The basis of anthropological and sociological theory that stratification of society leads to differential material, political, symbolic, psychosocial, and behavioral effects ultimately reflects aspects of power and its distribution that affect not only individual characteristics but also contextual characteristics of a population based on the area in which people live (Kawachi & Berkman 2003; Macintyre & Ellaway 2000; Diez-Roux 1998). Contextual characteristics of neighborhoods have been associated with all-cause mortality, 15

PAGE 33

infectious disease, birth outcomes, obesity, and lack of physical activity (Kawachi & Berkman 2003; Krieger et at. 2003). Research identifying negative exposures and their accumulation and exacerbation has been especially useful for elucidating effects of poverty, discrimination, residential segregation, and the differential distribution of health-promoting resources in poor communities and among disadvantaged populations (Collins & Schulte 2003; Fang et at. 1999; Collins & David 1990; Kleinman & Kessel 1987). Pathways by which area characteristics affect health are not always clear (Collins & Schulte 2003:228-230), although there are numerous hypotheses. Positive relationships between undesirable area characteristics (such as high percentages of poverty, unemployment, or female heads of household) and adverse birth outcomes are generally thought to be related to neighborhood stressors (Gorman 2005). Stressors might include residence in a violent community (Zapata et at. 1992), subjective psychophysiological reactions to a neighborhood (Collins et at. 1998), and racial or ethnic discrimination (Krieger 2000; Collins et at. 2000; LaVeist 1989). Undesirable neighborhoods may also fail to provide convenient access to healthy food sources (larger grocery stores versus convenience stores and fast food outlets), safe and inexpensive public transportation, safe places to walk and exercise, employment, and health care related resources such as clinics, hospitals, and medical professionals. In general, neighborhood disadvantage increases the risk of LBW independent of the individual's family situation (Sellstrom & Bremberg 2006; Collins & Schulte 2003; Robert 1999). But Finch et at. (2007) demonstrate that, after controlling for individual risk factors, the degree of immigrant orientation in Los Angeles neighborhoods moderates the expected increase in LBW associated with neighborhood deprivation, especially for foreign born mothers Social support (resources available to individuals and groups through connections in social networks) is sometimes suggested as a moderator of effects of individual or area characteristics. Social support can be an individual attribute as well as an area or even a "virtual" characteristic. 1 Social networks provide various aspects of support, including information and advice, emotional support, and access to material resources (Berkman & 1 Most studies focus on neighborhoods as "real estate," and networks as defined and confined by physical space (Berkman & Clark 2003:288). But social networks need not be geographic in today's social environment, where one may live in one neighborhood but find social support and networks through non-neighborhood work relationships, church affiliations, relationships with family located in another state or country, and even Internet-based social networks (dileonardo 1984:129-157). 16

PAGE 34

Clark 2003:289). One example of positive social support may be found in examining gay and lesbian networks that provide both safe places to get together and access to resources, especially in urban areas. In this same vein, it is possible that Mexican-born immigrants seek to live in neighborhoods with other recent immigrants from Mexico from whom they receive helpful social support (Finch et a/. 2007) or with whom they share cultural values, including language and practical solidarity that are protective of pregnancy. Several studies also have noted that the density of ethnic populations and extended kin networks is associated with better health outcomes such as child health status in Mexico (Kana'iaupuni et a/. 2005) and mental health in older Mexican Americans (Ostir eta/. 2003). Table 2.1, adapted from Lynch & Kaplan (2000:18-19), lists some measures of exposure that have been used in social epidemiological studies of health. Table 2.1. Measures of socioeconomic position Individual Level Measures Income Education Wealth Occupation Area Level Measures Occupational structure Educational structure Housing characteristics Poverty area Deprivation Population characteristics Access to resources Absolute income Income measured as % of official poverty level income Years of education Milestones of education, such as degrees Value of total assets Self-reported using population-specific scales of wealth Census classifications Score based on occupational prestige ratings Categorization based on managerial hierarchy of job % population in various occupational categories in area % population at various educational levels in area Age of housing, density of inhabitants per room, access to facilities, segregation % households below poverty-level income Unemployment, car/home ownership, overcrowding, crime %population subgroups (raciaVethnic) in area Resources such as hospitals, clinics, parks, private services Unexpected Deviances from the Social Gradient of Health -The Paradox In the United States, the expected social gradient of health reflects a kind of "majority minority'' population axis, along which poorer, less educated populations, which are also usually made up of racial or ethnic minorities, are assumed to (and often do) have poorer health outcomes (Collins & Schulte 2003:225). Given studies that associate lower income, lower educational attainment, lower social position, and residential segregation by income 17

PAGE 35

with poorer health outcomes (see Lynch & Kaplan 2000 for a general review), researchers note with interest when such patterns do not hold and when some groups appear to have better outcomes than would be predicted based on socioeconomic position outcomes that approach the reference (usually "majority") group (Chung eta/. 2003; Buekens eta/. 2000; Hummer eta/. 1999; Cobas eta/. 1996; Markides & Coreil 1986). It is useful to clarify the nature of this apparent paradox. It is not that a socioeconomically disadvantaged population necessarily has better health outcomes compared to the majority more advantaged population (although some studies have found such results). It is rather that some key health outcomes are more like those of the majority population than those of other disadvantaged populations, when their socioeconomic positions would suggest that their outcomes should be more similar (Hummer eta/. 2007), which then raises the question of what accounts for the paradox. Although evidence of an epidemiological paradox may have been identified as early as 1974 for a Hispanic population in the U.S. (Teller & Clyburn), most researchers credit Markides and Coreil with coining the term "Hispanic epidemiological paradox" in 1986. They reviewed studies of the health status of Hispanics from the southwestern United States that were conducted in the 1960s and 1970s and concluded that Hispanics enjoyed some health outcomes that were similar to non-Hispanic Whites in categories such as infant mortality, life expectancy, mortality from cardiovascular disease, mortality from major types of cancer, and measures of functional health, even while their socioeconomic position was more like that of most Blacks who, as a group, had worse health outcomes than Whites. But for indicators such as incidence of diabetes and infectious and parasitic diseases, their review showed that Hispanics had poorer health outcomes than Whites. Since then, there has been an explosion of work on the paradox in general and with respect to specific outcomes, subpopulations, and hypotheses to try to explain any deviance from the expected social gradient. Some studies have used national population databases (Hummer eta/. 1999; Singh & Yu 1996); others have examined locations with large numbers of immigrants from Puerto Rico (New York City) (Rosenberg eta/. 2005) and from Mexico and Central America (California) (Keleher & Jessop 2002; Fuentes-Afflick et a/. 1999; Zambrana et a/. 1997). An important result of these studies has been recognition of the heterogeneity of people classified as "Hispanic." Rates of LBW vary depending on the country of origin, with Cubans and Mexicans usually having lower rates of LBW than Hispanic mothers from other countries (Hummer eta/. 1999; Reichman & Kenney 1998). 18

PAGE 36

Even more interesting, the paradox has also been shown among some first generation (foreign-born) immigrants who have better outcomes than immigrants from the same country in their second and later generations in the U.S. The advantage of foreign birth seems to decay rather rapidly with the time spent in the United States (Madan et a/. 2006). These findings have led to a plethora of studies on immigration, assimilation, and acculturation (Acevedo-Garcia eta/. 2005; Kasirye eta/. 2005; Rosenberg eta/. 2005; Cho et a/. 2004; Gould eta/. 2003; Weigers & Sherraden 2001; Fuentes-Afflick et a/. 1999; Hummer eta/. 1999; Lansdale eta/. 1999; Cobras eta/. 1996; Singh & Yu 1996). Several of the studies reviewed by Markides and Coreil in 1986 are somewhat limited. Many looked only at individual-level characteristics and early studies were constrained by ambiguous ethnic designations. Until recently the U.S. Census identified Hispanic ethnicity using Spanish surname rather than self-identification as Hispanic/Latina or various sub-classifications within "Hispanic." Vital records contained a paucity of information (a problem that still exists, although to a lesser extent). Also many of the early studies used unadjusted incidence rates of health outcomes or relied on convenience samples. Since then, more researchers routinely include adjusted odds ratios and use census databases to avoid the need for sampling. Even today, however, many studies of the paradox focus only on the population's compositional attributes by including only individual-level characteristics, such as race/ethnicity, gender, age, marital status, health conditions, and health behaviors (Rosenberg eta/. 2005; Hummer eta/. 1999; Singh & Yu 1996) and most do not report interactions among variables used in the analysis. Fewer studies approach the problem from the perspective of area characteristics, seeking to explore whether a population's contextual characteristics influence health outcomes (Pickett et a/. 2005). Population studies of birth outcomes addressing Hispanics and in some instances foreign-born mothers from 1990 2007 are described in Appendix A. Few of these studies include SGA as an outcome, none includes LGA, and none addresses Colorado mothers. Traditional Explanations of the Hispanic Immigrant Paradox Several hypotheses have been proposed to explain unexpectedly better birth outcomes of Mexican-born mothers and U.S.-born women of Mexican origin. Among the most common explanations of these unexpected results are: (1) selective migration by healthy Mexicans [healthy (Cho et a/. 2004; Marmot et a/. 1984), 19

PAGE 37

(2) healthier diet and behaviors, especially lower rates of smoking and alcohol consumption among Hispanics generally, and especially among first-generation immigrants [healthy (Flores & Brotanek 2005; Kasirye et a/. 2005; Abraido eta/. 1999; Guendelman eta/. 1999), (3) acculturation/assimilation as a way of explaining the decline in better outcomes over time -as acculturation increases, women of Mexican origin increase their rates of smoking and drinking and consume a less healthy diet, and (4) social support among Hispanics during pregnancy (Landsale et a/. 1999; Rumbaut & Weeks 1996). Healthy Migrant Hypothesis The healthy migrant hypothesis may include elements from both ends of the agency -structure axis. Although migration to the United States from Mexico is motivated primarily by Mexico's poor economy, which is a structural rationale, self-selection for migration may reflect a form of agency. A number of researchers suggest that those who choose to migrate will have experienced better childhood health than those who remain in Mexico (Crimmins et a/. 2005) and that better health outcomes among first generation immigrants are a result of the migration of Mexicans who are healthier than those who stay behind (Cho eta/. 2004; Abraido-Lanza eta/. 1999; Marmot eta/. 1984). At first blush this hypothesis should be easy to test: compare rates of any given health outcome among Mexican immigrants to the U.S. with rates for those who stay in Mexico. For example, UNICEF estimates that the rate of LBW in Mexico in 2000 was 9% (UNICEF 2004). The rate of LBW in Mexico is not that different from the overall rate of 8.1% for all Hispanics in Colorado in 2000 (see Table 1.2). However, some researchers argue that one cannot compare rates of an outcome at this level without taking into account the context of each population, such as comparisons of health behaviors, risk profiles, social support, and other psychosocial effects (Finch, B., personal interview, February 27, 2007). Crude rates of outcomes do not account for individual and contextual effects and therefore one cannot immediately assume that any differences are or are not accounted for by the healthy migrant hypothesis. A direct test of this hypothesis, therefore, requires better birth outcome data on mothers living in Mexico than currently exist. Even when access to roughly comparable data is available, the results may not support the logical conclusion of healthy migrant selection. Crimmins et a/. were able to compare data from the U.S. National Health and Nutrition Examination Survey and similar 20

PAGE 38

data from the Mexican Health and Aging Survey to compare the height of Mexican immigrants to the U.S. with Mexicans who remained in their communities of origin. They found that immigrants are indeed selected for greater height after adjusting for socioeconomic status, but that same group of immigrants was shorter than people of Mexican origin who were born in the U.S. (2005). In another study of households in Mexico, Frank and Hummer analyzed data from the 1997 Encuesta Nacional de Ia Dinamica Demografica [ENADID] (2002). They found that membership in a "migranf' household (one in which at least one individual in the household took a migratory trip to the U.S. in the five years prior to the survey) is protective for LBW of an infant born in that household in Mexico after the migratory trip. The authors suggest that the protective factor of migration, so defined, was probably largely based on receipt of remittances from the migratory household member. This finding supports the general social gradient hypothesis -that access to more financial resources would be associated with better LBW rates. However, Frank and Hummer also split migrant households into two categories based on whether or not they received monetary remittances, and found that infants born into non-remittance migrant households in Mexico had a significantly lower risk of LBW compared with those born into remittance migrant households. The authors argue that some factor like "social remittances" may be associated with the better health outcomes of migrant families, but they have found no way to conceptualize or measure it using the ENADID data. Even more interesting is their finding of a "mini-epidemiological paradox'' within Mexico. Women in migrant households had riskier household socioeconomic profiles, but their infants had more favorable birth outcomes compared with infants in non-migrant households (2002:755). A number of studies have shown statistically significant results that foreign-born Hispanic women have better low weight-related birth outcomes, such as LBW and preterm birth (Wingate eta/. 2006; Acevedo-Garcia eta/. 2005; Gould eta/. 2003; Hummer eta/. 1999; Lansdale et a/. 1999). But Rosenberg et a/. found that, after adjusting for risk factors, the results did not support the healthy migrant hypothesis (2005). They concluded that the difference in outcomes reflected differential distribution of risk, not "healthier" migrants. Healthy Immigrant Hypothesis The healthy immigrant explanation finds support in studies showing that the socioeconomic risk profile of foreign-born immigrants is poor, but that their health behaviors are good immediately after they arrive in the United States (that is, they smoke and drink little and eat a healthier diet). This hypothesis "sounds in agency" because it is based on health 21

PAGE 39

behaviors, and suggests that immigrants follow their home cultural beliefs, which are thought to include healthier practices than those associated generally with life in the U.S. It does appear that female immigrants from Mexico as a group engage in fewer deleterious health behaviors such as smoking and drinking of alcohol when compared with non-Hispanic Whites (Meneses-Gonzalez et a/. 2006; Abrafdo-Lanza et a/. 2005). Some studies have found that Mexican-born women have healthier dietary intakes (Guendelman & Abrams 1995). These findings might support the positive outcomes noted in the first generation, which seem to decay rapidly with time spent in the U.S. (Montez & Eschbach 2008; Kasirye et a/. 2005; Neuhauser et a/. 2004). Kaplan et a/. (2004) demonstrated a dose-response relationship between obesity and length of residence in the U.S. Adjusting for multiple factors, those who lived in the U.S. for 15 or more years had an odds ratio of 4.0 for obesity compared to those who lived in the U.S. for fewer than 5 years. Reduction in the odds of better birth outcomes is suggested to be a result of acculturation processes such as adoption of unhealthy diet and sedentary lifestyle (Mainous eta/. 2008; Abrafdo-Lanza eta/. 2005). Barcenas eta/. (2007), however, found that length of residence in the U.S. is associated with risk of obesity, especially in Mexican-American women, but that degree of acculturation was not a predictor. Acculturation/ Assimilation For several reasons, defining the concept of "acculturation/assimilation" is complex and elusive. Acculturation and assimilation are sometimes conflated or parsed very finely to differentiate between the two concepts. Historically anthropologists and sociologists have used the concept of assimilation to address structural locations of immigrants as they move toward interacting with the more dominant social institutions of their new communities, whereas acculturation addresses changes in beliefs, values and practices at the individual level (Wingate & Alexander 2006; Weigers & Sherraden 2001 ). Regardless of definition, the idea that either process operates in a linear fashion does injustice to their complexity, especially with regard to Mexican immigrants. While in general Mexican-born immigrants sometimes experience better birth outcomes than later generations of Mexican descent born in the U.S., generational status alone is at best a gross measure of acculturation, which is more complex than just amount of time spent in the U.S. (Amaro & de Ia Torre 2002; Guendelman et a/. 1990). Moreover, acculturation itself is affected by exogenous factors such as the degree of, and need for, acculturation as well as the starting point of the immigrant when she arrives (Weigers & Sherraden 2001 ). Unlike immigration from many other countries, immigration from Mexico is not necessarily an "unknown" because there are 22

PAGE 40

historical and geographic links between Mexico and U.S. along their long shared border. Indeed, land that is now part of the U.S. was once the northern part of Mexico. In addition, there is substantial two-way immigration, so that immigration from Mexico cannot necessarily be seen as a permanent status (Weigers & Sherraden 2001 ). According to Partes et a/. {1985) education, age, poverty index, and size of community have important effects on acculturation. Even when researchers clearly define assimilation or acculturation, its operationalization is fraught with measurement difficulties. Either acculturation or assimilation as a process is like a black box that masks transition from one status to another, and it is unclear which dimensions or factors should be included and which are more important (Berry 1997). Acculturation may be a marker for values, beliefs, and lifestyle (including smoking and parity) (Scribner & Dwyer 1989). Some studies operationalize acculturation using the length of time immigrants have been in the country (Harley eta/. 2007), others look at the age at which immigrants came to the U.S. (Kasirye eta/. 2005), the language spoken (Ebin eta/. 2000), or other measures. Cobas et a/. (1996) criticize "an" acculturation hypothesis because, whatever the concept includes, they say it is multidimensional. They used structural equation modeling to address the conceptual and methodological weaknesses in acculturation theory, and found that not all components of acculturation had the same effect on LBW. They found that language spoken was a stronger predictor of smoking and LBW than ethnic identity, and that while acculturation influences LBW through diet and smoking, a significant direct effect of acculturation still persists. Therefore, more acculturated mothers were more likely to deliver LBW children, partly because they smoked. They also found that dietary intake was an intervening variable between acculturation and LBW status, with acculturation having a negative effect on dietary intake, and diet affecting LBW. The mechanisms that account for relationships between acculturation and health behaviors remain elusive (Abrafdo-Lanza eta/. 2005). In addition to the difficulty of operationalizing acculturation, however it is conceptualized, there is a dearth of population-wide individual-level data on acculturation. Birth and death records do not record the length of time lived in the U.S., age at immigration, date of immigration, or primary language. Instead, these data must be obtained from surveys that sample a population, usually one that is quite small. 23

PAGE 41

Social Support Some researchers argue that studies of acculturation focus too much on negative health behaviors (e.g., smoking and diet) and not enough on more general social support as a route to resources that improve outcomes (Weigers & Sherraden 2001 ). Guendelman eta/. agree and further suggest that social support variables may not correspond with degree of acculturation (1990). For example, protective factors for birth outcomes include strong cultural support for maternity, healthy diet, and marianismo (devotion to the maternal role) (McGlade eta/. 2004), which may exist for women regardless of their degree of acculturation. Social support may be facilitated by living in an area with a high concentration of culturally similar residents, which may provide better access to culturally appropriate prenatal care and other support during pregnancy that lead to better birth outcomes, either through social support pathways or through commonly-held cultural beliefs and practices. Finch eta/. (2007) found that foreign-born Hispanics living in Los Angeles had better birth outcomes than U.S. born-Hispanics, even when they lived in very deprived neighborhoods, so long as those neighborhoods had high concentrations of foreign-born residents. Another pathway to better birth outcomes may be the existence of protective cultural factors that may influence perceptions of resilience and risk and thus affect health outcomes. Bender and Castro conducted a qualitative study of Mexicans in North Carolina, testing the concept of resilience (universal capacity to prevent, minimize, or overcome damaging effects of adversity), as a "positive counterpart to 'risk factors"' (2000:156; see also Kawachi & Subramanian 2006). Bender and Castro found that protective factors may not directly promote positive health outcomes, but instead may buffer or increase resistance to negative events, through compensation (counteracting stressfuVadverse with external characteristics or external sources of support); challenge (stressful events in prior life can enhance competence); and immunization (protective factors temper the impact of stressors, but may not be evident absent stress). Among other findings, they identified as a positive theme the strength of social support enjoyed by the immigrants, who had strong nuclear and extended families, especially during pregnancy. These immigrants had larger numbers of friends and kin in North Carolina than the authors anticipated. Even small families had 8-10 members of extended family who would gather on weekends in the park for recreation and eating, as they did in their Mexican pueblos. Furthermore, pregnant women in the U.S. maintained ties to their mamas who were still in Mexico by relying on telephone calls. 24

PAGE 42

Cultural values may have negative effects as well. Culture shapes one's view of weight. Outside of the industrialized West, being large and round indicates positive health, and Hispanics are more likely than Whites to perceive overweight as closer to their ideal body size (Candib 2007). Ahluwalia et a/. find that Mexican American women are less likely to perceive that they are overweight (2007) and that less acculturation is associated with higher BMI among Mexican immigrants (Akresh 2007). Finally some researchers argue that the paradox is not real but is merely a statistical artifact (Palloni & Arias 2004). This particular view has gained adherents especially with respect to mortality and life expectancy studies. Palloni and Arias argue that mortality statistics of foreign-born immigrants are particularly suspect, because foreign-born individuals are especially likely to return home when they age or get sick, or to die. This "salmon effecf' is argued to account for any paradox in mortality statistics, especially among Mexican immigrants. Morales eta/. (2002) disagree, as do Hummer eta/. (2007) who report on a study that not only measures infant mortality among Mexican-born mothers, but also calculates how many infants (far too many to be plausible) would have to move back to Mexico to account for the paradox. The salmon effect cannot apply to weight-related birth outcomes, as the outcome is a measure of birth weight, which exists in the record An Alternative TheoryPolitical Economy Studies of health outcomes are situated in the broader context of theories of disease causation and health disparities. Disease is proximately caused by biomedical factors such as genetic inheritance and/or exposure to agents that affect the healthy working of the body. In regard to genetics, there is evidence that genetics plays a role in rates of diabetes among Mexicans. Those with more indigenous heritage have higher rates of obesity and diabetes (Martorell 2005). It is also interesting that immigrants from Mexico are not as exogamous as might be expected. In the 1990s Hispanic intermarriage rates with non-Hispanic Whites declined substantially (Lichter eta/. 2007). Based on these observations, one might expect that rates of diabetes and obesity and perhaps LGA would be comparable among Mexican born and U.S.-born mothers of Mexican origin. Intermediate and distal causes of disease include poverty, lack of education, poor nutrition, exposure to environmental insults, or lack of access to health care, all of which increase susceptibility to disease (Link & Phelan 1995; Millard 1994) and which fit within the ideas of the social gradient of health and political economy. Therefore a number of theorists suggest that models of disease causation and their opposite, models of disease prevention, 25

PAGE 43

must address macro-level processes (often economic) that affect individuals and their choices as well as individual health risks (Diez-Roux 1998). Studies that try to correlate a degree of assimilation or acculturation with diet, other health behaviors, and health outcomes necessarily focus on measures of assimilation and acculturation. While it is indeed likely that the social and cultural transitions after immigration to the U.S. affect diet, health behaviors, and health outcomes, stepping back to look at a larger frame of reference within which transitions occur may be more explanatory, especially if the outcomes are not uniformly positive. A succinct definition of political economy is elusive, because it is developed from a mix of intellectual and political movements (Roseberry 1988). It has its roots in Weber and Marx and at a minimum consists of a set of theories that center on Western economic and political domination of less developed societies and the large-scale effects of capitalism on specific countries, regions, or communities, approached from an historical perspective (Roseberry 1988). A political economy approach is fundamentally structural, positing that individual choices are framed and constrained by pervasive political and economic forces far from the individual. Candib (2007) diagrams pathways to obesity and diabetes, situated in a political economy perspective that appears applicable to Mexican migrants to the U.S. Rgure 1. P.ttn.ys to obesity and chbetes. GloMI!ulkln ObnogeM Eawlnnment POIII!fty: lnckntrializMion, Global tradt policy High cost of heh produce Childhood obesity mechanization. produce cheap tau Low iiCteS 10 good laod maHmedl and suga" lr0111 .... entrNimMIII __.llo Epidemic obesity corn lll!d W'f ,. No :late place 10 play Dl4bele Sdlooli wii!Qn Vascular diseMe DecrNHCI physital Third \Vorld food Iannen put out Socilll and family contms of W!Pight of high or lew Low-and environl!'ltl'ltS; phyliologOI I Urbilninltion l and psychologi
PAGE 44

consumed, even outside urban areas of Mexico (Jimenez-Cruz & Bacardi-Gascon 2004). According to Martorell, "Mexico is a country far along the nutrition transition" (2005:2). This nutritional transition within Mexico has led to higher rates of obesity and type II diabetes, with 58% of women of reproductive age either overweight or obese (Jimenez-Cruz & Bacardi Gascon 2004). In addition, there is evidence that malnutrition early in life followed by catch up growth among poor Mexicans (the same pattern seen among LBW and some SGA babies) is a high risk factor for obesity, diabetes, and cardiovascular disease in adulthood (Martorell 2005; Diamond 2003). Moreover, changes in Mexico's labor economy continue to create motivation to leave the country for the U.S. when people cannot find work in Mexico's cities or when they see better economic opportunities than exist in their villages (Partes & Bach 1985). Once in the U.S., immigrants recognize the inverse relationship between the cost of food and its energy density, with foods high in sugar and fats being the least expensive and the most energy dense (Carnethon 2008). Most acculturation studies try to define the construct and suggest that it operates over a generation or more to produce changes in diet and health behaviors such as smoking and weight gain. These changes are associated with poorer LBW outcomes in the second and third generations in the U.S. Yet it may be that an earlier, more personalized, nutritional transition occurs almost immediately upon immigration. This transition may explain lower rates of LBW and low weight outcomes and at the same time tend to increase LGA. When Mexican immigrants arrive in the U.S. they typically are poor. High density foods are cheaper and often more readily available; access to tropical fruits of the quality they are used to in Mexico is less (Yeh eta/. 2008; Drewnowski & Specter 2004). In addition, "fast food" is often viewed as high status in their native country-its low cost and high availability in the U.S. is attractive to immigrants. Himmelgreen eta/. (2007) found that changes in the diet that result in weight gain, less exercise, and eating fast food are occurring as early as two and a half years after immigration to the U.S. Moreover, they found that the proposed positive buffering of high immigrant density was not supported instead social isolation was the norm post immigration. Associating positive health outcomes with various facets of individual behavior and even neighborhood contextual factors likely represents oversimplification of a complex reality. All else being equal, regularly consuming a diet low in refined sugars and fats, living a lifestyle that includes physical exercise and less sedentary activity, refraining from smoking, 27

PAGE 45

and engaging in social and cultural activities that increase social support and cohesion result in better health outcomes generally, and better birth outcomes specifically. But centuries of social science research demonstrate that these behaviors are not completely voluntary. All of us live in larger contexts that affect both individual health behaviors and neighborhood contexts. Residents of Mexico and the U.S. live in a world economy that affects social conditions that in turn affect individual agency. The most common attempts at explaining any epidemiological paradox based on race, ethnicity, or immigrant status miss the mark because they are incomplete. More inclusive theories, such as political economy, need to be employed to place context in a much larger frame. These understandings may then allow more specific health behavior theories and cultural beliefs to inform interventions at multiple levels to improve birth outcomes. 28

PAGE 46

CHAPTER 3 RESEARCH DESIGN AND METHODS This study has two particular strengths compared with existing population studies of birth outcomes: the inclusion of a broader range of birth outcomes, with differentiation of high birth weight from low birth weight outcomes; and the use of mixed methods. Prior studies tend to focus on low birth weight (sometimes with the addition of very low birth weight) and premature birth. Few include small for gestational age as an outcome. More importantly, population studies have focused on the low-weight end of the spectrum of birth outcomes and ignored other weight-related outcomes. Broadening the scope of outcomes allows a stronger test of the hypotheses proposed to underlie population differences. In addition, the use of mixed methods allows for the collection of qualitative data to explore the ''why'' of differences that is not captured in the government-gathered official birth records. This study addresses four specific aims: to test the expected social gradient of health on birth outcomes by racelethnicity and to identify any paradoxical outcomes (Aim 1 ); to test the expected social gradient of health on birth outcomes by nativity of Hispanic mothers of Mexican origin and to explore the healthy migrant and healthy immigrant hypotheses as explanations for any paradoxically better birth outcomes experienced by Mexican-born mothers (Aim 2); to examine the association of neighborhood deprivation and immigrant orientation on birth outcomes of mothers of Mexican origin in two large Colorado counties (Aim 3); and to explore potential reasons for differential outcomes among Mexican-born mothers and U.S.-born mothers of Mexican origin (Aim 4). In brief, risk profiles are compared for a cohort of Colorado mothers and odds ratios of birth outcomes are calculated for two statewide populations: all mothers based on their race/ethnicity and mothers of Mexican-origin based on their place of nativity (Mexico or U.S.) using individual-level, compositional data from the state's official birth records. Separately, contextual neighborhood-level effects are examined to test whether neighborhood deprivation 29

PAGE 47

or immigrant orientation affects birth outcomes of mothers of Mexican origin in Adams and Denver Counties using data from the 2000 Decennial Census. All quantitative analysis is performed using SAS v. 9.1.3 (2003). Qualitative interviews with five key informants and ten recent mothers of Mexican origin provide insight into the quantitative results. Nested Mixed Method Design Mixed method designs include quantitative and qualitative approaches to a research question to minimize weaknesses and maximize strengths of each of the research paradigms (Medlinger & Cwikel2008; Morse & Field 1995:164). In general, qualitative methods are used to explore problems about which relatively less is known, whereas quantitative methods are used to test theories and hypotheses (Medlinger & Cwikel). In reality, neither paradigm alone is usually sufficient to answer complex questions that seek to uncover both outcomes and reasons underlying them (Korbin 2008). A review of 57 mixed method designs by Greene at a/. (1989) identified five purposes for using this approach: triangulation (corroboration of findings using different methods), complementarity (enhancing or clarifying results), development (using results from one method sequentially to inform and develop the next stage), initiation (developing new perspectives by highlighting conflicting findings), and expansion. Mixed methods can be employed in at least three ways: sequentially (each is a distinct study), in a nested design (one paradigm is paramount and the other is used as a supporting method of analysis), or in a fully integrated design (both paradigms are used concurrently) (Medlinger & Cwikel 2008). This study originally was proposed as a mixed method design using a development approach because it was expected that qualitative data would be needed to identify appropriate variables for inclusion in the quantitative analysis. However, based on the availability of variables in the birth record and prior literature, it was decided to perform the quantitative analysis first and use qualitative methods to complement and corroborate the quantitative results. The choice of a mixed method design became particularly useful to explore differential odds of LGA among mothers of Mexican origin in Aim 2. As a result, this study is a nested design, with the quantitative component dominant, using qualitative methods to complement the quantitative findings. Figure 3.1 depicts the nested design and shows that both methods inform each of the specific aims. 30

PAGE 48

Quantitative Methods Individual Level Aim 1 Race/Ethnicity Aim 2 Nativity Area Level Aim 3 Nativity Figure 3.1. SCHEMATIC OF RESEARCH DESIGN Quantitative Research Component The quantitative component of the study is a retrospective cohort design. For Aims 1 and 2 all singleton births recorded in Colorado's Vital Records for the years 2000 through 2005 are examined. For the contextual analysis of Aim 3, data on singleton births to women of Mexican origin residing in Adams and Denver counties are linked with census tract data from the 2000 U.S. Decennial Census. The original study design proposed evaluating four birth outcomes: low birth weight, preterm birth, small for gestational age, and infant mortality, all of which have several interrelating risk factors that focus on smaller size of the infant. Infant mortality was dropped as an outcome based on its very low incidence in Colorado compared to the other study outcomes and the questionable quality of cases reported as live births at very early gestation in the vital records dataset. In one of the earliest reviews of the apparent Hispanic "paradox," Markides & Coreil (1986) noted that Hispanics had lower incidences of some specific health outcomes such as infant mortality and breast cancer, but were not advantaged compared to the non-Hispanic White population for other conditions such as stomach cancer and diabetes. Delivering an infant that is large for gestational age is associated with maternal diabetes and also contributes to the infant's immediate and life course health risks, such as diabetes, chronic 31

PAGE 49

cardiovascular disease, and stroke, in ways that are similar to the life course effects of LBW, preterm birth, and SGA (Martorell 2005). LGA was therefore added to the study to broaden the scope of weight-related birth outcomes. Addition of LGA adds depth to the exploration of adverse birth outcomes because, although LBW is an international measure of population health, birth weight of >2500 grams is a somewhat simplistic measure for a favorable birth outcome given the complexity of the relationship of birth weight and both acute and chronic morbidities (Lubchenco 1971, 1972). In light of trends to overweight, obesity, and type II diabetes in the U.S., especially among Hispanics (Mainous eta/. 2008; Carnethon 2008; Martorell 2005), it is appropriate to include a more precise classification of weight-related birth outcomes that better captures health risk. Levels of Analysis Many studies examine the relationship between individual-level, compositional characteristics and health outcomes or between area-level, contextual characteristics and health outcomes. Results of studies at one level are sometimes misused to make statements about a different level, creating fallacies in logic. The ecological fallacy occurs when one draws inferences at the individual level based on area-level data; the atomistic fallacy draws inferences at the group level based on individual-level data. To avoid these valid criticisms, researchers doing single-level studies use data from and make inferences at the same level. But single-level studies alone, even when they avoid fallacies of cross-level interpretation, may miss the effect of mechanisms operating at multiple levels (Greenland 2002; Diez-Roux 1998). Contextual factors may constrain or enhance choices that individuals can make that affect their health (Diez-Rouz 1998). In a number of multilevel studies, including four studies on LBW, neighborhood socioeconomic status and social climate were shown to have a small (-10%) effect on child health outcomes (Sellstrom & Bremberg 2006) (see also Krieger eta/. 2003). Contextual variables are usually classified in one of two ways. Derived variables, sometimes referred to as aggregate variables, summarize the characteristics of the individuals in the group (Kawachi & Subramian 2006; Diez-Roux 1998). Examples include the percentage of the population in an area/neighborhood that is below a defined poverty level or the percentage of adult males in the area that is unemployed. This type of variable is used to describe group properties that are more than the sum of the individual characteristics. For example, Krieger eta/. (2003) demonstrated that a higher risk of multiple adverse birth outcomes in poorer neighborhoods is at least partly independent of large differences in 32

PAGE 50

maternal education at the individual level in urban areas of Canada. The second type of contextual variable, called an integral variable, describes a characteristic of an area that is not related to the aggregate characteristics of the individuals. Examples include the number of health clinics or the type of regulation, such as zoning, in an area. If variation is sufficient among areas, multilevel analysis can identify contextual predictors of outcomes using both individual and neighborhood variables (Kawachi & Subramanian 2006). It is an appropriate methodology when observations are clustered, exposures operate simultaneously at more than one level, or heterogeneity in exposure exists (Raudenbush & Bryk 2002). Various rather complex statistical methods permit simultaneous estimation of variability on the outcome at both levels and correction for any intraclass correlation at the area level (Sellstrom & Bremberg 2006). Some studies have linked individual level birth outcomes with area effects (Finch et a/. 2007; Gorman 2005; Pickett eta/. 2005; O'Campo et a/.1997). Most multilevel studies in the U.S. examine particularly poor health outcomes of Blacks to try to identify how much place of residence may contribute to their poor outcomes (Pickett eta/. 2005; O'Campo eta/. 1997). Fewer studies have applied area level analysis to study Hispanics or why they might enjoy better outcomes than would be expected based on their socioeconomic position (but see Finch eta/. 2007). In lieu of undertaking the complexity of simultaneous estimation of the contribution of individual and contextual level factors to the odds of the four birth outcomes, this study examines individual level characteristics in Aims 1 and 2 and area level characteristics in Aim 3, and interprets results at each level. Sample, Study Data, and Variables The CDPHE provided an historical observational dataset consisting of a census of de-identified records of all singleton live births occurring in Colorado for the years 2000 through 2005 on a CD-ROM in SAS format, pursuant to a written data agreement and after approval of the study by the Human Subjects Research Committee of the University of Colorado Denver as exempt under 45 CFR 46.101 (b)(4) and later as non-human subject research (Appendix B). Although it is possible that some births are not recorded in the vital records system, CDPHE states that its records account for over 99% of all births occurring in Colorado (Bol, K. personal interview, June 10, 2007). The sample, then, is a census, and need not be randomized (Stokes, Davis & Koch 2000:7-8). The dataset provides information on each birth by year, but omits the names of the infant and parents, the specific day and month of birth, and the date of mother's birth to 33

PAGE 51

preserve anonymity. The birth record includes demographic information and limited data on socioeconomic position of the parents, information on the weight and gestation of the infant, medical conditions associated with the pregnancy and delivery, and data reflecting certain maternal health behaviors. The census tract of the mother's residence and the specific altitude in feet above sea level of her address immediately before the birth are also associated with the birth record. CDPHE calculates and reports the adequacy of prenatal care using the Adequacy of Prenatal Care Utilization Index (APNCU), also known as the Kotelchuck Index, and flags births of infants weighing less than 2500 grams. Dependent Variables Four birth outcomes are the dependent variables for this study: LBW, preterm birth, SGA, and LGA. LBW is defined as an infant weighing less than 2500 grams at birth (approximately 5.5 pounds) (WHO 2006:36). Although CDPHE provided a flag for births <2500 grams, CDPHE's identification of LBW births was tested by writing SAS code to identify each birth <2500 grams and comparing those results to the flags provided by CDPHE. There were no differences between the two methods of identifying LBW infants. The Institute of Medicine's recent report on preterm birth considers a birth to be preterm if it occurs before 37 completed weeks of gestation (10M 2006). Each birth was coded as preterm or not based on the clinical estimate of gestation (in completed weeks) reported in the birth record. SGA is generally defined as birth weight of an infant below the 1oth percentile of weight for gestational age (Kiiegman & Das 2002; Alexander et a/. 1ggg). LGA is defined as birth weight of an infant above the goth percentile of weight for gestational age (Alexander eta/. 1ggg). There is no single national standard to determine percentiles of birth weight for gestation. This study uses the percentiles and associated birth weights, in grams, developed by Alexander et a/. (1ggg), which were based on a sample of g.6 million births to mothers living in the U.S. contained in the 1g94 1g96 U.S. Natality Files from the National Center of Health Statistics. The sample was racially/ethnically diverse, with approximately 61% non Hispanic Whites, 18% Hispanics, 15% Blacks, and the balance representing other raciaVethnic groups. Alexander's team calculated a percentile table for the total sample and separate tables for various racial/ethnic categories, including Hispanics. Using these tables, the 1oth and goth percentiles were calculated for each gestational age in weeks. Births below the 1 Olh percentile by gestation were coded as SGA and those above the goth percentile were coded as LGA. In addition, births that were neither SGA nor LGA were coded as appropriate for gestational age (AGA). For Aim 1 (births by race/ethnicity) percentile weights based on 34

PAGE 52

Alexander's entire, national sample population are used. For Aim 2 (births to mothers of Mexican origin categorized by mother's nativity) Alexander's percentile table for Hispanics is used. Independent Variables for Aims 1 and 2 The individual-level variables initially considered for inclusion in the analysis of Aims 1 and 2 are shown in Table 3.1. Also reported are the minimum and maximum values in the dataset and notes on their coding. CDPHE provided mother's age, previous pregnancies, altitude of residence, years of education, weight gain during pregnancy, cigarettes smoked per day and alcoholic drinks consumed per week during pregnancy, birth weight in grams, and clinical estimate of gestation as continuous variables. As described below, continuous variables were coded into categorical levels. For any categorical variable with more than two values, multiple levels were created. Table 3.1. Individual-level variables for Alms 1 and 2 Variable Values Operatlonallzatlon Year of infant birth 2000-2005 All 6 years = study cohort Age of mother 11-53 years :S19 20-34, Parity (calculated) Kleinman-Kessel Index First birth, low parity, high parity Altitude of mother's residence 3,375 11,893 feet above sea <5,000 ft, 1,000 ft intervals, level > 9000ft Education of mother 0-28 years <9, 9-11, Marital status Legitimate/illegitimate Married/unmarried Prenatal care (APNCU) Kotelchuck Index Inadequate, intermediate, adeauate, adeQuate plus Weight gain None/loss -95 lbs :S15, 16-40, Smoking during pregnancy 0-60 cigarettes per week Smoker or not Drinking alcohol during pregnancy 0-85 drinks per week Drinker or not Medical risk status 24 specific risks None or any one or more listed risks Race/ethnicity of mother Non-Hispanic White, Hispanic White, Hispanic, Black, Other White, Black and Hispanic Black, All Others Mother's nativity (Mexican origin) U.S.or Mexican-born U.S.-bom or Mexican-born Birth weight 113 -6492 arams Continuous Low birth weight (calculated) <2500 grams LBW or not Estimate of gestation -44 completed weeks Continuous Preterm birth (calculated) <37 completed weeks Premature or not Small for gestational age (SGA) Birth weight in grams <10'" SGA or not (calculated) I percentile Large for gestational age (LGA) Birth weight in grams >90m LGA or not (calculated) [percentile Appropriate for gestational age Birth weight in grams from 1 0'" AGA or not (AGA) (calculated) through golh percentile 35

PAGE 53

Causes of all four adverse birth outcomes in this study are multi-factorial and include demographic, social, behavioral, and medical risk factors (Ahlsson et a/. 2007; March of Dimes 2007; 10M 2006; Ehrenberg eta/. 2004; Kramer 1987; 10M 1985). Demographic and Socioeconomic Risk Factors Age of mother at the time of the birth of her infant has been shown to affect the likelihood of all four birth outcomes (Gould eta/. 2003; Singh & Yu 1996; 10M 1985). Various age categories are used in studies of pregnancy outcomes; in this study mother's age is categorized into "teen mothers" who are 19 years of age and younger, mothers who are ages 20-34, and "older mothers" who are age 35 and older. Teen mothers have higher risks of adverse reproductive outcomes (Fraser et a/. 1995), as do mothers over age 34 (Hansen 1986). The reference category for maternal age is 20 34 years old. Parity of the mother at the time of birth is another demographic risk factor. In general, first and low parity births are associated with LBW and SGA; higher parity is associated with greater weight gain during pregnancy and LGA (Olsen et a/. 2007). "Parity" is a term that is defined inconsistently in obstetrics practice and studies of birth outcomes (Beebe 2005). Parity is properly defined as the number of pregnancies completed past 20 weeks of gestation (not the number of living infants delivered). "Gravidy" is the number of pregnancies conceived, regardless of outcome (Beebe 2005). The data on previous pregnancies in Colorado's birth record are categorized as "previous pregnancies now living," "previous pregnancies now dead," and "other terminations." The intent of Colorado's previous pregnancy categories is to identify pregnancies completed past 20 weeks (whether issue are now living or dead) and to collect data on pregnancies ending before 20 weeks of gestation in "other terminations." However, the instructions for recording the information for these categories focus not on previous pregnancies but rather the number of infants. Therefore, it is possible that the categories used by CDPHE fail adequately to distinguish either viability or results of multiple gestations. For this study, parity is operationalized by summing the number of "previous pregnancies now living" and "previous pregnancies now dead" and using that number to form a base number for parity. Although this formulation is imperfect, it is a reasonable approach given the ambiguity in the collection of the data for the birth record in Colorado. From there, the variable for parity in this study is calculated using the Kleinman-Kessel index, which combines birth order and maternal age to account for the interaction effect of age and parity on outcomes, without producing collinearity between parity and age (Frisbie et a/. 1998; Kleinman & Kessel 1987). The Kleinman-Kessel Index 36

PAGE 54

categorizes parity into ''first births" at any age; "low parity," which is a second birth to mothers and up to a third birth to mothers 9000 feet. Residence before delivery at <5,000 feet is the reference category. Social factors that are related to adverse birth outcomes and that are recorded in the birth register are education of the mother and father, marital status, and adequacy of prenatal care. Education of the mother is categorized into three categories: less than high school education, some high school (grades 9-11 ), and greater than high school, because Mexican born immigrants often have only an elementary school education. Although there is often an association between more education and better health outcomes (Desai & Alva 1998; Link & Phelan 1995), some studies report contrary findings, where mothers of certain races/ethnicities with less education have better birth outcomes than mothers with more education (Acevedo-Garcia et a/. 2005; Gould et a/. 2003). More than 9.85% of cases are missing father's education in the dataset for Aim 1 and 13.03% of cases are missing father's education in the dataset for Aim 2. Accordingly, father's educational level is not included in the analysis. The reference category for mother's education is high school graduate and above. Marital status also contributes to birth outcomes. Being married or in a committed relationship is usually related to better birth outcomes (10M 1985), perhaps because it is a vehicle for access to more resources or social support (Williams et al. 2008; Schoenborn 37

PAGE 55

2004). Colorado records marital status as "legitimate" or "illegitimate" on the birth record, based on interview with the mother, and it includes a committed relationship in the category of "legitimate." Being married (legitimate) is the reference category for marital status. Although there is some suggestion that adequacy of prenatal care, as measured by various indices, may not correlate with birth outcomes (Alexander & Kotelchuck 2001; Fiscella 1995), U.S. and international policy encourages improvements in access to prenatal care (10M 2003). The Adequacy of Prenatal Care Utilization (APNCU), also known as the Kotelchuck Index, takes into account the length of gestation with the number of prenatal visits, thereby correcting for preterm births and distinguishing women who have more prenatal visits than the number recommended by the American College of Obstetricians and Gynecologists (and who likely are higher risk patients) (Kotelchuck 1994). Four categories are used to rate adequacy of care: inadequate, intermediate, adequate, and adequate plus. CDPHE calculates and provides data on the adequacy of prenatal care using the Kotelchuck Index. Adequate prenatal care is the reference group. Medical Risk Factors The birth record indicates whether the mother had certain medical risks during pregnancy and delivery. The following risk factors are associated with one or more of the birth outcomes of the study: anemia (hematocrit <30%/hemoglobin <1 Og/dL); cardiac disease; acute or chronic lung disease; gestational diabetes; pre-existing diabetes; genital herpes; hydramnios/oligohydramnios; hemoglobinopathy; chronic hypertension; pregnancy associated hypertension; eclampsia; incompetent cervix; previous infant weighing 4000+ grams at birth; previous preterm or SGA infant; renal disease; Rh sensitization; uterine bleeding; other risk factors; premature rupture of the membranes (>12 hours); abruptio placenta; placenta previa; other excessive bleeding; and seizures during labor. Medical risk status is treated as dichotomous: any one or more of the medical risks or none. The reference group is no medical risks. Three medical risks are highly associated with LGA: pre-existing diabetes; gestational diabetes; and having given birth to a previous infant weighing more than 4000 grams. This set of variables is examined separately from the more general medical risks associated with pregnancy for LGA. 38

PAGE 56

Behavioral Risk Factors Inadequate weight gain of the mother during pregnancy is associated with LBW, SGA, and sometimes preterm birth (10M 2006, 1985). Higher than recommended weight gain is associated with LGA (Dyer eta/. 2007). The Institute of Medicine adopted weight gain guidelines in 1990 based on BMI, but only weight gain in pounds is recorded in the Colorado birth record during the study period. For this study weight gain during pregnancy is coded into three categories: up to and including 15 pounds, 16 to 40 pounds, and more than 41 pounds, following studies by Frisbie and Song (2003). The Institute of Medicine is in the process of adopting new guidelines that will likely re-emphasize the importance of maternal BMI before pregnancy and recommend lower weight gain as a result of the trend to more overweight and obese mothers and concerns about the association between excessive weight gain and insulin overproduction in LGA infants (Barbour, L., personal interview, November 10, 2008). Weight gain between 16 and 40 pounds is the reference category. The effects of smoking and drinking on LBW and premature birth are well-recognized (10M 1985). The birth record includes the mother's report of the number of cigarettes smoked per day and number of alcoholic drinks taken per week during pregnancy. Smoking and drinking are each categorized using a dichotomous measure of smoking or not; drinking or not, with not smoking and not drinking as the reference categories. CDPHE collects data on smoking and drinking behaviors during pregnancy in two ways: from each mother's self report, which is recorded in the official birth register, and by survey using the Pregnancy Risk Assessment Monitoring System (PRAMS), a surveillance project of the CDC and state health departments (2006d). In 2004 and 2005 CDPHE identified discordance in reporting smoking and drinking between the two sources. For both years the birth record reported that about 1% of women drink alcoholic beverages and about 11% smoke during pregnancy. PRAMS survey reports for the same years estimated about 1 0% of women drank during pregnancy and 13% smoked. Using multivariate analysis, CDPHE tested eight predictive factors of discordance and, among other things, determined that Hispanic women were less likely than White women to report discordantly for either smoking or drinking, and Hispanic women report less smoking and drinking than Whites. For these reasons and because Hispanic ethnicity is the most important racial/ethnic category of interest, the data from the birth register on smoking and drinking are used in this study (Drisko, J., personal interview, October 13, 2007). In any event, using the lower numbers from the birth record tend to bias 39

PAGE 57

results to the null. If smoking or drinking are influential factors in any of the outcomes, their effects will be understated. Race!Ethnicity!Nativity Aim 1 compares Colorado's population of mothers using her identification of "race/ethnicity" to capture the significant influence of Colorado's Hispanic population. Mothers identify their "race" for the birth record. CDPHE uses the following categories to classify race: o White (includes Mexican, Cajun, Creole, Puerto Rican and all other Caucasian) o Black (Negro, Colored, Afro-American) o Indian (North American, Central American, South American, Alaskan, Canadian) o Chinese o Japanese o Hawaiian (includes part-Hawaiian) o Filipino o Other Asian or Pacific Islanders (Korean, Thai, Amerasian, Vietnamese, Chamorro) o Other non-white o Unknown, not stated, or not classifiable Mothers also identify their origin, which CDPHE records using the following categories: o Non-Hispanic o Mexican o Puerto Rican o Cuban o Central or South American o Other and unknown Hispanic Following Gonzalez-Quintero et a/. (2007), non-Hispanic women are classified by race and Hispanics are classified by origin. For this study, non-Hispanic White ('White") mothers are those who identify for the birth record as 'White" and who do not also identify Mexican, Puerto Rican, Cuban, Central or South American, or other and unknown Hispanic origin. "Hispanic" mothers are those who identify 'White" race and who also identify one of the .previously mentioned "Hispanic" categories. Black mothers are those who identify "Black" race and who also identify either "Hispanic" or "non-Hispanic." Only 360 mothers out of the total study population of 356,389 identified both "Black" and "Hispanic;" accordingly 40

PAGE 58

most mothers in the "Black" category are "Black non-Hispanic." Mothers who identify as Black and Hispanic could be analyzed in either the Black or Hispanic category. They are grouped with the "Black" category in this study because of their small number and based on a study that found that Black/Hispanics first identified themselves as Black, and identified Hispanic ethnicity only when the interviewer inquired directly about ethnicity, perhaps suggesting that Hispanic heritage was not as important to their identity as race (Baker eta/. 2006}. Mothers of any other race are categorized as "Other." For Aim 1, White mothers constitute the reference group. The birth record also reports the mother's nativity. The choice for state of birth includes each of the fifty United States and the District of Columbia, Puerto Rico, Guam, Canada, Mexico, a number of additional countries, and "remainder of the world." For Aim 2, mothers of Mexican origin are identified as Mexican-born or U.S.-born (born in any of the United States or District of Columbia}. No mothers of Mexican origin were born in either Puerto Rico or Guam. U.S.-born mothers of Mexican origin constitute the reference group for Aim2. Missing Data and Size of Study Population for Aim 1 The dataset of singleton births occurring in Colorado for the years 2000 2005 consists of 392,881 births. According to CDPHE, the birth record represents only "live births." Yet within the dataset are records of births of very low birth weight and very short gestation. CDPHE conducts quality control of the birth records in two ways when data appear to be outside of normal value ranges. It checks for data entry errors and it queries the hospitaVIocation of birth to inquire whether the data are correct as received (Bol, K., personal interview, June 10, 2007}. After quality control measures are exhausted, CDPHE retains records that nevertheless appear to be outside the range of viability. One reason for these anomalous records may be some parents' insistence on issuance of a live birth certificate (and subsequent death certificate} when a child is stillborn. From this dataset of 392,881 records, 254 "presumed non-viable" births that are recorded as weighing less than 400 grams and having completed fewer than 23 weeks of gestation were removed. These weight and gestation values follow the recommended cut points for consideration of non-initiation of resuscitation at birth, and represent the most widely circulated resuscitation guideline in the U.S. (American Academy of Pediatrics 2005:e1-10}. The removal of these presumed non viable births preserves births at very low birth weights that are extremely growth restricted, but which have longer gestation and could have been born alive. The distribution of births by 41

PAGE 59

race/ethnicity of the resulting dataset of 392,627 is shown in Table 3.2. Thirty-three records were missing any racial/ethnic identifier. Table 3.2. Original dataset by racalethnlcity Variable Race/Ethnicity Total 392,627 (100.00%) White Hispanic Black 239,921 118,130 17,213 (61.11%) (30.09%) (4.38%) Other Missing 17,330 33 (4.41%) (0.01%) Cases with any missing value for any of the variables of interest were successively deleted. Table 3.3 reports the number and percentage of cases missing any variable of interest. Table 3.3. Number and percentage of missing cases by variable for Aim 1 Variable Mother's altitude of residence Weight gain Adequacy of prenatal care Mother's education Smoking Drinking Parity Birth weight Estimated gestation Mother's age Mother's race Marital status Medical risk Total Number of Cases 13,414 11,374 9,232 4,520 1,785 1,076 346 67 66 54 33 0 0 41,940 % of Cases Missing 3.42% 2.90% 2.35% 1.15% 0.45% 0.27% 0.09% 0.02% 0.02% 0.01% 0.01% 0.00% 0.00% 10.68% After deleting cases with any missing variable of interest for Aim 1, the resulting dataset is 356,389 births (cases deleted equal 36,238). The number of cases deleted is lower than the total number of cases missing any variable shown in Table 3.3 because some cases were missing more than one variable. A total of 9.2% of cases were deleted from the original dataset. When availability of data is an exclusion criterion, the frequency of missing data for each variable is considered. If the amount of "overall missingness" is <10%, then the dataset is generally considered sufficient, even when differences between independent variables are significant (Burton & Altman 2004). For Aim 1, no single variable has more than 3.42% of 42

PAGE 60

missing values and the entire dataset is missing fewer than 10% of cases after exclusion of cases missing any variable of interest. The distribution of missing data by race/ethnicity in Table 3.4 shows that there are differential percentages of missing data by race/ethnicity; in particular, there are higher percentages of missing data among Hispanics relative to the percentage of births of Hispanics statewide (30.09%) for weight gain, education, and parity. To further address the question of whether and how much missing data might bias results, regressions were run using the original dataset with "Missing" shown as a separate category and using the dataset after deleting cases with missing variables. Comparisons of the odds ratios for each outcome, based on the fully adjusted model described at page 52 are shown in Figures 3.2 3.5. Odds ratios for each outcome are shown using the complete dataset followed by the odds using the dataset with missing cases removed for each outcome. For the low-weight associated outcomes, the largest drift occurs in the "Others" category; for Whites, Hispanics, and Blacks, the drift is small, especially for Hispanics, the main race/ethnicity of interest. For LGA, there is very little drift for any population. Accordingly, the dataset with missing cases deleted is appropriate for analysis. Table 3.4. Number and percentage of missing variables by racelethnlclty Total Race Variable 41,940 White Hispanic Black Other Missing {%} {%} {%} {%} {"'o} Altitude 13,414 8,993 (67.04) 3,556 (26.51) 356 (2.65) 498 (3.71) 11 (0.08) Weight gain 11,374 5,483 (48.21) 4,692 (41.25} 578 (5.08) 603 (5.30) 18 (0.16) Prenatal care 9,232 5,526 (59.86) 2,469 (26.74) 668 (7.24) 549 (5.95) 20 (0.22) Education 4,520 1 ,809 (40.02) 2,085 (46.13} 315 (6.97) 283 (6.26) 28 (0.62) Smoking 1,785 1,085 (60.78) 517 (28.96) 90 (5.04) 77 (4.31) 16 (0.90) Alcohol 1,076 621 (57.19) 325 (30.20) 69 (6.41) 44 (4.09) 17 (1.58) Parity 346 137 (39.60) 128 (36.99} 39 (11.27) 24 (6.94) 18 (5.20) Birth weight 67 33 (49.25) 18 (26.87) 6 (8.96) 4 (5.97) 6 (8.96) Gestation 66 55 (83.33) 4 (6.06) 0 (0.00) 1 (1.52) 6 (9.09) Age 54 21 (38.89) 16 (29.63) 2 (3.70) 1 (1.85) 14 (25.93) Race 33 NIA NIA N/A N/A NIA 43

PAGE 61

25 2 .11 2.16 2.25 2 _.,__Vftte VI 1.64 Q 175 -------l-lispfric iii 15 Bla:k a: 1.18 1.18 1.25 -a ... VI 0 075 05 0.25 LBWAIIData LBWSt udy Data Figure 3.2. AIM 1 LBW DRIFT OF ODDS 225 1.96 1.98 VI 1.75 .Q 1 5 iii 1.69 1.72 1.18 1.18 --+--W'Iite --l-liSpariC a: 125 Bla:k VI -atw , 075 0 05 025 SGA All Data SGA Study Data Figure 3.4. AIM 1 SGA DRIFT OF ODDS 25 225 VI 175 .Q iii 1 5 1.4 1 .42 a: 125 VI , Black .11 ..1.16 1.02 1 .01 -a"" 0 075 05 025 PrelermAII Data Preterm Study Data Figure 3.3. AIM 1 PRETERM DRIFT OF ODDS 25 2.25 15 a: 1.25 rn 1 0 075 0.5 025 r-0 95 0 68 n LGAAII Data 0.68 0 95 0 59 LGA Study Data Black ---M--Ol her Figure 3.5. AIM 1 LGA DRIFT OF ODDS Table 3.5 describes the final study population by race/ethnicity for Aim 1. Table 3.5. Final study population by race/ethnicity of mother-Aim 1 Study Population 356,389 Study Population by Race/Ethnicity White 219,029 (61.46%) Hispanic 106,291 (29.82%) Black 15,448 (4.33%) Missing Data and Size of Study Population for Aim 2 Other 15,621 (4.38%) Aim 2 replicates Aim 1, except that it compares birth outcomes of mothers of Mexican origin by nativity. Table 3.6 displays the original statewide dataset by nativity. Two hundred seven cases are missing mother's place of nativity. 44

PAGE 62

Table 3.6. Original dataset by place of nativity of mother Study Population 95,291 Study Population by Nativity U.S.-Born Mexican Origin 35,357 (37.1 0%) Mexican-Born Missing 59,727 (62.68%) 207 (0.22%) As with Aim 1, cases with any missing value for any of the variables of interest were successively deleted. Table 3.7 reports the number and percentage of cases missing any variable of interest from the dataset for Aim 2. Table 3.7. Number and percentage of missing cases by variable for Aim 2 Variable Weight gain Mother's altitude of residence Adequacy of prenatal care Mother's education Smoking Alcohol Parity Mother's age Birth weight Estimated gestation Marital status Medical risk Total Number of Cases % of Cases Missing Missing Variable 3,716 2,760 1,982 1,779 370 249 105 12 7 1 0 0 10,981 Variable 3.90% 2.90% 2.08% 1.87% 0.39% 0.26% 0.11% 0.01% 0.01% 0.00% 0.00% 0.00% 11.53% After deleting cases with any missing variable of interest for Aim 2, the resulting dataset consists of 85,755 births (cases deleted equal 9,536). The range of missing variables is 0-3.9%. As with Aim 1, the number of cases deleted is lower than shown in Table 3.7 because some cases had more than one missing variable. A total of 10.01% of cases were deleted from the original dataset of mothers of Mexican origin. To determine whether the missing data might affect the results of the analysis, the distribution of missing data by nativity is reported in Table 3.8. As can be seen, there are differential percentages of missing data by nativity; in particular, there are higher percentages of missing data among Mexican-born Hispanics relative to their percentage of births statewide (62.68%) for weight gain, altitude of residence, education, parity, and gestation. 45

PAGE 63

Table 3.8. Number and percentage of missing variables by nativity U.S.-Bom Mexican Nativity Variable Total Mexican Origin Born Missing 10,981 (%} (%} (%} Weight gain 3716 826 (22.23) 2,869 (77.21) 21 (0.57) Altitude 2760 949 (34.38) 1794 (65.00) 17 (0.62) Prenatal care 1982 785 (39.61) 1181 (59.59) 16 (0.81) Education 1,779 421 (23.66) 1325 (74.48) 33 (1.85) Smoking 370 204 (55.14) 162 (43.78) 4 (1.08) Alcohol 249 104 (41.77) 143 (57.43) 2 (0.80) Parity 105 21 (20.00) 81 (77.14) 3 (2.86) Age 12 2 (50.00) 6 (16.67) 4 (33.33) Birth weight 7 3 (43.86) 4(57.14) 0 (0.00) Gestation 1 0 (0.00) 1 (100.00) 0 (0.00) Nativity 207 N/A N/A N/A As with the dataset for Aim 1 to further address the question of whether and how much missing data may bias results, regressions of the fully adjusted models were run using the original dataset with "Missing" shown as a separate category and using the dataset with missing cases deleted. Figures 3.6 3.9 display the drift of odds ratios between the full dataset and the study dataset after removing cases with missing variables of interest. For LBW and preterm birth, odds for Mexican-born mothers are not significantly different from the odds for U.S.-born mothers. Given the small or non-existent drift in odds ratios for each outcome, the dataset without cases having missing data is used for analysis. 15 I 1 ---U.S.Borr t ........ Mexican 0.11 0.13 Born 1.5 1.: 1.ot 1.118 ...... .. ---U.S.Borr 1 1 MexicanBorn 0 0 Prater mAll Preterm Study LBWAIIData LBWStudy Data Data Data Figure 3.6. AIM 2 LBW DRIFT OF ODDS Figure 3.7. AIM 2 PRETERM DRIFT OF ODDS 46

PAGE 64

1.5 -1::.. ... 1.46 --+--U.S.Born 1.: 1 1 --+--U.S. Born \75 ...... ..,.74 Mexican Born 1.: .... 1 1 Mexican-Born 0 0 SGAAIIData SGAStudyData LGAAIIData LGAStudy Data Figure 3.8. AIM 2 SGA DRIFT OF ODDS Figure 3.9. AIM 2 LGA DRIFT OF ODDS The resulting study population for Aim 2 is a total of 85,755 births: 32,484 (37.88%) U.S.-born mothers of Mexican origin (born in any of the 50 states or the District of Columbia) and 53,271 (62.12%) Mexican-born mothers. Table 3.9 reports births for Aim 2 by year. The percentage of births by nativity is quite consistent throughout the study period. Table 3.9. Final study population by nativity of mothers of Mexican origin by year Aim 2 Study Population All Years 85,755 2000-2005 2000 2001 2002 2003 2004 2005 U.S.-Bom Mexican Origin 32,484 (37.88%) 4,606 (38.78%) 5,125 (37.57%) 5,553 (37.16%) 5,612 (37.43%) 5,753 (37.64%) 5,835 (38.86%) Contextual Variables for Aim 3 Mexican-Born 53,271 (62.12%) 7,271 (61.22%) 8,518 (62.43%) 9,389 (62.84%) 9,382 (62.57%) 9,530 (62.36%) 9,181 (61.14%) To examine the effect of neighborhood composition on birth outcomes, two scales used by Finch et a/. (2007), were calculated from information collected from the 2000 Decennial Census (Summary File 3). The Scale of Immigrant Orientation consists of: o % Mexican-born individuals living in the tract; calculated by dividing the number of Mexican-born individuals from Census Table PCT20 divided by the population in the tract o %non-citizens born in Mexico living in the tract; calculated by summing the numbers of individuals who are not citizens based on year of entry from Census Table PCT20 divided by the population in the tract 47

PAGE 65

o % linguistically isolated households speaking Spanish; calculated by dividing the number of linguistically isolated Spanish households from Census Table P20 by the number of households in the tract. A linguistically isolated household is one in which all members of the household 14 years old and over speak Spanish and have at least some difficulty with English. The Scale of Neighborhood Deprivation consists of: o % individuals living in the tract in poverty, calculated by dividing the number of individuals in poverty in 1999 in the tract by the number of individuals from whom poverty status is determined in the tract from Census Table P87 o % of households receiving public assistance income, calculated by dividing by the number of households receiving public assistance income in 1999 by the number of households in the tract from Census Table P64 o % female headed family households, calculated by dividing the number of female householders by the number of households in the tract from Census Table P9 o %males unemployed in the civilian work force reported in Census Table QT-P24 (no calculation necessary). Neither scale was normally distributed, so each was transformed. Neighborhood deprivation was transformed by taking the square root of the index value, following the practice of Finch et a/. (2007) and O'Campo et a/. (1997), which resulted in a normal distribution. Immigrant orientation was transformed by taking the natural log of the value, as that produced a better normal distribution than square root transformation. The validity of each scale was tested using Cronbach's Alpha statistic (Cronbach 1951 ). The closer the Cronbach score is to 1.0, the more reliable the generated scale is. Nunnaly (1978) indicates 0.7 to be an acceptable reliability coefficient, but lower thresholds are sometimes reported in the literature. Cronbach's alpha ranges from 0.7250.970 for all scales except Denver's neighborhood deprivation scale, which approaches 0.7 (value 0.674). Table 3.11 reports the characteristics of each index by county. Because the items in each scale are correlated, the items for each scale are summed and then averaged to create the scale values used in linear modeling. 48

PAGE 66

Table 3.1 0. Values of contextual scales for Adams and Denver counties 2000 Mean (SD) Minimum Maximum Cronbach's Alpha Adams Neighborhood Deprivation 7.99 .62 2.41 17.39 .725 Denver Neighborhood Deprivation 1 0. 78 .92 0.00 47.17 .674 Adams Immigrant Orientation 6.31 .09 0.00 34.90 .970 Missing Data and Size of Population for Aim 3 Denver Immigrant Orientation 10.97 .85 0.00 40.63 .931 Aim 3 analyzes the effects of neighborhood deprivation and immigrant orientation on birth outcomes in two of Colorado's largest counties, Adams and Denver Counties. These two counties were chosen because they have both a large percentage and number of residents of Mexican origin. Adams County, with 86 census tracts, has a mixed urban/rural composition; Denver, with 136 tracts, is urban. Urban/rural characterization is based on Summary File 3, Table H5 of the 2000 Census, which counts housing units as "urban" if they are inside urbanized areas or urban clusters or "rural," whether farm or non-farm. In 2000, 31.7% of Denver County's population identified as Hispanic/Latina; 28.2% of Adams County identified as Hispanic/Latina (Census 2000b, Table GTC-P6). CDPHE reports that the proportion of linguistically isolated households speaking Spanish increased by 171% statewide between 1990 and 2000, based on census data. The proportion of linguistically isolated households in Adams County increased by 416% and by 158% in Denver County during this same time period (CDPHE 2005). Most critics of contextual studies lament the necessity of using administrative subdivisions, such as census definitions, for area-level descriptions of "neighborhood" (Diez Roux 1998, 2000). Unfortunately, the use of secondary data from the birth record, which provides census tract information as the smallest geographic area unit, makes this problem unavoidable. The Census Bureau creates census tracts that consist of 1 ,500 8,000 (optimum 4,000) residents. Tracts are intended to represent relatively homogenous areas that conform to local perceptions of neighborhood (Lee & Marlay 2007). It has been shown that measures at the census tract level perform about equally to those at the smatter census block group level (Krieger et a/. 2003). Therefore, consistent with other population studies that examine area factors (O'Campo eta/. 1997; Finch eta/. 2007), this study uses census tracts as a proxy for neighborhoods. 49

PAGE 67

Subramanian eta/. recommend that at least twenty observations per geographic area are needed for area-level analysis (2003:105). One tract in Adams County had no births during the study period and it was therefore deleted. Six tracts in Adams County had fewer than twenty births. These were combined with nearby tracts by matching the quartile of neighborhood deprivation and the quartile of immigrant orientation to an adjacent or close tract, resulting in 79 tracts for analysis. In Denver County, one tract also had no reported births and it was deleted. Thirty-six tracts in Denver County had fewer than 20 births. These tracts were deleted from the analysis because it was not possible to combine them by matching on both neighborhood deprivation and immigrant orientation with an adjacent or nearby tract. The analysis of Denver County therefore includes 1 00 tracts. Unlike the multiple logistic regression modeling used in Aims 1 and 2, general linear modeling requires only the mother's tract of residence, nativity and outcome. Cases missing census tract or nativity were deleted for Aim 3. For Adams County, 435 cases (2.63%) were removed from the dataset. For Denver County, a total of 325 cases in 36 tracts were removed because matching was not possible within the a priori rules established. An additional171 cases were deleted because they were missing tract or nativity (0.7% of cases in the remaining 100 tracts). The final study populations for Adams and Denver Counties are described in Table 3.11. Table 3.11. Population of mothers of Mexican origin by nativity In Adams and Denver Counties 2000-2005 County Adams N = 16,107 Denver N = 23,332 U.S.-Bom Mexican Origin 5,733 (35.59%) 4,914 (21.06%) Mexican-Born 10,374 (64.41%) 18,418 (78.94%) Methods for Aims 1 and 2 Primary Character of County Mixed Urban The analytical tools for Aims 1 and 2 are univariate and bivariate descriptive statistics and multiple logistic regression. Frequencies of risk factors and birth outcomes describe mothers by race/ethnicity and nativity. Multiple logistic regression is used to predict each dependent birth outcome. As discussed in Chapter 2, according to the social gradient of health Hispanics and Blacks are expected to have demographic/social, medical, and health behavior risk profiles that are less advantageous than those of Whites, and therefore to have worse birth 50

PAGE 68

outcomes. U.S.-born mothers of Mexican origin are expected to have better birth outcomes than Mexican-born mothers based on the same social gradient of health. The null and alternative hypotheses for each risk factor are: H10 : each risk factor is independent of race/ethnicity/nativity H1A: each risk factor is related to race/ethnicity/nativity. The null and alternate hypotheses for incidence of birth outcomes without adjustment for risk factors are: H20 : each birth outcome is independent of race/ethnicity/nativity H2A: each birth outcome is related to race/ethnicity/nativity Pearson's chi-square is used to test the relationships between risk factors and race/ethnicity or nativity and each birth outcome and race/ethnicity or nativity. Unadjusted odds of any given birth outcome do not adequately account for individual level risk factors. Therefore, successive adjustments are made to examine the relationship between outcomes and race/ethnicity and nativity. H30 : race/ethnicity/nativity does not predict each birth outcome after adjusting for risk factors. H3A: race/ethnicity/nativity predicts each birth outcome after adjusting for risk factors. Model Building for Aims 1 and 2 Four nested logistic regression models are constructed for each outcome in Aims 1 and 2. Based on the existing literature and the availability of data from the birth record, the variables listed in Table 3.1 were selected as candidate variables. Model 1 tests the main effects of race/ethnicity or nativity in each aim. Each successive model adjusts for related variables as a block. Model 2 includes the main effects on each outcome from Model 1 and adjusts for demographic and social economic risk factors (age, parity of the birth, altitude of mother's residence immediately before the birth, marital status, level of mother's education, and adequacy of prenatal care). As described at pp. 36-39, reference categories are the lowest risk category of that variable. Model 2 tests the social gradient of health by examining whether there are differences in odds of a given outcome after adjustment for demographic and socioeconomic risk profiles. Model 3 includes the main effects and demographic and socioeconomic risk factors and adds the existence of one or more medical risks associated with pregnancy. For Aim 1 51

PAGE 69

this is a further test of the social gradient. For Aim 2, Model 3 tests the hypothesis that mothers born in Mexico are "healthier'' than mothers of Mexican origin born in the U.S. Although Model 3 cannot test the healthy migrant hypothesis directly, because it does not include data on birth outcomes for Mexican mothers who remain in Mexico, nevertheless it can assess the impact of pregnancy-associated medical risks on Mexican-born mothers. Model 4 includes the main effects, demographic and socioeconomic risk factors, medical risks, and adds health behaviors during pregnancy. For Aim 2, Model 4 tests the healthy immigrant hypothesis (whether Mexican-born mothers engage in healthier behaviors than mothers of Mexican origin who are born in the U.S.). To screen the candidate variables, univariate testing of outcomes and potential explanatory variables was conducted on the statewide datasets by race/ethnicity and by nativity. Main effects were tested by examining the relationship of each outcome variable with each explanatory variable using the Pearson chi square test. According to Hosmer and Lemeshow, variables with a p-value of less than .25 may be deleted (2000:86). This method may result in over-identification of potential variables, but further model testing techniques can then be used to eliminate marginal variables. No variables were deleted in this step for any outcome or aim. Finally, the relationship of each explanatory variable against each other explanatory variable was tested using the chi square measure of association to determine whether there might be any interactions among the explanatory variables. For Aim 1, each variable showed a statistically significant interaction with each other variable with a p-value <0.05. For Aim 2, each variable showed a statistically significant interaction with every other variable, except drinking* age. This screening provided additional information for further model building. For each outcome in Aims 1 and 2, each candidate variable from Table 3.1 was entered into the logistic regression using backward and forward selection. Variables that were not significant to the model were discarded, unless the literature suggested that the variable should be retained. Then, using interactive modeling, variables were entered into each model, one by one in order of significance, using forward selection modeling. Again, variables that were not significant to the model were discarded, unless the literature suggested that the variable should be retained. The significance of each variable added to the previous model was tested using the -2 log likelihood ( -2LL) test. Any variable that changed the -2LL score by an amount that was significant using Pearson's chi square of the difference between the -2LL statistic of each successive model was retained. Finally, each 52

PAGE 70

model was run with all of the model variables retained after testing, as a block, with forward selection of each interaction term. The significance of each interaction was tested using the -2LL test, and significant interactions were retained. Figure 3.10 depicts the steps in model building. Examination of Candidate Variables All variables significant at p=0.05 using chi square test; Each outcome each independent variable Q all variables retained Interactions All interactions significant at Each independent variable each other Q p=0.05 using chi square, except independent variable drinking*age, which is deleted Backward Selection of Candidate Variables Q Tested using chi square significance of difference of -2LL statistics Interactive Modelina Entry of Candidate Variables as block, Q Tested using chi square forward selection of each interaction significance of difference of fi -2LL statistics Final Model for Each Aim and Outcome Figure 3.10. SCHEMATIC OF MODEL BUILDING FOR AIMS 1 AND 2 The results of model building are described below by Aim and outcome. Aim 1 LBW. Drinking is not significant based on univariate tests, and is dropped. Five interactions are significant: medical risk*race, parity*marital status, prenatal care*weight gain, prenatal care*smoking, and smoking*age. o M1: Y (LBW) = a (intercept) + b,X, (race/ethnicity Hispanic) + (race/ethnicity Black)+ b3XJ (race/ethnicity Other)+ E (standard error) o M2: Y (LBW) = a + b,X, (race/ethnicity Hispanic) + (race/ethnicity Black) + baXa (race/ethnicity Other)+ b4'4(<20 years age)+ b5Xs(>34 years age)+ beXe (first birth) + b1X1 (high parity) + baXe (elevation 5000-5999) + bg)(g (elevation 6000-6999) + b,oX,o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b,2X12 (elevation >9000) + b,aX13 (school
PAGE 71

(unmarried)+ b,eX,s (inadequate pnc) + b11X11 (intermediate pnc) + b,aX,8 (adequate plus pnc) + b19-2aX19-23 (parity*marital status-5 levels)+ E o M3: Y (LBW) =a+ b,X, (race/ethnicity Hispanic)+ (race/ethnicity Black)+ b 3Xa (race/ethnicity Other)+ b4'4 (<20 years age)+ bsXs (>34 years age)+ bsXs (first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + bg)(g (elevation 6000-6999) + b1oX1o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b1aX1a (school 34 years age)+ bsXs (first birth) + bJX1 (high parity) + baXa (elevation 5000-5999) + bgXg (elevation 6000-6999) + b1oX1o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b1aX1a (school 34 years age) + bsXs (first birth) + b1X1 (high parity) + baXa (elevation 5000-5999) + bg)(g (elevation 60006999) + b,oX1o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b1aX1a (school 34 years age) + bsXs 54

PAGE 72

(first birth) + b1X1 (high parity) + b8Xe (elevation 5000-5999) + bg)(g (elevation 60006999) + b,oX,o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b,2X12 (elevation >9000) + b,aX,a (school 34 years age) + bsXa (first birth) + b1X1 (high parity) + b8Xa (elevation 5000-5999) + bg)(g (elevation 60006999) + b,oX,o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b,aX,a (school 34 years age)+ bsXa (first birth) + b?X1 (high parity) + b8Xa (elevation 5000-5999) + bgXg (elevation 6000-6999) + b,oX,o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b,2X12 (elevation >9000) + b,aX,a (school 34 years age)+ bsXs (first birth) + b?X1 (high parity) + b8Xa (elevation 5000-5999) + bg)(g (elevation 6000-6999) + b,oX,o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b,aX,a (school
PAGE 73

(adequate plus pnc) + b19-2sX19-2s (parity*marital status-7 levels) + b2sX2e (medical risk)+ b21-34X21-34 (race*medical risk-81evels) + E o M: Y (SGA) = a (intercept) + b,X, (race/ethnicity Hispanic) + b 2X2 (race/ethnicity Black)+ b3Xa (race/ethnicity Other)+ (<20 years age)+ bsXs (>34 years age)+ bsXe (first birth) + b?X1 (high parity) + beXe (elevation 5000-5999) + b 9 X 9 (elevation 6000-6999) + b,oX,o (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school 34 years age)+ beXe (first birth) + b1X1 (high parity) + baXe (elevation 5000-5999) + bgXg (elevation 6000-6999) + b,oX,o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school 34 years age)+ bsXe (first birth) + b?X1 (high parity) + baXe (elevation 5000-5999) + bgXg (elevation 6000-6999) + b,oX,o (elevation 7000-7999) + b,,x,, (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school 34 years age)+ beXe (first birth) + b1X1 (high parity) + baXe (elevation 5000-5999) + bgXg (elevation 6000-6999) + b,oX,o (elevation 7000-7999) + b11X11 (elevation 8000-8999) + b12X12 (elevation >9000) + b13X13 (school
PAGE 74

plus pnc) + b19X19 (medical risk) + (smoking) + (low weight gain) + (high weight gain) + E Appropriate for gestational age (babies born neither SGA nor LGA) was also tested for Aim 1 as a way to confirm the results of the more specific adverse outcomes. As with the other outcomes, drinking was not significant and was dropped from the models. There were two interactions: weight*parity and weight*marital status. Aim 2 LBW. Drinking is not significant based on univariate tests, and is dropped. Age and school do not add significantly to the model in backward selection, but as known contributors to LBW, they are retained. Only one interaction is significant: prenatal care*nativity. o M1: Y (LBW) =a (intercept)+ b,X, (Mexican-bam)+ E (standard error) o M2: Y (LBW) =a (intercept)+ b,X, (Mexican-bam) + (<20 years age) + bw (>34 years age)+ b4X.. (first birth)+ bsXs (high parity)+ bsXe (elevation 5000-5999) + b1X1 (elevation 6000-6999) + baXe (elevation 7000-7999) + bgXg (elevation 80008999) + b,oX1o (elevation >9000) + b11X11 (school 34 years age)+ b4X.. (first birth)+ bsXs (high parity)+ bsXe (elevation 5000-5999) + b1X1 (elevation 6000-6999) + baXe (elevation 7000-7999) + bgXg (elevation 80008999) + b,oX,o (elevation >9000) + b,,x,, (school 34 years age)+ b4)4 (first birth)+ bsXs (high parity)+ bsXe (elevation 5000-5999) + b1X1 (elevation 6000-6999) + baXe (elevation 7000-7999) + bgXg (elevation 80008999) + b1oX1o (elevation >9000) + b,,x,, (school
PAGE 75

is included because it is a main effect and a parent term for one of the interactions. Two interactions are significant: prenatal care*nativity and prenatal care*weight. o M1: Y (Preterm birth)= a (intercept)+ b,X, (Mexican-born)+ E (standard error) o M2: Y (Preterm birth)= a (intercept)+ b,X, (Mexican-born)+ (<20 years age)+ (>34 years age) + (first birth) + bsXs (high parity) + bsXe (elevation 50005999) + b1X1 (elevation 6000-6999) + beXe (elevation 7000-7999) + bg)(g (elevation 8000-8999) + b,oX,o (elevation >9000) + b,,x,, (school 34 years age) + (first birth) + bsXs (high parity) + bsXe (elevation 50005999) + b1X1 (elevation 6000-6999) + beXe (elevation 7000-7999) + bg)(g (elevation 8000-8999) + b,oX,o (elevation >9000) + b,,x,, (school 34 years age) + (first birth) + bsXs (high parity) + bsXe (elevation 50005999) + b1X1 (elevation 6000-6999) + bsXe (elevation 7000-7999) + bg)(g (elevation 8000-8999) + b,oX,o (elevation >9000) + b,,x,, (school 34 years age)+ (first birth)+ b5Xs (high parity)+ beXe (elevation 5000-5999) + b 7 X 7 (elevation 6000-6999) + b8Xe (elevation 7000-7999) + bg)(g (elevation 80008999) + b,oX,o (elevation >9000) + b,,x,, (school
PAGE 76

o M3: Y (SGA) = a (intercept) + b,X, (Mexican-bam) + b2X:! (<20 years age) + b3XJ (>34 years age)+ (first birth)+ bsXs (high parity)+ bsXs (elevation 5000-5999) + b1X1 (elevation 6000-6999) + beXa (elevation 7000-7999) + bg)(g (elevation 80008999) + b1oX1o (elevation >9000) + b,,x,, (school 34 years age)+ (first birth)+ bsXs (high parity)+ bsXs (elevation 5000-5999) + b1X1 (elevation 6000-6999) + beXa (elevation 7000-7999) + bg)(g (elevation 80008999) + b1oX1o (elevation >9000) + b,,x,, (school 34 years age)+ (first birth)+ bsXs (high parity)+ beXs (elevation 5000-5999) + b1X1 (elevation 6000-6999) + beXa (elevation 7000-7999) + bgXg (elevation 8000-8999) + b1oX1o (elevation >9000) + b,,x,, (school 9000) + b,,x,, (school 34 years age)+ (first birth)+ bsXs (high parity)+ beXs (elevation 5000-5999) + b1X1 (elevation 6000-6999) + beXs (elevation 7000-7999) + bgXg (elevation 8000-8999) + b,oX1o (elevation >9000) + b,,x,, (school
PAGE 77

As with Aim 1, AGA was also tested. Drinking was not significant and was dropped from the models. Only one interaction is significant, weight *parity. Significant standardized beta coefficients and odds ratios are reported for each outcome, model, and aim in Chapter 4. For models with interactions, the model was run substituting a newly created n-level interaction variable for the parent terms and the interaction to parse the effect of the interaction. The R2 statistic (for linear regressions) and the -2LL test (for logistic regressions) use differences in variance accounted for in successive models to test model fit. The c statistic tests how well a model predicts outcomes. The c statistic ranges from .51 to 1.0; with a c statistic of .51 representing random variation. The closer a c statistic is to 1 the better the fit. Each successive model should improve the fit. In addition, adequacy of fit is often tested using the Hosmer-Lemeshow statistic. The c statistic and Hosmer-Lemeshow statistic are reported for each model and outcome. For some models, where variables are retained notwithstanding their insignificant addition to the model, model fit is sacrificed in the name of previous research showing that the variables are important to the outcome. Model Building for Aim 3 Aim 3 examines the influence of neighborhood of the mother's residence immediately before birth to test the effect of neighborhood deprivation and immigrant orientation on birth outcomes by nativity in two Colorado counties. Neighborhood is operationalized by using census tract data from the 2000 Decennial Census. In Los Angeles County, California, Finch eta/. (2007) compared the probability of LBW by census tract for foreign-born and U.S.-born mothers of Mexican origin using hierarchical linear modeling. They showed that as neighborhood deprivation increased, the probability of LBW increased. They also showed that immigrant orientation moderated the probability of LBW for foreign-born mothers and lowered the probability of LBW when compared with the probability of LBW by neighborhood deprivation alone. For this study, the effect of the two contextual variables on each outcome is tested separately and independent of individual level effects. The hypotheses for Aim 3 are: H40 : immigrant-oriented neighborhoods have no effect on the four birth outcomes H4A: immigrant-oriented neighborhoods affect the likelihood of birth outcomes H50 : neighborhood disadvantage has no effect on the four birth outcomes 60

PAGE 78

HSA: neighborhood disadvantage affects the likelihood of birth outcomes For each of the two counties, the effects of neighborhood deprivation and immigrant orientation were modeled with the rate of each outcome in each tract using generalized linear modeling. First the transformed values for the contextual variables (square root for neighborhood deprivation and natural log for neighborhood immigrant orientation) were used to model area effects. The resulting models were then run using the actual values of the scales after centering at the mean. The number of births and the number of outcomes in combined tracts were summed and a new percentage of each outcome was calculated based on that new denominator to create the outcome rate for the combined tract. The scale values for combined tracts were also weighted based on births in the combined tracts. Y is the rate of each outcome by tract, modeled against the centered values for scale of immigrant orientation alone, the scale of neighborhood deprivation alone, the combination of the two scales, and the interaction of the two. The t statistic for each contextual variable are reported, as is the F statistic for each model. For each county, the general linear modeling formulae are: Y (rate of outcome by tract)= a+ b1X1 (neighborhood deprivation scale)+ Y (rate of outcome by tract)= a+ b1X1 (immigrant orientation scale)+ Y (rate of outcome by tract)= a+ b1X1 (neighborhood deprivation)+ b2X2 (immigrant orientation)+ Y (rate of outcome by tract)= a+ b1X1 (neighborhood deprivation)+ b2X2 (immigrant orientation)+ b3X3 (neighborhood deprivation immigrant orientation) + Qualitative Research Component Sample and Study Data "History" and much of social science is biased because it is constrained by the existence of what is collected in government files. Families and informal groups do not necessarily collect and maintain information that captures the lived experience (Boorstin 1987:7). Aim 4 is designed to explore the context surrounding the quantitative results, especially the finding of higher odds of LGA among Mexican-born mothers, by interviewing women of Mexican origin who had babies within the past year in Colorado. This component of the study was approved by the Human Subjects Review Committee of the University of 61

PAGE 79

Colorado Denver on an expedited basis under 45 CFR 46.102(i) as presenting minimal risk to research subjects. Waiver of documentation of consent pursuant to 45 CFR 46.1179(c)(2) was also approved so as to avoid any link between the interviewee and identifying information, which could present concerns for undocumented immigrants. Approvals from the University's Human Subjects Review Committee are provided in Appendix B. In contrast to quantitative studies, where the sample is chosen to address representation, generalization, replication, and detection of bias, qualitative research is processual (Morse 2008a) and need not, therefore, follow quantitative sampling requirements. Purposive and snowball sampling, also known as chain referral sampling, were used to identify mothers for qualitative interviews (Schensul et a/. 1999:240-244). Purposive sampling was accomplished by interviewing mothers receiving medical care at Salud Family Health Centers (Salud). Salud operates a number of clinics designed to serve the poor and near-poor, and unand under-insured, in various Colorado locations. Its client base includes many women of Mexican origin. Salud's medical director agreed to have its case managers ask recent mothers of Mexican origin attending the Brighton and Ft. Lupton clinics if they were interested in being interviewed for this study. If the woman said yes, clinic personnel provided the woman's first name and telephone number to the researcher to follow up and schedule an interview. Women who indicated interest in participating in the study were contacted and a face-to-face interview was scheduled. The consent was written in Spanish and English for comprehension at the 81 h grade level. A copy of the consent was provided to the interviewee in English or Spanish, at the preference of the interviewee, and read by or explained to each mother. Mothers were not asked to sign the consent, pursuant to the waiver of documentation authorized by the Human Subjects Review Committee. Each interviewee was paid $20 for a one hour interview. The interviews were held in a location chosen by the interviewee and were conducted in English or Spanish, at the interviewee's request. Snowball sampling was employed by asking each mother if she had a friend or relative who met the selection criteria who might be interested in participating. Two mothers were solicited this way. All interviews were audio-taped using a Sony digital recorder. Relevant portions of the interviews were transcribed and translated, as necessary. Ten women were interviewed; five U.S.-born mothers of Mexican origin, and five Mexican-born mothers. Interviews were conducted until it appeared that saturation on key themes of diet, exercise, health behaviors, and cultural beliefs, especially relating to LGA, was reached. 62

PAGE 80

The interviews were semi-structured and inquired into the mother's most recent birth and any other births, infant outcomes, her use of prenatal care, insurance status, her neighborhood, sources of social support, and her health behaviors before and during pregnancy. Special emphasis was placed on eliciting differences between life in Mexico and in the U.S. for Mexican-born mothers. In addition, five key informant interviews were conducted with professionals who had relevant insight into birth outcomes among Hispanics. One key informant has worked with poor Hispanic first-time mothers (primarily immigrants from Mexico) in the Nurse Family Partner Program. A second key informant is a certified nurse midwife, practicing in Colorado for many years in rural locations and with Denver Health. The third key informant is an obstetrician working with a primarily poor, Mexican population at Denver Health. The fourth is a physician-researcher who is an expert in diabetes during pregnancy at the University of Colorado Denver, who was interviewed by telephone. Finally, a woman who was a practicing physician in Mexico, who is now affiliated with La Clinica Campesina in Colorado as a nurse, provided insights on the differences between mothers' behaviors and birth outcomes in Mexico and in the U.S. All but the telephone interview were recorded using a Sony digital recorder and relevant portions transcribed. Field notes were also used to capture the researcher's observations and impressions of the interviewee as well as her surroundings. The consent documents and the interview guides are at Appendix C. Methods for Aim 4 Content analysis of interviews in qualitative research usually follows one of three approaches: conventional, directed, or summative (Hsieh & Shannon 2005). Through content analysis the researcher subjectively interprets text data using a systematic classification process of coding and identifying categories, themes, and patterns. Categories help to identify "whaf' is in the data and develop a taxonomy that identifies relationships between categories and subcategories. Themes are meanings that run through the data (Morse 2008b). Conventional content analysis is usually used to describe a phenomenon; codes are developed from the text without using preconceived categories or theoretical perspectives. Directed content analysis is used to validate or extend a theory; codes are developed using key concepts and operational definitions come from the theoretical framework applied in the study (Morse 2008b). Findings provide supporting and non supporting evidence for a theory. Summative analysis seeks to identify and quantify certain words or content in the text so as to understand the contextual use of the words or content to 63

PAGE 81

explore usage, not meaning. This study uses directed content analysis because the purpose of the qualitative analysis is to explore certain contradictory quantitative data in the context of theories that are being tested. After transcription and translation, the data from the interviews were analyzed for a priori categories and themes that centered on maternal weight before pregnancy, weight gain, eating and exercise habits while pregnant, sources of social support, and neighborhood characteristics. Each mother was also asked if she had insight into why Mexican-born mothers have higher risks of LGA. In addition to the a priori codes, certain themes emerged that were identified and ultimately organized around cultural beliefs and political economic theory, including cultural beliefs concerning diet and exercise during pregnancy, practice of Ia cuarentena after the birth, and male partners' attitudes and behaviors in the form of machista -protection of women partners. Several Mexican-born mothers discussed the economic difficulties of life in Mexico and in the U.S. after immigration, and the differences in lifestyle, including levels of energy expenditure and nutrition in Mexico and the U.S. Results of the qualitative interviews are reported in Chapter 5. 64

PAGE 82

CHAPTER 4 QUANTITATIVE ANALYSIS Aim 1 compares the odds of each outcome by race/ethnicity using unadjusted and adjusted odds ratios. Aim 2 compares odds of outcomes by nativity. For Aim 2, the successive models examine support for the healthy migrant or healthy immigrant hypotheses. Aim 3 examines the association of neighborhood deprivation and immigrant orientation with birth outcomes of mothers of Mexican origin in two large counties, based on Finch et a/.'s use of immigrant orientation as a general proxy for availability of social support. Aim 1 Comparison of Risk Factors by Race/Ethnicity Aim 1 tests whether racial/ethnic populations of mothers in Colorado differ in their risk profiles, and whether there is congruence between risk factors and birth outcomes. Based on census data for the U.S. and Colorado, Hispanics and Blacks should have demographic, social, medical, and health behavior risk profiles that are less advantageous than Whites. The null and alternative hypotheses for risk profiles are: H10 : each risk factor is independent of race/ethnicity. H1 A: each risk factor is related to race/ethnicity. Table 4.1 shows the frequency distribution of risk factors by race/ethnicity and the significance of differences using Pearson's chi-square as the test of the relationship between each risk factor and race/ethnicity. A risk factor is considered related to race/ethnicity if ps 0.05 (Gould et a/. 2003). In Colorado the frequency of each risk factor is related to ethnicity (p<0.0001 ). In addition, Hispanics experience higher frequencies of some risk factors than Blacks experience, which are underlined in Table 4.1. For these risk factors, Hispanics have worse risk profiles than Blacks. 65

PAGE 83

Table 4.1. Percent frequency distribution of risk factors by racelethnlclty 2000.2005 Characteristic Age of mother S19 20-34 9000 Education (mother) < 9 yrs 9-11 yrs 12+ yrs Prenatal care Inadequate Intermediate Adequate Adequate plus Marital status Married Unmarried Smoking No Yes Alcohol drinker No Yes Weight gained <151b 15-40 lb >401b Medical risk factors None One or more conditions LGA associated risk factors Gestational diabetes Preexisting diabetes Prev infant 4000+ grams Total 10.53% 75.37% 14.11% 42.14% 44 85% 13.01% 15.38% 64 40% 14.39% 3.51% 1.44% 0.88% 6 53% 15.60% 77.87% 13.46% 17 52% 42.93% 26.10% 73.93% 26.07% 91.66% 8.34% 99.07% 0 93% 10.51% 73.12% 16.37% 70.05% 29.95% 2.52% 0 40% 0 78% White 6.60% 75.87% 17.52% 44.39% 45.61% 10.01% 14.94% 60 54% 17.27% 4.10% 2.07% 1.07% 0.76% 7.57% 91.67% 8 51% 16.51% 46.06% 28.92% 82.18% 17.82% 89.98% 10.02% 98.91% 1.09% 7.91% 73.99% 18.09% 74.42% 25.58% 1.25% 0.22% 0.34% Hispanic 18.09% 74.40% 7.51% 37 23% 44.12% 18.65% 19 50% 68 14% 8.24% 2 97% 0 44% 0.71% 19.78% 32 62% 47.90% 22.67% 19.67% 37.14% 20 52% 59.81% 40 19% 94.97% 5.03% 99.36% 0.64% 71. 92% 12 53% 61.84% 38 16% 0 95% 0 14% Black 17.47% 73.08% 9.45% 40.24% 40.64% 19.12% 2.21% 83.60% 13.54% 0.40% 0.21% 0.04% 2.18% 17.78% 80.05% 19.37% 16.91% 37 69% 26.03% 47.48% 52.52% 89.68% 10.32% 99.07% 0.93% 13 79% 66 02% 14.56% 62.16% 37.84% 0.11% 0.02% 0.04% Other p-value 7.20% 77.16% 15.65% <0.0001 46.01% 43.37% 13.01% <0.0001 6.45% 74.02% 16.65% 2.00% 0.61% 0.27% <0.0001 3.58% 10.20% 86.22% <0.0001 14.26% 17.62% 43.48% 26.10% <0.0001 80.58% 19.42% <0.0001 94.81% 5.19% <0.0001 99.37% 0.63% <0 .0001 9.44% 76.00% 14.56% <0.0001 72.56% 27.44% <0.0001 0.20% 0.02% 0.20% < .0001 In accord with expectations based on census data, Hispanics have higher frequencies of social and medical risk factors than Whites (and sometimes higher than Blacks noted by an *) for: o Teen births* 66

PAGE 84

o High parity births o Lower education levels* o Unmarried status o Inadequate*, intermediate*, and adequate levels of prenatal care o Low weight gain* o One or more medical risk factors* o Previous infant 4000+ grams* Although Hispanics compare favorably with Whites for gestational and preexisting diabetes, Blacks and Others have much lower frequencies of the specific risk factors associated with LGA. Hispanics report the lowest frequency of smoking and the second lowest frequency of drinking alcohol during pregnancy (second to "Other''). Based on CDPHE's analysis of discordance in reporting of smoking and drinking behaviors during pregnancy (see page 39), smoking and drinking behaviors reported in vital records may understate the true rate for some groups, but the data are more likely to be reliable for Hispanics. Overall, as shown in Figure 4.1, the data on risk factors support a finding that Hispanics have poorer social, demographic, and medical risk factors compared to Whites, Blacks, and Others, but better self-reported smoking and drinking behaviors compared with Whites and Blacks. Risk factors specifically associated with LGA are not shown because their frequency is so low that they do not register on the scale of the graph. 100 90 80 70. 60 r i 50 40 30 20 10 0 # .<.: rtf WMe 1-tspanic D Black D Other Figure 4.1. DISTRIBUTION OF RISK FACTORS BY RACEIETHNICITY (r) indicates reference category 67

PAGE 85

Frequency of Adverse Birth Outcomes by Race!Ethnicity The theory of the social gradient of health, supported by the poorer risk profiles of Hispanics, suggests that they should have higher frequencies of adverse birth outcomes than Whites and Blacks. However, prior research reports that Hispanics have paradoxically better low-weight associated birth outcomes than Blacks and outcomes that approach the White majority population for LBW and preterm birth (Rosenberg et a/. 2005; Frisbie & Song 2003; Gould et a/. 2003; Singh & Yu 1996). Few studies report on SGA, although Gould et a/. (2003) report that foreign-born Mexican mothers have higher frequencies of SGA than U.S. born non-Hispanic Whites, and much lower frequencies than U.S-born Black or foreign-born Indian mothers. While previous research shows higher rates of diabetes in Hispanics than other population groups in the U.S. (Mainous et a/. 2008; Martorell 2005), no population studies of LGA were found. The null and alternate hypotheses for frequency of birth outcomes unadjusted for risk factors are: H20 : each birth outcome is independent of race/ethnicity. H2A: each birth outcome is related to race/ethnicity. Pearson's chi-square is used to test the relationship between each birth outcome and race/ethnicity. In accord with other research, the unadjusted frequency of each adverse birth outcome is related to race/ethnicity. However, Hispanics have disproportionately low frequencies of low-weight associated birth outcomes, as shown in Table 4.2. Despite having poorer demographic, social, and medical risk profiles, Hispanics are second to Whites in all categories of adverse birth outcomes, except LGA, where Hispanics are lower than the majority White population, but higher than Blacks and all Others. Figure 4.2 graphically demonstrates that, based on unadjusted frequencies, the epidemiological paradox exists for Hispanics in Colorado for LBW, preterm birth, and SGA. Rates of LGA are about the same for Hispanics and Whites, but Blacks and Others enjoy much lower rates of LGA than either Hispanics or Whites. Table 4.2. Percent frequency of LBW, pretenn birth, SGA, and LGA by race/ethnlclty 200Q-2005 Birth Outcome Total White Hispanic Black Other p-value LBW 6.73% 6.18% 6.83% 12.14% 8.48% <0.0001 Preterm births 7.31% 7.00% 7.40% 10.80% 7.66% <0.0001 SGA 12.13% 11.16% 12.51% 19.21% 16.12% <0.0001 LGA 5.07% 5.34% 5.00% 3.11% 3.67% <0.0001 68

PAGE 86

25 20-+-White -15 c ----Hispanic Gl u .. Black Gl 10 c. ----*Other 5 0 v CJ v -:.._0" q_<..0 Figure 4.2. UNADJUSTED FREQUENCIES OF BIRTH OUTCOMES BY RACE/ETHNICITY Odds Ratios of Birth Outcomes by Race!Ethnicity Unadjusted frequencies of outcomes do not take into account various factors that are expected to contribute to any given outcome. Multiple regression permits comparisons while adjusting for contributing factors. To compare odds of each of the four birth outcomes, multiple logistic regression for each dichotomous outcome was performed, creating odds ratios based on race/ethnicity alone (Model 1 ), and then adjusting successively for demographic and socioeconomic characteristics (Model 2), medical conditions associated with pregnancy (Model 3), and health behaviors available in the birth record (Model 4). The reference group is White mothers based on their superior risk profiles and the theory of the social gradient of health. The risk profiles of Hispanics and Blacks in Colorado predict that they should have poorer birth outcomes than non-Hispanic Whites. If Hispanics have odds of adverse outcomes similar to those of non-Hispanic Whites, the data suggest that a paradox may exist, at least for these outcomes at the individual/compositional level. H30 : race/ethnicity does not predict each birth outcome after adjusting for risk factors. H3A: race/ethnicity predicts each birth outcome after adjusting for risk factors. The results of the multiple logistic regression are reported and discussed separately for each outcome. 69

PAGE 87

Low Birth Weight Table 4.3 reports the summary odds of LBW, including confidence intervals at 95%, adjusting for each successive model. Also reported are the c statistic and the Hosmer Lemeshow Goodness of Fit statistic for each model. Table 4.3. Unadjusted and adjusted odds ratios (95"/o Cl) of LBW by race/ethnicity Race White Hispanic Black Other c statistic Hosmer-Lemeshow Model 2 Model 3 Model 1 Adjusted for Additional Race/Ethnicity Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95"/o) 1.00 1.00 1.00 1.11 (1.08-1.15) 1.07 (1.03-1.10) 1.15 (1.10-1.20) 2.10 (1.99-2.21) 1.91 (1.81-2.02) 2.09 (1.94-2.25) 1.41 (1.33-1.49) 1.46 (1.38-1.55) 1.62 (1.50-1.75) 0.535 0.687 0.704 0.9978 <.0001 <.0001 Model4 Additional Adjustment for Health Behaviors 1.00 1.18 (1.13-1.23) 2.16 (2.01-2.33) 1.64 (1.52-1.77) 0.733 0.0009 There is no monotonic increase or decrease in odds across the successive models, although in the fully saturated model the odds of LBW are slightly higher than in Model 1 (race/ethnicity alone). Hispanics continue to have the closest odds of LBW to those of Whites, notwithstanding their poorer risk profiles, and almost 1 00% better odds than Blacks. This particular model has a modest degree of discrimination (c statistic= 0.733) and the fit is not adequate, suggesting that, even with the high number of interactions (five), other factors account for much of the variation by race/ethnicity. 2.50 Non-lis panic White Ill 2.00 r----.2 ---Hispanic .. 1.50 Ill a: .. -Ill 1.00 Black "tl "tl 0 0.50 """""*"Other 0.00 2 3 4 Models Figure 4.3. LBW ODDS RATIOS BY RACEIETHNICITY AND MODEL 70

PAGE 88

Table 4.4 reports the significant standardized beta coefficients and odds ratios for each variable in the models for LBW by race/ethnicity. LBW by race/ethnicity has five interactions: parity*marital status, medical risk*race/ethnicity, smoking*age, prenatal care*weight gain, and prenatal care*smoking. Table 4.4. Estimated coefficients and odds ratios for LBW and raca/athnlclty Variable Model1 Model 2 Model 3 Model4 OR OR p OR OR Race White (r) 1.00 1.00 1.00 1.00 Hispanic 0.1076* 1.11(1.08.15) 0.06421 1.07(1.03-1.10) 0.1383* 1.15(1.10-1.20) 0.1650* 1.18(1.13-1.23) Black 0.7414* 2.1 0(1.99-2.21) 0.6489* 1.91(1.81-2.02) 0.7385 2.09(1.94-2.25) 0.7706* 2.16(2.01.33) Other 0.3408* 1.41(1.33-1.49) 0.3799* 1.46(1.38-1.55) 0.4802 1.62(1.50-1.75) 0.4944* 1.64(1.52.n) A e 20-34(r) 1.00 1.00 1.00 S19 0.08281 1.09(1.04-1.14) 0.0981* 1.10(1.05.16) 0.1794* 1.20(1.14-1.26) >35 0.1704* 1.19(1.14-1.23) 0.1223* 1.13(1.09-1.18) 0.0874* 1.09(1.05-1.14) Pa Low(r) 1.00 1.00 1.00 First 0.4326* 1.54(1.49-1.60) 0.4248* 1.53(148-1.59) 0.4473* 1.69(1.63-1.75) High 0.2572* 1.29(1.23-1.38) 0.2369* 1.27(1.20-1.34) 0.1673* 1.16(1.10-1.22) Elevation <5000 (r) 1.00 1.00 1.00 5000-5999 0.0371 1.00 0.0240 1.00 0.0796* 1.08(1.04-1.13) 6000-6999 0.0793t 1.08(1.03-1.14) 0.1080* 1.11(1.06-1.17) 0.1619* 1.18(1.12-1.24) 7000-7999 0.2435* 1.28(1.08-1.38) 0.1895* 1.21(1.12.31) 0.2228* 1.26(1.16-1.38) 8000-8999 0.3250* 1.38(1.24-1.55) 0.3284* 1.39(1.24-1.55) 0.4166* 1.52(1.38-1.70) >9000 0.5897* 1.80(1.60-2.04) 0.5977* 1.82(1.60-2.06) 0.6775* 1.97(1.73-2.24) Education HS grad (r) 1.00 1.00 1.00 No high school 0.0537 1.00 0.0173 1.00 -0.0323 1.00 Some high school 0.2284* 1.26(1.21-1.31) 0.2097* 1.23(1.19-1.28) 0.1395* 1.15(1.10.20) Marital status Married (r) 1.00 1.00 1.00 Not married 0.4385* 1.55(1.47.63 0.4143* 1.51(1.44.59) 0.3094* 1.36(1.29-1.44) Prenatal care Adequate (r) 1.00 1.00 1.00 Inadequate 0.5303* 1.70(1.63-1.78 0.4526* 1.57(1.50-1.84) 0.3874* 1.47(1.39-1.56) Intermediate -0.1007* 0.90(0.86-0.95) -0.1082* 0.90(0.86-0.94) -0. 13539* 0.86(0.81-0.91) Adequate plus 1.2532* 3.50(3.39-3.61) 1.2167* 3.38(3.27 -3.49) 1.2609* 3.53(3.39-3.67) Pa *marital Low*married (r) 1.00 1.00 1.00 111 *married 0.4326* 1.54( 1 .49-1.60) 0.4282* 1.53(1.8-1.59) 0.5225* 1.69(1.63-1.75) High*married 0.2572* 1.29(1.23-1.36) 0.2389* 1.27(1.20-1.34) 0.1472* 1.16(1.10-1.22) 1S\*unmarried 0.5694* 1.77(1.69-1.85) 0.5419* 1.72(1.64-1.80) 0.5769* 1.78(1.70-1.87) Low* unmarried 0.4383* 1.55(1.47-1.63) 0.4143* 1.51(1.44-1.59) 0.3094* 1.36(1.29-1.44) High*unmarried 0.6550* 1.93(1.80-2.06) 0.6266* 1.67(1.75-2.00) 0.4601* 1.58(1.48-1.70) Medical risk None (r) 1.00 1.00 One or more 0.7668* 2.15(2.8-2.23) 0.7208* 2.06(1.98.13) Medical risk*race White*no risk (r) 1.00 1.00 Hispanic no risk 0.1383* 1.15(1.10-1.20) 0.1650* 1.18(1.13-1.23) Black no risk 0.7385' 2.09(t .94-2.25) 0.7706' 2.16(2.01-2.33) Other no risk 0.4802* 1.62(1.50-1.75) 0.4944' 1.64(1.52-1.n) White w/risk 0.7668* 2.15(2.08-2.23) 0.7208* 2.06(1.98-2.13) Hispanic w/risk Black w/risk 0.6002* 1.82(1.74-1.91) 1.1213* 3.07(2.83-3.32) 0.6241* 1.88(1.78-1.96) 1.1829* 3.26(3.02-3.53) Other w/risk 0.9791* 2.66(2.42-2.93) 0.9711* 2.64(2.40-2.91) Smokin None r 1.00 Smoking 0.7564* 2.14(1.97-2.31) 71

PAGE 89

Table 4.4. Estimated coefficients and odds ratios for LBW and racafethnlclty (continued) Variable Wei h1 ain 16-40 r <16 Smokin a e No sm"20-34 (r) No smoke"<19 No smoke">35 Smoke"<19 Smoke"20-34 Smoke>35 Prenetal"welght Adeg"med wt (r) lnad"lowwt Inter" low wt Adeg"lowwt Adeg+"lowwt lnad"med wt lnter"med wt Adeg+"med wt lnad"high wt Inter" high wt Adeg"hlgh wt Adeg+"high wt Prenatal" smoke Adeg"no sm (r) lnad"no smoke Inter" no smoke Adeg+"no sm In ad" smoke Inter" smoke Adeg"smoke Adeq+"smoke (r) indicates reference category. p<.0001. tp <.001. Model1 Model2 Model3 OR OR 15 OR The contribution of some risk factors is not unexpected based on prior studies: Model4 15 OR 1.00 0.8510" 2.34(2.18-2.51) -Q.7491" 0.47(0.43-0.52) 1.00 0.1794" 1.20(1.14-1.28) 0.0874" 1.09(1.05-1.14) 0.4761" 1.61(1.55-1.67) 0.0221 1.00 0.4802" 1.62(1.40-1.86) 1.00 1.2008" 3.32(2.90-3.81) 1.7243" 5.61(4.7HI.68) 0.8510" 2.34(2. 18-2.51) 1.5836" 4.87(4.50-5.28) 0.7208" 2.06(1.81-2.34) 0.8407" 2.32(1.98-2.72) 0.7894" 2.20(2.06-2.35) -1.7368" 0.18(0.12-Q.28) -3.7448" 0.02(0.02-Q.02) -Q7491" 0.47(0.43-0.52) -Q.9581 0.38(0.37 -Q.40) 1.00 -Q.3334" 0.72(0.63-0.81) -Q.9948" 0.37(0.32-Q.43) 0.4715" 1.60(1.49-1.72) 0.5184" 1.68(1.43-1.97) -Q.1850 1.00 0.7384" 2.09(1.86-2.34) 0.9454" 2.57(2.28-2.91) o Odds of delivering an LBW infant increase for teen mothers (1.20) and older mothers (1.09) o First babies have higher odds (1.69) of being LBW o Odds of LBW show a monotonic increase as residence at elevation above 5000 feet increases (1.08-1.97) o Being unmarried increases the risk of LBW (1.36) o Mothers with inadequate (1.47) and adequate plus prenatal care (3.53) have increased odds of LBW o The presence of one or more medical risks associated with pregnancy is a marker for increased odds of LBW (2.06) o Smoking increases the odds of LBW by over 100% (2.14) 72

PAGE 90

o The odds of LBW with weight gain of less than 16 pounds increases by 134% (2.34) o Weight gain of >40 pounds is very protective against LBW gaining more than 40 pounds reduces the odds of LBW by 53% (0.47) Some results are unanticipated: o Low education (less than high school) is not significant o Intermediate prenatal care (one level below adequate) does not increase the odds of LBW (0.86) In the fully saturated model, the factors having the greatest predictive power for increased odds of LBW are adequate plus prenatal care (3.53), presence of one or more medical risks (2.06), being Black (2.16) smoking (2.14), and first parity (1.69). Interactions highlight the effect of some of these risks for LBW. The interaction of parity and marital status has a small impact on odds of LBW: first parity*married = 1.69, first parity*unmarried = 1. 78. Whereas intermediate prenatal care seems protective (0.86), the interaction of intermediate prenatal care and low weight gain increases odds to 5.61. The interaction of race and medical risk is most interesting. In Model 4, White mothers with at least one medical risk have 2.06 higher odds of LBW compared to White mothers with no medical risks (1.00 reference). Even though Hispanics have poorer medical risk profiles than Whites, Hispanics with at least one medical risk have an elevated odds of LBW compared to White mothers with no risk (1.82), but lower odds than those of Whites with medical risks (2.06). Hispanics with no medical risk compare favorably with Whites with no medical risk with odds of an LBW birth of 1.18. The ranking of odds of LBW by race/ethnicity for mothers with one or more risks places Hispanics first (1.82), then Whites (2.06), followed by Others (2.64), and Blacks (3.07). Preterm Birth Table 4.5 reports unadjusted and adjusted odds ratios of preterm birth by race/ethnicity. Each successive model improves the fit, with Model 4 having adequate fit and modest discrimination. Although Hispanics have higher frequencies of social and medical risk factors than Whites, and sometimes higher than Blacks, Hispanics' odds of having a preterm birth are close to those of Whites in Models 1, 2, and 3, and are no different than the odds of Whites in Model 4. 73

PAGE 91

Table 4.5. Unadjusted and adjusted odds ratios of preterm birth by race/ethnicity Race White Hispanic Black Other c statistic Hosmer-Lemeshow Model 2 Model 3 Model 1 Adjusted for Additional Race/Ethnicity Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1.00 1.00 1.00 1.06 (1.03-1.09) 1.08 (1.05-1.12) 1.03 (1.00-1.07) 1.61 (1.53-1.70) 1.54 (1.45-1.63) 1.45 (1.37-1.54) 1.10 (1.04-1.17) 1.17 (1.10-1.25) 1.16 (1.09-1.24) 0.518 0.735 0.753 0.9999 <.0001 0.0006 Odds not significantly different from 1.00. Model4 Additional Adjustment for Health Behaviors 1.00 *1.01 (0.98-1.05) 1.42 (1.34-1.51) 1.16 (1.09-1.23) 0.765 0.1892 Figure 4.4 shows the change in odds for preterm birth with the addition of risk factors in the successive models. For Hispanics and Blacks the movement of odds trends generally down with successive models; with Others the odds increase as risk factors are accounted for. 2.00 ...-----------. 1.50 -i I a: 1.00 I-III "CI 8 0.50 -1---------____.., 0.00 '------------,.-------' 2 3 4 Models -+-Non-lis panic WMe -Hispanic Black Figure 4.4. PRETERM BIRTH ODDS RATIOS BY RACE/ETHNICITY AND MODEL Table 4.6 reports the significant standardized beta coefficients and odds ratios for each variable in the models for preterm birth by race/ethnicity. Preterm birth by race/ethnicity has four interactions: prenatal care*medical risk, prenatal care*weight gain, parity*weight gain, and prenatal care*smoking. 74

PAGE 92

Table 4.8. Estimated coefficients and odds ratios for pretenn birth and racalethnlclty Variable Moc:lel1 Moc:lel2 Moc:lel3 Moc:lel4 OR OR OR OR Race Whitejrl 1.00 1.00 1.00 1.00 Hi!;!anic 0.0607" 0.0791" 1.08j1.05.12l 0.0317 1.00 0.0130 1.00 Black 0.4781" 1.61j1.53-1.70l 0.4291" 1.55j1.45.63l 0.3710" 1.45j1.37.54l 0.3507" 1.42j1.34.51l Other 0.0971" 1.1Q!1.04-1.17l 0.1582" 1.17j1.10-1.25l 0.1515 0.1457" A e 20jrl 1.00 1.00 1.00 S19 0.0264 1.00 0.0441 1.00 0.0757! 1.08j1.03-1.13l >35 0.1319" 0.0835" 1.09(1.05-1.13! 0.0749" 1.08j1.04-1.12l Peri Lowjrl 1.00 1.00 1.00 First 0.1823" 1.2Q!1.17.24l 0.1788" 0.2163" High 0.2886" 1.31j1.28-1.38l 0.2542" 1.29(1.24-1.34! 0.2260" 1.25j1.19-1.32l Elevation <5000 jrl 1.00 1.00 1.00 5000-5999 -<1.0083 1.00 -<1.0248 1.00 0.0083 1.00 6000-6999 -<1.0085 1.00 0.0176 1.00 0.0483 1.00 7000-7999 0.0701 1.00 0.0112 1.00 0.0261 1.00 8000-8999 0.1752! 0.1779! 1.20(1.07-1.34! 0.2287" >9000 0.15631.17j1.02-1.35l 0.1608 1.17j1.02-1.35l 0.2040t Education HS grad jrl 1.00 1.00 1.00 No high school 0.0587 1.00 0.0122 1.00 -<1.0250 1.00 Some high school 0.1504" 0.1277" 1.14j1.09-1.18l 0.0974" 1.1Q!1.08-1.15l Marital status Marriedjrl 1.00 1.00 1.00 Not married 0.1650" 1.18j1.14.22l 0.1454" 0.1225" Prenatal care Adeguate jrl 1.00 1.00 1.00 lnadeguate 0.8838" 2.42j2.31-2.53l 0.9714" 2.64j2.48-2.82l 0.9386" Intermediate -<1.0131 1.00 0.0211 1.00 -<1.0193 1.00 Adeguata E!IUS 1.8939" 6.65j6.42-6.88l 1.9620" 7.11j6.8Q-7.44l 1.9959" 7.36(7.0Q-7.74l Medical risk None jrl 1.00 1.00 One or more 0.8538" 2.35j2.21.49l 0.8181" 2.27j2.13-2.41l Prenatal"medrisk Adeg no risk jrl 1.00 1.00 lnadeg no risk 0.9714" 2.64j2.48-2.82l 1.2501" 3.5Q!2.94-4.14l Inter no risk 0.0211 1.00 -<1.2997 A!!!9;!: no risk 1.9620" 7.11 j6.8Q-7.44l 1.8348" lnadeQ wlrisk 1.4576" 1.6737" 5.33(4.50-6.32! Inter w/risk 0.7889" 2.1 0.3867! 1.47j1.2Q-1.81l Adeg wlrisk 0.8538" 2.35(2.21-2.49! 0.8181" 2.27j2.13-2.41l Ade9:!: wlrisk 2.5864" 13.28p2.66-13.93l 2.4390" 11 .48( 1 0.05-13.07! Smokl 1.00 0.3360" 1.4Q!1.27-1.54l 1.00 0.7471" 2.11 j1.93-2.31l -<1.5325" Prenatal"weight Adeg"med wt jrl 1.00 lnad"lowwt 0.7151" 73-2.42! lnter"low wt 1.2367" 3.44(2. 76-4.29! Adeg"lowwt 0.9215" 2.51 j2.27 78! Ade9:t"lowwt 0.8793" 2.41j2.09-2.78l lnad"med wt 0.2007 1.22j1.05-1.43l lnter"med wt 0.2804! 1.32j1.09-1.82l AdeQ+ "med wt 0.16131.18j1.04-1.33l lnad"high wt 2.0080" 7.45j7.33-7.57 Inter" high wt -1.4301" 0.24j0.24-
PAGE 93

Table 4.8. Estimated coefficients and odds ratios for preterm birth and racafethnlclty (continued) Variable Model1 Low*lowwt High"lowwt 111 "medwt 1 high wt Low*high wt High"high wt Prena1al"smoke Adeg"no sm (r) lnad"no smoke Inter" no smoke Adeq+"no sm In ad" smoke Inter" smoke Adeg"smoke Adeq+"smoke (r) indicates reference category p <.0001. tp <.001. t:p<.01 p<.05. OR Model2 Model3 OR OR As with LBW, some results are not surprising based on prior studies: Model4 p OR .Q.2n1t 0.76(0.65.Q.88 3.2330" 25.36(25.24-25.48) 1.00 0.3279" 1.39(1.27 -1 .51) .Q.1744" o.84(0.n.Q.91l 2.1889" 8.75(7.71-9.92) 0.2183" 1 .24(1.20-1 .29) 0.2260" 1.25(1.19-1.32) .().1573 0.85(0.75.().97) 0.25541 0.78(0.68.().89) 3.9465" 51.75(46.19-58.00) 1.00 .Q.5122" 0.60(0.55.().65) .Q.3250" 0.72(0.56.().94) 4.1705" 64.75(53.()0-79.10) 3.4360" 31.06(21.10-40.03) 0.4120" 1.51(1.30-1.75) 0.3360" 1.40(1.27-1.54) 0.1030t: 1.11(1.04-1.18) o The presence of medical risks is a marker for increased odds of preterm birth, raising odds by 127% (2.27) o Smoking increases the odds of preterm birth by only 40% compared with an increase of 114% for LBW, but smoking with inadequate prenatal care raise odds to 31.06 o Inadequate and adequate plus prenatal care are markers for increased odds of preterm birth (inadequate 2.56; adequate plus 7.36)-much higher increases than for LBW. Intermediate prenatal care alone is not significantly different from adequate care. o Being unmarried increases the odds of preterm birth slightly (1.13) o Both first birth (1.24) and high parity (1.25) increase the odds of preterm birth o The odds of preterm birth increase over 1 00% with weight gain of less than 16 pounds (2.11 ). Weight gain >40 pounds is protective against preterm birth gaining more than 40 pounds reduces the odds of preterm birth by 41% 76

PAGE 94

Some results are unanticipated: o Age has little effect on odds of preterm birth. Odds increase only 8% for both teen births and births to older mothers. o Altitude affects preterm birth only if mother's residence is above 8000 feet o No high school is the same as being a high school graduate in all models o Intermediate prenatal care (one level below adequate) is the same as adequate prenatal care o The interaction of high parity and high weight gain raises the odds to 51.75 o The interaction of adequate plus prenatal care and no smoking raises odds to 64.75, which is highly perplexing In the fully saturated model, the factors having the greatest predictive power for increased odds of preterm birth are being Black (1.42), having inadequate (2.56) or adequate plus (7 .36) prenatal care, gaining less than 16 pounds during pregnancy (2.11 ), the interaction of high parity*high weight gain (51.75), and the interaction of smoking*inadequate prenatal care (31.06). Although intermediate prenatal care alone is the same as adequate care, when intermediate care is combined with low weight gain, the odds of preterm birth rise to 3.44, and when it is combined with smoking, the odds rise to 1.51. The interaction of prenatal care and medical risk is striking. Needing adequate plus prenatal care combined with one or more medical risks is associated with 11.46 higher odds of preterm birth. Indeed, with the exception of intermediate care without any medical risks, having anything other than adequate care with no medical risks raises the odds of preterm birth substantially (from 1.47-11.46). Some interactions raise the odds even higher. Small for Gestational Age SGA has some overlap with LBW and preterm birth, but not all SGA babies are low birth weight or preterm. Indeed, the frequency of SGA babies is almost double the frequency of LBW in Colorado during the study period as reported in Table 4.2 above. Table 4.7 reports unadjusted and adjusted odds ratios of SGA by race/ethnicity. As with LBW and preterm birth, Hispanics have odds ratios closer to Whites (but in no model are the odds of SGA the same as Whites), while Blacks and Others have much higher odds of SGA. Hispanics continue to have the closest odds of SGA to those of Whites, notwithstanding their poorer risk profiles. This particular model has a poor degree of discrimination (c statistic = 0.634) and the fit is not adequate, suggesting that, even with the 77

PAGE 95

Table 4.7. Unadjusted and adjusted odds ratios (95% Cl) of SGA by race/ethnicity Race White Hispanic Black Other c statistic Hosmer-Lemeshow Model 2 Model 3 Model 1 Adjusted for Additional Race/Ethnicity Only Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1.00 1.00 1.00 1.14 (1.11-1.16) 1.05 (1.02-1.08) 1.11 (1.07-1.14) 1.89 (1.82-1.98) 1.73 (1.66-1.81) 1.80 (1.79-1.90) 1.53 (1.46-1.60) 1.53 (1.47-1.61) 1.60 (1.51-1.69) 0.535 0.593 0.597 0.9998 <.0001 <.0001 Model4 Additional Adjustment for Health Behaviors 1.00 1.18 (1.15-1.22) 1.98 (1.86-2.09) 1.72 (1.63-1.82) 0.634 0.0005 As shown in Figure 4.5, the change in odds for SGA moves in the same pattern for Hispanics and Blacks: odds decrease in Model 2 and increase in Models 3 and 4 to levels higher than all previous models. There is no change in odds between Models 1 and 2 for Others; odds increase in Model 3 and 4 to levels higher than previous models. 2.50 2.00 Ill 0 ;: 1.50 Ill a: Ill ... 1.00 0 0.50 0.00 -II 2 3 4 Models -+--N:>n-Hispanic White ----Hispanic Black Figure 4.5. SGA ODDS RATIOS BY RACEIETHNICITY AND MODEL Table 4.8 reports significant standardized beta coefficients and odds ratios for each variable in the models for SGA by race/ethnicity. 78

PAGE 96

Table 4.8. Estimated coefflclente and odde ratloe for SGA and racafethnlclty Variable Model1 Model2 Model3 Model4 OR OR OR OR Race White{r) 1.00 1.00 1.00 1.00 0.1298" 0.04661 1.05(1.02-1.06) 0.1005" 1.11(1.07-1.14) 0.1683" 1.16(1.15-1.22) Black 0.6384" 0.5478" 0.5855" 0.6804" 1.96(1.86-2.09) Other 0.4253" 0.4277" 0.4682' 0.5431" 1.72(1.63-1.82) Ae 20-34(r) 1.00 1.00 1.00 S19 0.0232 1.00 0.0262 1.00 0.0938" >35 0.0228 1.00 0.0097 1.00 1.00 Pa low{r) 1.00 1.00 1.00 First 0.3972" 0.3960" 0.4591" 1.58(1.54-1.63) H!gh 0.0442 1.05(1.Q0-1.09) 0.0379 1.00 1.00 Elevation <5000 {r) 1.00 1.00 1.00 5000-5999 0.0174 1.00 0.0157 1.00 0.05531 1.06(1.03-1.09) 6000-6999 0.1235" 0.1316" 0.1721" 7000-7999 0.2630" 1.30(1.23-1.38) 0.2495" 1.28(1.21.36) 0.2861" 1.33(1.26-1.41) 8000-8999 0.3843" 1.44(1.33-1.56) 0.3650" 1.44(1.33-1.56) 0.4246" >9000 o.n40" 2.17(1.98.37) o.n51" 2.17(1.98-2.38) 0.8334" 2.30(2.1 0-2.52) Education HS grad {r) 1.00 1.00 1.00 No high school O.o115 1.00 0.0047 1.00 1.00 Some high school 0.2123" 1.24(1.20-1.26) 0.2076 1.00 0.1369" Marital status Married(r) 1.00 1.00 1.00 Not married 0.3940" 1.46(1.43-1.54) 0.3856" 1.47(1.41-1.53) o.2n3 1.32(1.27-1.37) Prenatal care Adequate {r) 1.00 1.00 1.00 lnadeguate 0.1214" 0.1036" 1.11(1.07.15) 0.0301 1.00 Intermediate 1.00 1.00 0.97(0.94-1.00) AdeQuate !;!IUS O.t507" 0.1363" 1.15(1.12-1.18) 0.1279" Pa marital low"married {r) 1.00 1.00 1.00 1 married 0.3972" 1 .49(1.45.53) 0.3960" 0.4591" 1.58(1.54-1.63) High"married 0.0442 1.05(1.Q0-1.09) 0.0379 1.00 1.00 111"unmarried 0.5821" 0.5726" 1.7.7(1.71-1.84) 0.5697" 1.n(1.71-1.83) low" unmarried 0.3940" 1.46(1.43-1.52) 0.3856" 1.47(1.41-1.53) o.2n3 1.32(1.27.37) H!gh"unmarried 0.4046" 0.3951" 0.2427" 1.28(1.21-1.35) Medical risk None {r) 1.00 1.00 One or more 0.2666" 1.31(1.27-1.35) 0.2309" Medical risk"raca White" no risk {r) 1.00 1.00 no risk 0.1005" 1.11(1.07-1.14) 1.1683" 1.18(1.15-1.22) Black no risk 0.5855" 0.6804" 1.98(1.86-2.09) Other no risk 0.4682" 0.5431" 1.72(1.63-1.82) Whltewlrisk 0.2686" 1.31(1.27-1.35) 0.2309" 0.1687" 0.2079" 0.6922" 2.QQ!1.87-2.14) 0.7364" 0.5912" 1.81(1.67-1.98) 0.6712" 1.00 0.8383" 2.31 (2.22-2.41) 1.00 0.3045" 1.36(1.32-1.40) .{).6227" Smokln a e No sm"20 !rl 1.00 No smoke"<19 0.0938" No smoke">35 1.00 79

PAGE 97

Table 4.8. Estimated Coefficients and Odds Ratios tor SGA and RaceiEthnlclty (continued) Variable Model1 Model2 Model3 Mode14 P OR P OR p OR p OR Smoke"<19 -{).0032 1.00 Smoke"20-34 0.1505 1.00 Smoke >35 0.4433" 1.58(1.27-1.91) Smoki "race White"no smoke (r) 1.00 White"smoke 0.6878" 1.99(1.65-2.40) Hisp"smoke Black"smoke Other" smoke Hisp"no smoke Black"no smoke Other"no smoke (r) indicates relerenca category p <.0001. t p <.001. t:ps01 p <.05. 0.6258" 1.87(1.54-2.28) 0.38871 1.48(1.18-1.84) -1.1117" 0.33(0.32-{).34) 0.78(0.69-{).88) 1.0281' 2.8(2.55-3.06) 2.3260" 10.24(8.75-11.97) Risks for SGA are similar to those of LBW, but their effects are generally more attenuated. The contribution of some risk factors is not unexpected based on prior studies: o First babies have higher odds (1.58) of being SGA o As with LBW, SGA is sensitive to elevation, showing a monotonic increase with residence at elevation above 5000 feet (1.06-2.30) o Being unmarried increases the risk of SGA (1.32) o Adequacy of prenatal care is not strongly associated with SGA. Inadequate care in Model 4 is not significantly different than adequate care and intermediate care has slightly lower odds than adequate care (0.97). Even adequate plus prenatal care is not a strong marker for SGA (1.14) o The presence of medical risks associated with pregnancy has a relatively weak association with SGA (1.26) o Weight gain of >40 pounds is very protective against SGA gaining more than 40 pounds reduces the odds of SGA by 46% o Age was not significant in building the model for SGA by race/ethnicity, but it was retained based on the literature. Age is significant only in Model 4 and raises the odds of SGA minimally (1 0%) and only for teen mothers Nevertheless, some results are unanticipated: o Smoking increases the odds of SGA by a greater margin than it does for LBW (2.31 compared with 2.14) 80

PAGE 98

o The odds of SGA with weight gain of less than 16 pounds increases by only 36% com pared with 134% for LBW o Low education (less than high school) is not significant in any model; some high school is associated with slightly elevated odds of SGA (15%) o Intermediate prenatal care (one level below adequate) does not increase the odds of LBW (0.86) In the fully saturated model, the factors having the greatest predictive power for increased odds of LBW are race (Black 1.98, Other 1.72), residence above 8000 feet (1.53 2.30), and smoking (2.31 ). Interactions highlight the effect of some of these risks for SGA. While smoking is a risk factor for SGA in general, Whites who smoke have higher odds of delivering an SGA baby (1.99) than Hispanics who smoke (1.87) or Blacks (1.48). The results for the interaction of smoking and Other are perplexing. Other*non-smoking has an odds of 1 0.24, while Other* smoking has an odds of 0.33. These odds are a result of linear combinations of other variables. Repeated checks disclose no coding or computational error. The interaction of race and medical risk is similar to that for LBW. In Model 4, White mothers with at least one medical risk have slightly higher odds of SGA (1.26) compared to White mothers with no medical risks. Even though more Hispanics have a poorer medical risk profile than Whites, Hispanics with at least one medical risk have an elevated odds of SGA (1.23) compared to White mothers with no risk, but slightly lower odds than those of Whites with medical risks (1.26). The ranking of odds of SGA by race/ethnicity for mothers with one or more risks places Hispanics first (1.23), then Whites (1.26), Others (1.96), and Blacks (2.09), thus supporting an epidemiological paradox in favor or Hispanics for SGA. Large for Gestational Age Table 4.9 reports unadjusted and adjusted odds ratios of LGA by race/ethnicity. For LGA, Hispanics have equal odds ratios to those of Whites in Model 2, and have statistically significant lower odds ratios than Whites for Models 1, 3 and 4, although the differences are small, ranging from 5-7%. Blacks and Others have much lower odds ratios for LGA than either Whites or Hispanics ranging from 32% to 43% decreased odds of LGA, making LGA a White and Hispanic phenomenon. Although the fit is adequate, is discriminating power (0.650) is low. 81

PAGE 99

Table 4.9. Unadjusted and adjusted odds ratios (95% Cl) of LGA by racelethnicity Race Non-Hispanic White Hispanic Black Other c statistic Hosmer Lemeshow Model 2 Model 3 Model1 Adjusted for Additional Race/Ethnicity Only Demographic & Adjustment for Socioeconomic Medical Position Conditions Odds Ratios (Confidence Interval 95%) 1.00 1.00 1.00 0.93 (0.90-0.97) *0.97 (0.93-1.01) 0.95 (0.92-0.99) 0.57 (0.52-0.62) 0.63 (0.57-0.69) 0.61 (0.56-0.68) 0.68 (0.62-0.74) 0.68 (0.62-0.74) 0.68 (0.62-0.74) 0.522 0.602 0.605 0.9999 0.0403 0.0084 Odds not significantly different from 1.00. Model4 Additional Adjustment for Health Behaviors 1.00 0.95 (0.91-0.99) 0.59 (0.54-0.65) 0.68 (0.62-0.74) 0.650 0.1912 Figure 4.6 reports the odds ratios of LGA by race/ethnicity and model. Odds of LGA behave differently than odds of LBW, SGA, or preterm birth. As noted above, Whites have the highest odds of LGA, followed closely by Hispanics. Blacks and Others have much lower odds of LGA. For the category of Others, the odds remains the same across all models. 1.20 ltl 1.00 0 0.80 i a: 0.60 t:=== >E r---ltl "tl 0.40 "tl 0 0.20 0.00 2 3 Models -4 -+-LGA NonHispanic White -LGA Hispanic LGA Black Figure 4.6. LGA ODDS RATIOS BY RACE/ETHNICITY AND MODEL Table 4.1 0 reports the significant estimated coefficients and odds ratios for each variable. Unlike the situation with LBW, preterm birth, and SGA, there are no significant interactions among any of the variables, so the odds remain relatively constant across all models. In addition, adequacy of prenatal care was not significant in the model building for LGA, but that variable is retained nonetheless based on the literature. 82

PAGE 100

Table 4.1 0. Estimated coefflclenta and odd a ratloa for LGA and racalethnlclty Variable Model1 Model2 Model3 Model4 OR OR OR OR Race White(r) 1.00 1.00 1.00 1.00 Hil!l2anic -Q.0690' -Q.0353 1.00 -Q.0493 0.95(0.92-Q.99) -Q.0556! 0.95(0.91-Q.99) Black -Q.5645' 0.57(0.52-Q.62) -Q.4697' -Q.4671' 0.61 (0.56-Q.66) -Q.5268' Other -Q.3910' 0.68(0.82-Q.74) -o.3889' 0.66(0.62-Q.74) -Q.3904' 0.68(0.62-Q.74) -Q.3881' 0.66(0.62-Q.74) A e 20-34 (r) 1.00 1.00 1.00 :S19 -Q.4007' 0.67(0.62-Q. 72) -Q.3965' 0.67(0.63-Q.72) -Q.4396' 0.64(0.60-Q.69) >35 0.1844' 0.1744' 0.1812' 1.20(1.51-1.25) Pa Low(r) 1.00 1.00 1.00 First -o.4076' 0.67(0.64-Q.69) -Q.4089' 0.66(0.64-Q.69) -Q.4958' 0.61(0.59-Q.63) High o.onot 1.08(1.03-1.13) 0.0723! 1.08(1.03-1.12) 0.1189' Elevation <5000 (r) 1.00 1.00 1.00 5000-5999 -Q.0874' 0.92(0.88-Q.96) -Q.0930' 0.91(0.87-Q.95) -Q.1253' 0.88(0.85-Q.92) 6000 6999 -o.2357' -Q.2316' -Q.2659' 0.77(0.73-Q.81) 7000 7999 -Q.4849' 0.62(0.56-Q.66) -Q.5012' 0.61 (0.55-Q.67) -o.5254' 8000 8999 -o.7565' 0.47(0.40-Q.56) -o.7586' 0.47(0.40-Q.55) -Q.7972' 0.45(0.38-Q.53) >9000 -1.0688' 0.34(0.27-Q.44) -1.0693' 0.34(0.27-Q.44) -1.0971' 0.33(0.26-Q.43) Education HS grad (r) 1.00 1.00 1.00 No high school 0.1440' 0.12661 0.1869' 1.21(1.13-1.29) Some high school -o.06490.94(0.89-Q.99) -o.0745t -Q.0114 1.00 Marital status Married r Not married -Q.2589' 0.77(0.74-Q.81l -Q.2650' 0.77(0.74-Q.BO) -Q.2251' o.8Q!O.n-o.83l Prenatal care Adeguate (r) 1.00 1.00 1.00 lnadeguate -Q.1158' -Q.1371' 0.87(0.83-Q.92l -o.0738t Intermediate 0.0016 1.00 0.00027 1.00 0.0095 1.00 Adeguate elus -Q.0419 0.96(0.92-1.00) -Q.0635t 0.95(0.91-Q.98) -Q.0449 Medical risk None (r) 1.00 1.00 One or more 0.1723' 0.1865' 1.21(1.17-1.25) Smokin 1.00 -o.8560' 1.00 -Q.2649' 0.77(0.72-Q.81) 0.8281' 2.29(2.21-2.37) (r) indicates reference category p <.0001. t p <.001. p:S01 p<.05. As is expected from the medical literature discussed in Chapter 1 at pages 6-8, o Higher age and high parity increase odds of LGA (1.20 for higher age; 1.13 for high parity) o Like LBW and SGA, elevation has a dose-response relationship with LGA, except the effect is inverse: as altitude increases, odds of LGA decrease (0.88 0.33) o The presence of medical risks increases odds of LGA by 21% 83

PAGE 101

o Smoking is not associated with higher odds of LGA: mothers who smoke during pregnancy have 57% lower odds of LGA o High weight gain raises odds of LGA by 129% (2.29); low weight gain is protective, lowering odds of LGA by 23% There are some surprising results. o In Model 4 some high school is the same as high school graduate o In contrast to LBW, preterm birth, and SGA, being unmarried is protective (odds of LGA decline to 0.80) o Neither inadequate nor adequate plus prenatal care is a marker for LGA odds of LGA are lower by 4-7%; intermediate prenatal care is not statistically different than adequate care In the fully saturated model, the factor having the greatest predictive power for increased odds of LGA is high weight gain (OR 2.29). The most predictive protective factors are first birth (OR 0.61 ), increasing elevation of residence before birth (OR 0.88 0.33), low weight gain (OR 0.77), and being unmarried (OR 0.80). Because gestational diabetes and delivering a previous infant of 4000 or more grams are specific predictors of LGA, it is instructive to examine the distribution of these two factors by race/ethnicity and to run the fully saturated model replacing the more generic medical risk category with gestational diabetes and previous large infant. Hispanics have the highest frequency of previous infants weighing 4000+ grams, and the second highest frequency of gestational diabetes (second to Whites). Interestingly, this model is slightly more discriminating with a c statistic of 0.661 (compared with 0.650) and the predictive power of both specific risk factors far exceeds the predictive power of medical risks generally: gestational diabetes 2.23 (2.07-2.39); previous infant 4000+ grams 5.42 (4.96-5.93). The odds of LGA by race/ethnicity, however, change very little. Odds for Hispanics fall slightly (0.95 to 0.93) as do odds for Others (0.68 to 0.66). Odds for Blacks remain the same (0.59). Discussion of Aim 1 Hispanics have worse risk profiles than Whites or Blacks for the frequency of teen births, low educational attainment, inadequate and intermediate prenatal care, low weight gain, the presence of one or more medical risk factors, and the specific risk of having a previous infant who weighed more than 4000 grams at birth. In addition, Hispanics have a risk profile that is inferior to Whites for low and high parity and recommended weight gain. In contrast, Hispanics report the lowest frequency of smoking. Overall, the data on risk factors 84

PAGE 102

suggest that Hispanics have poorer social, demographic, and medical risk factors compared to other races but better self-reported smoking behavior compared with non-Hispanic Whites and Blacks. Yet, after controlling for all risk factors, Hispanics have odds of LBW and SGA that are more similar to Whites (and much lower odds than Blacks). In light of the disparity of socioeconomic position between Hispanics and Whites and the somewhat worse social profile of Hispanics compared with Blacks, an odds ratio of 1.18 is considered to reflect a relatively positive outcome. Hispanics also have the same odds of preterm birth (and much lower odds than Blacks), and slightly lower odds of LGA than Whites have, although odds of LGA for Whites and Hispanics are much higher than for Blacks and Others. Table 4.11. Comparison of fully adjusted odds ratios of birth outcomes by racelethnicity LBW Preterm Birth SGA LGA White 1.00 1.00 1.00 1.00 Hispanic 1.18 (1.13-1.23) *1.01 (0.98-1.05) 1.18 ( 1.15-1.12) 0.95 (0.91-0.99) Odds not significantly different from 1.00. 2.5 2 Ill .!i! 1.5 .. Ill a: 8 0.5 0 LBW Preterm Birth Outcome SGA Black 2.16 (2.01-2.33) 1.42 (1.34-1.51) 1.98 (1.86-2.09) 0.59 (0.54-0.65) LGA Other 1.64 (1.52-1.77) 1.16 (1.09-1.23) 1.72 (1.63-1.82) 0.68 (0.62-0.74) -+-White -Hispanic Black """""*"""" Other Figure 4.7. FULLY ADJUSTED ODDS RATIOS BY RACE/ETHNICITY In addition to suggesting that the epidemiological paradox exists for Hispanics, Aim 1 confirms the positive linear association of elevation with LBW and SGA and the negative linear association of elevation with LGA. Education has little effect on any of the outcomes a result that contradicts the usual association of higher education and better health, but which 85

PAGE 103

is consistent with studies showing that education is not necessarily related to better outcomes among Hispanics (Acevedo-Garcia eta/. 2005; Gould et a/. 2003). Prenatal care is also less clearly associated with outcomes. Adequate plus care is a marker for all poor birth outcomes, although most markedly for preterm birth. Hispanics enter prenatal care later than other groups, but inadequate and intermediate levels of care are less clearly negatively associated with heightened adverse outcomes. The interaction of race/ethnicity and medical risk for both LBW and SGA show Hispanics with lower odds of the respective outcomes, followed in ascending order by Whites, Others, and Blacks. Although the paradox can be said to exist for LGA, because Hispanics and Whites have about the same odds, both are well above the odds for Blacks and Others. Hispanics have slightly lower frequency of gestational diabetes than Whites, but that may be a result of the fact that more Hispanics have inadequate and intermediate prenatal care, which may reflect under-diagnosis of gestational diabetes in that population. More surprising is the protective effect of being unmarried on LGA. Resource-based theories of social determinants of health suggest that being married improves outcomes, but this is not the case for LGA. Examining only those non-interacting factors that increase odds of the outcome by 50.0% or more, Table 4.12 indicates which factors influence each outcome at that level. Table 4.12. Comparison of contributing risk factors by birth outcome and racelethnlclty LBW Preterm Birth SGA LGA Race/Ethnicity Black Race/Ethnicity Other Adeguate + PNC Inadequate PNC Low weight gain High weight gain Smoking Medical risks First birth Elevation As a check on the effect of SGA and LGA on birth outcomes, appropriate for gestational age (AGA) was also tested (babies born neither SGA nor LGA). Because AGA is a desired outcome, odds ratios lower than 1.00 indicate less favorable outcomes. As can be seen in Table 4.13, Hispanics are second to Whites in their odds of having an AGA baby, while Blacks and Others are much more likely to have a baby that is not appropriate for gestational age. These results confirm, again, that the epidemiological paradox exists for Hispanics in Colorado. 86

PAGE 104

Table 4.13. Comparison of fully adjusted odds ratios of AGA by race/ethnlcity AGA White 1.00 Hispanic 0.96 (0.93-0.98) Black 0.72 (0.69-0.75) Other 0.78 (0.75-0.82) This model has adequate fit (Hosmer Lemeshow 0.3804), but little discriminating value (c=0.564). Aim 2 Aim 2 examines birth outcomes of mothers of Mexican origin by her place of birth {nativity), using the same candidate variables as are used in Aim 1. Comparison of Risk Factors of Mothers of Mexican Origin by Nativity Aim 2 tests whether the place of birth of mothers of Mexican origin in Colorado affects their risk profiles, and whether there is congruence between risk factors and birth outcomes. Based on the 2000 Census, Mexican-born mothers are poorer and less well educated than U.S.-bom mothers of Mexican origin. Prior population studies report that foreign-born Mexican immigrants to the U.S. enjoy unexpectedly better outcomes for low weight-associated birth outcomes. No national studies of LGA as an outcome for mothers of Mexican origin were found. As with Aim 1, the null and alternative hypotheses for each risk factor are: H40 : each risk factor is independent of race/ethnicity. H4A: each risk factor is related to race/ethnicity. Table 4.14 reports frequencies of each risk factor by nativity and the significance of any differences using Pearson's chi-square. A risk factor is considered related to nativity if ps 0.05 {Gould eta/. 2003). Mexican-born mothers have higher frequencies of social risk factors for low weight-associated outcomes and two LGA-specifi.c medical risk factors than U.S.-born Hispanics of Mexican origin, shown by underlining in Table 4.14, in the following categories: o Older mothers o Much lower education levels o Inadequate and intermediate levels of prenatal care o Presence of one or more medical risk factors o Low weight gain o Gestational diabetes o Previous 4000+ gram infant 87

PAGE 105

In contrast, Mexican-born mothers report lower frequency of smoking, drinking alcohol during pregnancy, high weight gain, and preexisting diabetes. They are also more likely to be married, and they have lower frequencies of teen births and first births. As shown in Table 4.14, the frequency of each risk factor is significantly different by nativity (p<.0001 for all factors except for preexisting diabetes p=0.0198). Figure 4.8 displays the same information using a bar graph. Table 4.14. Percent frequency distribution of risk factors of mothers of Mexican origin by nativity 2D00-2005 Characteristic Total U.S.-Bom Mexican-Born p-value Age of mother :S19 17.85% 24.37% 13.88% 20-34 75.06% 69.99% 78.15% 7.09% 5.64% 7.97% <0.0001 Parity First 36.67% 39.70% 34.81% Low 44.65% 40.20% 47.36% High 18.68% 20.10% 17.82% <0.0001 Altitude (ft sea level) <5000 20.09% 30.19% 13.93% 5000-5999 68.51% 57.47% 75.24% 6000-6999 8.32% 10.38% 7.06% 7000-7999 2.03% 1.38% 2.43% 8000-8999 0.38% 0.30% 0.43% >9000 0.67% 0.29% 0.91% <0.0001 Education (mother) < 9 yrs 22.57% 4.01% 33.89% 9-11 yrs 33.72% 31.23% 35.24% 12+ yrs 43.71% 64.76% 30.88% <0.0001 Prenatal care Inadequate 23.59% 18.62% 26.62% Intermediate 19.56% 16.21% 21.60% Adequate 36.90% 39.24% 35.47% Adequate Plus 19.95% 25.93% 16.31% <0.0001 Marital status Married 61.27% 52.66% 66.53% Unmarried 38.73% 47.34% 33.47% <0.0001 Smoking No 96.08% 92.66% 98.77% Yes 3.92% 8.34% 1.23% <0.0001 Alcohol drinker No 99.50% 99.13% 99.72% Yes 0.50% 0.87% 0.28% <0.0001 Weight gained <151b 15.61% 12.98% 17.22% 15-40 lb 72.45% 71.39% 73.09% >401b 11.94% 15.63% 9.69% <0.0001 88

PAGE 106

Table 4.14. Percent frequency distribution of risk factors of mothers of Mexican origin by nativity 2000-2005 (continued) Characteristic Total U.S.-Born Mexican-Born Medical risk factors None 62.84% 29.37% 41.91% One or more 37.16% 70.63% 58.09% Gestational diabetes 3.25% 2.57% 3.67% Preexisting diabetes 0.45% 0.52% 0.41% Prev infant 4000+ gr 1.28% 0.40% 1.81% A'ev 4000+ g A'eexist Diabetes I Ges t Diabetes Hgh Weight .. I I Rec Weight (r) Low Weight Drinker Srroker r 11.1edical Rsk Adequate + RIC I .. Adequate (r) RIC 0 lnterrrediate RIC I Mexican-Born ii hadequate RIC a: I U.S.-Born Uurarried HS Grad (r) Some HS I NoHS r-I I Hgh Parity Low Parity (r) Rrst Birth >34 --I I 20-34 (r) Teen Birth l....i.. I I 0 10 20 30 40 50 60 70 80 90 Percentage Figure 4.8. DISTRIBUTION OF RISKS BY NATIVITY (r) designates reference category 89 p-value <0.0001 <0.0001 0.0198 <0.0001

PAGE 107

Frequency of Adverse Birth Outcomes by Nativity Although the risk profile of Mexican-born mothers is something of a mosaic, the social gradient of health and their generally poor risk profiles suggest that they should have higher rates of the adverse birth outcomes under study than U.S.-born Hispanics of Mexican origin. However, prior research shows that Hispanics, particularly foreign-born Hispanics, have paradoxically better LBW and SGA outcomes than U.S.-born women Hispanics (Rosenberg eta/. 2005; Frisbie & Song 2003; Gould eta/. 2003; Singh & Yu 1996). The null and alternate hypotheses for incidence of birth outcomes unadjusted for risk factors are: H50 : each birth outcome (LBW, preterm birth, SGA, and LGA) is independent of nativity. HSA: each birth outcome (LBW, preterm birth, SGA, and LGA) is related to nativity. Pearson's chi-square is used to test the relationship between each birth outcome and nativity. On the basis of frequencies of risk factors by nativity, particularly education, less than adequate prenatal care, and low weight gain, it is surprising that Mexican-bam mothers fare so well when examining the frequencies of LBW, preterm births, and SGA. However, the paradox does not hold for LGA, for Mexican-born mothers have a higher frequency of LGA, even while they also have higher frequencies of low weight gain and lower frequencies of high weight gain. These unadjusted frequencies indicate the importance of including LGA in the analysis. While Mexican-born mothers demonstrate the "paradox" with respect to LBW, preterm birth, and SGA, they do not demonstrate the paradox with respect to LGA. Table 4.15. Percent frequency of LBW, pretenn birth, SGA, and LGA by nativity 2000.2005 Birth Outcome All Mexican U.S.Bom Mexican-Born p-value Origin Mexican Origin LBW 6.44% 7.67% 5.69% <0.0001 Preterm births 7.23% 8.19% 6.64% <0.0001 SGA 12.94% 15.32% 11.49% <0.0001 LGA 6.14% 4.69% 7.02% <0.0001 90

PAGE 108

18 16 Jt 14 12 / \ --+--U.S.-Born .. / ..... \ J 10 '8 ....- \ Mexican-6 Born 4 2 0 cPt-0t-v q-.0 Figure 4.9. UNADJUSTED FREQUENCIES BY NATIVITY Odds Ratios of Birth Outcomes by Nativity As was done for Aim 1, multiple logistic regression for each dichotomous outcome was performed using the same nested model design. Model 1 produces odds ratios based on nativity alone. Adjusting for demographic and socioeconomic characteristics in Model 2 accounts for variables contributing to the social gradient. In Model 3 medical conditions associated with pregnancy are added, to account for medical risks and to test, indirectly, the healthy migrant explanation for any paradox. Finally, in Model 4, the pregnancy-associated health behaviors of smoking and weight gain are added, to examine the healthy immigrant hypothesis The reference group is U.S.-born mothers of Mexican origin. The risk profiles of Mexican-born mothers suggest that they should have poorer birth outcomes than U.S-born Hispanic mothers (Table 4.14). If nativity is independently associated with birth outcomes and if Mexican-born mothers have odds of adverse outcomes close to or lower than those of U.S.-born mothers of Mexican origin, the data suggest that the paradox does exist, at least for these outcomes at the compositional level. H60 : nativity does not predict each birth outcome (LBW. preterm birth, SGA, and LGA) after adjusting for risk factors. H6A: nativity predicts each birth outcome after adjusting for risk factors. Summary results are reported for each outcome in Tables 4.16-4.19. Low Birth Weight Table 4.16 reports the summary odds of LBW, including confidence intervals at 95%, for each model. Also reported are the c statistic and the Hosmer-Lemeshow Goodness of Fit statistic. 91

PAGE 109

Table 4.16. Unadjusted and adjusted odds ratios (95% Cl) of LBW by nativity Nativity U.S.-Bom Mexican Mexican-Born c statistic Hosmer-Lemeshow Model1 Nativity Only Model2 Adjusted for Demographic & Socioeconomic Position Model3 Additional Adjustment for Medical Conditions Odds Ratios (Confidence Interval 95%) 1.00 1.00 1.00 0.73 (0.69-0.77) *0.95(0.85-1.07) *0.91 (0.81-1.02) 0.539 0.669 0.681 0.1166 0.1331 Odds not significantly different from 1.00. Model4 Additional Adjustment for Health Behaviors 1.00 *0.93 (0.83-1.05) 0.710 0.5107 Mexican-born mothers have lower odds of LBW than mothers of Mexican origin who are born in the U.S. before any adjustments. In Models 2, 3, and 4, after adjusting for all factors, Mexican-born mothers have the same odds as U.S.-born mothers of Mexican origin. Despite their less advantageous risk profiles, Mexican-born mothers have the same odds of LBW U.S.-born mothers of Mexican origin, demonstrating the existence of the paradox based on these individual level characteristics. This model has a modest degree of discrimination (c statistic= 0.710) and adequate fit. 1.20 1.00 --. . 0 0.80 i a: 0.60 Mexican-"D '; "D 0.40 Born 0 0.20 0.00 1 2 3 4 Models Figure 4.10. LBW ODDS RATIOS BY NATIVITY AND MODEL Table 4.17 reports the significant standardized beta coefficients and odds ratios for each variable in the models of LBW by nativity. Only one interaction is significant for LBW by nativityprenatal care*nativity. 92

PAGE 110

Table 4.17. Estimated coefficients and odds ratios for LBW and nativity Variable Model1 Model2 Model3 Model4 OR OR OR OR Nativ U.S.-bom !rl 1.00 1.00 1.00 1.00 Mexican-born -Q.3205 -Q.0518 1.00 -Q.0999 1.00 -Q.0688 1.00 A e 20-34 !rl 1.00 1.00 1.00 S19 0.0588 1.00 0.0675 1.00 0.1056! 1.11!1.03-1.20! >35 0.2535" 0.20341 0.1723! Pa Low!rl 1.00 1.00 1.00 First 0.3342" 1 .4Q! 1.31-1 .49! 0.3317" 0.4259" High 0.1966" 1.22!1.13-1.32! 0.1869" 1.21!1.11-1.31! 0.1340! Elevation <5000 !rl 1.00 1.00 1.00 5000-5999 0.0483 1.00 0.0216 1.00 0.0498 1.00 6000-6999 -Q.0521 1.00 -o.0278 1.00 0.0117 1.00 7000-7999 0.4551" 1.58!1.32-1.88! 0.4394" 1.55!1.30-1.85! 0.4578" 1.58!1.32-1.89! 8000-8999 0.5072! 0.5249! 1.7Q!1.16-2.47l 0.5711! 1.77!1.21-2.60! >9000 0.6758" 1.97!1.14-1.29! 0.7166" 2.05!1.55-2.70! 0.8078" Education HS grad!rl 1.00 1.00 1.00 No high school 0.1082 1.11!1.02-1.21! 0.0826 1.00 0.0433 1.00 Some high school 0.11641 1.12!1.14-1.29! 0.0971! 1.1Q!1.03-1.18l 0.0769 1.08!1.01-1.16! Marital status Married!rl 1.00 1.00 1.00 Not married 0.1923" 1.21!1.14-1.29! 0.1819" 1.2Q!1.13-1.27l 0.1540" 1.17!1.10-1.24! Prenatal care Adeguate !rl 1.00 1.00 1.00 InadeQuate 0.6775" 1.97!1.74-2.23! 0.6138" 1.85!1.63-2.09! 0.5294" 1.7Q!1.50-1.93l Intermediate -o.oo1o 1.00 -o.0014 1.00 -o.0248 1.00 Adeguate !!Ius 1.2818" 1.2438" 1.2393" 3.45!3.11-3.64! Adeg"U.S.-bom !rl 1.00 1.00 1.00 1.00 lnadeg"U.S.-bom 0.6775" 1.97!1.74-2.23! 0.6138" 1.85!1.63-2.09! 0.5294" 1.7Q!1.50-1.93ll lntermed"U.S.-bom -o.oo1o 1.00 -o.0014 1.00 -o.0248 1.00 Ade91;"U.S.-bom 1.2818" 1.2438" 3.47!3.12-3.85! 1.2393" 3.45!3.11-3.84! lnadeg"Mexican-bom 0.1202 -o.0032 1.00 -Q.1239 0.88!0. 78-1.00! lnterm"Mexican-bom -Q.2728" 0. -o.3380" 0.71 !0.62-Q.82l -Q.3592" 0.7Q!0.61-Q.60l Adeg"Maxican-bom -Q.0518 1.00 -Q.0999 1.00 -Q.0688 1.00 Ade91;"Maxican-bom 1.1398" 3. 1 1.1043" 3.02!2.70-3.37! 1.1263" 3.08!2.78-3.45! Medical risk None !rl 1.00 1.00 One or more 0.4998" 1.65!1.58-1.75! 0.4592" 1.58!1.49-1.68! Smokin 1.00 0.6706" 1.96(1.75-2.19! 1.00 0.7027" 2.02!1.89-2.16! -Q.8424" 0.43(0.38-Q.48) (r) indicates reference category p<.0001. tp<.001. :tps01 p<.05. The contributors and markers for LBW by nativity are the same as those for LBW by race/ethnicity, although the order of their importance and magnitude vary somewhat. 93

PAGE 111

o Odds of delivering an LBW infant increase for teen mothers (1.11) and older mothers (1.19) o First babies have higher odds (1.53) of being LBW o Odds of LBW show a monotonic increase as residence at elevation above 7000 (1.58 2.24) o Being unmarried increases the risk of LBW ( 1.17} o Mothers with inadequate (1.70) and adequate plus prenatal care (3.45) have increased odds of LBW o The presence of medical risks associated with pregnancy is a marker for increased odds of LBW (1.58) o Smoking increases the odds of LBW (1.96) o The odds of LBW with weight gain of less than 16 pounds increases by 102% (2.02) o Weight gain of >40 pounds is very protective against LBW gaining more than 40 pounds reduces the odds of LBW by 57% (0.43) As with LBW by race/ethnicity some results are unanticipated: o Low education (less than high school) is not significant o Intermediate prenatal care (one level below adequate) does not increase the odds of LBW In the fully saturated model, the factors having the greatest predictive power for increased odds of LBW are adequate plus prenatal care (3.45), presence of one or more medical risks (1.58), smoking (1.96}, and first parity (1.53). The one interaction, adequacy of prenatal care by nativity, supports the paradox in favor of Mexican-born mothers. Mothers reporting adequate prenatal care have the same odds regardless of nativity (1.00). For those reporting inadequate care, U.S.-born mothers have more than 80% greater odds of LBW than Mexican-born mothers (1.70 compared with 0.88). Mexican-born mothers with intermediate prenatal care are 30% less likely (0.70) to have an LBW baby than U.S.-born mothers with adequate or intermediate care (1.00 for both}. And even with adequate plus prenatal care, a marker for a more complicated pregnancy, Mexican-born mothers have lower odds compared with U.S.-born mothers of Mexican origin (3.08 compared with 3.45). Preterm Birth Table 4.18 reports unadjusted and adjusted odds ratios of preterm birth by nativity. Interestingly, Model 2 has adequate fit and modest discrimination (c statistic = 0.711), but although Models 3 and 4 increase discrimination somewhat, the fit drops to inadequate, 94

PAGE 112

suggesting that medical risk and health behaviors are not particularly relevant to the overall fit of the models. Table 4.18. Unadjusted and adjusted odds ratios (95% Cl) of preterm birth by nativity Nativity U.S.-Born Mexican Mexican-Born c statistic Hosmer-Lemeshow Model1 Nativity Only Model2 Adjusted for Demographic & Socioeconomic Position Model3 Additional Adjustment for Medical Conditions Odds Ratios (Confidence Interval 95%) 1.00 1.00 1.00 0.80 (0.76-Q.84) 1.14 (1.01-1.30) *1.08 (0.96-1.23) 0.527 0.711 0.726 0.2214 0.0016 Odds not significantly different from 1.00. Model4 Additional Adjustment for Health Behaviors 1.00 *1.08 (0.95-1.23) 0.743 0.0082 Figure 4.11 shows the change in odds rations for preterm birth across models. After increasing from 0.80 in Model 1 to 1 .14 in Model 2, in Models 3 and 4, Mexican-born mothers have the same odds of preterm birth as those mothers of Mexican origin born in the U.S. 1.20 ........ 1.00 . Ill _, U.S.-Born 0 0.80 ii a: 0.80 Ill ---Mexican'a 'a 0.40 Born 0 0.20 0.00 1 2 3 4 Modele Figure 4.11. PRETERM BIRTH ODDS RATIOS BY NATIVITY AND MODEL Table 4.19 reports the significant standardized beta coefficients and odds ratios for each variable in the models for preterm birth by nativity. Preterm birth by nativity has two interactions: prenatal care*nativity and prenatal care*weight. 95

PAGE 113

Table 4.111. Eetlmated coettlclenta end odde retloe for preterm birth end nativity Variable Mod811 Mod812 Mod813 Mod814 I! OR I! OR I! OR I! OR Nativi U.S.bom !rl 1.00 1.00 1.00 1.00 Mexican-born .().2272" 0.8Q!0.76-35 0.2684" 1.31!1.19.441 0.2096" 0.18731 1.21!1.09.331 Pa Low!rl 1.00 1.00 1.00 First 0.1705" 0.1678" 1.18!J.11-1.26I 0.2444" High 0.2274" 0.2179" 0.1808" 1.2Q!1.11.291 Elevation <5000 !rl 1.00 1.00 1.00 5000-5999 0.0239 1.00 .().0043 1.00 0.0189 1.00 6000-6999 .().17901 .().15051 -9000 0.1291 1.00 0.1678 1.00 0.2320 1.00 Education HS grad !rl 1.00 1.00 1.00 No high school 0.0565 1.00 0.0905 1.00 .().0029 1.00 Some high school 0.1114! 1.12!1.05.191 0.0905 1.1Q!1.03-1.171 0.0784 Marital status Married !rl 1.00 1.00 1.00 Not married 0.1220" 1.13-1.07.201 0.11111 1.12!1.05.181 0.10001 1.11!1.04.171 Prenatal care Adeguate !rl 1.00 1.00 1.00 lnadeguate 1.0212" 2. 78!2.44-3.171 0.9482" 2.58!2.26-2.951 0.9978" 2.27!2.35-3.131 Intermediate .().0394 1.00 .().0401 1.00 .().0887 1.00 Adeguate 1.8356" 6.27!5.61-7.011 1.818r 6.1 1.8560" Adeg'U.S.bom !rl 1.00 1.00 1.00 1.00 1.00 lnadeg"U.S.bom 1.0212' 2.78!2.44.171 0.9482" 2.58!2.26.951 0.3734 1.45!1.13-1.881 lntermed"U.S.bom -
PAGE 114

Table 4.19. Estimated coefficients and odds ratios for preterrn birth and nativity (continued) Variable Model1 p OR lnad"high wt Inter" high wt Adeg"high wt Adeq+"high wt (r) indicates reference calegory p <.0001. tp<.001. tps01 p <.05. Model2 Model3 p OR p OR Model4 p OR -1.3547" 0.26(0.25-Q.27) 4.9105" 0.01(0.01-Q.01) -o.6399" 0.53(0.42-Q.67) 0.3733" 1.45(1.39-1.52) Some of the results are consistent with those of preterm birth by race/ethnicity. For example smoking, marital status, education, altitude, and parity have similar effects. However, there are some differences. o Teen births do not increase the odds of having a preterm birth among the population studied in Aim 2 o The presence of medical risks associated with pregnancy increases the odds of preterm birth by 72%, but this increase is quite a bit lower than the increase for race/ethnicity (127%) As might be expected, interactions provide the most interesting information. Inadequate and adequate plus prenatal care alone are markers for increased odds of preterm birth (inadequate by 127%; adequate plus by 546%), these are somewhat smaller increases in odds than by race/ethnicity. Intermediate prenatal care alone is not significantly different from adequate care. However the interaction of prenatal care and weight gain shows that intermediate prenatal care combined with low weight gain creates an odds ratio of 6.46. Low weight gain in combination with any level of prenatal care, including adequate, increases odds of preterm birth by 2.24 to 6.46. The interaction of prenatal care and nativity shows that Mexican-born women with either adequate or inadequate prenatal care have the same odds as the reference category-U.S.-born mothers with adequate care; Mexican-born mothers with inadequate care (1.00) compare favorably with U.S.-born mothers with inadequate prenatal care (1.45). Although both groups have roughly the same odds if they have adequate plus prenatal care (U.S.-born 3.26 compared with Mexican-born 3.14), Mexican born mothers' odds are lower. 97

PAGE 115

Small for Gestational Age Table 4.20 reports unadjusted and adjusted odds ratios of SGA by nativity. Table 4.20. Unadjusted and adjusted odds ratios (95% Cl) of SGA by nativity Nativity U.S.-Born Mexican Mexican-Born c statistic Hosmer-Lemeshow Moclel1 Nativity Only Moclel2 AdJusted for Demographic & Socioeconomic Position Model3 Additional Adjustment for Medical Conditions Odds Ratios (Confidence Interval 95%) 1.00 1.00 1.00 0.72 (0.69-Q.75) 0.73 (0.70-Q.77) 0.73 (0.70-0.76) 0.540 0.586 0.587 0.0007 0.0025 Moclel4 Additional Adjustment for Health Behaviors 1.00 0.74 (0.71-Q.78) 0.614 0.4307 As shown in Figure 4.12, Mexican-born mothers have much lower odds of SGA than their U.S.-born counterparts. There are no significant interactions, which is one reason the odds ratios remain so stable across the models. The fully saturated model has low discrimination (c statistic= 0.614) but adequate fit. Although Mexican-born mothers have higher frequencies of social risk factors than U.S.-born mothers of Mexican origin, in all cases Mexican-born mothers have 26-28% lower odds of having an SGA baby than Mexican mothers born in the U.S., supporting the existence of the epidemiological paradox in favor of Mexican-born mothers in Colorado. Models Figure 4.12. SGA ODDS RATIOS BY NATIVITY AND MODEL 98

PAGE 116

Table 4.21 reports significant standardized beta coefficients and odds ratios for each variable in the models for SGA by nativity. Table 4.21. Estlma1ed coefficients and odda ratloa for SGA and na11vlty Variable Model1 Nativi U.S.-bom !rl Mexican-born -0.3323" A e 20-34 !rl S19 >35 Pa Low!rl First High Elevation <5000 !rl 5000-5999 6000-6999 7000-7999 8000-8999 >9000 Education HS grad !rl No h!gh school Some high school Marital status Married r Not married Prenatal care Adeguate !rl lnadeguate Intermediate Adeguate E!IUS Medical risk None !rl One or more Smoki (r) indicates reference category p<.0001. t p <.001. tp S01 p <.05. OR 1.00 0. 72!0.69-0. 75) Model2 Model3 OR OR 1.00 1.00 -0.3100" -0.3174" 1.00 1.00 O.On1t 1.o8p.o2-1.14l 0.0784! 1.08!1.02-1.15) -0.0192 1.00 -0.0268 1.00 1.00 1.00 0.3400" 1.41!1.34-1.47) 0.3392" 1.4Q!1.34-1.47) -0.0013 1.00 -0.0034 1.00 1.00 1.00 0.0275 1.00 0.0207 1.00 0.1283! 1.14!1 .05-1.23) 0.1313! 0.3676" 1.47!1.29-1.69) 0.3841" 1.47!1.28-1.68) 0.4428! 1.56!1.17-2.06) 0.4461! 0.8236" 2.28!1.87-2.78) 0.8306" 2.30(1.88-2.80) 1.00 1.00 0.0450 1.00 0.0406 1.00 0.0730! 1.08!1.02-1.13) 0.0690! 1.07!1.02-1.13) 0.1529" 1.17p.12-1.22l 0.1509" 1.00 1.00 0.0950! 1.10!1.04-1.16) 0.0828! 0.0173 1.00 0.0156 1.00 0.1405" 1.15!1.09-1.22) 0.1369" 1.15!1.09-1.21) 1.00 0.0873" Model4 OR 1.00 -0.2979" 1.00 0.10181 1.11p.05-1.17) -0.0493 1.00 1.00 Q.4005" -0.0393 1.00 1.00 0.0361 1.00 0.1536! 1.17!1.06-1 .27) 0.3968" 0.4744! 1.61!1.20-2.15) 0.8704" 1.00 0.0178 1.00 0.0554 0.1276" 1.00 0.0178 1.00 -0.0025 1.00 0.1297" 1.14!1.08-1.20) 1.00 0.0640! 1.07!1.02-1.11) 1.00 0.6501" 1.92!1.76-2.09) 1.00 0.3371" 1.4Q!1.33-1.48) -0.6325" 0.53(0.49-0.57) The model for SGA is uncomplicated by interactions. Similarities and differences with respect to prior analyses include: o In contrast to the results for SGA by race/ethnicity, teen births are significant across all models, although the increase in odds is small (1.11) 99

PAGE 117

o There is a dose-response relationship between increasing altitude of residence and increased odds of SGA in all models starting at 6000 feet (1.17 2.39), compared with race/ethnicity, where the increase in odds began above 5000 feet. o Smoking increases the odds of SGA by 92%, lower than for race/ethnicity (131 %) o First birth increases the odds of SGA by 49% o Low weight gain is associated with SGA, increasing odds by 40%; high weight gain reduces odds by 47%, the same pattern that is observed by race/ethnicity o The presence of medical risks increases the odds of SGA by only 7% o Education does not have the expected effect; having less education does not increase odds of SGA In Model 4 having inadequate or intermediate prenatal care is not statistically different from adequate prenatal care; adequate plus prenatal care is associated with SGA but it raises the odds by only 14%. These results suggest that SGA may not have as clear predictor conditions as LBW and preterm birth may have, and that it may not be diagnosed during prenatal visits and then followed more closely resulting in adequate plus prenatal care. The factors having the greatest predictive power for increased odds of SGA are birth above 7000 feet (1.49 increasing to 2.39 at 9000 feet and above), smoking during pregnancy (1.92), and first birth (1.49). Large for Gestational Age Table 4.22 reports unadjusted and adjusted odds ratios of LGA by nativity. Here the paradox does not hold for Mexican-born mothers. In marked contrast to odds ratios for LBW, preterm birth, and SGA, Mexican-born mothers have 45% higher odds of delivering an LGA baby than U.S.-born mothers of Mexican origin in the fully saturated model. Table 4.22. UnadJusted and adJusted odds ratios (95% Cl) of LGA by nativity Nativity U.S.-Bom Mexican Mexican-Born c statistic Hosmer-Lemeshow Moc:le11 Nativity Only Moc:lel2 AdJusted for Demographic & Socioeconomic Position Moc:lel3 Additional AdJustment for Medical Conditions Odds Ratios (Confidence Interval 95%) 1.00 1.00 1.00 1.53 ( 1.44-1.63) 1.43 ( 1.34-1 .54) 1.41 ( 1 .32-1.52) 0.548 0.616 0.619 0.0111 0.0325 100 Moc:lel4 Additional AdJustment for Health Behaviors 1.00 1.45 (1.35-1.55) 0.649 0.8893

PAGE 118

Figure 4.13 reports the odds ratios of LGA by nativity. Aim 1 showed that Hispanics as a group have slightly lower odds of LGA compared with Whites. But when the Hispanic population is limited to Hispanics of Mexican origin and then compared by nativity, it can be seen that LGA is a phenomenon of recent immigration. 1.80 1.60 1.40 ......... -... 1.20 U.S.-Born ..2 1i 1.00 a: 0.80 Mexican-'a 'a 0.60 Born 0 0.40 0.20 0.00 1 2 3 4 Models Figure 4.13. LGA ODDS RATIOS BY NATIVITY AND MODEL Table 4.23 reports the significant standardized beta coefficients and odds ratios for variable in the models for LGA by nativity. Table 4.23. Estimated coefficients and odds ratios for LGA and nativity Variable Model1 Model2 Model3 Model4 OR OR OR OR Nativi U.S.-bom !rl 1.00 1.00 1.00 1.00 Mexican-born 0.4278" 0.3593" 0.3453" 1.41!1.32-1.52} 0.3692" 1.45!1.35-1.55} A 20-34!rl 1.00 1.00 1.00 S19 -0.4332" 0.65!0.58-0. 72! -0.4302" 0.65!0.59-0. 72! -0.4625" >35 0.3009" 1.35!1.23-1.481 0.2851" 1.33(1.21-1.48} 0.3121" 1.37!1.24-1.50} Pa Low!rl 1.00 1.00 1.00 First -0.3918" 0.68!0.63-0.73! -0.3933" 0.68!0.63-0.73} ..o.4no 0.62!0.58-0.67! High 0.0481 1.00 0.0415 1.00 0.0759 1.08!1.00-1.16} Elevation <5000 !rl 1.00 1.00 1.00 5000-5999 -0.1439" 0.87!0.81-0.93! -0.1627" 0.85!0.79-0.91! -0.1784" 0.84(0.78-0.90! 6000-6999 -0.2965" 0.74!0.66..Q.84} -0.2919" 0.75!0.66..Q.85l -0.3284" 0.75!0.84-0.82! 7000-7999 -0.6162" -0.6250" 0.54(0.42-0.69} -0.8424" 8000-8999 -1.0765! 0.34!0.17 -0.69! -1.0706! -1.0813! 0.34(0.17 -0.691 >9000 -1.2345" -1.2219" 0.30(0.17 -0.511 -1.2848" 0.28!0.16-0.49} Education HS grad !rl 1.00 1.00 1.00 No high school 0.0720 1.00 0.0612 1.00 0.0926 1.1 Qi1.02.191 Some high school 0.0624 1.00 0.0523 1.00 0.0842 1.00 Marital status Married !rl 1.00 1.00 1.00 Not married -0.1455" 0.87(0.81-0.92) -0.1500" 0.86(0.81-0.92) -0.1433" 0.87(0.81-0.92) 101

PAGE 119

Tabla 4.23. Estimated coefficients and odds ratios for LGA and nativity (continued) Variable Modal1 Modal2 Modal3 Model4 @ OR @ OR @ OR @ OR Prenatal care Adequate (r) 1.00 1.00 1.00 Inadequate ..0.1494" 0.86(0.80..0.93) ..0.1762' 0.84(0.78..0.91) ..Q.1034t 0.90(0.83..().98) Intermediate 0.0039 1.00 ..0.0011 1.00 0.0268 1.00 Adequate plus ..0.0364 1.00 ..0.0398 1.00 ..0.0285 1.00 Medical risk 1.00 1.00 0.1858' 1.20(1.14.28) 0.2009' 1.22(1.15-1.30) 1.00 ..0.5167' 0.60(0.48..0.74) 1.00 -.4083' 0.67(0.61..0.73) 0.7928' 2.21 (2.05-2.38) No sm"20-34 (r) 1.00 No smoke'<19 ..0.5384' 0.63(0.57..0.70) No smoke"35 + 0.3338' 1.37(1.24-1.50) Smoke"<19 ..0.9513' 0.48(0.29..0.78) Smoke'20-34 ..0.6575' 0.60(0.46..0.72) Smoke"35+ (r) indicates reference category p <.0001. t p <.001. tps01 p<.05. ..0.2983 In contrast to LGA by race/ethnicity, where Hispanics have about the same odds of LGA as Whites, the second most influential predictor of LGA among women of Mexican origin, after excessive weight gain during pregnancy, is nativity in Mexico (1.45). Interestingly, most of the other variables have about the same influence whether one examines LGA by race/ethnicity or by nativity, suggesting that factors related to nativity may be explanatory. o As with race/ethnicity, low age and first parity are protective for LGA (both by about 35%); being 35 or older raises odds by 37% (20% by race/ethnicity) o As with race/ethnicity, elevation has an inverse dose-response relationship with LGA; as elevation increases, odds of LGA decrease (odds decrease by 16-72%) o The presence of medical risk increases odds of LGA by 22% (21% by race/ethnicity) o High weight gain increases the odds of LGA by 121%, about the same as by race/ethnicity o Education as an aggregate variable is not significant; only in Model 4 lack of high school education increase the odds of LGA (by 1 0%) and only at <.05 significance. 102 1.00

PAGE 120

o As with race/ethnicity, being unmarried is protective against LGA, reducing odds by about 13% (for race/ethnicity, being unmarried is slightly more protective reducing odds by 20%) o Inadequate prenatal care is associated with 10% lower odds of LGA; intermediate and adequate plus prenatal care are not significantly different than adequate prenatal care o Smoking is associated with lower odds of LGA by 33% (57% by race/ethnicity) In Model 4, the factors having the greatest predictive power for increased odds of LGA are high weight gain (2.21) and nativity in Mexico (1.45). Replacing the category of general pregnancy-associated medical risks with the specific risk factors of gestational diabetes, preexisting diabetes, and having a previous infant weighing more than 4000 grams lowers the odds of LGA for Mexican-born mothers slightly compared to the model with the general medical risks (1.42 compared with 1.45), but the point estimates for the specific medical risks then become the highest contributors to odds of LGA. o Gestational diabetes increases the odds of LGA by 163% o Preexisting diabetes increases the odds of LGA by 357% o Previous infant weighing more than 4000 grams at birth increases the odds of LGA by 293% Table 4.24 presents a summary comparison of the fully saturated models using presence of one or more general medical risks and LGA specific risks. Using specific risks improves the discrimination slightly (although the effect is still low) and retains adequate fit. Table 4.24. Model 4 adJusted odds ratios (95"/o Cl) of LGA by nativity Nativity U.S.-Bom Mexican Mexican-Born c statistic Hosmer-Lemeshow Model 4 Model 4 General Medical Risk Specific Medical Risks for LGA Odds Ratios (Confidence Interval 95%) 1.00 1.45 (1.35-1.55) 0.649 0.8893 1.00 1 .42 ( 1.32-1 .53) 0.668 0.3409 Given the higher risk of Mexican-born mothers of having an LGA baby, it is appropriate to explore factors that might explain these unexpected results. As can be seen from Figure 4.14, the frequency of LGA by nativity by year within the study period shows a 103

PAGE 121

pattern whereby Mexican-born mothers consistently have about 2% greater frequency of LGA babies than U.S.-born mothers, except for 2005, when the gap narrows to about 1.7%. 8 7 ........... .. -. 6 5 - Mexican--....._......-c Born I 4 IL 3 2 1 0 '!,-# .... #' '!,-# Year of Infant Birth Figure 4.14. FREQUENCY OF LGA BY NATIVITY BY YEAR Figures 4.15 4.17 compare demographic risk factors thought to affect the rate of LGA: weight gain, age, and parity. As each increases, LGA increases, and the frequency of LGA for Mexican-born mothers increases more dramatically than for U.S.-born mothers. 14 12 10 ----1'/exican-J 8 Born 8 . fl / -+-U.S. Born _, 4 ...2 0 Low Medium Hgh (rei) Weight Gain Figure 4.15. FREQUENCY OF LGA BY WEIGHT GAIN 12 10 8 -----IVexican-i . ... Born 6 /"" _____...._U.S. Born 4 v 2 -0 <20 20-34 35+ (ref) Age Categorlee Figure 4.16. FREQUENCY OF LGA BY AGE 104

PAGE 122

10 8 .... ,. .... .. . --+--Mexican-. .. 6 Born c I #. 4 .. ..,.,Born .. 2 0 Rrst Low Hgh (ref) Parity Figure 4.17. FREQUENCY OF LGA BY PARITY As shown in Figure 4.18, although Mexican-born mothers have higher frequencies of gestational and preexisting diabetes and a previous 4000+ gram infant, the differences between the two population groups are quite constant. 8 7 6 -5 i l 4 3 2 1 0 ... ... ... Gest Pre-exist Prev Diabetes Diabetes 4000+ gram Fisk Factors I -M!xican! Born 1 ----U.S. Born 1 Figure 4.18. FREQUENCY OF LGA-SPECIFIC RISKS Discussion of Aim 2 In accord with previous studies, the population of "Hispanic" mothers in Colorado is not homogeneous. Looking at race/ethnicity alone misses the disproportionate impact of LGA on Mexican-born mothers. Stratifying by Mexican origin and then by nativity shows that Mexican-born mothers contribute to the lower odds for SGA but have higher odds LGA, which tends to be masked if only race/ethnicity is considered. 105

PAGE 123

J ca a: 8 Table 4.25. Fully adjusted odds ratios of birth outcomes by nativity LBW Preterm Birth SGA LGA U.S.-Bom Mexican Origin 1.00 1.00 1.00 1.00 Mexican-Born *0.93 *1.08 0.74 1.45 Odds not significantly different from 1.00. 1.6 1.4 1.2 ...... ...... 0.8 0.6 ..... . . ..... ..... +------------_ ....;''.:..o......:=.:..--------! I Mexican-Born I 0.4 0.2 0 LBW Preterm SGA LGA Birth Outcome Figure 4.19. FULLY ADJUSTED ODDS RATIOS BY NATIVITY Aim 2 demonstrates that the epidemiological paradox exists for Mexican-born mothers at the low weight end of the birth outcome spectrum, but not for LGA. It also confirms the positive linear association of elevation with LBW and SGA and the negative linear association with LGA. As with Aim 1, education has little effect on any of the outcomes, a result that conflicts with studies that associates higher levels of education with better birth outcomes as part of the social gradient. This is especially evident for mothers of Mexican origin, who have even lower educational levels than Hispanics generally. Also consistent with Aim 1, prenatal care is less obviously associated with outcomes, as adequate and intermediate levels of care are less clearly associated with heightened adverse outcomes. Mexican-born mothers have higher frequencies of LGA with higher weight gain, higher parity, and higher age of pregnancy than U.S.-born mothers of Mexican origin. It is worth noting that the reports of lower frequencies of preexisting diabetes in the birth record for Mexican-born mothers (as compared to higher reported frequencies of gestational diabetes and previous infants weighing more than 4000 grams) suggest that preexisting 106

PAGE 124

diabetes may not always be diagnosed in Mexican-born mothers. In light of Mexican-born mothers' higher frequencies of inadequate and intermediate prenatal care, it is likely that they have limited access to health care outside of pregnancy; hence diabetes that is identified during prenatal care and reported as gestational may in some instances be undiagnosed preexisting diabetes. Complementing the discussion of adverse outcomes is consideration of presumed "healthy birth weighr babies those who are born appropriate for gestational age. For the population of all mothers of Mexican origin, 80.92% have AGA babies. Mexican-born mothers have a slightly higher frequency of AGA babies (81.49%) compared with U.S.-born mothers (79.98%). Table 4.26 reports the odds of having an AGA baby using the fully saturated Model 4 with the one interaction for AGA in this population weight*parity. The results of logistic regression for the fully-saturated model show that Mexican-born mothers are 8% more likely to have an AGA baby than U.S.-born mothers of Mexican origin. Thus, although Mexican-born mothers have lower odds of having SGA babies and higher odds of having LGA babies, they also have higher odds of having AGA babies compared to U.S.-born mothers in Colorado. Table 4.26. Comparison of fully adjusted odds ratios of AGA by nativity AGA U.S.-Bom Mexican Origin 1.00 Mexican-Born 1.08 (1.04-1.12) Model has adequate fit (Hosmer-Lemeshow 0.1458), but little discriminating value c=0.540. This study was designed with nested models to test whether any inferences might be drawn about the healthy migrant and healthy immigrant hypotheses. This study does not test the healthy migrant hypothesis directly, because adequate data are not available on rates of the birth outcomes under study among Mexican mothers who remain in Mexico. Nevertheless, those who suggest that Mexican-born mothers who migrate to the U.S. are healthier than those Mexican-born mothers who do not migrate miss the point. It is indeed possible that Mexican-born mothers who migrate to the U.S. may be healthier than those who remain in Mexico, but one particularly appropriate comparison for purposes of the epidemiological paradox is the population of U.S. resident mothers of Mexican origin not those who remain in Mexico. If the healthy migrant hypothesis explains differences in outcomes, odds ratios should change between Models 2 and 3 when medical risk factors are taken into account. In the same manner, if the healthy immigrant hypothesis explains 107

PAGE 125

differences in outcomes, the odds should change between Models 3 and 4 after health behaviors relating to smoking and weight gain are taken into account. Mexican-born mothers have a worse demographic and socioeconomic profile than U.S-born mothers as measured by the frequency of adequate prenatal care, level of education, and higher age at delivery (Table 4.14). Table 4.27 reports the odds of Mexican born mothers compared with the reference group of U.S.-born mothers of Mexican origin by outcome and by model, using general medical risks associated with pregnancy. Table 4.28 reports odds of LGA using adjustments for LGA-specific medical risks in Models 3a and 4a. Differences based on the nature of medical risks are small. Table 4.27. Fully adJusted odds ratios (95% Cl) by nativity and model Nativity LBW Preterm Birth SGA LGA Model1 Nativity Only Model2 AdJust for Demographic & Socioeconomic Position Model3 AdJust for General Medical Conditions Odds Ratios (Confidence Interval 95%) 0.73 (0.69-0.77) 0.80 (0.76-0.84) 0.72 (0.69-0.75) 1.53 (1.44-1.63) *0.95(0.85-1.07) *0.91 (0.81-1.02) 1.14 (1.01-1.30) *1.08 (0.96-1.23) o.73 (0.7o-o.n) o.73 (0.7o-0.76) 1.43 (1.34-1.54) 1.41 (1.32-1.52) Odds not significantly different from 1.00. Model4 AdJust for Health Behaviors *0.93 (0.83-1.05) *1.08 (0.95-1.23) 0.74 (0.71-0.78) 1.45 (1.35-1.55) Table 4.28. Fully adJusted odds ratios (95% Cl) by nativity for LGA using LGA-speclflc medical risks Nativity LGA Model1 Nativity Only Model2 AdJust for Demographic & Socioeconomic Position Model3a AdJust for Medical Conditions Specific to LGA Odds Ratios (Confidence Interval 95%) Model4a AdJust for Health Behaviors 1.53 (1.44-1.63) 1.43 (1.34-1.54) 1.38 (1.29-1.48) 1.42 (1.32-1.53) Model 2 adjusts for demographic characteristics and socioeconomic position, and thus puts Mexican-born women and U.S.-born women of Mexican origin on the same level for the further tests of the healthy migrant and healthy immigrant hypotheses. Model 3 adjusts for general medical risks associated with pregnancy and delivery. One might argue that Mexican-born mothers are somewhat "healthier'' than U.S.-born mothers as measured by the frequency of one or more general medical risks associated with pregnancy because 58.1% of Mexican-born mothers have one or more medical risks 108

PAGE 126

compared with 70.6% borne by U.S.-born mothers. The healthy migrant hypothesis would suggest that odds would change after adjustment for medical risks if Mexican-born mothers are healthier. However, the odds of LBW and SGA remain the same for Mexican-born mothers between Models 2 and 3 when medical risks are adjusted for. They decrease 4% for preterm birth to the same level as U.S.-born mothers and decrease 2% for LGA using the general medical risk model. However, Mexican-born mothers have higher frequencies of medical risks specific to LGA gestational diabetes and previous infants weighing 4000 or more grams. Odds for Mexican-born mothers drop 5% between Models 2 and 3a. The difference in odds after adjusting for these risks remains smallbetween 2-4% depending on outcome and medical risk model. Finally, in the fully-saturated model, after the addition of health behaviors associated with pregnancy weight gain and smoking the odds stay the same for preterm birth and increase only 1% for SGA. The odds increase 2% for LBW but the increase is not statistically significant. The odds increase slightly (4% in both Models 4 and 4a) for LGA, suggesting that the healthy immigrant hypothesis does not overly affect the odds of these outcomes, at least as measured by these two behavioral factors. In short, neither the healthy migrant nor the healthy immigrant hypotheses is supported using this design for measuring the influence of pregnancy related health factors (healthy migrant) or health behaviors (healthy immigrant) for the population of mothers of Mexican origin in Colorado. Before discussing the effect of the contextual variables on outcomes, it is worth testing whether the paradox persists for mothers of Mexican origin at geographic levels below the state. Table 4.29 and Figure 4.20 compare the odds ratios of each of the four outcomes for Mexican-born mothers in Adams County, Denver County, and statewide, with U.S.-born women of Mexican origin as the reference group. All models have adequate fit. For LBW, preterm birth, SGA, and LGA the results for the two counties and the state are consistent, although mothers of Mexican origin in Denver County have higher odds of LGA than mothers in either Adams County or statewide. In addition, the confidence intervals for LBW, preterm birth, and LGA increase at the county level. For AGA, whereas statewide Mexican-born mothers have 8% higher odds of having an AGA baby, in the two counties their odds are the same as U.S.-born mothers. The paradox continues to exist in Adams and Denver counties for low weight related birth outcomes, but not for LGA. 109

PAGE 127

Table 4.29. Comparison of fully adjusted odds ratios of birth outcomes of Mexican-born mothers In Adams and Denver Counties and statewide Adams County Denver County LBW o.89 (0 67-1.17) .19 (0.90-1.57) Preterm Birth o. 96 (0.71-1.30) .1.26 (0.95-1.67) SGA 0 .70 (0.63-0.79) 0 75 (0.67-0.84) LGA 1 .54 (1.30.82) 1.86 (1.55-2.23) AGA 08 (0.99.18) 02 (0 94.12) Odds ratio not significantly different from 1 .00. 2 .00 1 80 1.60 1.40 0 1.20 "i a: 1.00 "0 0.80 "0 0 0 60 0 40 0 20 0.00 LBW A'eterm Birth SGA Outcomes LGA AGA Statewide o.93 (0 83 05) .08 (0. 95-1.23) 0.74 (0. 71-0 78) 1.45 (1.35-1.55) 1.08 (1.04-1 12) -Denver ....._state Figure 4 20. ODDS OF OUTCOMES FOR MEXICAN -BORN MOTHERS IN ADAMS AND DENVER COUNTIES AND STATEWIDE Aim3 Aim 3 is designed to examine the role of neighborhood on birth outcomes of mothers of Mexican origin, specifically the effect of neighborhood deprivation and immigrant orientation. Neighborhood is defined as census tract based on the 2000 Decennial Census. As described more fully in Chapter 3, pages 47-48, neighborhood deprivation is operationalized for each tract using data from the 2000 Census by summing and averaging the percent of: individuals living in poverty households receiving public assistance income female headed family households males unemployed in the civilian work force 110

PAGE 128

Immigrant orientation is operationalized by summing and averaging the percent of: Mexican-born individuals non-citizens born in Mexico linguistically isolated households These values were centered at the mean for the linear modeling. are: Following Finch et a/. (2007), the hypotheses for contextual influence on birth outcomes H70 : immigrant-oriented neighborhoods have no effect on the four birth outcomes H7 A: immigrant-oriented neighborhoods affect the likelihood of birth outcomes H80 : neighborhood disadvantage has no effect on the four birth outcomes HBA: neighborhood disadvantage affects the likelihood of birth outcomes To test the effect of contextual conditions on outcomes, the rate of each birth outcome in each tract is modeled against each contextual variable separately, both variables, and their interaction using generalized linear modeling. In Adams County, the index of immigrant orientation is correlated with neighborhood deprivation with Pearson's coefficient of 75375. In Denver County the correlation is less pronounced, with a Pearson's coefficient of .51979. Influence of Neighborhood Deprivation and Immigrant Orientation on Outcomes As discussed in more depth in Chapter 3, in Adams County there were no births in one tract for the entire study period and fewer than twenty births in five tracts. The five tracts with fewer than twenty births were matched on both the quartile of neighborhoods deprivation and immigrant orientation and combined with a matching tract, leaving 79 tracts, and 16,107 births for analysis. The number of births per tract after combination ranged from 20-936. Table 4.30 summarizes the range of frequencies of each outcome in Adams County. Table 4.30. Outcomes In Adams County Outcome Frequency Range Frequency In Tracts Countvwlde LBW N=1071 1.82-13.33% 6.65% Preterm Birth N=1201 0-13.00% 7.46% SGA N=2044 5.11 35.00% 12.69% LGA N=1040 0-13.33% 6.46% 111

PAGE 129

Table 4.31 summarizes the results of the general linear modeling on outcomes in Adams County. In Adams County, neither contextual variable influences rates of LBW, preterm birth, or SGA. For LGA, however, neighborhood deprivation and the interaction of neighborhood deprivation and immigrant orientation are significant at the 0.10 level, which some researchers suggest is appropriate, because contextual variables are generally considered to contribute less to outcomes than individual level factors (Sellstrom & Bremberg 2006; Pickett & Pearl 2001 ). Neighborhood deprivation is marginally significant with respect to LGA in Adams County (p=0.09). As neighborhood deprivation increases 10% from the average measure of deprivation, the rate of LGA increases 1.4%. Immigrant orientation alone is not significant. However, the interaction of deprivation and immigrant orientation on LGA in Adams County is marginally significant (p=0.10). Plotting hypothetical values of the interaction shows that when immigrant orientation is low, as deprivation increases so does LGA; when immigration orientation is higher, increasing deprivation decreases LGA. These results, although weak, suggest that immigrant orientation may moderate the effect of neighborhood deprivation in Adams County when there are high levels of immigrant orientation in the tract. Table 4.31. Estimated coefficients for neighborhood deprivation and Immigrant orientation In Adams County Outcome Model1 Model2 Model3 Model4 Neighborhood Immigrant Deprivation & Interaction of Deprivation Deprivation Orientation Immigrant Orient & Immigrant Orient F statistic F statistic F statistic F statistic t statistic Estimate t statistic Estimate t statistic Estimate t statistic Estimate p-value p-value p-value p-value LBW F=0.03 F=0.87 F=O.n 0.1065 F=1.16 .().0348 deprlv 1=0.82 0.0140 1=-
PAGE 130

hundred tracts with 23,332 births are analyzed. The number of births per tract ranges from 21 902. Table 4.32 summarizes the range of frequencies of each outcome in Denver County. Table 4.32. Outcomes In Denver County Outcome Frequency Range LBW Preterm Birth SGA LGA N=1594 N=1928 N=2802 N=1483 In Tracts 0-13.13% 0-19.05% 2.27 25.00% 0-11.11% Frequency Countywide 6.83% 8.26% 12.01% 6.36% In Denver County, LBW and SGA are influenced by neighborhood deprivation. For each 10% increase from average deprivation in Denver County, the rate of LBW increases 0.9% from the average in the county and the rate of SGA increases 1.3%. Contextual variables have no influence on preterm birth or LGA. Table 4.33. Estimated coefficients for neighborhood deprivation and Immigrant orientation In Denver County Outcome Model1 Model2 Model3 Model4 Neighborhood Immigrant Deprivation & Interaction of Deprivation o8prlvatlon Orientation Immigrant Orient & Immigrant Orient F statistic F statistic F statistic F statistic t statistic Estimate t statistic Estimate t statistic: Estimate t atatlstlc Eatlmate p-value p-value p-value p-value LBW F=3.49 F=1.12 F=1.74 0.0868 F=1.36 1=.67 0.0910 t=.06 0.0311 depriv I= 1.53 0.0049 1=-.0057 immigrant t=-<>.15 p:0.07 ns ns ns Prelerm birth F=1.62 F=0.09 F=0.88 0.0850 F-0.64 1=1.27 0.07207 1=-<>.31 0.0104 depriv t=1.29 -<>.0152 t=-<>.41 -<>.0034 immigrant 1=-<>.39 ns ns ns ns F=3.81 F=0.13 F=2.04 0.1616 F=1.34 SGA 0.1331 0.0152 depriv I= 1 99 -<>.0004 1=1.90 1=-<>.36 immiarant-<>.69 -<>.0336 t=-<>.04 P=0.08 ns ns ns F=0.18 F=1.90 F=1.00 F=0.73 LGA deprtv 1=-<>.34 -<>.0183 -<>.0031 1=0.40 0.0188 1=1.38 0.0383 immiaranl t=1.35 0.0438 1=.45 ns ns ns ns Contextual variables affect birth outcomes Adams and Denver County differently. Differences in the character of the counties may influence these results. Adams County is a mixed urban and rural county with a population density per square mile of 305; Denver is urban, with a population density of 3,617. Only LGA is affected by contextual variables in 113

PAGE 131

Adams County where it is weakly influenced by neighborhood deprivation and by the interaction of deprivation and immigrant orientation where high levels of immigrant orientation moderate the effect of high neighborhood deprivation. In Denver County, which is more densely populated, LBW and SGA are weakly affected by neighborhood deprivation. Finch et at. (2007) reported a monotonic moderation of immigrant orientation on rates of LBW as neighborhood deprivation and immigrant orientation increased, using multi-level hierarchical modeling. It is not clear why results for LBW in Colorado are not more fully consistent with those in Los Angeles. Adams County is much less densely populated than either Los Angeles County (2,344 per square mile according to the 2000 Census) or Denver, and that may suggest why these contextual variables have no effect on LBW in Adams County. Just as different regions of the U.S. were settled by different religious groups in the 1600s and 1700s (Phillips 1999; 80-122), residents of various Mexican states settle in specific parts of the U.S. Los Angeles may have denser networks of immigrants from these Mexican states and cities of origin compared with Colorado. It may also be that the multilevel hierarchical study by Finch et at. was able to capture more variation than this study's design. Discussion of Quantitative Analysis The design of this study sought to expand the spectrum of previous weight-related birth outcome studies with the addition of LGA. The results of Aim 1 show that the epidemiological paradox exists with respect to LBW, preterm birth, and SGA and race/ethnicity. It also exists with respect to LGA, but both Whites and Hispanics have higher odds than less advantaged Blacks or Others. Technically this is a paradox based on the reference group of White mothers, but LGA behaves differently than LBW, preterm birth, and SGA by race/ethnicity. In addition, the study design sought to test the existence of the social gradient. Both Hispanic and Mexican-born mothers have much worse socioeconomic profiles than Whites. Adding medical risks in Model 3 sought to elucidate the influence of pregnancy-associated medical risks. In Aim 1, Hispanics have worse medical profiles that those of Whites. In Aim 2, the object was to test whether Mexican-born immigrants are "healthier'' than their U.S.-born counterparts. Although Mexican-born mothers have a better general medical risk profile (one or more of general medical risks associated with pregnancy), their LGA-specific risk profiles are worse, suggesting that they are not necessarily "healthy'' migrants. Finally, adding smoking and weight gain into Model 4 tested the healthy immigrant hypothesis, whereby Mexican-born mothers are postulated to have "healthier'' behaviors which explain their 114

PAGE 132

favorable outcomes results. None of the fully saturated models had superior discriminating power. Nevertheless, the lack of change in odds from Models 2 to 3, and from Models 3 to 4 suggests that the proposed explanations in the literature do not provide powerful rationales for differences in weight-related birth outcomes of Hispanics generally or mothers of Mexican origin specifically who live in Colorado. The results of the analysis also raise questions about the influence of certain variables on these outcomes. Education was uniformly not explanatory, even though it is a key factor in socioeconomic position and the theory of the social gradient. The information in the birth record on education levels is probably accurate as it is consistent with census data showing low levels of education among Mexican immigrants. Smoking is a strong predictor of LBW and SGA. Colorado PRAMS data suggest self-report of smoking by Hispanics is fairly accurate, but under-reporting may be greater for other population groups. The low prevalence of smoking among Hispanic mothers likely accounts for an even greater share of LBW and SGA than birth registry data suggest. In addition, the low rate of smoking among Hispanics removes the antagonistic effect of smoking on LGA. Altitude operates on LBW, SGA and LGA just as predicted. There is a monotonic increase in low weight outcomes with increasing altitude, and decrease for LGA. The low rate of drinking reported in the birth registry probably explains why it is insignificant in all model building. The influence of measures of adequacy of prenatal care is a puzzle. Intermediate care is almost always not significant or the same as adequate care. The between inadequate, intermediate, and adequate care are necessarily arbitrary in the Kotelchuck index. Perhaps, as Fiscella (1995) and Alexander & Kotelchuck (2001) discuss, the value of prenatal care remains poorly measured. Given the late entry into prenatal care by Hispanics and Mexican-born mothers, however, it may be that medical risks are not adequately diagnosed. Their prenatal care may also be interrupted by inability to pay. The interaction of weight gain and prenatal care show that low weight gain with any level of prenatal care raises the odds of preterm birth by 2.24 to 6.46 among women of Mexican origin, so prenatal care should not be written off as irrelevant. Area measures of neighborhood deprivation and immigrant orientation had weak association with outcomes. The difference in population density between Adams and Denver Counties probably explains why LBW and SGA are not affected by contex1ual variables in Adams but are in Denver. Most interesting are the results for LGA. Just as incidence of LGA does not behave in the same way as LBW, preterm birth, and SGA for mothers of Mexican 115

PAGE 133

ong1n based on compositional characteristics, neither does the influence of contextual variables on LGA behave as they appear to on LBW and preterm birth. Contextual variables have only weak influence in both counties, but even there, the results are not consistent. Only in Adams County is there is a weak moderating effect of immigrant orientation on the negative effect of neighborhood deprivation on LGA. The qualitative portion of the study sheds some light on the importance of neighborhood on mothers of Mexican origin in Colorado. 116

PAGE 134

CHAPTER 5 QUALITATIVE RESULTS The qualitative portion of this study complements the quantitative results and is designed to be directive: based on the statistical results, certain matters were explored with key informants and recent mothers of Mexican origin. In light of quantitative findings suggesting that mothers of Mexican origin have odds of LBW, preterm birth, and SGA that are similar, regardless of their place of nativity, but that Mexican-born mothers have higher odds of LGA than their U.S.-born counterparts, the interviews focused on contributors to LGA diet, exercise, weight gain, diabetes, sources of support including the neighborhood, cultural beliefs concerning exercise and diet while pregnant, and with Mexican-born women, differences between life in Mexico and the U.S. Recent Mothers Ten mothers of Mexican origin were interviewed; five were born in the U.S. and five were born in Mexico. All women were either married to or in a committed relationship with a man of Mexican origin; several U.S.-born mothers were married to men born in Mexico. Mothers ranged in age from 18 to their early 30s. Three were first-time mothers; the others had between two and four children. Five of the mothers had full time jobs, one worked part time in a medical clinic, and four were not employed in the formal economy. One woman had worked for a nursing home, but lost her job after her baby was born prematurely and she spent a lot of time with him in the hospital (about five weeks). She wants to return to school to earn an LPN degree. The women entered prenatal care between eight weeks (two were planning on getting pregnant and started care as soon as they had a positive test) and four months into the pregnancy. Some women recalled being tested for diabetes, and all said the test was normal. One mother, though, probably had preexisting diabetes. She spent her seventh month in the hospital after premature rupture of the membranes, and was put on a strict diet and given insulin shots when her blood sugar rose. Three of the interviews with mothers were conducted in Spanish; the balance were conducted in English. Three mothers were interviewed at their place of work during a lunch break; six were interviewed in their 117

PAGE 135

homes; and one was interviewed outside one of the Salud clinics. Pseudonyms are assigned to the recent mothers. Anna, Isabel, Sara, Linda, and Marta were born in Mexico. Juanita, Graciela, Carmen, Yolanda, and Alicia were born in the U.S. Table 5.1 describes certain characteristics of the women interviewed. Table 5.1. Selected characteristics of women Interviewed Characteristic Mexican-Born (N=S) U.S.-Bom (N=S) Age 18 -low 30s 19-low 30s Employment full-time outside home 1 4 First time mother 2 1 2-4 children at time of interview 3 4 Weiaht aain >40 or hiah BMI 2 2 Medical risk/problem pregnancy 2 0 Key Informants Five key informant interviews were conducted with professionals who work in various capacities with Hispanic mothers. One key informant works with poor, mostly Hispanic, first time mothers in the Nurse Family Partnership Program in Colorado. A second key informant is a certified nurse midwife, practicing in Colorado for many years in rural locations and with Denver Health, the public health hospital in the city and county of Denver. The third key informant is an obstetrician working with a primarily poor, Mexican population at Denver Health. The fourth is a woman who was a practicing physician in Mexico, who now works as a nurse with La Clinica Campesina, a public health clinic for the poor, many of whom are of Mexican origin, with clinics in Adams, Boulder and Denver counties. Her experience in both countries provided insights on the differences between mothers' behaviors in Mexico and in the U.S. Each of these women is immersed in providing health care and counseling to women of Mexican origin in Colorado. The final interview was with a physician-researcher who is an expert in diabetes in pregnancy at the University of Colorado Denver. All key informant interviews were conducted in English; four were conducted at various locations convenient to the interviewee (office, coffee shop, home); one person was interviewed by telephone. Diet and Exercise During Pregnancy Diet All mothers except one described their diet as primarily "Mexican" eggs, beans, tortillas, ham, enchiladas, rice, and chile. Several mentioned cooking the way their mothers 118

PAGE 136

did or cooking to please their husbands. Their diets were influenced by cultural preference for Mexican foods and some mentioned a belief that it was not only permissible, but necessary, to eat a lot during pregnancy, a view that was reinforced by their mothers or mothers-in-law. [We eat] green chile with potatoes and meat; beans, rice, the basic Mexican meal. Juanita The number one thing we have on our plate, all the time, is beans. Beans have to be there . it's weird, but food doesn't taste the same if we don't have beans. That's our main dish. Some days it's just flour tortillas with beans and cheese and chile. Graciela The way my mom cooks is the way I cook. Mostly everything I make is Mexican foods like enchiladas. . Like yesterday was my husband's birthday, so I made mole, which is a real tradition of Mexico, green chile, red chile, rice and beans, sopa. Alicia All but one mother described a relatively healthy diet for their most recent pregnancy, citing cereal, fruits, vegetables, salad, milk, and avoiding (for the most part) fast foods, although there were times when lunch consisted of pizza, burritos from a local restaurant, or hamburgers. Their staples, however, leaned heavily to carbohydrates in the form of beans, tortillas, and rice, and their diet typically included various fried foods, such as enchiladas and tostadas. Their sopas (soups) contain pasta-like bits of fried flour with onion, garlic, and tomatoes. Sara's diet was particularly caloric; she consumed four to six meals a day, eating tacos and tostadas at home with her children and institutional meals (meat and potatoes) with the residents at the nursing home where she worked. Two U.S.-born mothers mentioned receiving WIC assistance during their pregnancy and said they ate foods from WIC that they probably wouldn't have bought or eaten on their own, such as eggs, peanut butter, milk, and cheese. With my second child, I had a lot of problems and I didn't really eat-at all. So he was only 6 pounds when he was born he was underweight. And I'm sure that was the reason why. And when I got pregnant with this one, I was also on WIC. So they introduced me to the nutrition, and what's healthy and what's not healthy, like the sugars, and that I needed to drink more milk and cheese, and stuff like that. So I kind of balanced myself out. ... So ... it was a big change in diet between my second child and the third. Carmen 119

PAGE 137

Key informants reported that a diet high in carbohydrates and fats and fast foods characterized the Mexican families they worked with. The OB from Denver Health talked about the Diabetes Clinic where she emphasizes diet and logging dietary intake. She sees a direct correlation of the number of tortillas mothers have eaten and their blood sugar levels. Staples, they have the same kinds of staples [here as in Mexico]. They have a hard time finding fresh tortillas, but they eat home-made beans, rice, and some are conscious of eating the way they did in Mexico. And others, they're conscious of trying to make as much money as they can, and so they are working before the baby comes and they're driving around and they stop for fast food. Public health nurse The normal diet in Mexico is tortillas, chile and a big soda. Soda is available, even in rural areas. And also we are pretty fond of dessert ... In the old towns they would slaughter pigs, and then butcher the meat and make lard and use that for cooking. My father-in-law would do that. He was diabetic and continued to do it all the same .... [And we keep it up in the U.S.] Former Mexican physician Both mothers and key informants discussed the cultural belief that pregnancy is a time to eat for two. Mothers also indicated that their mothers urged them to "eat for the baby." Carmen's mother would tell her "you have to eat; because it's not for you, it's for your baby." Marta suggested that women in Mexico eat what they want when they are pregnant without regard to the health benefits of the food, eating foods cooked with lard. Key informants reported similar experiences with pregnant moms. They think they have to eat for two. So people start increasing the calories, both in Mexico and in the U.S .... They say "I have to eat; I have to eat for two," but they eat this type of carbs, this much protein [and] they just increase the amounts .... Portion sizes are so different here compared to Mexico. Former Mexican physician This idea that when you're pregnant you have to eat a lot probably comes from the sense of having had food insecurity in the past and you want to be sure that this baby comes out all right, so you eat a lot during your pregnancy. Public health nurse 120

PAGE 138

Exercise/Energy Expenditure Among all the mothers interviewed, exercise levels during pregnancy were relatively low. Two mothers responded positively when asked about exercise during pregnancy; Alicia would walk after dinner and Yolanda walked two miles each day until her seventh month. Then she said she slept until noon and sat around the last three months of her pregnancy doing nothing. Others described walking a few blocks once or twice a month (Juanita) or not working out or doing anything other than what was required at their jobs (Carmen and Juanita). Anna said she would walk a little but that her neighborhood (a trailer park) was not that pretty or friendly. Graciela, born in the U.S., explained that she stopped exercising during her most recent pregnancy because she had a previous miscarriage. I was active before I was pregnant I was going to Curves every day it wasn't something like really major but I had been going and once we started planning the baby-because before this baby, I had a miscarriage . . and I got this gut feeling that I was pregnant and then I stopped going . . My doctor said wait 'til you're three months and you can start up again. But I honestly didn't want to because I had that miscarriage already and I didn't want anything to happen to it. Gracie Ia Key informants affirmed that, in their experience, exercise is not as much a part of the lifestyle of Mexican-born mothers after they first come to the U.S. Whereas in Mexico people eat to work, here women in particular are less likely to have jobs and they feel physically isolated or afraid of being out and about as they were in Mexico. In addition, cultural norms work to dampen exercise when a woman is pregnant. When they move here from Mexico they may be isolated, and they aren't working in the fields or walking everywhere, as they did in Mexico, and so their base rate of exercise is much less. In Mexico people eat to have enough energy to work. OB A lot of the moms lived in Mexico in areas where they lived in the cal/e, in the streets, and they would go outside all the time ... and they would go to the market on foot and if they were in a rural area they were outside walking around. When they come to the U.S. they're stuck, because they don't necessarily drive or have cars. Public health nurse When I take care of women here, that's a huge problem -that they don't exercise ... the exercise they might get is walking their kids to and from school. Midwife 121

PAGE 139

In Mexico, as soon as they are pregnant, they don't work as hard. People say "you don't have to run." Former Mexican physician And yet, no group is homogeneous. Marta, born in Mexico, refused to give in to morning sickness or tiredness while pregnant. She read several books about pregnancy and decided that she would be just as active as she was before becoming pregnant. This did not extend to working out, but she convinced herself that she would not allow her pregnancy to slow her down. Maternal Weight and Weight Gain Although weight gain is the most commonly used predictor of weight-related birth outcomes, more recent research suggests that maternal BMI immediately before pregnancy rather than weight gain is a stronger marker for poor birth outcomes (Casey eta/. 1997). The quantitative results show that Hispanics generally, and mothers born in Mexico, have higher rates of low weight gain during pregnancy. Observations by key informants corroborate the literature with respect to LGA. The most important predictor of LGA is maternal weight, not weight gain during pregnancy. You could have high BMI before pregnancy, gain just a little weight, and smoke, and end up having an AGA baby even though all signs point to macrosomia. Diabetes researcher It may be that Mexican-born moms gain less weight during pregnancy, but that they enter pregnancy at heavier weights. BMI is high among Mexican women in the Diabetes Clinic some have a BMI between 40 49. OB Gain although it's not very predictive because a lot of these women don't gain much weight, but especially if they are overweight to start, it's key. Midwife Mothers interviewed gained varying amounts of weight. Table 5.2. Weight gain of mothers Interviewed and birth weight of Infanta Mexican-born Parity Age Maternal Weight Birth Weight Complications Weight Gain {Pound-Ounce) Anna 1 18 98 52 6-8 None Isabel 3 30s 198 31 7-7 Induced at 34 weeks Marta 1 20s 110 20 6-12 None Sara 4 30s n/a n/a 3-0 C-section at 32 weeks Linda 2 20s 123 39 7-8 Born at 38 weeks 122

PAGE 140

Table 5.2. Weight gain of mothers Interviewed and birth weight of infants (continued) U.S.-born Parity Age Maternal Weight Birth Weight Complications Weight Gain (Pound-Ounce} Graciela 3 25 140 35 7-1 None Carmen 3 Late 20s 148 42 8-5 Induced at 39 weeks Alicia 2 Late 20s n/a 22 6-1 None Juanita 2 Mid 20s 150 29 7-2 None Yolanda 1 24 140 45 7-10 None Two mothers, both born in Mexico, had complications. At the time of her interview, Isabel appeared to have high BMI. She said she had been tested for diabetes during pregnancy and the results were negative. Nevertheless, when she delivered her third child at 34 weeks of gestation, he already weighed 7 pounds 7 ounces. At this weight, 3,373.7 grams, he was almost LGA for his gestation (3,595 grams) according to Alexander eta/. (1999). If she had carried him to term, he would probably have been LGA, because the fetus typically gains several pounds during the last trimester. Sara has four children. Both of her daughters were large -9 pounds 14 ounces and 1 0 pounds. She needed a c-section to deliver her 1 0-pound daughter. During her most recent pregnancy she ate four to six meals a day. She was hospitalized at six months with premature rupture of her membranes. Her blood sugar was very high, and she was put on a strict diet and given insulin shots. The baby was born eight weeks early and weighed only 3 pounds (1360 grams definitely an LBW baby and also SGA). Although she said her glucose levels were fine after the birth, mothers with pre existing diabetes often have small babies, while mothers with gestational diabetes tend to have LGA babies. Body Image Some studies have noted differing views of ideal body image across cultures (Candib 2007, Ahluwalia eta/. 2007). This was also noted by two key informants. And some of them who have high BMis ... And also, it's amazing to me ... they see it as a positive thing. A big body for a woman is what is desired. It's desirable, that looks good. I worked in a clinic it was 95% Latina and most of them were immigrants. The staff was also Mexican, bilingual. .. We would talk and they were "That's how my man wants me to be; this is what looks good; and you skinny ladies look terrible." It's a different, completely different culture. They're in health care, they know it's not healthy to be overweight, but it's difficult for them to believe that or overcome those beliefs, values that they have. Midwife 123

PAGE 141

There is this switch from being somehow a young woman to being a mother, and the switch in body image it's OK to be overweight, to care a little less about how you keep yourself, and that you are the cook. You make all the food, you feed everybody, and food becomes the center of your life in a way. Public health nurse Whereas being married is generally protective for birth outcomes at the low-weight end of the spectrum, the quantitative results showed an unexpected protective effect on LGA of being unmarried. This finding is likely consistent with the idea of the married mother as the "cook" as noted above and perhaps also consistent with some Hispanic men's preference for larger women. Sara linked this preference with "machista." Mexican men are very machista. They're always like really possessive, they want to have their woman stay home; some won't let you work. Some they just want, like if you are out at the street, they don't let you look any where else or they get mad. When asked if she agrees that Mexican men like larger women, Sara said it has to do with control over the woman. I think it's they think, because they don't want nobody else to look at you and they like them to get like that I think .... I was telling my baby's dadI was telling him that I want to start doing exercise because I was getting chunky, and he said "No, you're looking better like that. You look good like that. You don't need to do anything." Smoking and Drinking In accord with the literature and results of the quantitative results, the mothers interviewed said they did not smoke during pregnancy. A couple said they smoked occasionally or "maybe had three cigarettes a day'' before getting pregnant but once they got pregnant they stopped. A few said members of their households smoked, but they always smoked outside. When asked why they didn't smoke during pregnancy, they all said it wouldn't be good for the baby, and a couple said it wasn't good for them either. Mothers didn't drink alcoholic drinks during pregnancy either, although a few said they had before they were pregnant or afterwards, if they weren't nursing. The public health nurse agreed that smoking was minimal. My impression after working with these families is that women are careful; they really don't smoke. I'm in their homes and I don't smell it; they don't smoke. Public health nurse 124

PAGE 142

Other Cultural Beliefs About Pregnancy I asked each woman if she practiced /a cuarentena after birth. All said they did to one degree or another. I then asked them to describe the practices they engaged in. The answers ranged from "no sex for 40 days" to relatively elaborate rituals for the first forty days after giving birth, usually encouraged strongly by her mother and/or mother-in-law. The most elaborate descriptions came from U.S.-born mothers. Oh yeah, it was like a big deal for my mom and my mother-in-law. More on my mother-in-law; since she's a bit older than my mother. I mean, she told me stories about she had six kids -and in all her six kids and /a cuarentena, she never showered, and to be honest, I only did it for eight days. And just so I wouldn't like, I didn't want to get her upset. My mom said she bathed after she had the baby ... You have to cover your back and if you don't take care of yourself in those 40 days, your milk production won't be good if you don't cover your back. You're not supposed to be out in the cold, 'cause you're in the healing process. They were really strict. With my first one it was even worse, these other ones, they would let me at least go to the kitchen. My first one I was just in the bedroom for 40 days and I was ''This is just ridiculous. Mom, I think I could go to the living room!" She'd say "No, just stay in your room because it's warm." .. The second one was a little bit easier, but I do believe in that, Ia cuarentena. Gracie Ia La cuarentenaI take that seriously in a woman. And my mom, being there telling me a woman should always take care of herself before and after birth. And she would tell me "Don't go nowhere, you're not supposed to be out and about, you know, because, how would she say it "hemorragia "kind of like bleeding to death and so she would tell me. So I really took care of myself. I really highly believe in taking care of your body after-you and the baby as well. I would stay home. And my mom would say "Don't go up and down the stairs, don't be outside when there is wind. If you have sit there all day, then sit there all day, and take care of your baby." So I, so when a baby is born, you have to see the doctor within two days, so that was the only time I went out. I went out for his check ups and came straight home. I was at home the whole time. . She would tell me to, well in the tub you have to step over and she would always tell me like when you lift up your leg and all that, you need to be very careful [not to let air go up the birth canal]. But I would take a shower every day for that for the same reason that I was breastfeeding at that time, and to be clean because you are breastfeeding the baby. I didn't eat different. I ate the same. I just didn't do any house chores. Carmen 125

PAGE 143

SourcesofSocmiSuppon All mothers were asked what they would do or whom they would call if they were six months pregnant and they hadn't felt the baby move for a while and were getting scared. None mentioned getting support from their neighborsall said they would seek support from members of their familywhether or not that person was close by or from their medical provider. I'd call the doctor. Right away. And if the clinic wasn't available, I would call the hospital, like the nurse line and get information. And I would also turn to my mom. So I would call several people, all on the same day, so I would know what is going on. I would also call my sisters. Carmen My mother, my sister-in-law, a friend at work. Juanita Mi suegra (mother-in-law) before even talking to my mother. I might go to the clinic or hospital if there was a problem. Anna My mom .... And then my mother-in-law lives with me so she's an older woman and she was raised in Mexico so she knows a lot of home remedies, and my mom's like that too. Gracie Ia When asked generally where they got support generally, mothers mentioned parents, in-laws, siblings, and husbands. One mother mentioned that since her miscarriage, she had become more spiritual and went to church more. I kind of started going a lot closer to church after I had my miscarriage and then after I had this baby and I kind of felt I needed more spiritual so I would say I pray a lot more. Gracie Ia The striking "cultural zero" in all the interviews was the unimportance of the neighborhood. No one spontaneously mentioned their neighborhood or neighbors as sources of support or even much in the way of socializing (Kutsche 1998:10). Mothers occasionally mentioned that they knew a neighbor or two, but not well enough to call on for help. My neighborhood is a trailer park. I don't like it. Anna [My neighborhood] it's not very big and everybody works, so by the time I get home or they get home before me, and I don't really see my neighbors ... I think I could [get support] if I needed it. I know that my 126

PAGE 144

neighbors would be there, but I'm never home. The hours I'm never home, they're not home when I'm home. Carmen I can probably say I don't have a real neighborhood. I live way out in the country. My nearest neighbor is probably, maybe a mile away. My only neighbor-he's not very nice. Graciela I talk a lot to my neighbor over here. She's an old lady, she's really nice, but friends? No. Sara Political Economy and Birth Outcomes All mothers were asked if they had any ideas why women of Mexican origin tended to have healthy birth outcomes at the low-weight end of the spectrum but why recent immigrants had worse LGA outcomes than mothers of Mexican origin who were born in the U.S. Answers ranged from differences in cultural beliefs to pregnancy diet to economic difficulties in Mexico and the U.S. to big changes in diet and exercise after they moved to the U.S. Although not always stated explicitly, lack of resources in general framed their experiences. Four mothers had lived with in-laws or parents during one or more of their pregnancies and spoke of the difficulty generally of making it in the U.S. Others talked about how hard life in Mexico is compared to the U.S. I walked in the park at the rec center while I was pregnant, but didn't use the rec center for any type of exercise. [although not stated, it seemed that the reason was lack of money to use the rec center] Anna I know that being in the U.S. for a Mexican and try to make it here in the United States is a lot harder. .. to buy anything in general it's harder than over there [in Mexico]. I have heard that from my mother-in-law that from her point, her boys, they were all large babies, you know, bigger than normal. And she had them over there. And when I had my kids, except for the most recent, the other two were smaller and she would tell me, how come I wasn't eating enough because my baby wasn't growing. Carmen The adjustment to practical exigencies of life in the U.S. creates changes in diet and energy expenditure almost immediately. Key informants spontaneously talked about economic pressures to migrate and the abrupt changes in diet, activity patterns, and isolation of recent immigrants, all of which are likely to contribute to LGA babies. 127

PAGE 145

I'm not so convinced that it's the healthier women that come, because a lot of the time it's impoverished women that come; they are the ones that are willing to take the risk because the poverty levels in their families in Mexico are so desperate that they have to leave. They leave behind their kids .... It used to be the male from Mexico who comes to support the family, but now it's more, the whole character is changing face. A lot of women who're coming over to work and getting jobs related to childcare and cleaning ... the women are often sending money home and a lot of them still have their kids back in Mexico with their parents .... A lot of it is just to seek employment and help improve their family's situation in Mexico. Midwife I think most come because the economy is really bad; so mostly poor people immigrate. Former Mexican physician The more physically active ones have hotel jobs, they make beds, and then clean houses, and they work in fast food joints. It's becoming harder and harder for them to find work because people are asking for social security numbers. And so informal labor, like caring for each other's children while one of them goes out for work. Public health nurse Sara described her life in Mexico and after immigration to the U.S. She and her sister stayed in Mexico with her grandparents for three years while her parents and two younger siblings worked to make enough money to bring Sara and her sister to the U.S. I lived in a little village [in Mexico]. I used to go to school. After school I would 'cuz I was with my grandma my parents came over here first. Life over there is so different. They don't pay good on the job or anything like that. So I used to help my grandma after school. To clean, to cook, to feed the chickens, and cows. I used to help milk the cows. I was sometimes hungry. It was hard. When my parents came over here it was a little bit better, but when we were all there [in Mexico] it was hard sometimes and we were hungry. . My parents used to buy shoes that we both could use, that way sometimes I used his shoes and sometimes he would. It was hard. It was really hard. So that's why they came over here so they could give us a better life. Sara described the difference in diet between Mexico and the U.S. Her mother worked two jobs and her father worked late hours. As the oldest child, she was in charge of the children and helped with cleaning and cooking after they arrived in the U.S. In Mexico we used to eat beans, fruit, potatoes. Sometimes, not a lot of times, we used to eat meat, not so much. Fruit, a lot of fruit; vegetables not that much. My grandma had a lot of like peaches, apples, in trees, like that, and we used to eat that. When we came here, we used to eat more 128

PAGE 146

like junk food. But not there-more like homemade food. Especially when you are at school you go "Oh, let's go get a hamburger, nachos, pizza." Or like on the weekends, we were all together, junk food. "Oh I don't feel like cooking, let's go get some junk food." We would be tired during the week, cooking, and stuff like that Friday, Saturday and you don't want to do anything, you buy junk. Key informants noticed these same changes the abrupt nutritional transition, coupled with changes in energy expenditure and increased social isolation observations also made by some researchers such as Himmelgreen eta/. (2007). These elements, added to a traditional Mexican diet, provide the predicate for higher maternal weight at pregnancy and higher odds of LGA among Mexican-born women. Food insecurity leads to energy dense food diets because such foods are cheaper in the U.S. It has an impact in terms of food security ... a lot of times when they first come, there is food insecurity because they just don't have money to buy food. But then when they start working and get a little more settled, then I think you see the changes in their diet. And I think there is less food security in Mexico than they end up having once they get settled here. .. It's partly a lack of understanding of nutrition and a change of diet when they come here and they eat very differently than they did at home because they have all these cheap, cheap fast foods that they can get and so it's partly educational deficits and partly economic reasons why they eat high fat, high calorie foods. Public health nurse When they come, they start adopting the American lifestyle and start eating carbs, because in Mexico poor families mostly have carbs in the form of tortillas. Here they start eating doughnuts and pizza. Former Mexican physician Changes in energy expenditure and social isolation also contribute to higher maternal weight. And they're in the trailer park and the trailer park is far from the supermarket and so their husbands drive them where they need to go and other than that they are at home. And they're also nervous about walking around because it's an unfamiliar area. .. They're socially isolated, they don't know many people, and they are physically isolated in their homes and of course culturally isolated because they have left a lot of familiarity behind. [and] that leads to all sort of lifestyle changes around physical activity and diet. Public health nurse Some people come from small towns; they don't have washing machines, so they do their wash by hand [on a washboard]. [When they are in the U.S.] they are staying at home, waiting for the husband to come home at 129

PAGE 147

night, and when he comes, they start eating and eating. And when she stays at home, taking care of the kids, and in the evenings they are afraid to go out. Former Mexican physician So you look at health disparities in some populations, you see what foods are available to them in their neighborhoods, for example, where do they go they go to fast food places because food is cheap and they don't go to large supermarkets and buy fresh fruits and vegetables because they don't see them as being cheap, so they go to the 7-11 and you can buy a lot of food cheaply but it's soda and junk food. . that's one of the shifts access to more processed foods. .. A lot of them tell me they don't feel safe exercising, going out and exercising in the areas they live in or they don't have access they can't even go to the neighborhood rec center because they don't have $10. Because they have small children, they don't have anyone to leave their children with. I think there's this idea that Hispanic families are these tight-knit, extended, large families that really support one another, and I think that can be true, but many of these new immigrant women are here alone, and they are very isolated. Midwife Discussion of Qualitative Interviews The observations of key informants represent a convergence of opinion that is consistent with the medical literature, social and political aspects of immigration and the effects of immigration on lifestyle. Maternal BMI before pregnancy and the presence of gestational diabetes are predictors of LGA. Abrupt change in diet combined with access to more and cheaper energy-dense foods, changes in daily energy expenditure, cultural beliefs about "eating for two" and "not running" when pregnant, and isolation characterize many of the Mexican immigrants the key informants see in the course of their work. They also suggest that escape from poverty in Mexico is the primary motivating factor for immigration. Based on the key informant interviews, I anticipated sharper differences between the mothers born in Mexico and those born here. To a surprising degree, though, there was little discrepancy overall between the two groups of women. All mothers considered themselves rooted in Mexican culture, at least with respect to diet and certain beliefs of post-pregnancy practices. Every woman was married to or in a committed relationship with a man of Mexican origin, and several of the U.S.-born women were married to men who had been born in Mexico, consistent with a strong degree of endogamy. All mothers also generally eschewed exercise during pregnancy because they felt they got enough exercise in their daily routines, because they were concerned about exercise being harmful to their pregnancy, or because they believed it was permissible to be less active or more "lazy" (Yolanda). 130

PAGE 148

Neighborhood was decidedly not a source of support or even a meaningful construct from which to seek support for these women, in line with the quantitative results showing little association of immigrant orientation on birth outcomes. However, social isolation seemed greater for women born in Mexico. It is likely that three of the Mexican-born mothers didn't drive I met them in their homes or they arranged for someone to drive them to meet me. These three mothers also asked to conduct the interview in Spanish (although each had some facility in English). In these senses, they may have been more socially and linguistically isolated, although only the youngest mother seemed so in the interview and that may have been because she was only 18. She referred often to her mother-in-law as the person to whom she looked for advice, usually with clear subservience in her tone. All mothers had some family within the Front Range area either in the same town or within fifty miles of where they lived, with whom they described friendly relations. In-laws were cited more often than even parents as sources of support of all types (close ties, sources of advice, and encouragement for Ia cuarentena). Isabel summed it up in a torrent of Spanish, without interruption, about why she thought Mexican-born women have more LGA babies. Although there are some cultural beliefs at work (women wanting to be rounder, eating carbohydrates to provide nutrition to the baby), her focus is on the structural constraints within the larger political system that affect Mexican immigrants (poverty, isolation, depression, the inability to get around without a car). Isabel is one of eleven children, and migrated to the U.S. when she was 24. It is possible she was describing herself to some extent, because she alternated between "we" and ''they." Perhaps because we like food too much. Mexican women in the U.S., we are inclined to eat more at meals and to get fat. And to be good for the baby, you should be more round, more fat. Women in Mexico eat more foods like lentils, beans, favas, all to provide nutrition to the baby. The majority of [Mexican] women in the U.S. have problems with money; economically and there are many women who are not economically well off. Therefore it is more difficult to feed ourselves and to live well and to have a good environment for the baby. And sometimes we can't go to the doctor because we don't have money for the appointment. And I think all of that affects many things and emotionally too. You feel alone here. You don't feel you have support. Even if you have family in the U.S., sometimes you don't feel much support. There are women who are here alone. Alone! Sometimes they don't have help getting to the clinic, or they don't understand, so they don't go. And many people do not know that you can get food stamps to eat. When you are pregnant and alone your are emotional. You feel sad. Melancholy. And your self esteem drops. And sometimes you don't have anything to eat and you don't take care of yourself. The difference in Mexico, is, for example, most people have their 131

PAGE 149

house. And they don't have payments hanging over them. So I believe these things contribute to a person's emotional state. Also in Mexico, if you don't have money you don't lose your house. And to get around you don't need a car or a truck. It's not difficult there, you can walk, or take the bus. 132

PAGE 150

CHAPTER 6 DISCUSSION This study sought to determine whether an epidemiological paradox exists for four birth outcomes related to the weight of the infant (LBW, preterm birth, SGA, and LGA) among Hispanic mothers giving birth in Colorado during the period 2000 2005. The present study compared Hispanic mothers with mothers of other races/ethnicities and compared Mexican born mothers with U.S.-born mothers of Mexican origin to identify individual-level contributors to those outcomes. This study also sought to determine whether contextual/area level factors contribute to birth outcomes among mothers of Mexican origin in Adams and Denver Counties and used qualitative methods to enrich interpretation of quantitative findings. Previous studies of the epidemiological paradox focused on low weight related birth outcomes only. Favorable odds of low birth weight, preterm birth, and infant mortality led to considerable controversy over the existence of an epidemiological paradox and various proposed hypotheses to explain these results. Many explanations seem unsatisfying, either because they are too simplistic or too narrow to provide insight into the complex factors that affect birth outcomes. Four key findings emerged from this study. First, an epidemiological paradox exists for Hispanics in Colorado with respect to low birth weight, preterm birth, small for gestational age status, and large for gestational age status. Despite having worse social and medical profiles than non-Hispanic White mothers, Hispanic mothers have similar odds of each of the four birth outcomes. Second, the epidemiological paradox also exists for Mexican-born mothers compared with U.S.-born mothers of Mexican origin for LBW, preterm birth, and SGA. The paradox does not, however, exist for LGA, where Mexican-born mothers have much higher odds of LGA than U.S.-born mothers of Mexican origin. Third, the hypotheses offered to explain the paradox, are, indeed, unsatisfying. The results of the study do not support the healthy migrant or healthy immigrant hypotheses. Finally, and unexpectedly, neighborhood measures of immigrant orientation and neighborhood deprivation do not influence the likelihood of outcomes as strongly as certain other studies have shown. 133

PAGE 151

Limitations This study has several limitations. First, it is cross sectional and therefore cannot measure outcomes across generations or determine causation. Second, important data that are lacking in the birth record would strengthen the study. Age at migration, length of residence of foreign-born mothers, and primary language spoken would be informative for categorizing mothers as well as for creating predictors for behaviors and outcomes. Maternal weight and BMI before pregnancy would be very useful for understanding LGA. Finally, if it is believed that social neighborhoods are important influences on health outcomes, reliance on census tracts must cede to definition of neighborhoods by the people who live there, or perhaps it is time to move away from neighborhood analysis and focus instead on social networks of the type mothers in this study suggest are meaningful to them. Is There an Epidemiological Paradox in Weight-Related Birth Outcomes? The quantitative results of this study demonstrate that the epidemiological paradox exists in Colorado for Hispanics. Hispanics have the same odds of preterm birth and lower odds of LGA compared with non-Hispanic White mothers. Hispanics have 18% higher odds of LBW and SGA than White mothers, but these odds are much lower than those of Black mothers whose SES is similar to that of Hispanics. For LGA, while Hispanics have odds 5% lower than those of Whites, their odds are 36% higher than those of Blacks. Hispanic, Black, and Other mothers have lower odds of having an AGA baby that is, their likelihood of having an appropriate weight for gestational age is less than the majority White population, although the odds for Hispanic mothers are only 4% lower than those of Whites. Table 6.1. Comparison of fully adJusted odds ratios by race/ethnlclty White 1.00 1.00 1.00 1.00 1.00 Hispanic 1.18 *1.01 1.18 t 0.95 t0.96 Black LBW Preterm Birth SGA LGA AGA No statistical difference from odds for White mothers. t Statistically significant at p so.os. 2.16 1.42 1.98 t0.59 t0.72 Other 1.64 1.16 1.72 t0.68 t0.78 The paradox also exists for low weight-related outcomes for Mexican-born mothers compared with U.S.-born mothers of Mexican origin, even though Mexican-born mothers have lower SES than U.S.-born mothers. In contrast to results based on race/ethncity, 134

PAGE 152

however, the paradox does not hold for LGA, because Mexican-born mothers have 45% higher odds of LGA than U.S.-born mothers of Mexican origin. Mexican-born mothers also have higher (better) odds of having an AGA baby (1.08). Table 6.2. Comparison of fully adjusted odds ratios by nativity LBW Pretenn Birth SGA LGA AGA U.S.-Bom Mexican Origin 1.00 1.00 1.00 1.00 1.00 *No statistical difference from odds for U.S.-bom mothers. t Results statistically significant p :S0.05. Mexican-Born *0.93 *1.08 t0.74 t1.45 t1.08 This study confirms the results of many previous population studies with respect to low weight outcomes. The intense and continuing public health emphasis on LBW and preterm birth is understandable, given the large immediate costs, both personal and to the healthcare system, of caring for LBW and preterm babies, and the longer-term health consequences, such as cardiovascular disease and diabetes, that are associated with being born early or small. It is somewhat surprising that less public health emphasis has been placed on SGA babies, who share the increase in risk for negative long-term consequences with LBW babies, in light of the high frequency of SGA births in Colorado (12.13%), which is almost double the rate of LBW. SGA has multifactorial causal contours, and is likely undiagnosed during prenatal care. Given the life course complications from SGA, it deserves more study as an adverse outcome. More important, however, studying outcomes only at the low weight end of the spectrum misses an important public health risk that of LGA. Overall, the public health focus on LBW has created a false sense of well-being for babies born weighing more than five and a half pounds. All four outcomes in this study are associated with higher risks for developing obesity and metabolic disorders, especially diabetes, later in life. Certainly the focus of population studies on low weight birth outcomes has left LGA out of the public eye and policy formulation, and even suggested that various populations may be healthier than they really are (Amaro & de Ia Torre 2002, Borrell 2005). In a recent meeting of agencies interested in health baby outcomes in Colorado, a funder mentioned that her agency was considering allocating fewer resources to programs focusing on Hispanics because of their 135

PAGE 153

"better" birth outcomes. Information on the results of this study for LGA was a surprise to the audience. The "Hispanic Paradox" is mistakenly reported even in the mass media, together with cultural stereotypes explaining it. Women who have recently arrived from Mexico have bigger, healthier babies than more affluent non-Hispanic white natives. That's because strong family and social networks support these pregnant women, reminding them what to eat and do. But the longer they stay, and the more assimilated they become, the more bad habits they acquire and the more problems their subsequent babies have. (Brooks 2006). Perhaps most distressing is Colorado's recent campaign to reduce low birth weight with a social marketing message "A Healthy Baby is Worth the Weight," which is not appropriate messaging to mothers of Mexican origin. Although the state is now rethinking its campaign to better align with recommendations to eat a healthy diet and to continue to exercise during pregnancy, the existence of a three year state-wide campaign that focused on gaining weight during pregnancy shows the policy emphasis on low weight outcomes to the exclusion of others. It is tempting to suggest that the better outcomes at the low weight end of the spectrum for Hispanics and mothers of Mexican origin merely represent a shift to LGA outcomes. But the analysis of odds of AGA babies, at least for mothers of Mexican origin, belies that notion because they have better odds of an AGA baby. Although it is possible that a mother with high maternal weight entering pregnancy who smokes will deliver an AGA baby when all signs point to LGA (Barbour, L., personal interview, November 1, 2008), the increase in LGA babies among Mexican-born mothers is not explained by their odds of AGA. It is possible, however, that these mothers have lower odds of LBW and SGA because they have more AGA and LGA babies. Figure 6.1 shows the histograms of birth weight in grams for Mexican-born mothers (above) and U.S.-born mothers of Mexican origin {below). The right shift of the population curve for Mexican-born mothers is consistent with their higher odds of LGA babies and slightly larger babies overall, but it is unlikely to explain the much higher risk of LGA among Mexican-born mothers. Nevertheless, shifting the curve even slightly to the left to would improve outcomes for Mexican-born mothers (Rose 1992). 136

PAGE 154

,.,._, .. 0 ,.. i ,a, 10.0 .. o a :200 aoo ,400 :aeoo .3QOO .-.co 4400 ..co eGICIO eaoo '7'400 11!!1000 eaoo :aGIOO aeoo .. Figure 6.1. BIRTH WEIGHT IN GRAMS OF MEXICAN-BORN MOTHERS (ABOVE) WITH U.S.-BORN MOTHERS Do the Hypotheses in the Literature Explain the Paradox? How Should Health Be Measured? Since two of the hypotheses that try to explain paradoxically positive outcomes are based on the relative health of a population, it is useful to ask how one should measure "health" in this context. The healthy migrant and the healthy immigrant hypotheses suggest that foreign-born women are either healthier overall (healthy migrant) or engage in healthier behaviors, notably with respect to diet, smoking, and weight gain during pregnancy, and are therefore healthier during their pregnancy (healthy immigrant). The epidemiological paradox literature equates "health" with better specific outcomes, while recognizing that immigrant populations have better outcomes for some health conditions, but not for others. Existing literature about birth outcomes has focused on low weight birth outcomes, such as LBW and preterm birth, with a few studies including SGA as an outcome. At a minimum, it is reasonable to suggest that healthy birth outcomes based on weight should include outcomes along the full range of weight-related outcomes, including LGA and perhaps AGA. Another way to measure health is to examine the risk factors associated with various outcomes and determine whether different populations have different health risk profiles. 137

PAGE 155

Hispanic mothers have worse medical risk profiles than Whites (and to a large degree worse than those of Blacks). Mexican-born mothers have higher odds of specific medical risks associated with LGA babies (gestational diabetes, pre-existing diabetes, and previous 4000+ gram infant) than their U.S.-born counterparts. By these measures, neither Hispanics nor Mexican-born mothers can be said to be "healthier" than their comparison groups in the U.S. Healthy Migrant Hypothesis The healthy migrant hypothesis posits that healthier people migrate to the U.S., and that this selection bias explains the paradoxically better outcomes of Mexican-born mothers. Although this study does not test the healthy migrant hypothesis directly, it is possible to critique this explanation. While it is undoubtedly accurate to suggest that those who make the migration to the U.S from Mexico are sufficiently able to make the trip, else they would not be here, that fact alone does not mean they are "healthier" than those who do not migrate, nor does it mean that they make the choice to migrate based on their health. Moreover, the explanation fails on logic alone. To explain birth outcomes in the U.S., the appropriate comparison is between Mexican-born mothers in the U.S. and other U.S. mothers, not a comparison with mothers in Mexico, who are irrelevant to rates of birth outcomes in the U.S. Thus, although they do not speak to the relative health of Mexican women who immigrate and those who do not, the data show that Colorado's Mexican-born mothers are more likely than U.S.-born mothers of Mexican origin to have LGA related medical risks, so they cannot be painted broadly with the "healthier" brush. In addition, the very small differences on odds for each outcome between Models 2 and 3 (where medical risks are entered into the predictive model) suggest that underlying health related to pregnancy does not explain differences in outcomes. For these reasons alone, the healthy migrant explanation falls short. Moreover, studies of immigration from Mexico to the U.S. identify various aspects of political economy, writ large, to explain the motivation to migrate (Portes & Bach 1985). Their individual decisions to migrate (agency) are spurred almost exclusively by structural economic factors. Key informants described the changing face of migrants from Mexico and the poor economic conditions that cause them to come here. I'm not so convinced that it's the healthier women that come, because a lot of time it's impoverished women that come; they are the ones that are willing to take the risk because the poverty levels in their families in Mexico are so desperate that they have to leave. Midwife 138

PAGE 156

Another Colorado mother born in Mexico described the cultural beliefs about food, the harsher economic conditions, and despair of Mexican-born mothers in the U.S. The majority of women in the U.S. have problems with money; economically and there are many women who are not economically well. Therefore it is more difficult to feed ourselves and to live well and to have a good environment for the baby. And sometimes we can't go to the doctor because we don't have money for the appointment. And I think all of that affects many things -and emotionally too. You feel alone here. You don't feel you have support. Even if you have family in the U.S., sometimes you don't feel much support. There are women who are here alone. Alone! Healthy Immigrant Hypothesis The healthy immigrant hypothesis suggests that immigrants engage in healthier behaviors, at least for some period of time, and then the positive outcomes decay over time spent living in the U.S. The data on Colorado mothers support the observation that Mexican born mothers smoke less than U.S.-born mothers of Mexican origin, by a factor of 6. But, Mexican-born mothers are more likely to gain less weight and less likely to gain excessive weight during pregnancy than Whites or U.S.-born mothers of Mexican origin. Yet Mexican-born mothers have much better outcomes than U.S.-born mothers of Mexican origin at the low-weight end of the spectrum (which is consistent with not smoking but not consistent with low weight gain) and they have higher odds of LGA (even though they gain less weight). Weight gain during pregnancy does not provide a complete picture of the effect of weight on the birth outcomes in this study. If a mother begins her pregnancy overweight or obese, even if she gains a normal amount of weight during the pregnancy, she may be at risk for LGA. At least one of the key informants noted that her Mexican-born population of mothers was generally overweight and often obese key risk factors for LGA and gestational diabetes. These observations call into question the validity of "healthy immigranf' behaviors. The slight differences in odds between Models 3 and 4 (where smoking and weight gain are added into the predictive models), suggest that the healthy immigrant hypothesis does not explain differences in outcomes. Smoking is a significant factor in increasing odds of LBW and SGA, and should be avoided. Low weight gain is associated with LBW, preterm birth, and SGA and high weight gain is associated with LGA. However, focusing just on weight gain, rather than maternal weight before pregnancy and the quality of the weight gain is simplistic and does not speak to Mexican-born mothers, for whom pre-pregnancy weight and LGA are significant concerns. Discussions with key informants and mothers describe cultural 139

PAGE 157

beliefs of eating for two during pregnancy and cultural values of their male partners favoring rounder and heavier women. Perhaps most important, differences in availability and quality of food in Mexico and the U.S., abrupt changes in energy expenditure, and social and linguistic isolation likely contribute to a mini-nutritional transition upon immigration. These factors speak more to choices constrained by structure than to immigrants arriving with individually-based and realized healthy behaviors. Neighborhood Effects Neighborhood effects, especially immigrant orientation, might be a gross measure of availability of social support. As Finch et a/. (2007) demonstrated in Los Angeles, neighborhoods consisting of immigrant ethnic enclaves might moderate the effect of neighborhood deprivation on LBW. In contrast, contextual effects are weak in the two counties measured in this study. Neighborhood deprivation and the interaction of deprivation and immigrant orientation have a marginally significant influence on LGA in Adams County. Immigrant orientation may slightly moderate the negative effect of deprivation on LGA there. In Denver County, only LBW and SGA are weakly influenced by neighborhood deprivation which acts to increase slightly the rate of these two outcomes. Immigrant orientation has no effect. Interviews with mothers in Colorado strikingly confirmed that neighborhood is not a meaningful construct for them nor is it a proxy for social support. It is possible that the slight influence of immigrant orientation in Adams County has little to do with social support but instead merely reflects residential housing patterns among immigrants. Political EconomyA Broader Perspective Even with a broader view of health one that includes individual and area factors ascribing the reasons for an epidemiological paradox to selective migration of "healthier'' migrants, cultural influences that (may) support healthier lifestyles, or the effect of neighborhood enclaves that (may) provide social support does not explain the complexity of human behavior and health outcomes, not even for the few birth outcomes of this study. Instead it may be useful to step back and examine the structural effects of political economy. Behavior is complex and informed by multiple, often contradictory, cultural and health beliefs. Political economy can be used to explain those global economic conditions that underlie the genetic condition of humans in the current era, the reasons for migration from one economic environment to another, the dietary and cultural changes associated with immigration to a different culture, and the structural constraints on individual behavior in any political 140

PAGE 158

environment in which one lives. While such sweeping statements sound distal to birth outcomes that also have some genetic and biological influences, it is this broader view that provides a framework for the epidemiological findings of this study. Only then might it be helpful to suggest interventions to modify health behaviors that pertain to the specific outcomes of specific populations at risk. Significance The public health importance of this study centers on its revelation of important current health realities and insight into the structure of health disparities. By including LGA, it broadens the range of weight-related outcomes so that any paradox does not mask other negative outcomes. In the same vein, it shows how important it is not to make policy decisions based on the traditional view of the paradox that misses both heterogeneity of the population and a broader view of outcomes. It is hoped that Colorado will adopt culturally appropriate messaging about maternal weight and weight gain during pregnancy that addresses the specific risks and needs of Hispanic women, who represent 30% of the singleton births in Colorado. The broader political economic perspective on LGA suggests that reliance on individual interventions or social marketing alone will be insufficient. To the extent that Mexican-born immigrants are constrained by structural barriers to better health outcomes, such as poverty, lack of access to food stamps and healthy foods, real social support, especially for Mexican women living alone in the U.S., and social and linguistic isolation, broader approaches to pre-pregnancy and pregnancy assistance will be needed. And not to be lost in a study of the ''whys," it is important to remember that as the odds of positive low weight outcomes among Mexican-born women decay over time, the health of the largest subpopulation in the U.S. will deteriorate rapidly in coming years and set the stage for trans-generational perpetuation of ill health related to adverse health outcomes. Indeed, it may be time to eschew focus on paradoxical outcomes and instead work to improve birth outcomes for all populations. Even those who may temporarily enjoy paradoxically better outcomes today for LBW, preterm birth, and SGA also represent its disappearance later their own childbearing years or in the next generation. And the focus on low weight-associated outcomes obscures LGA where there is no immigrant paradox. Accordingly, several next steps to expand on this study might include policy recommendations for Improved data collection relating to causes of LGA in PRAMS (Pregnancy Risk Assessment Monitoring System). Although questions about pre-existing and 141

PAGE 159

gestational diabetes are included in PRAMS, there are no detailed questions about nutrition (nothing about fruits, vegetables, methods of cooking [fried, baked, steamed, etc.]), or questions about depression generally or during pregnancy (as compared to post-partum) Greater of fruits and vegetables on the WIG list of approved foods Improved prenatal and inter-conceptual counseling, especially for recent immigrants from Mexico. In addition, LGA may be an outcome that affects immigrants from countries other than Mexico. The political economy approach suggests that the conditions that may lead to increased risk of LGA in Mexican immigrants may apply equally to other immigrants. A national population study of LGA among foreign-born immigrants is warranted. 142

PAGE 160

APPENDIX A SELECTED STUDIES 143

PAGE 161

Appendix A. Selected Individual-level studies Author/Year Cobas eta/. (1996) Singh & Yu (1996) Frisbie eta/. (1998) Data HHANES1980s 1985-1987 NCHS linked birth/infant death files 1989-1991 NCHS linked birth/infant death files Hummer eta/. 1989-1991 NCHS (1999) linked birth/infant death files Fuentes-Afflick 1992 CA birth et a/. certificates (1999) Lansdale eta/. Puerto Rican (1999) Maternal and Infant Health Study 19941995 Hessol & 1990-1993 CA Fuentes-Afflick linked birth/infant (2000) death files Buekens eta/. 1994 US birth (2000) Certificates Chung JH eta/. 1997-2002 (2003) Retrospective Cohort Memorial Health Care S tern Frisbie & Song 1995-1997 NCHS (2003) Unked birth/infant death files compared with 1989-1991 Iiies Gould eta/. (2003) 1995-1997 CA linked birth/infant death records Outcomes LBW LBW Preterm birth Infant mortality LBW Preterm birth Infant mortality Infant mortality VLBW1 MLBW LBW Infant mortality LBW Infant mortality Birth weight LBWby gestation LBW Preterm birth Infant mortality VLBW MLBW Preterm birth SGA<3% IM Resulta 1. Acculturation factors differentially affect LBW 2. Acculturation affects LBW through diet and smoking 3. Language is more important than ethnic ID on acculturation 4. Independent of dietary intake, acculturated women are more likely to have LBW babies 1. Foreign-born status associated with reduced risk of IM and LBW for, among others, Mexicans 1. Paradox applies to Hispanic subpopulations for IM adjusted for LBW and preterm birth 1. Nativity affects pregnancy outcome by racelethnicity, when large % is foreign-born 2. Mexican-born have less risky health profiles than US-born (especially re smoking) 3. BW and preterm birth are intervening variables in IM 1. VLBW OR 0.93 latina/White 2. MLBW OR 1.0 latina/White 3. VLBW OR 1.0 foreign/US-born (ns) 4. MLBW OR 0.93 foreign/US-born 1. Recent immigrants experience fewer stressful life events; less likely to engage in negative health behaviors during pregnancy 2. Recent immigrants have better outcomes than earlier immigrants or US-born of Puerto Rican descent 1. LBW OR 0.98 latina/White 2. IM OR 0.88 latina/White 3. Some risk factors for LBW vary from risk factors for IM, e.g., age <18 = less LBW but higher IM 1. Lower LBW of Mexican American blc fewer small, preterm babies 2. But mean BW lower for Mexican American than White and overall preterm BW higher than for Whites, which may represent misclassification 1. Descending ranking of infants by BW for gestation =Whites, Hispanics, Blacks 2.Differences in LBW due to differences in size at birth because of gestation 1. All groups showed increased rates of adverse birth outcomes but decreased rate of IM 2. Mexican Americans had higher preterm birth rate than Whites in both time periods, but lower rates of IM than Whites 1. Despite high risk demographic profile, Mexican-born did not have elevated levels of LBW or neonatal mortality 2. But Asian Indians, with lower risk demographic profile had high levels of LBW, SGA, fetal mortality 3. Higher education, better prenatal care and private insurance was protective for White and African American but not Mexican or Asian Indian mothers -"dual paradox" 144

PAGE 162

Appendix A. Selected Individual-level studies (continued) Author/Year Data Outcomes Cho eta/. 1989-1994 National Self-reported (2004) Health Interview overall health Survey Daily activity limitations #days in bed Rosenberg et 1996-1997 NYC LBW a/. (2005) birth files Acevedo1998 Detail Natality LBW Garcia dataset eta/. (2005) Appendix A. Selected multi-level studies Author/Year O'Campo eta/. (1997) Data 180 Census tracts in Baltimore linked to birth certificate data 1985-1989 Johnson et a/. CO 1992-1994 (1999) Birth certificate data Pear1 eta/. (2001) 1994-1995 birth records 18 CA hospitals Gorman (1999) 1990 linked birth and death files Reagan and Salsberry (2005) Sellstrom and Bremberg (2006) 1979-1998 National Longitudinal Survey of Youth 1979 cohort, native born only Review of multiple level studies Outcomes LBW BWas continuous variable Preterm birth BWas continuous variable LBW Preterm birth (very and moderate) LBW Child behavior Child injuries Child maltreatment Finch eta/. (2007 2000 birth records LBW for Los Angeles Coun CA Results 1. Hispanic subgroup differentials wide 2. Foreign nativity = more favorable outcomes, supporting healthy migrant hypothesis 1. Positive outcomes of foreign-born largely due to more favorable distribution of behavioral risk factors 2. Nativity is significant predictor of LBW only among Mexicans (OR 0.6) and "other Hispanics" (OR 0.7) 1. Interaction between racelethnicity and nativity (foreign-born) significant 2. Protective effect of foreign-born for Blacks and Hispanics 3. Inverse relationship between low education and LBW for foreign-born Hispanics Reeulte 1.Some neighborhood characteristics directly associated with and had interactions with higher odds of LBW among Blacks 2. Individual risk factors for LBW behaved differently depending on characteristics of neighborhood 1. Male unemployment has larger effect than crime rates 2. No particular effects for Latino population 1 Less favorable neighborhood = lower BW for Black and Asian 2. No consistent relationship of neighborhood for foreign-born or US-born L.atinas or Whites. 3. BW increased with less-favorable neighborhoods among foreign-born L.atinas in high poverty or high unemployment neighborhoods 1. Differences by racelethnicity are function of individual and area characteristics 2. Negative relationship between % foreign-born in county and LBW 3. Specific contextual variables vary by racelethnicity 1. Neighborhood disadvantage highly sensitive across racelethnicity, depending on measure 2. Direct effect of cumulative income inequality only for Hispanics; Female head of household frHisp. very preterm ft 1. Risk of LBW increased 1 0%+ for mothers in disadvantaged neighborhoods 1. Protective effect of Hispanic immigrant co-residence at neighborhood level, especially for foreign-born 145

PAGE 163

APPENDIX B HUMAN SUBJECTS APPROVALS 146

PAGE 164

Univel'lllty of Colorado at Denver nd Health Selene Center Human Subje(:b Reeeardl CommittH Institutional Rlllliew Board Downtown Denver Campua Boll 120, P.O Box173384 Denver. Colorado 80217-3364 Phone: Fax: 303-55e-33n DATE: October 5, 2006 TO: Sharon Devine FROM: Debbie Kellogg. HSRC Chair SUBJECf: Human Subjects Research Protocol H2007-0J8-The Latina Epidemiological Paradox in Colo111do Your protocol has been approved as exempt under CFR Title 45 Part 46.101.b. This approval is good for up to one year from this date Your responsibilities as a researcher include: If you make changes to your research proloool or design you should contact the HSRC so that we can determine if your exempt status continues You an: responsible for maintaining all documentation of consent. Unless specified differently in your protocol all data and consents should be maintained for three yeal'9. If you should encowtter adverx human subjects issues, please contact us immediately. If your research continues beyond one year from the above date, contact the HSRC for an extension. The HSRC may audit your documents at any time. Thank you for submitting yom protocol and good luck with your research Campuses. Downtown Denver Fitzsimons at Aurora Ninth and Colorado 147

PAGE 165

OO'>,...TOIVN OfN\'1:1< Lvo
PAGE 166

Univel'sity of Colorado Denver Domatcnm C'UDpu 0311212008 Cartlflcate of Approval Investigator: Sponsor( a): Subject: Sharon Devine HSRC Protocol 2008-110 Initial Review (APP001) 2nd AppfuYal Dale: 11 Marctl 2008 Expiration Date: 11 March 2008 Expedited Calegofy: 7 Tllfe: 1380 LIIINTence Street, Sulta 300 Campua Box 120 P.0.8001 173364 Denver. co 80217-3364 WH'I' DO -IIORH IN MEJaC0 0ELNeR HEALTHIR BABES THAN WOWIN OF MeC1CNC ORIGIN lORN IN THE U.S? Appruvallndudes: Protocol lnveatlgator 1 Form(a). 2 Advertlsement(s) 1 Questionnalnt(a) Watver of Conaent AI HaRe ,......,_. ,,_....,.,. mu .. -.p1J with the lolowlng: For 1M durdan any ohatge In !he ....--UI wwJit:tt ...-tlonn mUit by lhe HSRC befarelmplam.-...on ollie clwlgel. Use orly a copy ol .. HlllC lllgReCI end dDed eo-t 8lldlar AAaerit FCIIIIL Tile lie ruponalbllly lOr oblliininQ flam .. SUbjecla .,'*""-! Co...-.r' IIPII"Mid by 11W HSRC The HSRC REQUIRES lhel b 8ubjecl be s;-a c:opy of the endlol...c 1om. ear-wwJit:tt forms n.-lnctudll .,. and lilltlphDIIe ......,., of .. Prvvllle n-.1!1.,.,. ....-... eulljecta w1t11 certltled tn11181811an o1 the c--mdl Auenl Form in the IM.Cijed'a Tha *o t..w lie .-panaibilty for lnfanoolng the H8RC lmiMdl...., of_, Unanllalpeled Pnoblema lhllt .., rolfatad lo the unapeclld per HSRC Poley and P!ooedu.._ Ob..., H8RC epproqllor .. ..,...._.., queellolwllra or ...,..,. bdn .,._ .,..... .......... ...,... c-118ulng Review ID ,._ appawil oftllle projecl wlll*l 12-manl'l J81oc1 flam lhe II*...._ ollwwM lildic*d In the I'INWw qde IIDd below. If you 1ww llak pnliDoal, IP8dfla Clelillll w11 be ouii!Md In IIIIa ...... Nan-campi.._ wllh Conllnulng ANew wll _.In ._ lilnnlnatlon of IIIIa n. prgjec;t ha been aa9*' .... HSRC Continuing Rr#lew Cycle: 12 months We wll -.Ill yau eora.utng Revt.w Fonn to be oompiiMCI priar ID lhe due dele. Maty Geda, MSH 20111-110 P ..... : 8 149

PAGE 167

Univea,.ity of Colorado Denver ('uapu 1 J80 Lawrence Slreel. Sulle 300 CIIITipuS Bo 120 P.O.ao. 173364 osn-. co 80217-3364 08/1912008 Certificate of Approval Investigator: Sponsor: Subject: ApprOYBI Date: Approval Includes: Sharon Devine Ucd Heallh And BehaYioral Sciences Dept HSRC Protocol2008-110 Amendment Review (PAM002) 2nd 18 August 2008 Title: WHY DO WOMBIIIOAN IN IIEICICO DELIVER HEAL THIER IIAIEB THAH MElOCAN ORIGIN lOAN IN THE U.S.? 1 Consent Form(a) 1 Amendment(s). 1 Quntlonnaire(a) Protocol Amendment Approved Dated: Description: Ma')' G.U, MSN Rrmad03/05 0712412008 00:00:00 Prokxxll ArnendmentfConsent FOfTil Addition/Questionnaire Addition: Thill amendment adds lniBrViewa with up to 4 key Informants. Included are an addiOonal conaent fonn and additional topic:s lil!lt al for the key informant il1biilliews. Two of the key lnfor'IMntl are iclentltlad: 1) Lila W819l, a Registered NurM working with the Nurse Home Vi*lg Program in Boulder County. The Nwse Home VIsiting Program. devaloped by Dr. Dftkl Olda (UC DIII'Mtr), Ylsb at risk mothe,.. for up to 2 yaara and Ia deelgned to help mothers take better an of thermleiYea and their bablell. Lisa's WOIIt Ia primarily will young Hispanic ITIOthera. 2) S1aphanie Goodrrwl, CNM, PhD, who has worked aa a midwle with Hispanic mothers. I have not Identified one or two additional potential key lnforrn.-.. They wll be people who 'MJr1( with mottwn. No information about .-.y particular other ex mollen will be solciled in the lnteMewa. This amendment wil not change the number af !IUbjecta in my original application becauH at this time, I do not antlclpale reaching the total number of subfecta already-approved (up to 50). \Y VYonne Keii-Guenther, ( TonyRobirwiOfl 200&-111 Penel: 8 150

PAGE 168

of Colorado Dt>nvn '/ DO\rntO\Ql C'amp"s 08/19/2008 MIIIYGeda,I\IISN Revised 03105 Protocol with highlighted revisions Consent for Kay lnfonnanm Discussion Guide for Kay Informant Interviews Tony Robinson 151 1380 Street. Suite 300 Campus Box 120 P.O.Box 17338<1 Denver. CO 80217 2001-110 Pwoel: a

PAGE 169

APPENDIX C SOLICITATION GUIDE, INTERVIEW GUIDES, CONSENTS 152

PAGE 170

Solicitation Script Approved by Humans Subjects Review Committee March 11, 2008 Would you like to participate in a study about differences in the size of babies at birth? Salud is helping Sharon Devine find women of Mexican origin for a research study about their pregnancy experiences. This study is being conducted by Sharon through the University of Colorado Denver. If you o are at least 18 years old o were born in Mexico or born in the U.S and of Mexican origin o have had a baby in the past year o are willing to be interviewed for an hour I will give Sharon your telephone number and she will call you to see if you want to be in the study. If you agree to be in the study you will be paid for your time. Is it OK for me to give Sharon your first name and phone number? She will call you if you are interested. You can say no and it will not affect your access to Salud's services. 153

PAGE 171

t,Le gustarfa participar en un estudio sobre las diferencias de tamaiio de los bebes? Salud esta ayudando a Sharon Devine a encontrar mujeres mexicanas para un estudio sobre sus experiencias durante el embarazo. Sharon conduce este estudio con Ia Universidad de Colorado Denver. Si usted tiene por lo menos 18 aiios es Mexicana o de origen mexicano (Sharon quiere hablar con madres que nacieron en Mexico o en los Estados Unidos) tuvo un bebe el aiio pasado esta de acuerdo en hablar con Sharon por no mas de una hora le dare su mimero de telephono a Sharon y ella le llamara para preguntarle si quiere participar en Ia entrevista. Siesta de acuerdo en participar, Sharon leva a pagar por su tiempo. Por favor dfgame si le puedo dar a Sharon su nombre y numero de telefono. Ella le llamara por telefono y preguntara cuando y d6nde Ia puede ver. Usted decide si participar o no. Si no quiere participar en el estudio, va a seguir recibiendo atenci6n medica en Ia clfnica Salud como de costumbre. 154

PAGE 172

Topics for Semi-Structured Interview After Consent Process These are not the exact questions, but the topics and questioning approaches. Approved by Human Subjects Review Committee March 11, 2008 Introduction Going to ask questions, maybe share stories, want to know about your pregnancy, neighborhood, where/from whom you got support while pregnant. No right answers, just want you to tell me what is comfortable about your pregnancy. Audio tape so I don't miss anything. You can stop at any time; you can decide not to answer any question that you would rather not answer. Confmn whether born in Mexico or born in U.S. and of Mexican origin. If born in Mexico at what age did you come to U.S.Ihow long have you been in U.S.? Repeat that do not need to know full name, address, or immigration status. About the baby How long ago was your baby born? Boy? Girl? Other children? First pregnancy? All babies living? Remember if baby born early (premature)? How much baby weighed when born (approximate is OK). Any complications with birth? Probe: c section? gestational diabetes, hemorrhage, other? Things going well now? Health behaviors What kinds of foods did you eat while pregnant? Probe for vegetables, home cooked, prepared (like from grocery store), fast foods What did you eat on a typical day? Breakfast/lunch/dinner. Did you smoke while you were pregnant? Before pregnant? If yes, probe how far along when found out pregnant, continue to smoke? Less? More? If no, why didn't you smoke? 155

PAGE 173

Did others in household smoke while you were pregnant, when in same room? How much? Did you drink alcohol (including beer) while you were pregnant? Before pregnant? If yes, probe how far along when found out pregnant, continue to drink? Less? More? If not, why didn't you drink? Did you gain a lot of weight while pregnant? Any idea how much? Exercise/activity about same before pregnant as when pregnant? Do you remember when you ftrst went to the doctor/clinic after you became pregnant for prenatal care? Probe: how far along in the pregnancy? Diabetes? Check in I'm going to ask questions next about your neighborhood and the types of support you received during your pregnancy. Are you comfortable continuing? Need a break? Neighborhood How would you describe your neighborhood? Probe: where generally-city/more rural? Types of people who live there? Mostly people of Mexican origin? More recent immigrants than not? Probe for how much/%. Feel safe in neighborhood? Friends or family in neighborhood? Would you describe your neighborhood as poor? While pregnant did you spend most of your time in your neighborhood? How long have you lived in your neighborhood? Support Tell me/list the people who gave you support while you were pregnant-not their actual names, but, for example, my sister, who lives next door, or my mother, who lives in Mexico -I would call her each week, or my husband, or whatever. Elicit list. Ask which ones provide which types of support-probe: information? rides? maternity clothes? baby clothes/equipment? financial support? emotional support? Groups of people, church? Which ones most important to you? Why? 156

PAGE 174

An alternative if the listing is not eliciting responses would be to tell a story and ask if the woman were in this situation, where would she find solace/support. Example: You are 6 months pregnant. You can't remember when you felt the baby last move and you are getting worried. Would you talk to someone about your fears? Whom? Example: You are pregnant and your family lives far away, except your husband. This is your first child and you don't know what to expect. You are feeling overwhelmed. What would you do? Probe for people, groups, church, as sources of support. Familiar with concept of machismo/marianismo? Was marianismo a strong value during your pregnancy? How would you say it was expressed? By whom, examples? Did you practice Ia cuarentena after pregnancy? Probe-completely, partially? Did you know you would do Ia cuarentena before the baby was born? Did Ia cuarentena make you less anxious about your pregnancy? Can you think of anything else you think would explain how you coped with being pregnant and who/what was helpful? Sisters Do you have any sisters who live in Mexico? Have they had babies in Mexico? If so, tell me about their babies size, premature? How old were they when they had the/each baby? Finding that mothers born in Mexican mothers are very "healthy" when look at low weight gain outcomes that is, much less low birth weight babies, babies that are premature, or that are small given their gestational age. BUT they have more large for gestational age babies. And with women of Mexican origin who were born in the US, the pattern is the opposite. What do you think is happening? notice this based on your family or acquaintances? Significant changes in behavior between US and Mexico that are important and significant to them? What do you think is happening? Thank you 157

PAGE 175

Thank you very much for your help. Make payment. This research is for my dissertation, which will result in a long written report of my study. When it is finished (in about a year!), I will give a copy if it to Salud, in case you would like to see it. 158

PAGE 176

Consent Form Approval Date: Approved by HSRC March 11, 2008 Valid for Use Through: Study Title: Why do Women Born in Mexico Deliver Healthier Babies in the U.S. than Women of Mexican Origin Born in the U.S.? Principal Investigator: Sharon Devine HSRC No: 2008-110 Version Date: Version No: 1 You are being asked to be in a research study. This form provides you with information about the study. A member of the research team will describe this study to you and answer all of your questions. Please read the information below and ask questions about anything you don't understand before deciding whether or not to take part. Why is this study being done? The goal of this study is to learn about the difference in size of babies. You are being asked to be in this study because you did all the following. You had a baby within the last year. You are of Mexican origin. You got health services from Salud. Up to 50 people will be in the study. What happens if I join this study? If you join the study, you will answer some questions. This should take less than one hour. You will be asked how you took care of yourself while you were pregnant. You will be asked whether you smoked or drank alcohol while you were pregnant. You will also be asked to tell me about your neighborhood. I will ask you questions about the types of people who live there, and the support the area provided during your pregnancy. In the interview, I will share with you some stories about pregnancy. I will ask you whether your experience was the same or not. 159

PAGE 177

Consent Form Approval If you have sisters living in Mexico who have had children, you will be asked to tell me a little about what happened. Were their babies born early or small? I will also ask how old you sister was when the baby was born. What are the possible discomforts or risks? Answering my questions may make you uncomfortable. You may be upset or sad if your pregnancy was difficult or you felt that you did not have much support, embarrassment. You may also feel guilty you smoke or drank during your pregnancy. Finally, you may be embarrassed if you felt you did not have enough money to get care while you were pregnant. What are the possible benefits of the study? There is no direct benefit to you for being in this study. We hope to learn how to increase the number of healthy babies born to women of Mexican origin. Who is paying for this study? This research is being paid for by the University of Colorado Denver. Willi be paid for being in the study? Will I have to pay for anything? You will be paid $20 for participating in the interview. It will not cost you anything to be in the study. Is my participation voluntary? Taking part in this study is voluntary. You have the right to choose not to be in this study. If you choose to be in the study, you have the right to stop at any time, or to refuse to answer specific questions. If you refuse or decide to withdraw later, you will not lose any benefits or rights to which you are entitled. If you choose not to be in the study, you will still be able to get health services from Salud. 160

PAGE 178

Consent Form Approval Who do I call if I have questions? The researcher carrying out this study is Sharon Devine. You may ask any questions you have now. If you have questions later, you may call Sharon at 303-556-6797. You may have questions about your rights as someone in this study. You can call Sharon Devine with questions. You can also call the Human Subject Research Committee (HSRC). You can call them at 303-315-2732. Who will see my research information? Information from the interview may be looked at by others. They are: o Federal agencies that monitor human subject research o Human Subject Research Committee o The researchers o The funder (the Health and Behavioral Sciences program at the University of Colorado Denver). People who work at Salud will not know if you agree to be in the study. People who work at Salud will not see information from the interview that identifies you. You will not be asked to give us your name, specific address, or your immigration status. The results from the research may be shared at a meeting or in published articles. Your name will be kept private when information is presented. Audio recording You may choose to speak either English or Spanish during your interview. The interview will be audio recorded and then written in a paper document. The audio file will be deleted after it is transferred to paper and all copies of the interview will be kept in a locked office file or on a secure computer that can be accessed only by the researchers. Three years after the study is concluded, the paper and computer copies of the interviews will be destroyed. 161

PAGE 179

Consent Form Approval Agreement to be in this study I have read this paper about the study or it was read to me. I understand the possible risks and benefits of this study. I know that being in this study is voluntary. I choose to be in this study: I will get a copy of this consent form. Date: __ Consent form explained by: ________ Date: ___ Investigator: ______________ Date: ___ 162

PAGE 180

Consent Form Approval Fecha: Approved by HSRC March 11, 2008 Esta forma es valida hasta: Titulo del estudio: (,Por que las mujeres mexicanas tienen bebes mas saludables que las mujeres de origen mexicano que nacen en los Estados Unidos? lnvestigadora Principal: Sharon Devine Numero HSRC: 2008-110 Fecha de esta versi6n: Versi6n Numero: 1 Se le esta invitando a participar en un estudio de investigacion. Esta forma le da informacion sabre este estudio. Una persona del equipo de investigacion le va a describir de que se trata este estudio y respondera a sus preguntas. Por favor lea Ia informacion que se presenta a continuacion y pregunte cualquier cosa que no entienda antes de decidir si quiere participar o no. (.Por que estamos hacienda este estudio? Porque queremos aprender sabre las diferencias de tamano de los bebes. La estamos invitando a que participe en el estudio porque: Tuvo un bebe en este ano que paso Es mexicana Se atendio en una cllnica Salud Hasta 50 mujeres van a ser parte de este estudio. (.Que pasa si partlcipo en este estudlo? Si participa en el estudio, va a contestar algunas preguntas. La entrevista no tamara mas de una hora. Se leva a preguntar sabre como se cuido mientras estaba embarazada. Se le va a preguntar si fumo o tomo alcohol cuando estaba embarazada. Algunas preguntas son sabre ellugar donde vive. Le voy a hacer preguntas sabre las diferentes personas que viven en Ia misma area, y el apoyo que recibio durante su embarazo. En Ia entrevista, voy a contarle historias de algunos embarazos y le voy a preguntar si su experiencia fue Ia misma o no. 163

PAGE 181

Consent Form Approval Si usted tiene hermanas en Mexico que tienen hijos, le voy a pedir que me cuente un poco sobre sus bebes, si los bebes de sus hermanas nacieron antes de tiempo o muy pequenos. Tambien le voy a preguntar que edad tenlan sus hermanas cuando sus hijos nacieron. (,Cuales son los posibles riesgos del estudio o Ia posibilidad de que me sienta incomoda? Responder estas preguntas podrla hacerle sentir inc6moda. Puede ser que las preguntas le causen tristeza si usted tuvo un embarazo diflcil o si no recibi6 mucho apoyo o si su embarazo fue causa de verguenza. Puede ser que sienta culpa si fumo o tom6 alcohol durante su embarazo. Tal vez sienta verguenza si no tuvo suficiente dinero para ir al medico a checarse cuando estaba embarazada. (,Cuales son los posibles beneficios de este estudio? Usted no obtendra ningun beneficia por participar en este estudio. Nosotros esperamos aprender c6mo aumentar el numero de bebes saludables de mamas de origen mexicano. (,Quien esta pagando por este estudio? Esta investigaci6n es pagada por Ia Universidad de Colorado en Denver. (,Me pagaran por participar en este estudio? (, Tengo que pagar algo por partlcipar? Le pagaremos $20 por participar en Ia entrevista. No le costara nada ser parte de este estudio. (,Que pasa si no quiero partlcipar? (,Es mi partlcipacion voluntaria? Participar en este estudio es completamente voluntario. Usted decide si quiere o no formar parte. Si decide ser parte del estudio, puede parar Ia entrevista en cualquier momenta, o puede no contestar algunas preguntas si no quiere. Si no quiere participar o si una vez que empecemos decide parar 164

PAGE 182

Consent Form Approval Ia entrevista, usted no perdera ningun beneficia o derecho que le corresponda. Si no quiere participar en el estudio, puede seguir recibiendo atencion medica en una cHnica Salud. ;.,A quien le llamo si tengo preguntas? La investigadora a cargo de este estudio es Sharon Devine. Puede hacer cualquier pregunta que usted tenga ahora. Si tiene preguntas despues, puede llamar a Sharon al 303-556-6797. Puede ser que tenga preguntas sobre sus derechos al participar en este estudio. Puede llamar a Sharon Devine con preguntas. Tambien puede llamar al Comite de Sujetos Humanos que participan en investigacion (Human Subject Research Committee) al 303-315-2732. ;.,Quien podra ver Ia informacion de esta investigaci6n? Las instituciones o personas que podrfan ver Ia informacion de las entrevistas son: o Agencias federales que se encargar de vigilar Ia investigacion con sujetos humanos o El Comite de lnvestigacion con Sujetos Humanos o Los investigadores o El donador ( el program a de Salud y Ciencias de Ia Conducta de Ia Universidad de Colorado en Denver). El personal de Salud no va a saber si usted aprueba ser parte del estudio. El personal de Salud no podra ver informacion de las entrevistas que Ia pueda identificar. No le vamos a pedir su nombre, direccion o su estado migratorio. Los resultados de esta investigacion podran ser difundidos en conferencias o publicados en artfculos de revistas. No se usara su nombre cuando se presente informacion en publico. 165

PAGE 183

Consent Form Approval Grabacion de audio de las entrevistas Puede elegir si prefiere hablar en ingles o en espaiiol durante Ia entrevista. La entrevista sera grabada para poder ponerla por escrito mas tarde. La grabaci6n sera borrada en cuanto Ia entrevista se encuentre por escrito y las capias de las entrevistas se guardaran en un archivero bajo llave o en una computadora con medidas de seguridad de manera que solo los investigadores las puedan ver. Tres aiios despues de que se termine el estudio tanto las capias en papel como los archives de Ia computadora seran destruidos. Acuerdo para participar en este estudio He lefdo este documento sabre el estudio o alguien me lo ha lefdo. Entiendo los posibles riesgos y beneficios del estudio. Se que participar en el estudio es voluntario. Deseo participar en el estudio: voy a obtener una copia de esta forma de consentimiento. Fecha: ____ Forma de consentimiento explicada por: _______ Fecha: __ lnvestigador(a): _______________ Fecha: __ 166

PAGE 184

Topics for Discussion in Key Informant Interviews 2008-110 Approved by HSRC August 18, 2008 1. Description of quantitative findings In accord with expectations and the "Hispanic paradox," Mexican born mothers in Colorado have much better (lower) odds than U.S. born mothers of Mexican origin for low birth weight, small for gestational age, and preterm birth. Mexican-born mothers have much higher odds of LGA than U.S. born mothers of Mexican origin. 2. The "healthy immigranf' hypothesis suggests that women who migrate to the US are "healthier'' than those who are born in the US. What does healthier mean in this context? 3. What factors might account for the higher rate of LGA among Mexican born mothers? Diet Exercise Cultural values (and how expressed) 167

PAGE 185

Consent Form Approval Date: Approved by HSRC August 18, 2008 Valid for Use Through: Study Title: Why do Women Born in Mexico Deliver Healthier Babies in the U.S. than Women of Mexican Origin Born in the U.S.? Principal Investigator: Sharon Devine HSRC No: 2008-110 Version Date: Version No: 2 (Key Informants) You are being asked to be in a research study. Why is this study being done? The goal of this study is to learn about four birth outcomes low birth weight, preterm birth, small for gestational age, and large for gestational age among women who have given birth in Colorado during the years 2000 2005. You are being asked to be in this study because you have experience working with mothers who have recently given birth in Colorado. Up to 50 people will be in the study. What happens if I join this study? If you join the study, you will answer some questions. This should take less than one hour. You will be asked if you have insights, based on your experience, into findings that suggest that: In accord with expectations and the "Hispanic paradox," Mexican-born mothers in Colorado have much better (lower) odds than U.S.-born mothers of Mexican origin for low birth weight, small for gestational age, and preterm birth. Mexican-born mothers have much higher odds of LGA than U.S.-born mothers of Mexican origin. What are the possible discomforts or risks? Answering my questions may make you uncomfortable if you are unable to shed light on the findings. You will not be asked to divulge or discuss 168

PAGE 186

Consent Form Approval personally identifiable information about any of your present or past patients or clients. What are the possible benefits of the study? There is no direct benefit to you for being in this study. We hope to learn how to increase the number of healthy babies born to women of Mexican origin. Who is paying for this study? This research is being paid for by the University of Colorado Denver. Willi be paid for being in the study? Will I have to pay for anything? You will not be paid for participating in the study. It will not cost you anything to be in the study. Is my participation voluntary? Taking part in this study is voluntary. You have the right to choose not to be in this study. If you choose to be in the study, you have the right to stop at any time, or to refuse to answer specific questions. If you refuse or decide to withdraw later, you will not lose any benefits or rights to which you are entitled. Who do I call if I have questions? The researcher carrying out this study is Sharon Devine. You may ask any questions you have now. If you have questions later, you may call Sharon at 303-556-6797. You may have questions about your rights as someone in this study. You can call Sharon Devine with questions. You can also call the Human Subject Research Committee (HSRC). You can call them at 303-315-2732. Who will see my research information? Information from the interview may be looked at by others. They are: o Federal agencies that monitor human subject research o Human Subject Research Committee o The researchers o The funder (the Health and Behavioral Sciences program at the University of Colorado Denver). The results from the research may be shared at a meeting or in published articles. 169

PAGE 187

Consent Form Approval Audio recording The interview will be audio recorded and transcribed. The audio file will be deleted after it is transferred to paper and all copies of the interview will be kept in a locked office file or on a secure computer that can be accessed only by the researchers. Three years after the study is concluded, the paper and computer copies of the interviews will be destroyed. Agreement to be in this study I have read this paper about the study and I understand its possible risks and benefits. I know that being in this study is voluntary. I choose to be in this study: I will get a copy of this consent form. Date: __ Printed Name: ________ Signature: __________ 170

PAGE 188

BIBLIOGRAPHY Abrafdo, A., Dohrenwend, B., Ng-Mak, D., & Turner, J. (1999). The Latino mortality paradox: a test of the "salmon bias" and healthy migrant hypotheses. American Journal of Public Health, 89(10), 1543-1548. Abrafdo-Lanza, A., Chao, M., & Florez, K. (2005). Do healthy behaviors decline with greater acculturation?; Implications for the Latino mortality paradox. Social Science & Medicine, 61, 1243-1255. Acevedo-Garcia, D., Soobader, M., & Berkman, L. (2005). The differential effect of foreignborn status on low birth weight by race/ethnicity and education. Pediatrics, 115 (1 ), e29-e30. Ahlsson, F., Gustafsson, J., Tuvemo, T., & Lundgren, M. (2007). Females born large for gestational age have a doubled risk of giving birth to large for gestational age infants. Acta Paediatrica, 96, 358-362. Ahluwalia, I.B., Ford, E.S., Link, M., & Bolen, J.C. (2007). Acculturation, weight, and weight related behaviors among Mexican Americans in the United States. Ethnicity & Disease, 17, 643-649). Akresh, I.A. (2007). Dietary assimilation and health among Hispanic immigrants to the United States. Journal of Health and Social Behavior, 48 (Dec), 404-417. Alexander, G.R. & Kotelchuck, M. (2001 ). Assessing the role and effectiveness of prenatal care: history, challenges, and directions for future research. Public Health Report, 116(4), 306-316. Alexander, G.R., Kogan, M.D., & Himes, J.H. (1999). 1994-1996 U.S. singleton birth weight percentiles for gestational age by race, Hispanic origin, and gender. Maternal and Child Health Journal,3 (4), 225-231. Amaro, H. & de Ia Torre, A. (2002). Public health needs and scientific opportunities in research on Latinas. American Journal of Public Health, 92 (4), 525-529. American Heart Association and American Academy of Pediatrics. (2005). 2005 American Heart Association guidelines for cardiopulmonary resuscitation (CPR) and emergency cardiovascular care (ECC) of pediatrics and neonatal patients: neonatal resuscitation guidelines. Pediatrics, 117 (5), e1-1 0. Baker, D.W., Cameron, K.A., Feinglass, J, Thompson, J.A., Georgas, P., Foster, S., Pierce, D., & Hasnain-Wynia. R. (2006). A system for rapidly and accurately collecting patients' race and ethnicity. American Journal of Public Health, 96 (3), 532-537. 171

PAGE 189

Balcazar, H., Krull, J., & Peterson, G. (2001). Acculturation and family functioning are related to health risks among pregnant Mexican American women. Behavioral Medicine, 27, 62-60. Barcenas, C.H., Wilkinson, A.V., Strom, S.S., Gao, Y., Saunders, K.S., Mahabir, S., Hernandez-Valera, M.A., Forman, M.A., Spitz, K.R., & Bondy, M.L. (2007). Birthplace, years of residence in the United States, and obesity among Mexican American adults. Obesity, 15(4), 1043-1052. Barker, D. (1998). In utero programming of chronic disease. Clinical Science, 95 (2), 115-128. Barker, D. (2001 ). A new model for the origins of chronic disease. Medical Health Care Philosophy, 4 (1 ), 31-35. Barker, D. (2002). Fetal programming of coronary heart disease. Trends in Endocrinology Metabolics, 13 (9), 364-368. Basch, P. (1999). Textbook of International Health. New York: Oxford University Press. Baumeister, L., Marchi, K., Pearl, M., Williams, R., & Braveman, P. (2000). The validity of information on "race" and "Hispanic ethnicity" in California birth certificate data. Health Services Research, 35 (4), 869-883. Beebe, K.R. (2005). The perplexing parity puzzle. Nursing for Women's Health, 9 (5), 394-399. Bender, D., & Castro, D. (2000). Explaining the birth weight paradox: Latina immigrants' perceptions of resilience and risk. Journal of Immigrant Health, 2 (3), 115-173. Berkman, L., & Clark, C. (2003). Neighborhoods and networks: the construction of safe places and bridges. In I. Kawachi & L. Berkman (Eds.), Neighborhoods and Health (pp.288-302). Oxford, UK: Oxford University Press. Berkman, L., & Glass, T. (2000). Social integration, social networks, social support, and health. In L. Berkman &.I. Kawachi (Eds.), Social Epidemiology(pp. 137-173). Oxford UK: Oxford University Press. Berry, J.W. (1997). Immigration, acculturation, and adaptation. Applied Psychology: An International Review, 46 (1 ), 5-68. Boorstin, D.J. (1987). Hidden History: Exploring our Secret Past. New York, NY: Random House, Inc. Breier, B.H., Vickers, M.H., lkenasio, B.A., Chan, K.Y. & Wong, W.P.S. (2001). Fetal programming of appetite and obesity. Molecular and Cellular Endocrinology, 185, 73-79. 172

PAGE 190

Buekens, P., Notzon, F., Kotelchuck, M., & Wilcox, A. (2000). Why do Mexican Americans give birth to few low-birth-weight infants? American Journal of Epidemiology, 152 (4), 347-351. Borrell, LN. (2005). Racial identity among Hispanics: implications for health and well-being. American Journal of Public Health, 95 (3), 379-381. Brooks, D. (2006, March 30). Immigrants to be proud of. The New York Times. Retrieved November 15, 2008, from The New York Times website: http://select.nvtimes.com/2006/03/30/opinion/30brooks.htm I? r= 1 &scp= 1 &sq=David %20Brooks%201mmiqration%20March%2030%202006&st=cse. Burton, A. & Altman, D.C. (2004) proposed guidelines for reporting missing covariate data. British Journal of Cancer, 91 (1 ), 4-8. Candib, L. M. (2007). Obesity and diabetes in vulnerable populations: reflections on proximal and distal causes. Annals of Family Medicine, 5 (6), 547-556. Carnethon, M.A. (2008). Diabetes prevention in US ethnic minorities: Role of the environment. Journal of American Dietetic Association, 108 (6), 942-944. Casey, B.M., Lucas, M.J., Mcintire, D.O. & Leveno, K.J. (1997). Pregnancy outcomes in women with gestational diabetes compared with the general obstetric population. Obstetrics & Gynecology, 90 (6), 869-873. Centers for Disease Control & Prevention. (2002a). Births: final data for 2000. National Vital Statistics Reports 50 (5): Tables 43, 44, 46. Retrieved January 3, 2007, from the CDC's website: http://www.cdc.gov/nchs/data/nvsr/nvsr50/nvsr50 05.pdf). Centers for Disease Control & Prevention. (2006a). America's children in brief: key national indicators of well being, 2006. Table HEALTH 7: infant mortality: death rates among infants by detailed race and Hispanic origin of mother, selected years 1983-2003. Retrieved January 4, 2007 from the CDC's website: http://www.childstats.gov/americaschildren/tables/health7.asp. Centers for Disease Control & Prevention. (2006b). Births: final data for 2004. National Vital Statistics Reports 55 (1): Table 32. Retrieved January 3, 2007 from the CDC's website: http://www.ded.gov/nchs/data/nvsr/nvsr55/nvsr55 01.pdf. Centers for Disease Control & Prevention. (2006c). Eliminating disparities in infant health. Fact Sheet. Retrieved January 9, 2006 from the CDC's website: http://www.cdc.gov/omh/AMH/factsheets/infant.htm. Centers for Disease Control & Prevention. (2006d). Pregnancy Risk Assessment Monitoring System (PRAMS) Methodology. Retrieved April 15, 2007 from the CDC's website: http://www .cdc.gov/pram s/methodology. htm. Chadwick, E. (1842). Report of an enquiry into the sanitary conditions of the labouring population of Great Britain. London, UK: Poor Law Commission. 173

PAGE 191

Cho, Y., Frisbie, W., Hummer, A., & Rogers, A. (2004). Nativity, duration of residence, and the health of Hispanic adults in the United States. International Migration Review, 38 (1), 184-212. Chung, J., Boscardin, W., Garite, T., Lagrew, D., & Porto, M. (2003). Ethnic differences in birth weight by gestational age: At least a partial explanation for the Hispanic epidemiological paradox? American Journal of Obstetrics & Gynecology, 189 (4), 1058-1062. Clark, L. (2002). Mexican-origin mothers' experiences using children's health care services. Western Journal of Nursing Research, 24 (2), 159-179. Cobas, J., Balcazar, H., Benin, M., Keith, V., & Chong, Y. (1996). Acculturation and low birthweight infants among Latino women: A reanalysis of HHANES data with structural equation models. American Journal of Public Health, 86 (3), 394-396. Collins, J., & David, A. (1990). The differential effect of traditional risk factors on infant birthweight among blacks and Whites in Chicago. American Journal of Public Health 80, 679-681. Collins, J., David, A., Symons, A., Handler, A., Wall, S., & Andes, S. (1998). African American mothers' perceptions of their residential environment, stressful life events, and very low birthweight. Epidemiology, 9, 286-289. Collins, J., David, A., Symons, A., Handler, A., Wall, S., & Dwyer, L. (2000). Low-income African American mothers' perception of exposure to racial discrimination and infant birthweight. Epidemiology, 11(3), 337-339. Collins, J., & Schulte, N. (2003). Infant Health: Race, Risk, and Residence. In I. Kawachi & L. Berkman (Eds.), Neighborhoods and Health (pp. 223-232). Oxford, UK: Oxford University Press. Colorado Department of Public Health and Environment (2000a). Tipping the scales: weighing in on solutions to the low birth weight problem in Colorado. Denver, CO: Colorado Department of Public Health & Environment. Colorado Department of Public Health and Environment (2000b). Vital statistics. Retrieved January 4, 2007, from the CDPHE website: www .cdphe.state.co/hs/statebirthtables2000f2. pdf. Colorado Department of Public Health & Environment. (2000c). Vital statistics Retrieved January 4, 2007, from CDPHE's website: www.cdphe.state.co/hs/county 2000/Colorado OOb.pdf. Colorado Department of Public Health & Environment. (2000d). Technical notes. Retrieved March 10, 2007, from CDPHE's website: http://www.cdphe.state.co.us/hs/2000 appendix 1.pdf. Colorado Department of Public Health & Environment. (2004a). Vital statistics. Retrieved January 4, 2007, from CDPHE's website: www.cdphe.state.co/hs/vs/2004/b15.pdf. 174

PAGE 192

Colorado Department of Public Health & Environment (2004b). Vital statistics. Retrieved January 4, 2007, from CDPHE's website: http://www.cdphe.state.co.us/hs/vs/2004/b11.pdf. Colorado Department of Public Health and Environment. (2004c). Vital statistics. Retrieved January 4, 2007, from CDPHE's website: http://www .cdphe.state.co. us/hs/vs/2004/Colorado 2004. pdf. Colorado Department of Public Health and Environment (2005a). Racial and ethnic disparities in Colorado. Denver, CO: Colorado Department of Public Health & Environment. Colorado Department of Public Health and Environment (2005b). The health status of Colorado's maternal and child health population. Retrieved March 7, 2007, from CDPHE's website: www.cdphe.state.co.us/ps/mch/healthstatus2005.pdf. Crimmins, E.M., Soldo, B.J., Kin, J.K., & Alley, D.E. ((2005). Using anthropometric indicators for Mexicans in the United States and Mexico to understand the selection of migrants and the "Hispanic paradox." Social Biology, 52(3-4), 164-177. Cronbach, L.J. (1951 ). Coefficient alpha and the internal structure of tests. Psychometrika, 16, 297-334. Day, J. (2001). National population projections. Retrieved March 30, 2007, from the U.S. Census website, http://www .census.gov/population/www/pop-profile/natproj. htm I. Desai, S. & Alva, S. (1998). Maternal education and child health: is there a strong causal relationship? Demography, 35, 71-81. Diamond, J.(2003). The double puzzle of diabetes. Nature, 423(5), 599-602. Diez-Roux, A. (2000). Multilevel analysis in public health research. Annual Review of Public Health, 21, 171-192. Diez-Roux, A. (1998). Bringing context back into epidemiology: variables and fallacies in multilevel analysis. American Journal of Public Health, 88 (2), 216-222. Di Leonardo, M. (1984). The varieties of ethnic experience. Ithaca, NY: Cornell University Press. Dollberg, S., Marom, A., Mimouni, F.B. & Yeruchimovich, M. (2000). Normoblasts in large for gestational age infants. Archives of Disease in Childhood Fetal Neonatal, Edition 83, F148-F149. Donato, K., Kana'iaupuni, S., & Stainback, M. (2003). Sex differences in child health: effects of Mexico-US migration. Journal of Comparative Family Studies, 34 (3), 455-477. Drake, A.J. & Walker, B.A. (2004). The intergenerational effects of fetal programming: non genomic mechanisms for the inheritance of low birth weight and cardiovascular risk. Journal of Endocrinology, 180, 1-16. 175

PAGE 193

Drewnowski, A. & Specter, S.E. (2004). Poverty and obesity: the role of energy density and energy costs. American Journal of Clinical Nutrition, 79, 6-16. Dubowitz, T., Acevedo-Garcia, D., Salkeld, J., Lindsay, A.C., Subramanian, S.V., & Peterson, K.E. (2007). Lifecourse, immigrant status and scculturation in food purchasing and preparation among low-income mothers. Public Health Nutrition, 104 (4), 396-4-4. Dyer, J.S., Rosenfeld, C.R., Rice, J., Rice, M., & Hardin, D.S. (2007). Insulin resistance in Hispanic large-for-gestational neonates at birth. Journal of Clinical Endocrinology & Metabolism, 92 (1 0), 3836-3843. Ebin, V., Sneed, C., Morisky, D., Rotheram-Borus, M., Mangusson, A., & Malotte, C. (2000). Acculturation and interrelationships between problem and health-promoting behaviors among Latino adolescents. Journal of Adolescent Health, 28, 62-72. Ehrenberg, H.M., Mercer, B.M., & Catalano, P.M. (2004). The influence of obesity and diabetes on the prevalence of macrosomia. American Journal of Obstetrics and Gynecology, 191 964-968. Eriksson, J., Forsen, T., Tuuomilehto, J., Osmond, C., & Barker, D. (2001). Early growth and coronary heart disease in later life: longitudinal study. British Medical Journal, 322, 949-953. Fang, J., Madhavan, S., & Alderman, M. (1999). Low birth weight: race and maternal nativityimpact of community income. Pediatrics, 103 (1), e5-e10. Finch, B., Lim, N., Perez, W., & Do, D. (2007). Towards a population health model of segmented assimilation: the case of low birth-weight in Los Angeles. Sociological Perspectives, 50 (3), 445-468. Fiscella, K. (1995). Does prenatal care improve birth outcomes? A critical review. Obstetrics & Gynecology, 85 (3), 468-479. Flores, E., & Armijo, C. (2001). Colorado's Latino population grows. Research Brief, 1. Denver, CO: Latina/a Research & Policy Center. Flores, G., & Brotanek, J. (2005). The healthy immigrant effect: a greater understanding might help us improve the health of all children. Archives of Pediatric and Adolescent Medicine, 159, 295-297. Frank, A., & Hummer, A. (2002). The other side of the paradox: the risk of low birth weight among infants of migrant and nonmigrant households within Mexico. International Migration Review, 36 (3), 7 46-765. Fraser, A., Brockert, J., & Ward, A. (1995). Association of young maternal age with adverse reproductive outcomes. New England Journal of Medicine, 332, 1113-1117. Frisbie, W., Forbes, D., & Hummer, A. (1998). Hispanic pregnancy outcomes: additional evidence. Social Science Quarterly, 79(1), 149-169. 176

PAGE 194

Frisbie, W., & Song, S. (2003). Hispanic pregnancy outcomes: Differentials over time and current risk factor effects. The Policy Studies Journal, 31 (2), 237-252. Fuentes-Afflick, E., Hessol, N., & Perez-Stable, E. (1999). Testing the epidemiological paradox of low birth weight in Latinos. Archives of Pediatric and Adolescent Medicine, 153, 147-153. Gorman, B. (1999). Racial and ethnic variation in low birthweight in the United States: individual and contextual determinants. Health & Place, 5, 195-207. Gould, E. (2006). Health insurance eroding for working families: employer-provided coverage declines for fifth consecutive year. EPI Briefing Paper 175. Retrieved May 14, 2007, from the Economic Policy Institute website: http://www.epinet.org/content.cfmlbp175. Gould, J., Madan, A., Qin, C., & Chavez, G. (2003). Perinatal outcomes in two dissimilar immigrant populations in the United States: a dual epidemiological paradox. Pediatrics, 111 (6), e676-e682. Greenland, S. (2002). A review of multilevel theory for ecologic analyses. Statistics in Medicine, 21, 389-395. Guendelman, S. & Abrams B. (1995). Dietary intake among Mexican-American women: generational differences and a comparison with white non-Hispanic women. American Journal of Public Health, 85, 20-25. Guendelman, S., Buekens, P., Blondel, B., Kaminski, M., Notzon, F., & Masuy-Strubant, G. (1999). Birth outcomes of immigrant women in the United States, France, and Belgium. Maternal and Child Health Journal, 3 (4), 177-189. Guendelman, S., Gould, J., Hudes, M., & Eskenazi, B. (1990). Generational differences in perinatal health among the Mexican-American population:findings from HHANES 1982-1984. American Journal of Public Health, 80, 61-65 (Supplement). Hansen, J.P. (1986). Older maternal age and pregnancy outcome: a review of the literature. Obstetrics and Gynecology Survey, 41 (11 ), 726-742. Hales, C.N. & Barker, D.J.P. (2001). The thrifty phenotype hypothesis. British Medical Bulletin, 60, 5-20. Hales, C.N. & Ozanne, S.E. (2003). For debate: fetal and early postnatal growth restriction lead to diabetes, the metabolic syndrome and renal failure. Diabetologia, 46 (7), 1013-1019. Hamlin, C. & Sheard, S. (1998). Revolutions in health : 1848, and 1998? British Medical Journal, 317, 587-591. Hediger, M.L., Overpeck, M.D., Kuzmarski, R.J., McGlynn, A., Maurer, K.R. & Davis, W.W. (1998). Muscularity and fatness of infants and young children born smallor large for-gestational-age. Pediatrics, 102, e60. 177

PAGE 195

Hessel, N., & Fuentes-Afflick, E. (2000). The perinatal advantage of Mexican-origin Latina women. Annals of Epidemiology, 10, 516-523. Hill, J.O., Wyatt, H.r., & Melanson, E.L. (2000). Genetic and environmental contributions to obesity. Medical Clinics of North America, 84 (2), 333-346. Himmelgreen, D, Daza, N.R., Cooper, E., & Martinez, D. (2007). "I don't make soups anymore": pre-to post-migration dietary and lifestyle changes among Latinos living in west-central Florida. Ecology of Food and Nutrition, 46, 427-444. Hosmer, D.W. & Lemeshow, S. (2000). Applied Logistic Regression (2nd ed.). New York, NY:John Wiley & Sons. Hsieh, H-F. & Shannon, S.E. (2005). Three approaches to qualitative content analysis. Qualitative Health Research, 15, 1277-1288. Hummer, R., Biegler, M., deTurk, P., Forbes, D., Frisbie, W., Hong, Y., et al. (1999). Race/ethnicity, nativity, and infant mortality in the Unites States. Social Forces, 77 (3), 1083-1118. Hummer, R., Powers, D., Pullum, S., Gossman, G., & Frisbie, W. (2007). Paradox found (again): infant mortality among the Mexican origin population in the United States. Demography, 44 (3), 441-451. Institute of Medicine. (1985). Preventing low birthweight. Washington, D.C.: National Academies Press. Institute of Medicine. (2006 pre-publication). Preterm birth: causes, consequences, and prevention. Washington, D.C.: National Academies Press. Institute of Medicine. (2003). Improving birth outcomes: meeting the challenge in the developing world. Washington D.C.: National Academies Press. Jensen, G., & Moore, L. (1997). The effect of high altitude and other risk factors on birthweight: independent or interactive effects? American Journal of Public Health, 87, 1003-1007. Johnson, T., Drisko, J., Gallagher, K., & Barela, C. (1999). Low birth weight: a women's health issue. Women's Health Issues, 9 (5), 223-230. Jimenez-Cruz, A. & Bacardi-Gascon, M. (2004). The fattening burden of type 2 diabetes on Mexicans. Diabetes Care, 27 (5), 1213-1215. Jung, C.G. (1980). Psychology and Alchemy: Collected Works of C.J. Jung (Vol. 12) (R.F.C. Hull, Trans.). Princeton, NJ: Princeton University Press. Kana'iaupuni, S., Donato, K., Thompson-Colon, T., & Stainback, M. (2005). Counting on kin: Social networks, social support, and child health status. Social Forces, 83 (3), 11371164. 178

PAGE 196

Kaplan M., Huguet, N., Newsom, J., & McFarland, B. (2004). The association between length of residence and obesity among Hispanic immigrants. American Journal of Preventive Medicine, 27 (4), 323-326. Kasirye, 0., Walsh, J., Romano, P., Beckett, L., Gardia, J., Elvine-Kreis, B., et al. (2005). Acculturation and its association with health-risk behaviors in a rural Latina population. Ethnicity & Disease, 15, 733-739. Kassel, J. (1964). Social science theory as a source of hypotheses in epidemiological research. American Journal of Public Health, 54, 1482-1488. Kawachi, 1., & Berkman, L. (Eds.). (2003). Neighborhoods and Health. Oxford, UK: Oxford University Press. Kawachi, 1., & Subramanian, S. (2006). Measuring and modeling the social and geographic context of trauma: a multilevel modeling approach. Journal of Traumatic Stress, 19 (2), 195-203. Kelaher, M., & Jessop, D. (2002). Differences in low-birthweight among documented and undocumented foreign-born and US-born Latinas. Social Science & Medicine, 55, 2171-2175. Kliegman, R., & Das, U. (2002). Intrauterine growth retardation. In A. Fanaroff & R. Martin (Eds.), Neonatal-Perinatal Medicine: Diseases of the Newborn (pp. 228-262). St. Louis, MO: Mosby. Kleinman, K., & Kessel, S. (1987). Racial differences in low birth weight-trends and risk factors. New England Journal of Medicine, 317 ( 12), 7 49-753. Kotelchuck, M. (1994). The adequacy of prenatal care index: its US distribution and association with low birthweight. American Journal of Public Health, 84, 1486-1489. Kramer, M. (1987). Determinants of low birth weight: methodological assessment and meta analysis. Bulletin of the World Health Organization, 65 (5), 663-737. Kramer, M. (2003). The epidemiology of adverse pregnancy outcomes: an overview. Journal of Nutrition, 5 (Suppl 2), 1592S-1596S. Krieger, N., Chen, J., Waterman, P., Rehkopf, D., & Subramanian, S. (2003). Race/ethnicity, gender, and monitoring socioeconomic gradients in health: a comparison of area based socioeconomic measures-the public health disparities geocoding project. American Journal of Public Health, 93 (10), 1655-1671. Krieger, N., & Gordon, D. (1999). Letter to the editor re: use of census-based aggregate variables to proxy for socioeconomic group: evidence from national samples. American Journal of Epidemiology, 150 (8), 892-894. Kutsche, P. (1998). Field ethnography: a manual for doing cultural anthropology. Upper Saddle River, NJ: Prentice Hall. 179

PAGE 197

Lansdale, N., Oropesa, R., Llanes, D., & Gorman, B. (1999). Does Americanization have adverse effects on health?: stress, health habits, and infant health outcomes among Puerto Ricans. Social Forces, 78 (2), 613-641. LaVeist, T. (1989). Linking residential segregation to the infant mortality race disparity in U.S. cities. Sociological Research, 73, 90-94. Lee, B.A. & Marlay, M. (2007). The right side of the tracks: affluent neighborhoods in the metropolitan United States. Social Science Quarterly, 88 (3), 766-789. Lee, P.A., Chernausek, S.D., Hokken-Koelega, A.C.S., & Czernichow, P. (2003). International small for gestational age advisory board consensus development conference statement: management of short children born small for gestational age, Aprii24-0ctober 1, 2002. Pediatrics, 111, 1253-1261. Lichter, D.T., Brown, J.B., Qian, Z., & Carmalt, J.H. (2007). Marital assimilation among Hispanics: evidence of declining cultural and economic incorporation. Social Science Quarterly, 88 (3), 7 45-765. Link, B., & Phelan, J. (1995). Social conditions as fundamental causes of disease. Journal of Health and Social Behavior, 35 (Extra Issue), 80-94. Lubchenco, L.O., Searls, D.T. & Brazie, J.V. (1972). Neonatality nortality rate: relationship to birth weight and gestational age. Pediatrics, 81 (4), 814-822. Lubchenco, L.O. & Bard, H. (1971 ). Incidence of hypoglycemia in newborn infants classified by birth weight and gestational age. Pediatrics, 47, 831-838. Lynch, J., & Kaplan, G. (2000). Socioeconomic position. In L. Berkman & I. Kawachi (Eds.), Social Epidemiology (pp. 13-35). Oxford, UK: Oxford University Press. Macintyre, S., & Ellaway, A. (2000). Ecological approaches: rediscovering the role of the physical and social environment. In L. Berkman & I. Kawachi (Eds.), Social Epidemiology (pp. 332-348). Oxford, UK: Oxford University Press. Macintyre, S., & Ellaway, A. (2003). Neighborhoods and health: an overview. In I. Kawachi & L. Berkman (Eds.), Neighborhoods and Health (pp. 20-44). Oxford, UK: Oxford University Press. Madan, A., Palaniappan, L., Urizar, G., Wang, Y., Fortmann, S., & Gould, J. (2006). Sociocultural factors that affect pregnancy outcomes in two dissimilar immigrant groups in the United States. Journal of Pediatrics, March, 341-346. Mainous, A.G., Diaz, V.A., & Geesey, M.E. (2008). Acculturation and healthy lifestyle among Latinos with diabetes. Annals of Family Medicine, 6 (2), 131-137). March of Dimes. (2007). Professionals & researchers. Low birth weight. Retrieved March 9, 2007, from March of Dimes website: http://www .marchofdimes.com/professionals/681 1153.asp. 180

PAGE 198

Markides, K., & Coreil, J. (1986). The health of Hispanics in the southwestern United States: an epidemiological paradox. Public Health Reports, 101, 253-265. Marmot, M., Adelstein, A., & Bulusu, L. (1984). Lessons from the study of immigrant mortality. Lancet, 112, 1455-1457. Marmot, M., Kogevinas, M., & Elston, M. (1987). Social/economic status and disease. Annual Review of Public Health, 8, 111-137. Martorell, R. (2005). Diabetes and Mexicans: Why the two are linked. Preventing Chronic Disease. Retrieved November 20, 2008, from CDC website: http://www .cdc.gov/pcd/issues/2005/jan04 01 OO.htm. McGlade, M., Saha, S., & Dahlstrom, M. (2004). The Latina paradox: an opportunity for restructuring prenatal care delivery. American Journal of Public Health, 94 (12), 2062-2065. Medlinger, S. & Cwikei,J. Spiraling between qualitative and quantitative data on women's health behaviors: a double helix model for mixed methods, Qualitative Health Research, 18, 280-293, 2008. Meneses-Ganzalez, F, Romieu, I., Salgado de Snyder, N., Camargo-Bohorquez, C., Hennessy, T. & Schenker, M. (2006). Socioeconomic status, workforce and determinants of health among Mexican immigrant women in the U.S. Journal of Epidemiology, 17 (6 Supp), S385-386). Merson, M., Black, R., & Mills, A. (Eds.). (2001 ). International Public Health: Diseases, Programs, Systems, and Policies. Gaithersburg, MD: Aspen Publishers, Inc. Miech, R., Kumanyika, S., Stettler, N., Link, B., Phelan, J., & Chang, V. (2006). Trends in the association of poverty with overweight among US adolescents 1971 2004. Journal of the American Medical Association, 295 (20), 2385-2393. Millard, A. (1994). A causal model of high rates of child mortality. Social Science & Medicine, 38, 253-268. Modanlou, H.D., Komatsu, G., Dorchester, W., Freeman, R.K. & Bosu S. (1982). Large-for gestational-age neonates: anthropometric reasons for shoulder distocia. Obstetrics & Gynecology, 60, 417-423. Montez, J.K. & Eschbach, K. (2008). Country of birth and language are uniquely associated with intakes of fat, fiber, and fruits and vegetables among Mexican-American women in the United States. Journal of the American Dietetic Association, 108 (3), 473-480). Morales, L., Lara, M., Kington, R., Valdez, R., & Escarce, J. (2002). Socioeconomic, cultural, and behavioral factors affecting Hispanic health outcomes. Journal of Health Care for the Poor and Underserved, 13(4), 447-503. 181

PAGE 199

Morse, J.M. (2008a). 'What's your favorite color? Reporting irrelevant demographics in qualitative research. Qualitative Health Research, 18, 299-300. Morse, J.M. (2008b). Confusing categories and themes. Qualitative Health Research, 18, 727-728. Morse, J .. M. & Field, P.A. Qualitative Research Methods for Health Professionals (2d ed.). Thousand Oaks, CA: Sage, 1995. Murray, C., & Lopez, A. (Eds.). (1996). The Global Burden of Disease. Cambridge, MA: Harvard University Press. National Institute of Nursing Research, National Institute of Child Health & Human Development, & National Institute of Dental & Craniofacial Research (2003). Reducing preterm and low birth weight in minority families. PA-04-027. Washington, D.C. Neuhauser, M., Thompson, B., Coronado, G., & Solomon, C. (2004). Higher fat intake and lower fruit and vegetables intakes are associated with greater acculturation among Mexicans living in Washington state. Journal of the American Dietetic Association, 104 (1), 51-57. Niermeyer, S., Wells, C., Williford, D., Honigman, B., Moore, L., Asmus, 1., Shupe, A., Jacobellis, J., Lezotte, D., Egbert, M. (2006). Analysis of low birth weight, high altitude and smoking using geographic information systems. Platform paper presented at the Pediatric Academic Societies' Meeting: May 2006, San Francisco, CA. Nunnaly, J. (1978). Psychometric Theory. New York, NY:McGraw Hill. O'Campo, P., Xue, X., Wang, M., & Caughy, M. (1997). Neighborhood risk factors for low birthweight in Baltimore: a multilevel analysis. American Journal of Public Health, 87(7), 1113-1118. Oken, N., & Gillman, G. (2003). Fetal origins of obesity. Obesity Research, 11 (4), 496-506. Olsen, S.F., Halldorsson, T.l., Willett, W.C., Knudsen, V.K., Gillman, M.W., Mikkelsen, T .. Olsen, J., & the NUTRIX Consortium (2007). Milk consumption during pregnancy is associated with increased infant size at birth: Prospective cohort study. American Journal of Clinical Nutrition, 86, 11 04-111 0. Ostir, G., Eschbach, K., Markides, K., & Goodwin, J. (2003). Neighbourhood composition and depressive symptoms among older Mexican Americans. Journal of Epidemiology of Community Health, 57, 987-992. Palloni, A., & Arias, E. (2004). Paradox lost: explaining the Hispanic adult mortality advantage. Demography, 41(3), 385-415. Palloni, A., & Morenoff, J. (2001). Interpreting the paradoxical in the Hispanic paradox. Annals of the New York Academy of Sciences, 954, 140-174. 182

PAGE 200

Pearl, M., Braveman, P., & Abrams, B. (2001). The relationship of neighborhood socioeconomic characterists to birthweight among 5 ethnic groups in California. American Journal of Public Health, 91 (11), 1808-1814. Phillips, K. (1999). The Cousins' Wars: Religion, Politics, & the Triumph of Anglo-America. New York, NY: Basic Books. Pickett, K., Collins, J., Masi, C., & RG, W. (2005). The effects of racial density and income incongruity on pregnancy outcomes. Social Science & Medicine, 60, 2229-2238. Pickett, K., & Pearl, M. (2001). Multilevel analyses of neighbourhood socioeconomic context and health outcomes: a critical review. Journal of Epidemiology and Community Health, 55, 111-122. Portes, A., & Bach, A. (1985). Latin Journey: Cuban and Mexican Immigrants in the United States. Berkeley, CA: University of California Press. Gonzalez-Quintero, V.H., Tolaymat, L., Luke, B., Gonzalez-Garcia, A., Duthely, L., O'Sullivan, M.J., & Martin, D. (2007). Outcome of pregnancies among Hispanics. Journal of Reproductive Medicine, 51 (1), 10-14. Raudenbush, S., & Bryk, A. (2002). Hierarchical Linear Models: Applications and Data Analysis Methods (2hd ed.). Thousand Oaks, CA: Sage Publications. Reagan, P., & Salsberry, P. (2005). Race and ethnic differences in determinants of preterm birth in the USA: broadening the social context. Social Science & Medicine, 60, 1952-1957. Reichman, N.E. & Kenney, G.M. (1998). Prenatal care, birth outcomes and newborn hospitalization costs: patterns among Hispanics in New Jersey. Family Planning Perspectives, 30(4), 182-187 & 200. Ricketts, S., Murray, E., & Schwalberg, A. (2005). Reducing low birthweight by resolving risks: results from Colorado's Prenatal Plus Program. American Journal of Public Health, 95 (11), 1952-1957. Robert, S. (1999). Socioeconomic position and health: the independent contribution of community socioeconomic context. Annual Review of Sociology, 25,489-516. Rose, G. (1992). The strategy of preventive medicine. New York, NY: Oxford University Press. Rose, G., & Marmot, M. (1981 ). Social class and coronary heart disease. British Heart Journal, 45, 13-19. Roseberry, W. (1988). Political economy. Annual Review of Anthropology, 17, 161-185. 183

PAGE 201

Rosenberg, T., Raggio, T., & Chiasson, M. (2005). A further examination of the "epidemiologic paradox": birth outcomes among Latinas. Journal of the National Medical Association, 97 (4), 550-556. Rumbaut, R., & Weeks, J. (1996). Unraveling a public health enigma: why do immigrants experience superior perinatal health outcomes? Research in the Sociology of Health Care, 13 (B), 337-391. SAS. (2003). Version 9.1.3. Cary, NC: SAS Institute Inc. Schensul, S., Schensul J.J., & LeCompte, M.D. (1999). Essential Ethnographic Methods: Observations, Interviews, and Questionnaires. Walnut Creek, CA: Altamira Press. Schoenborn, C.A. (2004). Marital status and health: United States, 1999-2002. Advance Data from Vital & Health Statistics 351, 1-33. Retrieved September 24, 2008, from the website of NCHS: www.cdc.gov/nchs/data/ad/ad351.pdf. Scribner, R. (1996). Editorial: paradox as paradigm -the health outcomes of Mexican Americans. American Journal of Public Health, 86 (3), 303-305. Scribner, R., & Dwyer, J. (1989). Acculturation and low birthweight among Latinos in the Hispanic HANES. American Journal of Public Health, 79, 1263--1267. Sellstrom, E., & Bremberg, S. (2006). The significance of neighbourhood context to child and adolescent health and well-being: a systematic review of multilevel studies. Scandinavian Journal of Public Health, 34, 544-554. Singh, G., & Yu, S. (1996). Adverse pregnancy outcomes: differences between USand foreign-born women in major US racial and ethnic groups. American Journal of Public Health, 86 (6), 837-843. Smith, J., & Edmondston, B. (1997). The New Americans. Washington, D.C.: National Academy Press. Sonfield, A. (2007). The impact of anti-immigrant policy on publicly subsidized reproductive health care. Guttmacher Policy Review, 10 (1), 7-11. Stokes, M.E., Davis, C.S., & Kock, G.G. (2000). Categorical Data Analysis Using the SAS System (2nd ed.). Cary, NC: SAS Institute, Inc. Subramanian, S., Jones, K., & C, D. (2003). Multilevel methods for public health research. In I. Kawachi & L. Berkman (Eds.), Neighborhoods and Health (pp. 65-111 ). Oxford, UK: Oxford University Press. Sullivan, L., Dukes, K., & Losina, E. (1999). Tutorial in biostatistics an introduction to hierarchical linear modeling. Statistics in Medicine, 18, 855-888. Surkan, P.J., Hsieh, C-C., Johansson, A.V.L., Dickman, P.W. & Cnattinguis, S. (2004). Reasons for increasing trends in large for gestational age births. Obstetrics & Gynecology, 104, 720-726. 184

PAGE 202

Syme, S., & Berkman, L. (1876). Social class, susceptibility, and sickness. American Journal of Epidemiology, 104, 1-8. Teller, C., & Clyburn, S. (1974). Trends in infant mortality. Texas Business Review, 48, 240246. United Nations (2008). The millennium development goals report. New York, NY: United Nations. Retrieved October 8, 2008, from the United Nations website: http://mdgs.un.org/unsd/mdq/Resources/Static/Products/Proaress2008/MDG Report 2008 En.pdf. U.S. Census. (2000a). PCT-1. 100 percent data; corrected counts. Retrieved March 10, 2007, from the U.S. Census website: http://www.census.gov/Press Release/www/2001/sumfile2.htm I. U.S. Census. (2000b). Census 2000 Summary File 1 (SF1) 100 Percent Data for Colorado. QT-P9, Hispano or Latino by Type, DP-1 and GTC-P6. Retrieved November 13, 2006, from the U.S. Census website: http://factfinder.census.gov. U.S. Census. (2000c). American FactFinder, Colorado, by selected county. Retrieved November 15, 2006, from the U.S. Census website: http://factfinder.census.gov/home/saff/main.html? lang=en. U. S. Census. (2000d). Hispanic population in the U.S. Current population reports. Retrieved March 30, 2007, from the U.S. Census website: http://www .census.gov/prod/2001 pubs/p20-535.pdf. U.S. Census. (2000e). Population, Housing Units, Area, and Density:2000. SF1, GCT-PH1. Retrieved Nobember 11, 2008, from the U.S. Census website: http://factfinder.census.gov/. U.S. Census. (2002b). Hispanic population in the U.S. Current population reports. Retrieved March 30, 2007, from the U.S. Census website: http://www .census.gov/prod/2003pubs/p20-545.pdf. U.S. Census. (2002c). Census 2000 gazeteer. Retrieved June 10, 2007, from thhe U.S. Census website: http://www .census.gov/qeo/www/qazetteer/places2k.htm I. U.S. Census. (2007). State and county quickfacts. Retrieved March 10, 2007, from the U.S. Census website: http://quickfacts.census.gov/qfd/states/08000.html. U.S. Department of Health & Human Services. (2003). Healthy people 2010. Vol. II. Retrieved March 10, 2007, from the Department of Health and Human Services website: http://www .healthypeople .gov/Default. htm. U.S. Department of Health and Human Services. (2000). Healthy People 2010 (2"d ed.). Understanding and improving health and objectives for improving health (2 vols.). Washington, DC: U.S. Government Printing Office. 185

PAGE 203

UNICEF. (2004). Low birthweight: country, regional and global estimates. Retrieved April6, 2007, from the World Health Organization website: http://www .childinfo.orqlareas/birthweiqht/LBW WHO UN ICEF%202000. pdf. Weigers, M., & Sherraden, M. (2001 ). A critical examination of acculturation: the impact of health behaviors, social support and economic resources on birth weight among women of Mexican descent. International Migration Review, 35 (3), 803-839. Wingate, M.S. & Alexander, G.R. (2006). The healthy migrant theory: variations in pregnancy outcomes among US-born immigrants. Social Science and Medicine, 62, 491-498. Winkleby, M.A., Jatulis, D.E., Frank, E., & Fortmann, S.P. (1992). Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease. American Journal of Public Health, 82, 816-820. World Health Organization. (2006). Reproductive health indicators-guidelines for their generation, interpretation and analysis for global monitoring. Retrieved March 1 0, 2007, from the World Health Organization website: http://www.who.int/reproductive health/publications/rh indicators/guidelines.pdf. Yang, Q., Greenland, S., & Flanders, W. (2006). Associations of maternal ageand parity related factors with trends in low-birthweight rates: United States, 1989 through 2000. American Journal of Public Health, 96 (5), 856-861. Yeh, M-C., Ickes, S.B., Lowenstein, L.M., Shuval, K., Ammerman, A.S., Farris, R.. & Katz, D.L. (2008). Understanding barriers and facilitators of fruit and vegetable consumption among a diverse multi-ethnic population in the USA. Health Promotion International, 23 (1 ), 42-51. Zambrana, R., Scrimshaw, S., Collins, N., & Dunkel-Schetter, C. (1997). Prenatal health behaviors and psychosocial risk factors in pregnant women of Mexican origin: The role of acculturation. American Journal of Public Health, 87 (6), 1022-1026. Zapata, B., Rebolledo, A., Atalah, E., Newman, B., & King, M. (1992). The influence of social and political violence on the risk of pregnancy complications. American Journal of Public Health, 82 (5), 685-690. 186