Citation
Adherence to recommended follow-up on screening mammography in a multiethnic sample of Colorado women

Material Information

Title:
Adherence to recommended follow-up on screening mammography in a multiethnic sample of Colorado women
Creator:
Strzelczyk, Jadwiga
Publication Date:
Language:
English
Physical Description:
219 leaves : ; 28 cm

Subjects

Subjects / Keywords:
Breast -- Cancer -- Treatment -- Colorado ( lcsh )
Breast -- Radiography -- Colorado ( lcsh )
Patient compliance -- Colorado ( lcsh )
Breast -- Cancer -- Treatment ( fast )
Breast -- Radiography ( fast )
Patient compliance ( fast )
Colorado ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 202-219).
General Note:
Department of Health and Behavioral Sciences
Statement of Responsibility:
by Jadwiga (Jodi) Strzelczyk.

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:
49664942 ( OCLC )
ocm49664942
Classification:
LD1190.L566 2001d .S77 ( lcc )

Full Text
ADHERENCE TO RECOMMENDED FOLLOW-UP
ON SCREENING MAMMOGRAPHY
IN A MULTIETHNIC SAMPLE OF COLORADO WOMEN
by
Jadwiga (Jodi) Strzelczyk
M.Sc., University of Lodz, Poland, 1968
M.S., State University of New York at Stony Brook, 1982
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Health and Behavioral Sciences
2001


2001 Jadwiga (Jodi) Strzelczyk
All rights reserved.


I
This thesis for the Doctor of Philosophy
degree by
Jadwiga (Jodi) Strzelczyk
has been approved
by
John Lewin
I
I
hhu. /S cWOf
Date
I


Strzelczyk, Jadwiga (Ph.D., Health and Behavioral Sciences)
Adherence to Recommended Follow-up on Screening Mammography
in a Multiethnic Sample of Colorado Women
Thesis directed by Professor David Tracer
ABSTRACT
The advantage of detecting breast cancer in its early stages may be lost in
women who do not regularly participate in screening as well as those who do not
adhere in a timely manner to recommended follow-up after abnormal
mammograms. While numerous studies have investigated factors that influence
mammography utilization, little is known about adherence to follow-ups.
This study employed a specifically developed Adherence to Follow-up
Model and methodological triangulation. The primary dependent variables were
rates of adherence and timeliness of follow-up among a multiethnic sample of
13,857 Colorado women. Quantitative methods were used to examine socio-
demographic and clinical factors as predictors of adherence. Qualitative methods
were also used to explore possible psychosocial, cognitive and cultural predictors
of adherence behavior in a smaller multiethnic sub-sample.
While ethnic minorities overall were less likely than white women to adhere
to recommended follow-up in a timely manner, substantial differences in rates and
timeliness of follow-up were observed among and within ethnic groups even after
the influence of age, education, health insurance and family history of breast
cancer were controlled. Results of this study also suggest that clinical factors,
mammography finding and invasiveness of the recommended follow-up play
important roles in adherence. Among all the psychosocial, cognitive and cultural
factors cited by women regarding health and its maintenance, mammography and
its findings, and the importance and urgency of following up, fear of breast cancer
appeared to be the universal emotion shared by all subjects. This fear manifested
IV


itself, however, in varied responses that ranged from expecting immediate attention
by health professionals to non-adherence. No specific ethnic patterns were evident
from the women's narratives about the importance and urgency of follow-up.
This study lends important insights into the factors that influence adherence
among women of different ethnic groups. These insights will help to advance efforts
to provide patient-centered quality care and more effective breast cancer control,
as well as point to productive areas for future research.
This abstract accurately represents the content of the candidates thesis.
I recommend its publication.
Signed
v


DEDICATION
I dedicate this work to my Mother.


ACKNOWLEDGEMENT
Many thanks to faculty of the Health and Behavioral Sciences and
Anthropology Departments for triggering my interest in research aiming at a (setter
understanding of the social and behavioral underpinnings of health and for
guidance in the transition from "hard sciences" paradigm. Not only are you
wonderful teachers and scholars, you seem to have an inexhaustible supply of
energy and ability to bolster independent thinking combined with a sincere caring
attitude toward students.
Thanks to the dissertation committee for their expertise and support in the
preparation of this work. Very special thanks to Dr. Tracer, my advisor, for his first-
rate leadership and rigorous emphasis on the substance behind conclusions of
research findings. Thanks to Dr. OBrien-Gonzales for her clarification of my initial
skepticism regarding simultaneous use of quantitative and qualitative research
methods. Many thanks to Dr. Dignan for trusting me with the precious data set and
believing in my ability to make sense of the complexities of information it contains.
Thanks to Dr. Bryant for her diligence, reviews of the multiple versions of the
manuscript and for her constructive criticism. Finally, thanks to Dr. Lewin for clinical
information vital to this work and for sharing mammography trade secrets.
I am indebted to my husband Marian (Mark) for his gentle prodding and
assuming many additional tasks in the past four years; to our sons, Rob and Martin,
and their wives who offered unconditional love and understanding that, due to time
constraints, I could not be in New York as frequently as I would have wanted to,
and to my little granddaughters who provided so much needed joy and laughter.
Among all friends, who helped me persevere, two ladies deserve special thanks,
Krystyna Madajewicz and Judy Edwards-Prasad.
Thank you all!


CONTENTS
Figures .......................................................... xii
Tables ............................................................ xiii
Chapter
1. Introduction .................................................. 1
1.1 Study Objectives and Specific Aims ............................ 1
1.1.1 Preliminary Study ............................................. 1
1.1.2 Hypotheses and Research Questions ............................. 2
1.1.3 Research Aims ................................................. 3
1.2 Background..................................................... 6
1.2.1 Racial/ethnic Differences in Cancer Incidence and Mortality:
Emphasis on Breast Cancer .................................... 6
1.2.2 Approaches to Breast Cancer Control:
Factors that Influence Efficacy of Screening ............... 13
1.2.3 Issues Related to Follow-up on Abnormal Screening Results .... 19
1.3 Significance ................................................. 22
2. Literature Review ............................................ 24
2.1 Adequacy of Follow-up on Mammography ......................... 24
2.1.1 Appropriateness of Medical Services .......................... 24
2.1.2 Untimely Adherence to Follow-up Recommendations .............. 26
2.2 Models Suitable in Studies of Adherence Behavior ............. 29
2.2.1 The Health Belief Model (HBM) ................................ 30
2.2.2 Health and Health Behavior Models ............................ 36
2.2.3 The Multicultural Understanding Model ........................ 45


3. Research Design and Methodology ........................... 48
3.1 Research Design ........................................... 48
3.1.1 The Adherence to Follow-up Model .......................... 48
3.1.2 Quantitative Study Design ................................. 51
3.1.3 Qualitative Study Design .................................. 55
3.2 Research Methodology ...................................... 57
3.2.1 Quantitative Methods ...................................... 57
3.2.2 Qualitative Methods ....................................... 67
3.2.3 Reconciliation of Methods ................................. 71
4. Quantitative Data Analysis and Results .................... 73
4.1 Participation in Screening Mammography:
Socio-demographic Characteristics of the Sample ......... 74
4.1.1 Dynamic Character of Participation by Various Groups ...... 75
4.1.2 Interactions of Race/ethnicity with Other Characteristics 79
4.1.3 Section Summary and Discussion ............................ 82
4.2 Follow-up Recommendations and Overall Adherence:
Socio-Demographic Characteristics.......................... 83
4.2.1 Differences in Follow-up Recommendations .................. 83
4.2.2 Differences in the Overall Adherence to Follow-up ......... 85
4.2.3 Interactions of Race/ethnicity with Other
Socio-Demographic Characteristics ......................... 86
4.2.4 Screening versus Adherence: Discussion of Results ......... 89
4.3 Mammography Results and Types of Follow-up: Adherence
of the Cohort with Possibly Abnormal Mammograms ......... 94
4.3.1 Possibly Abnormal Mammography Results:
Differences Among Groups .................................. 94
4.3.2 Differential Adherence Rates by Invasiveness of Follow-up 98


4.3.3 Interactions of Race/ethnicity with Other
Socio-demographic Characteristics ......................... 102
4.3.4 Adherence when Mammography Results Indicate Possibly
Abnormal Mammogram versus Overall Adherence:
Discussion of Results...................................... 106
4.3.5 Section Summary and Discussion ............................ 109
4.4 Adherence of the Cohort with Probably Benign
Result Category ........................................... 110
4.4.1 Adherence Rates and Timeliness ............................ 110
4.4.2 Interactions of Race/ethnicity with Other
Socio-Demographic Characteristics ......................... 114
4.4.3 Section Summary and Discussion ............................ 116
4.5 Adherence to Follow-up After Abnormal Mammograms .......... 118
4.5.1 Comparison of Adherence Rates and
Timeliness of Follow-up ................................... 118
4.5.2 Interactions of Race/ethnicity with Other
Socio-Demographic Characteristics ......................... 121
4.5.3 Adherence After Probably Benign Findings versus
Follow-up After Abnormal Mammograms ....................... 124
4.5.4 Section Summary and Discussion ............................ 127
4.6 Chapter Summary and Discussion ............................ 131
4.6.1 Summary ................................................... 132
4.6.2 Limitations and Conclusion of the Quantitative Study ...... 136
5. Qualitative Data and Results .............................. 137
5.1 In Their Own Words: Life Health Events Stories .......... 138
5.1.1 Stories of Younger (<50 Years of Age) Women ............... 138
5.1.2 Stories of Older (50 Years of Age and Over) Women ......... 151
x


5.2 Chapter Summary ............................................ 163
5.2.1 Characteristics of Participants ............................ 163
5.2.2 Meanings of Health and Well-Being .......................... 164
5.2.3 Mammography Screening Attitudes ............................ 164
5.2.4 Follow-up on an Abnormal Mammogram ......................... 164
6. Conclusions ................................................ 167
6.1 The Adherence to Follow-up Model ........................... 168
6.1.1 Predisposing Factors ....................................... 168
6.1.2 The Critical Event ......................................... 169
6.1.3 Modifying Factors .......................................... 171
6.1.4 Likelihood of Action ....................................... 173
6.1.5 Usefulness of the Framework ................................ 175
6.2 Summary and Implications ................................... 176
6.3 Limitations ................................................ 178
6.4 Future Work ................................................ 179
6.5 Conclusions ................................................ 180
Appendix ........................................................ 181
Glossary of Terms ............................................... 198
Bibliography
202


FIGURES
Figure
2.1 The HBM as predictor of preventive health behavior ........... 32
2.2 The initial version of Andersens Behavioral Model of 1960s... 39
2.3 The Preventive Health Model .................................. 43
2.4 Multicultural Understanding Model ............................ 47
3.1 The Adherence to Follow-up Model............................... 50


TABLES
Table
1.1 Age-adjusted BC incidence rates, BC mortality rates and average
annual percentage changes (APC), by race/ethnicity for 1990-1997... 10
1.2 Age-adjusted (<50 years) BC incidence and mortality rates in
younger Black and White women in 1990 & 1997) .................. 11
1.3 Age-specific BC incidence and mortality rates, 1993-1997......... 12
1.4 Breast cancer stage distribution in 1997.......................... 13
4.1 Changes in the composition of screening cohort in the opening
and closing years of the study period........................... 76
4.2 Comparison of proportions (%) of screening participants possessing
characteristics of interest within racial/ethnic groups with the
proportion of White women with corresponding characteristics.. 80
4.3 Socio-demographic characteristics of screening participants, of
women who received follow-up (F/U) recommendations, and of
women who adhered to recommendations within 12 months......... 84
4.4 Comparison of proportions (%) of adherent women possessing
characteristics of interest within racial/ethnic groups with the
proportions of White women with corresponding characteristics. 88
4.5 Comparison of proportions (%) of screened (S) and adherent (A)
women possessing characteristics of interest within racial/ethnic
groups........................................................ 90
4.6 Socio-demographic characteristics of women whose
mammographic results indicated possible abnormality; proportions
of women in each result category................................ 96
xiii


4.7 Adherence rates (%) after possibly abnormal mammogram by
type of follow-up procedure among socio-demographic groups....... 99
4.8 Comparison of proportions (%) of adherent women whose
mammograms were possibly abnormal, possessing characteristics
of interest, within racial/ethnic groups with the proportions for White
women with corresponding characteristics; all types of F/U,
non-invasive (Nl), and invasive (I) F/U)......................... 103
4.9 Comparison of proportions (%) of adherent women overall (A), and
proportions of women who adhered following possibly abnormal
mammogram (PA) by characteristics of interest within racial/ethnic
groups with the proportions of White women...................... 107
4.10 Adherence rates, means with standard deviation (SD), and median
time intervals, with Interquartile Ranges (IQR) from probably benign
screening mammogram to recommended F/U diagnostic procedure
by socio-demographic characteristics of the sample............... 111
4.11 Comparison of proportions (%) of women within racial/ethnic groups
possessing characteristics of interest with the proportions for White
women with corresponding characteristics who were adherent after
possibly benign mammography result.............................. 115
4.12 Adherence rates, means with SD, and median time intervals with
IQR from index abnormal mammogram to recommended F/U
diagnostic procedure by socio-demographic characteristics of the
sample.......................................................... 120
4.13 Comparison of proportions (%) of adherent women whose
mammograms were abnormal possessing characteristics of interest
within racial/ethnic groups with the proportions of White women. 122
4.14 Comparison of proportions (%) of adherent women when
mammography result was probably benign (PB) with proportions
of women who were adherent following abnormal mammogram (AB)
by characteristics of interest within racial/ethnic groups...... 125
xiv


4.15 Adherence rates, means and median time intervals to follow-up,
hazard ratios (HR), and probability values, for women with
abnormal screening mammograms................................ 128
4.16 Hazard ratios, HRs, and probability values (p) in multivariate
adherence models for racial/ethnic groups, compared with
White women (HR=1) with abnormal mammograms........................ 130


1. Introduction
Numerous studies have investigated factors that facilitate participation of
women in screening mammography but few have examined adherence to
recommended follow-up. Additionally, a review of the literature indicates that
studies that did consider some aspect of adherence to follow-up generally limited
their scope to two racial groups; Black versus White comparisons. To date, no
study has examined differences in and predictors of the rates and timeliness of
adherence to recommended follow-up in a large multiethnic sample of American
women. This study attempts to close that gap.
1.1 Study Objectives and Specific Aims
The data set for this study consisted of records of 167,232 screening
participants who self-reported their race/ethnicity as: White (82.79%), Black
(3.34%), Hispanic (11.04%), Asian (1.56%), Native American (0.58%) or Other
(0.69%). In preparation for this dissertation, a preliminary investigation was
conducted.
1.1.1 Preliminary Study
The aim of the preliminary study was to examine socio-demographic
characteristics of women as possible predictors of their participation in screening
mammography over an 8-year period, 1990-1997. The quantitative analysis of data
considered race/ethnicity, age, level of education, health insurance status, and
family history of breast cancer among first-degree relatives (FHBC) as unique
factors that may influence screening behavior. Additionally, this study examined the
composition of screening participants in 1990 and 1997, and compared the
proportions of women with corresponding characteristics in the opening and the
1


ending years of the study period.
In an effort to learn more about women in this sample and factors that may
have influenced their participation in screening, further examination of their data
continued as follows. The aim of this phase of the preliminary study was to
investigate qualitative interactions of race/ethnicity with each of the considered
socio-demographic characteristics as possible predictors of adherence to follow-up.
Stratifying data on screening participants within each race/ethnicity facilitated an
assessment of directionality of proportion of women in dichotomous categories of
the remaining socio-demographic characteristics, i.e., it allowed a determination
whether the proportion of women with less education was lower than the proportion
of women with more education within each ethnic group. Results of this phase of
the preliminary study were utilized in the comparisons of proportions of women with
characteristics of interest who participated in screening with the proportions of
women who were adherent to follow-up recommendations.
1.1.2 Hypotheses and Research Questions
Based on the limited body of literature on the subject of adherence to follow-
up on mammography, differences in the proportions of diverse women who
adhered and in the timeliness of that adherence were anticipated but their ranges
and significance were unknown. The findings of the preliminary study suggested
that socio-demographic characteristics of women serve as good predictors of
screening behavior. Its results indicated that qualitative interactions of race/ethnicity
with other socio-demographic factors are also good determinants of screening
behavior.
It seemed plausible to hypothesize that socio-demographic characteristics
of screening participants might also predict adherence behavior. The basic
research questions of the quantitative study were: Are there differences in
adherence among racial/ethnic groups that are larger than would be expected by
chance alone? Are there differences in the proportions of women who receive
follow-up recommendations? Is the result of mammography, severity of the
2


finding in particular, related to the rate of adherence among various groups; and if
so, how significant are these differences? Does the invasiveness of a follow-up
procedure play a role in adherence behavior? Are there differences in the
timeliness of adherence among various groups, when the mammography finding
suggests a malignancy; and if so, are these differences significant?
While it was expected that some explanations of differences in adherence
behaviors would be derived from the quantitative study, there was little doubt that
statistical findings would not be sufficient to elucidate the reasons that these
differences exist. A qualitative study was therefore launched to complement the
quantitative study.
The objectives of the qualitative study were to explore the meanings of
health, the value of health maintenance and disease prevention, the understanding
of the importance of breast cancer screening in general and mammography in
particular, the meaning of an abnormal mammogram, and the understanding of
recommendation to follow-up in a small multiethnic sample of women. The
qualitative inquiry was based on the hypothesis that such meanings and
understandings may be among psychosocial and/or cultural factors that facilitate or
deter adherence.
1.1.3 Research Aims
The specific aims of this research were:
Aim 1: To examine differences among ethnically diverse groups of women in rates
of overall adherence to follow-up recommendations, by pursuing the following sub-
aims:
determine whether there are differences in follow-up
recommendations among various groups of women,
compare the overall adherence levels of women possessing socio-
demographic characteristics of interest with adherence levels of
appropriate reference groups,
3


examine qualitative interactions of race/ethnicity of adherent
(overall) women and compare proportions of women possessing
characteristics of interest with the proportions of appropriate
reference groups, i.e., compare the proportion of adherent Hispanic
women who reported FHBC with the proportion of White women
with FHBC,
compare proportions of adherent (overall) women possessing
characteristics of interest with corresponding proportions of
screening participants.
Aim 2: To examine differences in adherence rates among diverse groups of
women with possibly abnormal screening mammograms by pursuing the following
sub-aims:
determine whether there are differences among various ethnic
groups in the proportions of those having possibly abnormal
mammograms,
examine differential adherence rates among diverse women whose
mammograms were possibly abnormal controlling for invasiveness
of the follow-up procedure,
examine qualitative interactions of race/ethnicity of adherent women
whose mammograms were possibly abnormal with their other
characteristics; compare proportions of women possessing
characteristics of interest with the proportions of appropriate
reference groups,
compare proportions of adherent women whose mammograms were
possibly abnormal and who possessed characteristics of interest
with corresponding proportions of all adherent women who were
recommended to follow-up (overall adherence).
Aim 3: To examine adherence rates and timeliness of follow-up for the cohort with
a probably benign mammography results by pursuing the following sub-aims:
4


examine differential adherence rates among various groups of
women in the cohort with probably benign findings,
examine differences in the timeliness of adherence among the
various groups of women in this cohort,
examine qualitative interactions of race/ethnicity of adherent women
in this cohort with their other characteristics as possible predictors of
adherence.
Aim 4: To determine the extent of differences in the rates and timeliness of
adherence among diverse women whose screening mammograms were abnormal,
by pursuing the following aims:
examine differential adherence rates among diverse women whose
mammograms were abnormal,
examine differences in the timeliness of adherence among diverse
women in this cohort,
examine qualitative interactions of race/ethnicity of adherent women
in this cohort with their other characteristics as possible predictors of
adherence,
compare proportions of adherent women possessing characteristics
of interest, whose mammograms were abnormal with corresponding
proportions of women whose mammograms were probably benign,
to examine the influence of severity of mammography result on
adherence rates.
These quantitative analyses considered clinical variables, in addition to
socio-demographic factors, as potential predictors of adherence, both unique and in
qualitative interactions with womens characteristics. The clinical variables included:
American College of Radiology screening mammography assessment categories
(BI-RADS system is described in the Glossary of Terms), and the invasiveness of
recommended follow-up procedures.
5


Aim 5: To identify the range of psychosocial, cognitive and cultural factors that, in
women's own words, may have facilitated or prevented adherence to
recommended follow-up.
Semi-structured, focused interviews were conducted with a small multiethnic
sample of women representing various socio-demographic groups. In accordance
with stated objectives of this study, the following meanings and understandings
were explored in an effort to link them to adherence behavior. This was
accomplished by asking open-ended questions and allowing women to talk freely
about:
the meaning of health and well-being,
the value of health maintenance and disease prevention,
the understanding of the importance of breast cancer screening in
general and mammography in particular,
the meaning of an abnormal mammogram,
the understanding of recommendation to follow-up, and
about factors that, in womens own words, influenced their decision
to follow-up.
1.2 Background
This section contains a review and discussion of what is known about
racial/ethnic differences in cancer incidence and mortality with emphasis on breast
cancer (BC), approaches to BC control including consideration of factors that
influence adequacy and efficacy of screening, and issues related to follow-up on
abnormal screening mammograms.
1.2.1 Racial/ethnic Differences in Cancer Incidence
and Mortality: Emphasis on BC
Epidemiologic data show that certain racial and ethnic groups exhibit
significant differences in cancer risk and cancer mortality (Kinzler & Vogelstein
1996). As incidence rates alone give an incomplete picture of the burden this
6


disease imposes on human populations, various other measures such as
prevalence, mortality, and survival rates are commonly reported in the literature.
1.2.1.1 Cancer Statistics: Racial and Ethnic Variations
A review of recent cancer statistics (Ries et al. 1997) indicates that, relative
to Whites, Black Americans experience cancer incidence rates that are
approximately three times higher for esophageal cancer; twice as high for multiple
myeloma, liver, cervical, and stomach cancers; and about fifty percent higher for
cancers of the oral cavity and pharynx, larynx, lung, prostate and pancreas. The
incidence of chronic lymphatic leukemia and premenopausal BC is also higher in
Black Americans. In contrast, White Americans have higher incidence rates for
melanoma, leukemia, lymphoma, and cancers of thyroid, endometrium, bladder (in
males), testis, and brain, as well as postmenopausal BC.
Other racial/ethnic groups also differ from Whites in their rates of various
cancers. Hispanics, for example, have generally lower overall cancer rates than
White or Black Americans, but rates differ substantially among Hispanics of
different races and ancestry (Perera 1996). A comprehensive 9-year assessment
of variations in cancer incidence among South Florida Hispanic women indicates
that race has the same relative effect as that observed for non-Hispanics; the
observed White versus Black cancer rate ratio (RR) was 1.27 for Hispanics, and
1.26 for non-Hispanics (Trapido et al. 1994). Age-standardized breast cancer
incidence rates (per 100,000) reported in this study were as follows: 100.9 for
White Non-Hispanic versus 72.7 for White Hispanic women (RR=0.72, 95% Cl:
0.7, 0.8), and 70.7 for Black Hispanic versus 59.5 for Black Non-Hispanic women
(RR=0.84, 95% Cl: 0.7, 1.1). These variations in cancer risk may reflect differences
in environmental exposure, socio-economic and socio-demographic factors, and
medical care access and utilization, as well as innate biological factors (Perera
1996). The recent rise in traditionally low BC rates experienced by the descendants
of Asian immigrants to the U.S. provides strong evidence that environmental
factors affect cancer patterns (Hedeen et al. 1999, Ursin et al.1999). After several
7


!
generations, BC rates in Asian American women are closer to those prevailing in
the U.S. White population than to those in Asians in their countries of origin (Ries
et al. 1997, Hedeen et al. 1999). Among the biological factors that might explain
higher cancer risks in certain racial/ethnic groups, variations in the prevalence of
genetic traits affecting carcinogen metabolism and DNA repair appear to be
relevant (Perera 1997). However, it is clear that no one ethnic or racial group is
exempt from genetic susceptibility to the effects of carcinogens, nor is any group at
universally higher risk than another.
Cancer statistics studied over time for various racial/ethnic groups provide
a wealth of information on general trends. However, the interpretation of these
statistics is not straightforward. The investigation of trends in survival rates for all
cancers combined must take into account the constantly changing mix of types of
cancer and different fatality rates related to various types of cancer. Technological
advances in diagnostic and therapeutic procedures need to be considered as well.
Early detection and changes in diagnostic criteria are likely to increase incidence
rates for specific types of cancer. A shift toward earlier detection of disease,
generally in a more treatable stage, termed stage migration effect (Saffrin et al.
1995, Zahl 1997, Noldus et al. 2000), along with improvements in cancer treatment
may ultimately lead to higher survival rates (Roukos et al. 1996, Vicini et al. 1997).
The role of these factors in the interpretation of observed trends may be totally lost
if one considers only numerical changes in the overall incidence and mortality
rates.
1.2.1.2 BC Statistics and Trends:
Racial and Ethnic Variation
This research examines behaviors that may influence BC survival:
screening and adherence to recommended follow-up of six racial/ethnic groups of
women who participated in the Colorado Mammography Project (CMAP) during the
1990-1997 period. When reviewing cancer statistics in this Chapter, emphasis is
placed on the racial/ethnic trends during the corresponding time period.
8


"The Annual Report to the Nation on the Status of Cancer, 1973 -1997, with
a Special Section on Colorectal Cancer" (Ries et al. 2000), which was jointly issued
by the American Cancer Society (ACS), the National Cancer Institute (NCI), the
North American Association of Central Cancer Registries (NAACCR), and the
Centers for Disease Control and Prevention (CDC), including the National Center
for Health Statistics (NCHS), is the main source of statistics presented here.
The recent changes in cancer rates, observed for the period of this study,
are encouraging. The joint report shows that rates of new cancer cases and deaths
in the U.S., for all cancers combined as well as for most of the top ten cancer sites,
declined between 1990 and 1997. Incidence rates for all cancers combined
decreased on average 0.8% per year during this 8-year period. This trend,
although not statistically significant, may represent a change in the pattern of
steadily increasing incidence rates during 1973 to 1992 period. The report also
shows that the overall cancer mortality rates declined 0.8% for the period from
1990-1997. The report analyzed data for White, Black, Asian/Pacific Islander
(A/PI), American indian/Alaska Native (AI/AN), and Hispanic populations. Again,
large differences in cancer incidence rates by race and ethnicity were noted with
highest rates for Black Americans in the top-four cancer site categories except for
BC. Four sites (lung, prostate, breast, and colon/rectum) accounted for over fifty
percent of all new cancers; they were also the leading causes of cancer deaths for
every racial/ethnic group.
BC is the most common cancer among women in the Western world.
Among American women, BC is currently the number one cancer site regardless of
race/ethnicity. BC incidence rates rose steadily since 1973, leveled off in the late
1980s, and showed little change in the 1990s (Ries et al. 2000). This disease
occurs in both women and men, but is quite rare among men. As less than 1% of
all breast carcinoma cases occur in males (Landis et al. 1998), there are too few
cases to explore racial/ethnic differences. This study, therefore, considers BC in
women only.
During 1990 -1997, the age-adjusted incidence of BC in Black American
women was lower than that observed for White women, but the highest among
9


minority women, as shown in Table 1.1. A large body of literature has attempted
to explain these differences. Some of the possible factors considered in studies
included: age at onset of menarche and menopause, differential use of hormones,
specifically the long- term use of oral contraceptives or estrogen replacement
therapy (SEER 1995, ICRF 1996); frequencies of aborted pregnancies (Daling et
al.1994, Krieger et al.1997); socio-economic and situational factors such as
poverty (Krieger 1993, Breen et al.1996), residing in highly polluted environments
(Rundle et al. 2000) as well as behavioral factors such as smoking (SEER 1995).
While important links were found, the authors generally exercised caution
regarding definite conclusions and recognized the need for further studies.
Table 1.1 Age-adjusted BC incidence rates, BC mortality rates and average annual
percentage changes (APC), by race/ethnicity for 1990 -1997 (from Ries et al. 2000).
Race/ethnicity Incidence rate APC Mortality rate APC
Overall 109.7 +0.4 25.6 -2.1
White 114.0 +0.3 25.3 -2.4
Black 1002 +0.7* 31.4 -0.2
Hispanics* 68.9 -1.0 15.1 -1.2*
AI/AN 45.4 +4.8 12.1 +1.9
A/PI 74.6 +2.2* 11.2 -1.4
Rates are per 100,000 and age-adjusted to the 1970 U.S. standard population.
* The APC is significantly different from zero (p<0.05)
* Hispanics are not mutually exclusive from other racial/ethnic groups
As shown in Table 1.1, the overall BC incidence rates slightly increased
during the 1990-1997 period, 0.4% annually on average, with the largest increase
noted for American Indian/Alaskan Natives (4.8%). The increases for Asian/Pacific
Islander (2.2%) and Black American (0.7%) women reached statistical significance.
A decrease in BC incidence rates is observed for Hispanic (1.0%) women.
Racial/ethnic patterns of mortality due to BC for the 1990-1997 (Table 1.1)
differ from those observed for BC incidence with the highest death rate occurring in
10


Black women. While significant improvement in BC mortality occurred during this
period of 1990s, evidenced in the overall annual percentage rate (APC) decrease
by 2.1%, differences among racial/ethnic groups persisted. Disparity is seen for all
minority groups when compared with White women. The 2.4 percent average APC
decrease during the 8-year period (over 4% in the 1995 -1997 period) for White
women was not matched for minority women. Along with the largest increase in BC
incidence rates, American Indian/Alaskan Native women experienced an increase
in death rates (1.9%) and mortality rates in Black women were roughly stable
(0.2% decrease). During this period, BC was not the number one cause of cancer
deaths in American women, it ranked second after lung cancer (Ries et al. 2000).
An exception was observed for Hispanic women; this is in spite of a significant
decrease in BC mortality rate (1.2%) in this group.
The difference in age-adjusted BC mortality rates between Black and White
(31.4 versus 25.3 per 100,000 for 1990-1997) does not give a complete picture;
age appears to be an important confounding factor to consider in these
comparisons. While age-adjusted BC incidence rates in younger (<50 years of
age) Black women are only slightly higher than those in younger White women,
large excesses in BC mortality rates exist in this age group (Table 1.2).
Table 1.2 Age-adjusted (< 50 years) BC incidence and mortality rates in younger Black
and White women in 1990 & 1997 (from Ries et al. 2000)
White (<50) Black (<50) Overall (<50)
Incidence rates Year of diagnosis 1990 33.6 35.6 33.5
1997 31.9 32.0 32.0
Mortality rates Year of death 1990 5.9 8.8 6.1
1997 4.6 8.2 5.0
Rates are per 100,000 and are adjusted to the 1970 U.S. standard population.
11


A closer look at age-specific rates indicates that BC occurs more frequently
in very young (20 39 years of age) Black women, as compared with their White
counterparts. Corresponding age-specific BC mortality rates in very young Black
women are at least twice as high as those observed for White women, in every age
group in this category (Table 1.3). One possible explanation of some of the excess
mortality could be the far more aggressive nature of premenopausal, as compared
with postmenopausal breast cancers (Berg 1995, Hiatt & Pasick 1996), however,
this should be true for White women as well.
Table 1.3 Age-specific BC incidence and mortality rates, 1993 -1997 (from Ries et al.
2000)
White Black Overall
Incidence rates Age at diagnosis
20-24 1.1 2.9 1.4
25-29 7.8 11.3 8.1
30-34 24.0 32.0 24.8
35-39 57.4 64.4 58.4
Mortality rates Age at death
20-24 0.1 0.3 0.1
25-29 0.9 2.2 1.1
30-34 3.5 7.8 4.0
35-39 9.3 17.5 10.2
Rates are per 100,000 and are adjusted to the 1970 U.S. standard population.
This excess mortality among younger Black women could also be linked to
the fact that, compared to White women, a large percentage of their breast cancers
are diagnosed at a later, generally less treatable stage (Miller et al. 1996). An
example of BC stage distribution is depicted in the Table 1.4. Being Black or
Hispanic has been identified as a risk factor for late stage BC diagnosis as have
lower socio-economic status (SES), younger age, and earlier year of BC diagnosis
(Burgess et al.1998). However, controlling for lower SES levels among Black and
Hispanic women does not appear to change the effect of race/ethnicity on the late
12


Table 1.4 Breast cancer stage distribution in 1997 (from Ries et al. 2001)
BC Stage White Black
BC In-situ 16.4% 18.6%
Stage I 40.6% 27.8%
Stage II 27.8% 31.6%
Stage III 5.2% 7.6%
Stage IV 3.2% 5.8%
Stage Unknown 6.7% 8.6%
stage BC diagnosis (Richardson et al. 1992, Chen et al. 1994, Lannin et al. 1998).
BC staging is described in the Glossary of Terms.
One of every 3 new cancer cases in American women is cancer of the
breast. Based on current trends, 1 woman in 8 (12.5%) will have BC in her lifetime
and 1 in 33 will die from it. It is estimated that there will be more than 40,000
deaths from this disease in the year 2001 in the U.S. (Greenlee et al. 2001).
These estimates beg the question of whether some of these deaths can be
prevented?
1.2.2 Approaches to BC Control:
Factors that Influence Efficacy of Screening
At one time, receiving a diagnosis of BC was considered to be a virtual
death sentence. However, in recent decades, new approaches to treatment such
as improved techniques in surgery, radiation therapy and chemotherapy,
application of combinations of these three modalities by multidisciplinary teams
and sentinel node mapping, have significantly improved the rates of BC cure.
Recent declines in BC death rates for American women suggest that improved BC
management and breast care in general, ranging from prevention to early detection
through screening to treatment, are having beneficial effects.
13


1.2.2.1 Screening Techniques
and Recommended Frequencies
The major screening procedures for the detection of BC are clinical breast
examination (CBE) and mammography. Numerous studies support the value of
mammography and of CBE as effective means of detecting BC, particularly in
women over 50 years of age (Moskowitz 1983, Eddy et al. 1988, Kerlikowske et al.
1995, Fisher 1989, Alexander et al. 1999). Although the breast self-examination
(BSE) has been widely recommended, its benefit has not yet been conclusively
demonstrated (Fisher 1989, US DHHS 1991, Grady 1992).
Consistent evidence that BC mortality is reduced by the use of
mammography has been demonstrated in multi-national controlled trials, both
randomized and non-randomized. (Smart et al. 1995, Bjurstam et al. 1997, Feig
1997). Several important issues, however, remain unresolved and are the subjects
of controversy. These include the age at which mammography should begin in
order to confer a reduction in mortality, the intervals at which the screening should
be performed, the appropriate number of views, and the degree to which adding
CBE offers further improvement (Costanza 1992, Smith 2000).
Recommendations regarding screening are not consistent among various
health organizations. Most organizations recommend that women age 20 and older
should perform BSE every month and that annual CBEs should begin no later than
at 40 years of age (Fisher 1989). Honest differences of opinion exist among major
organizations regarding the optimal frequency of screening mammography. The
American Cancer Society (ACS) and the American College of Radiology (ACR)
both recommend that, for most women, mammography screening should begin not
later than at age 40, continue every one to two years between ages 40 and 49, and
annually thereafter. For women at high risk for BC, screening may begin sooner
(ACS 2000). Other organizations, including the National Cancer Institute (NCI),
recommend that annual mammography screening should begin at 50 years of age
(NCI Breast Cancer Consortium 1990, NIH Consensus Statement 1997).
The controversy over the age at which screening mammography should
commence hinges on the balance between cost and benefit at younger ages.
14


Although the cost of a screening mammogram is small when compared with many
other medical procedures, and although mammography performed at appropriate
intervals is a covered benefit under most insurance and government sponsored
programs such as Medicare and Medicaid, providers are not adequately
reimbursed (Brill 2000, Smith 2000). BC occurs less frequently among women in
their 40s than among women aged 50 and older. Thus, it is argued, few cancers
will be detected in this group of women at a high total cost and strain on provider
facilities. Others counter with the assertion that, if cancer is detected and cured in a
younger patient, we have saved more good years of life compared with the number
of years for someone diagnosed at an older age.
There has also been debate about the relative benefit of screening
mammography in elderly (over 65 years of age) women. While it is true that
cancers do occur more frequently in older women, the argument goes, their
remaining years of life are fewer (Brill 2000). Also, at some point, women are at
less risk of dying from BC than from other diseases. Based on this argument, some
professional organizations such as the European Organization for Research and
Treatment of Cancer (EORTC) and the Breast Cancer International Research
Group (BCIRG) recommend screening at longer (up to 3 year) intervals in elderly
women (Aapro 2000, Brill 2000).
While women pay attention to mammography screening debates, most
misunderstand them, according to recent findings (Woloshin et al. 2000). In one
national survey, few women identified scientific evidence regarding the benefits of
mammography screening as the source of the controversy. In contrast, almost half
of the 503 surveyed women thought that the controversy was about money. About
one third of the comments about costs explicitly asserted that insurers, in order to
save money, did not want to pay for mammography screening.
1.2.2.2 Quality of Mammography
Mammography is one of the most difficult forms of medical radiography,
subject to competing demands of high spatial resolution (ability of image receptors
15


to resolve small structures), high contrast, and low radiation dose. High spatial
resolution is required to see fine details in the breast, including the detection and
characterization of small calcifications. High contrast is needed to improve the
detection of low contrast lesions, especially on the background of glandular tissue
found in the breast. On the other hand, low radiation dose is desired to minimize
the risk of BC that could result from subjecting patients to high-level exposures.
The quality of mammography has improved steadily as this technique has
evolved over the past three decades. Improvements in X-ray units and image
receptors have not only provided higher resolution and contrast but have also
made it possible to reduce the typical average glandular radiation dose from a two-
view mammogram to 1-3 mGy (milligray), less than one-tenth the dose from
mammography in the 1970s (Hendrick 1993). The goal is to achieve the highest
possible sensitivity and specificity (see Glossary of Terms). The implementation of
federal mammography standards, mandated by the Mammography Quality
Standards Act (MQSA) passed by the U.S. Congress in 1996, comes close to
providing all essential elements of the plan for high quality mammography for
American women.
1.2.2.3 Factors that Affect Use of Mammography
In an effort to increase access to and utilization of mammography, the U.S.
Congress passed the Breast and Cervical Cancer Mortality Prevention Act (PL101-
135) in 1991. As a result of this act, cancer control programs at the state level
expanded Medicare coverage to include mammograms for women over 65 at 2
year intervals. By 1992, most states adopted legislation requiring that private
insurers provide some form of insurance coverage for mammograms, and most
states provide Medicaid coverage for the indigent segment of the population (Berg
1995). That year, 23.5 million mammograms were performed at a cost of about
$2.5 billion (Bassett et al. 1994). However, restrictions in eligibility and the low level
of provider reimbursement by government-sponsored programs have had a
negative impact on access (Horsch & Wilson 1993). In addition to medical
16


coverage issues, social and behavioral issues related to mammography utilization
also became apparent (Romans et al. 1991). Such issues include understanding
the benefits of screening, understanding screening guidelines, and attitudes of
patients and physicians.
In order to determine attitudes and practices concerning mammography,
the Jacobs Institute of Women's Health, with grants from the NIH and the ACS,
conducted a comprehensive national Mammography Attitudes and Usage Study
(MAUS) for women 40 years of age and older, in 1990 and 1992. The two surveys
indicated a number of positive trends in screening behaviors. Results showed that
the percent of women who have ever been screened increased from 64% in 1990
to 74% in 1992, as did the percent of those who underwent screening regularly in
compliance with the screening guidelines, an increase from 31% in 1990 to 41% in
1992. These increases, however, did not occur uniformly among all groups of
women. The major beneficiaries, the MAUS study showed, were White women,
women with at least a high school education, and women with incomes of or
greater than $25,000 per year. In contrast, little change in screening behavior was
noted among Black women. Both income and educational level were strongly
(positively) associated with the likelihood of having a mammogram. Another
significant factor was age. Younger women, aged 40 49 and 50 59, appeared to
be more likely to utilize mammography than those 60 and older. In 1992, 78% of
women aged 40 49 and 82% of women in 50 59 age bracket reported having
had at least one mammogram; the proportion of women aged 60-69 was 71% and
only 67% of women over 70 reported ever having received a mammogram.
Among all surveyed women over 40 who had never had a mammogram,
the most frequently mentioned reasons were absence of a family history of breast
cancer, cost, and the lack of a doctor's recommendation (Romans, 1993). Although
the percentage of women who adhered to ACS guidelines increased, less than half
of women aged 40 and older reported that they actually obtain mammograms
regularly. The greatest compliance was again in the 40 49 and 50 59 age
groups, with lower rates among women 60 and older. Compliance with the ACS
17


screening guidelines increased for both White and Black women: 42% of White
women were following the guidelines in 1992 as compared with 32% in 1990; 34%
of Black women complied with the guidelines in 1992, up from 26% in 1990
(Romans 1993).
The MAUS study underscored the importance of a physician referral in
motivating women to obtain a mammogram. In the 1992 survey, 76% of women
who had a mammogram reported that their physicians recommended it; 24% of the
women decided on their own to have one, and among these women, almost one in
four reported that her decision was influenced by a friend or the media. The MAUS
study explored important aspects of public education about BC. There was a high
level of agreement among all women surveyed in 1992 that BC detected in its
earliest stages is highly curable and that mammograms are effective in detecting
these cancers in asymptomatic women.
Women who had never had a mammogram and those who did not follow
the ACS guidelines were more likely than those women who followed the
guidelines to state that cost and concerns over radiation exposure were barriers to
their receiving mammography screening. The same groups of women were also
twice as likely to agree with a statement that a woman need not worry about BC if
no one in her family has had it. Other barriers to receiving a mammogram were
revealed in this survey: nearly a third of women who had never had a mammogram
agreed with the statement: / would not be comfortable asking my doctor to refer
me for a mammogram if he or she did not bring it up. Over 40% of these women
agreed with the statement that mammography is important only for women who
feel a lump or have other symptoms of breast cancer.
Although efforts to increase mammography utilization rates have been
successful, data accumulated in the above surveys identify further challenges in
changing both patients' and physicians' attitudes and behaviors. It appears that
public health education has been most successful in reaching younger, more
affluent, and generally healthy women. The next challenge is to reach out, in a
significant way, to those groups of women who are less easily reached because of
18


barriers such as poverty, culture, limited access to health care, less awareness of
the need for mammography, and older age (Romans, 1993). There is also a need
to remind physicians about the importance of referring their patients for
mammography screening.
Good primary care includes screening for BC among appropriate age
groups of female patients, and virtually all expert groups recommend such
screening. During the past decades, these recommendations translated into
increasingly higher percentages of women receiving BC screening (Anderson &
May 1995). By 1995, a national study found, 70% of women aged 40 years and
older reported receiving a mammogram in the previous 2 years (MMWR 1997). It is
likely that the increasing amount of BC screening in the U.S. is a contributing factor
to the drop in BC mortality between 1990 and 1997.
1.2.3 Issues Related to Follow-up
on Abnormal Screening Results
The very nature of screening mammography service delivery is rather
unique. The procedure is often performed on healthy women who participate in
screening as part of their regular health maintenance. The exam may cause
discomfort or pain, and it carries a risk of discovering the potentially fatal disease.
The extent of BC screening a woman receives can vary according to the specialty
of her physician. A recent study by Finison and colleagues (Finison et al. 1999)
looked at Medicare Part B claims for 1993 and 1994 in New England and found
that, overall, 55% of women aged 65 to 69 years had received mammography in
the previous two years; 78% of these women obtained their breast care from
gynecologists, 67% from internists, 58% from family physicians, 47% from general
practitioners, and 41% from other specialists. Although primary care physicians are
increasingly recommending routine BC screening, these data suggest that there
still is room for improvement, particularly for older women.
Like any screening or diagnostic procedure, mammography has its
limitations; it may fail to detect BC. A mistakenly negative (normal) mammogram
19


may result in delay of appropriate care when there are clinical signs of abnormality.
Films may be interpreted inaccurately. Patients may not receive mammography
results, or received recommendations for diagnostic testing may not be followed.
Painful mammograms may discourage women from future participation in regular
screening (D'Orsi & Debor 1995). A false positive (abnormal) mammogram creates
stress and a temporary decrease in quality of life but does not deter women from
participation in subsequent screening (Gram et al. 1990). With BC screening rates
on the rise, attention is turning to issues related to follow-ups on abnormalities
detected during screening.
1.2.3.1 Frequency of Abnormal Screening Results
Determining the frequency of abnormal BC screening results is complicated
by the fact that patients may receive more than one kind of test, often all three:
CBE, BSE, and mammography. The frequency of abnormal test results, which
generally require non-routine follow-up, is much better documented for
mammography than for the other two types of examination. Earlier studies (Sickles
1991) indicated that between 6% and 18% of women who underwent routine
screening mammography required some type of follow-up. In a national study of
mammography centers (Brown et al. 1995), 11% of mammograms required a
follow-up. Others studies (Keriikowske et al. 1996, Elmore et al. 1998) performed
at various centers found that 6% to 7% of screening mammograms produced
abnormal findings. For CBEs, in Elmore et al. study (1998), 3.7% (409 out of
10,965) of examinations had findings recorded as abnormal over a ten year
period. The most recent study, performed by the Iowa Breast and Cervical Cancer
Early Detection Program researchers (Schootman et al. 2000), found that 11%
(351 out of 3,198) of CBEs and mammograms combined were recorded as
abnormal. Although it is not known how frequently BSE produces an abnormal
result, anecdotal evidence indicates that women often discover palpable lesions.
20


1.2.3.2 What Constitutes an Abnormality?
A breast abnormality is not always palpable and may be free of other
symptoms such as pain or nipple discharge; there is no perfect method to detect all
types of breast abnormality. When established as a result of screening, it is subject
to accuracy of the test (sensitivity and specificity) and standards of comparison. A
review of CBE efficacy (Barton et al. 1999) reported a pooled sensitivity of CBEs in
several studies of BC screening of about 54%, with a specificity of about 94%. This
is in contrast to mammography, which has sensitivity of 76% to 88% in 50-59 year
old women, reported in controlled randomized trials (Fletcher 1995), and a
specificity of 89% to 93% (Brown et al. 1995, Elmore et al. 1998). One reason for
relatively low sensitivity of CBE is the lack of standardized examination technique.
All too often, each physician has his or her own method of conducting an
examination. If standardized, thorough CBE could lead to improved accuracy of
this technique. It thus seems prudent that primary care physicians should work to
improve their CBE skills. In one study, 80% of physicians reported that they felt a
need to improve their abilities in breast lump detection (Fletcher et at. 1985). A
subtler problem concerns the classification of an abnormality found on CBE. The
commonly used two-grade approach, consisting of normal (not suspicious of
cancer) and abnormal (suspicious of cancer) categories, may be too crude to
capture the clinical variability and subtleties of abnormality.
In contrast, radiologists specializing in mammography have developed a
system to grade the severity of abnormalities found on mammograms. The five-
category BI-RADS system of the American College of Radiology (ACR 1998),
first introduced in 1992, has been widely adopted to promote more uniform
evaluation of mammograms. The five complete assessment categories are: 1 -
Negative, 2" Benign Finding, 3 Probably Benign Finding, 4 Suspicious
Abnormality, and 5 Highly Suggestive of Malignancy. There is also an
incomplete assessment category 0, which generally calls for an Additional
Imaging Evaluation. The BI-RADS system is described in detail in the Glossary
of Terms. It should be noted that, while this system aids in the elimination of
21


ambiguous mammography reports, the evaluation of mammograms is not trivial.
Sources of common confusion relate not only to the general usage of the lexicon
by practitioners but also to some features of it. It is often not possible to use a
single descriptor to characterize the finding, particularly for certain types of
calcifications and margin lesions (ACR 1998). The interpretations may also vary
among radiologists (Ciccone et al. 1992, Skaane et al. 1997, Berg et al. 2000).
Regardless of the means of finding an abnormality, screening will not be
effective without adequate follow-up diagnostic studies. This research aims to
examine this aspect of mammography. The next section provides a synopsis of this
project's salience.
1.3 Significance
Much has been published on how delay in evaluating breast symptoms
impacts breast cancer mortality (Charlson 1985, Machiavelli et al. 1989, Neave at
al. 1990, Tabar et al. 1992, Afzelius et al. 1994), but there is an apparent void in
our knowledge on patient-related factors that contribute to delays in obtaining
follow-up care after abnormal mammogram by racially/ethnically diverse
asymptomatic women. This study aims to explore possible predictors of patient-
related delays in six racial/ethnic groups of women: White, Black, Hispanic, Asian,
Native American, and Other.
Quantitative methods, which emphasize isolating the phenomenon under
study from its context, were utilized in the examination of socio-demographic and
clinical factors as predictors of rate and timeliness of follow-up. A qualitative
approach, which emphasizes the context of the phenomenon, was utilized to
explore other factors, such as psychosocial, cognitive, and cultural. Data obtained
from interviews with subjects are used to present womens narratives about their
concerns regarding mammography, breast care, and health decisions. The
methodological triangulation, utilized in this research, offered an effective means of
investigation of the barriers that may deter women from adherence to
recommended follow-ups.
22


The information obtained from this study provides the first known data of
this kind regarding both the delays to follow-up and their predictors in a large
cohort (n=17,358) of asymptomatic women who received follow-up
recommendations to obtain additional diagnostic procedures after their first
mammographic screening.
Successful identification and deeper understanding of the predictors of and
barriers to adherence may serve as a foundation for the development of
interventions aimed at patient-centered quality breast care. This, in turn, may lead
to potentially more effective breast cancer control in women who have no
symptoms of this insidious disease.
23


2. Literature Review
A review of literature pertinent to this study is presented in this chapter.
The following topics are discussed: adequacy of follow-up on mammographic
exams, adherence of patients to screening and prescribed treatments, pertinent
health behavior theories and models, and constructs utilized in the Adherence to
Follow-up Model developed for this study.
2.1 Adequacy of Follow-up on Mammography
2.1.1 Appropriateness of Medical Services
Appropriate diagnostic services are essential components of a follow-up
on questionable findings of any type of screening. The medical community strives
to achieve uniformity of recommendations given to a patient. Defining, measuring,
and classifying appropriateness has been delineated by Brook and colleagues
(Brook et al.1990):
A service is provided appropriately when its benefits (expected
positive consequences) exceed its costs (expected negative
consequences)... Based on literature review, we have delineated
three approaches by which to judge whether a service is
appropriate... Benefit-risk approach... a service is judged to be
appropriate if the health benefits to the patient exceed the negative
health consequences ...by a wide enough margin so that the
procedure is worth doing... Benefit-cost approach... extends the
benefit-risk approach by considering the resource cost of providing
the service... Implicit approach... standards forjudging
appropriateness may have to be used, but no clear, explicit
definitions are specified... Each of these approaches... can be
used to classify a service as being underused, misused, or
overused (p. 3).
In case of mammography, specific guidelines were issued in 1995 by a joint
committee of the Society of Surgical Oncology, the Commission on Cancer of the
24


American College of Surgeons, and the Centers for Disease Control and
Prevention (Cady et al. 1995). Recently, these guidelines were further improved
by a Canadian consensus effort (The Steering Committee on Clinical Practice
Guidelines for the Care and Treatment of Breast Cancer, 1998).
Specific diagnostic tests must be completed for the evaluation and
management of such breast problems as a palpable mass or a non-palpable
mammographic abnormality (Cady et al. 1995). In the case of a solid breast mass,
for instance, triple diagnosis is suggested. If a mass is interpreted as benign by all
three methods, follow-up CBE, mammogram, and fine-needle aspiration,
diagnostic accuracy approaches 99% (Donegan 1992, Osuch et al. 1998). The use
of subsequent diagnostic procedures is imperative if there are abnormal findings
on CBE and a normal mammogram. Similarly, careful measures are required in the
management of situations when women with abnormal mammography results have
clinically normal breasts (Stems 1995).
A common dilemma facing clinicians is how to approach a patient-
discovered mass that is confirmed by CBE but not visualized on a mammogram
(Osuch et al. 1998). Some breast cancers may are missed mainly because it is not
possible to capture the posterior portion of the breast on the film; a cancer cannot
be visualized against a background of dense tissue. Such an event can also occur
as a result of misinterpretation of the film by the radiologist (Svane et al. 1993). In
the Physician's Insurance Association of America (PIAA 1995) study, more than
half of the women who brought successful claims for failure to diagnose BC had
self-discovered masses that failed to impress their physicians on clinical
examination; 80% of these women had normal or probably benign mammogram
results. On the other hand, mammography is capable of revealing large numbers of
non-palpable BC cases; over 15% of all malignancies detected by mammography
in women with palpable masses are non-palpable breast cancers in a different area
of the breast or in the other breast (Rosen et al. 1999).
To enhance survival following detection of BC, all women who have
abnormal BC screening results need to receive appropriate diagnostic services in a
25


timely fashion. A delay in establishing a diagnosis and initiating treatment can result
in a more advanced stage of the disease and thus a worse outcome (Afzelius et al.
1994). Delayed diagnosis also accounts for the most expensive and most common
medico-legal claims against physicians (Osuch et al. 1998).
Causes of untimely follow-up on mammography remain largely unknown.
There are three main types of delay that should be considered: patient delay,
provider delay, and system delay (Caplan & Helzlsouer 1992/93, Facione 1993).
Among clinical and health care system-related factors, screening results, access
time to specialty clinic, the number of diagnostic events per patient visit, and
operating room access time, have been identified to play a role in an urban public-
hospital setting (Wall et al. 1998). Two comprehensive studies that evaluated
reasons for untimely follow-up (Mandelblatt et al. 1993, McCarthy et al. 1994)
suggest that the majority of untimely follow-up is due to patient delay.
2.1.2 Untimely Adherence to Follow-up
Recommendations
While much has been published on what deters women from participating in
screening and from undergoing evaluation of breast symptoms (Vernon et al. 1985,
Richardson 1992, Caplan & Helzlsouer 1992/93, Facione 1993, Lannin 1998), little
is known about factors influencing women's decision of whether and when to
adhere to a recommended follow-up after mammography. There are few published
data on the proportion of women with abnormal mammographic examinations who
do not have timely follow-up, and on the factors that may influence this occurrence.
Definitions of what is considered timely vary, with published reports
measuring time from index abnormal screening mammography to first subsequent
evaluative test or to final test/definitive diagnosis (Chang et al.1996). The
appropriate time span for such follow-up has not been established, but
recommended follow-up intervals generally depend on the category (severity) of
abnormality (Kerlikowske 1996). Screening mammography studies that report a
prevalence of untimely follow-ups also use different methods of classifying the
26


abnormality of interest. Some studies look at complete assessment BI-RADS
categories 4 Suspicious Abnormality, and 5" Highly Suggestive of Malignancy
(Brown et al. 1995, Kerlikowske 1996), while others include the incomplete
assessment category 0 Need Additional Imaging Evaluation (Chang et al 1996,
Webber et al 1996), or examine category 3" Probably benign only (Helvie et al.
1991, Sickles 1991).
McCarthy and colleagues (McCarthy et al. 1994, McCarthy et al. 1996)
investigated categories of follow-up exams as possible predictors of patients' timely
or untimely compliance with scheduled follow-up. They found that 18.1% (in the
sample of n=1,249) of women with abnormal screening mammograms did not
obtain a diagnostic exam within three months of due date and thus had an
inadequate follow-up. Among women with follow-up recommended in four to six
months, 36.8% had inadequate follow-up; 7.2% of women who were recommended
for an immediate follow-up did not comply. Women with low income and no history
of a previous mammogram were at the highest risk for inadequate follow-up.
Chang and colleagues (Chang et al. 1996) examined time intervals from the
date of index abnormal (including incomplete category "0") mammogram to final
disposition (diagnosis) for 317 diverse women and found that these intervals were
significantly longer for Non-white women (median time 19 days) compared with
White women (median time 12 days). The difference persisted after adjusting for
patient age, family history of breast cancer, report of palpable mass, and income.
The authors also stated that the analysis specifying Non-white racial/ethnic groups
found that each group (African American, Latina, and Asian) had significantly
longer log time to first diagnostic test compared with White women (p. 1399);
median times were not provided in the article. Race/ethnicity was recorded based
on clinic staff ascertainment. Age was found not to be a significant independent
predictor of timeliness of follow-up on abnormal mammograms.
Age and race as predictors of untimely follow-up on abnormal
mammogram, along with mammographic interpretation, and type of tracking
system, were also examined by Kerlikowske (1996). The rate of timely follow-up,
27


defined as occurring within 8 to 12 weeks after index abnormal mammography,
ranged from 69% to 99%. Elderly women (>65 years of age), those with lower
socio-economic status, and those who had been instructed to obtain repeated
evaluations in four to six months had the highest proportion of untimely follow-up.
Differential follow-up rates by type of mammography results are the subject
of a study by Webber and colleagues (Webber et al. 1996). The authors examined
1,202 cases where a follow-up on mammography was required. They found that
the overall adherence was the highest (89%) for the youngest (<29 years of age)
group while elderly women had the lowest rate (68%) of follow-up. When ethnicity
was examined as a factor, overall follow-up rates did not significantly differ by race.
Descriptively, Black women had the highest follow-up rates with Native American
women having the lowest rates (not provided in the article). When mammography
results indicated suspicious abnormality or were highly suggestive of malignancy,
there were no significant differences by age. However, for these mammographic
result categories, the rates of adherence differed significantly by race. In the
sample of 292 women, 79.1% of women adhered with Black women reaching
92.3%, followed by White (75.3%), Hispanic (74.1%), and Asian/Pacific Islander
(64.1%); Native American women again had the lowest follow-up rate of 37.5%
(proportions of each racial/ethnic group in the sample were not provided).
Among women with no follow-up within three months of abnormal
mammography screening, 83% (of n=149) recalled being notified of test results
(Kerlikowske 1996). Of these women, 53% had the understanding that their
mammographic examination was normal and that no further action was required.
Such women may not have followed-up because they did not understand their test
result, because their physician told them it was unnecessary, or because of low
perceived risk of breast cancer, forgetfulness, low perceived importance of follow-
up, or insufficient access to facilities that perform diagnostic evaluations.
Mandelblatt and colleagues (Mandelblatt et al. 1993) examined the
adherence to follow-up among poor, elderly, Black women. They found that in the
sample of 491, nearly one-third of women failed to complete follow-up on abnormal
28


mammography or clinical breast exam. The authors also found that follow-up was
delayed by nonspecific fears, a feeling of being too old for treatment, the desire to
not want to know if something is wrong, and the perception that nothing was
bothering them. Fear of being diagnosed with cancer has also been cited as a
reason for lack of compliance with follow-up on abnormal Pap smears (Lerman et
al. 1991 [a]).
Depending on the kind of abnormality and which test indicated it, adequate
follow-up might include a more or less invasive procedure: a repeat examination
which may include compression or magnification views, surgical consultation,
ultrasonography, excisional or core biopsy, or fine-needle aspiration (Cady et al.
1995, Schootman et al. 2000). The invasiveness of a follow-up procedure may also
play a role in womens decision to adhere (Sickles 1991).
2.2 Models Suitable in Studies of Adherence Behavior
A review of literature on the adherence of patients to screening and on
compliance with prescribed treatments indicates that the Health Belief Model
(HBM) provides a useful framework to study these phenomena. Behaviors explored
within this framework include adherence to BC (Thomas & Fick, 1995, Bowen et al.
1997) and prostate cancer screening (Plowden 1998, Myers et al. 2000), and
compliance with coronary heart disease exercise (Mirotznik et al. 1995),
antiretroviral therapy (Mostashari et al. 1998), tuberculosis treatment (Pablos-
Mendezet al. 1997), schizophrenia treatment with neuroleptics (Bebbington 1995),
and malaria prophylaxis regimens (Abraham et al. 1999). The versatility of the
model, its disease-avoidance orientation, and the possibility of utilizing its selected
components, make this model suitable for such studies. The Adherence to Follow-
Up Model, developed for this project and described in Section 3.1 of Chapter 3,
utilized elements of the HBM and its phenomenological foundation.
29


2.2.1 The Health Belief Model (HBM)
2.2.1.1 Historical Context of the HBM Creation
The HBM grew out of a set of independent, applied research problems with
which Public Health Service (PHS) investigators were confronted in the 1950s. In
the post WWII years, the PHS was for the most part oriented toward the
prevention, not the treatment, of disease. During that same time, there was a
widespread failure of people to accept disease preventives or screening tests for
early detection of asymptomatic disease (Rosenstock 1974). These measures
included at first tests for tuberculosis and TB prevention programs, then cervical
cancer screening, followed by prevention programs for dental disease, rheumatic
fever, polio and influenza. Any useful theory had to have an avoidance of disease
orientation and explain preventive health behavior; it had to provide insight
regarding behaviors of individuals who were not currently suffering because of a
disease.
The researchers who worked cooperatively in the development of the HBM
were strongly influenced by theories of Kurt Lewin, particularly by the psychological
theory of goal setting in the level-of-aspiration situation. Level-of-aspiration is
defined as ...the level of future performance in a familiar task, which an individual,
knowing his level of past performance in that task, explicitly undertakes to reach
(Frank 1935, p. 119). This concept considers the conflict an individual faces in
deciding whether to attempt tasks that appear difficult to achieve or to be satisfied
with more-certain success at easier tasks. An extensive theoretical exposition of
goal setting in this context, undertaken by Lewin and others (Lewin et al. 1944)
incorporates the concepts of valence, subjective probability and force, deriving from
Lewin's general theory of behavior (Lewin 1935). The level-of-aspiration theory
assumes that the psychological condition of a person exhibiting his/her level of
aspiration is characterized as a choice situation, where choice is determined by the
valence (positive or negative attraction), which degrees of difficulty within the same
type of activity have for that person. Selected level of performance allows for the
30


possible outcomes of reaching ("success") or not reaching ("failure") that level. The
individual is influenced not only by the attractiveness of success (positive valence)
at the level of performance, but also by the subjective probability of being able to
attain that level (Lewin et al. 1944). In summary, this theory hypothesizes that
behavior depends mainly upon two variables: the value placed by an individual on a
particular outcome, and the individual's estimate of the likelihood that a given action
will result in that outcome. The more general version of this theory states that paths
to action are selected depending upon possible outcomes (values), the needs of
the person, and the barriers to the goal region (Maiman & Becker 1974).
The HBM creators, all trained social psychologists, exhibited strong
phenomenological orientations. They shared the view that it is the world of the
perceiver that determines his/her decisions and actions, not the physical
environment of the individual, except for the representation of the physical
environment in the mind of the individual. A later version of this theory would
include a heavy component of motivation along with the perceptual world of the
behaving individual. Also, in Lewinian tradition, the theory focused on the current
situation (ahistorical orientation) confronting the behaving individual rather than his
or her prior experiences (Rosenstock 1974). The original HBM, formulated by Drs.
Godfrey Hochbaum, Stephen Kegeles, Howard Leventhal, and Irwin Rosenstock is
presented in Fig. 2.1.
2.2.1.2 Basic Features of the HBM
The HBM model postulates that, in order for an individual to take action to
avoid a disease, he needs to believe that he is personally susceptible to it, that the
occurrence of the disease will affect some component of his life with at least
moderate severity, and that taking a particular action will in fact be beneficial in
terms of reducing his susceptibility to the condition or, if the disease occurs,
reducing its severity.
Additionally, the model assumes that preventive action will not entail
overcoming important psychological barriers such as cost, inconvenience, pain or
31


INDIVIDUAL PERCEPTIONS
MODIFYING
FACTORS
LIKELIHOOD OF
ACTION
Fig 2.1 The HBM as predictor of preventive health behavior, adapted from Becker, 1974.
embarrassment (Rosenstock 1966). With respect to having a test for early
detection of a disease, the same factors were deemed necessary. Another
assumption of the model is that the individual believes that he could have a disease
32


even in the absence of symptoms (Rosenstock 1960).
Further advancements in the development of the HBM involved weighing
benefits against psychological and other barriers or costs of the proposed action,
perceived by the individual (including effort or work involved in taking action).
A stimulus or cue to action is also included (Becker 1974). Whether its origin is
internal (i.e. perception of bodily stress) or external (i.e. mass media
communications, personal knowledge of someone affected by the condition), the
stimulus must be present to trigger the appropriate health behavior. This
expectancy-value approach to health behavior emphasizes the attractiveness of
the incentive value of the health action or goal to the individual in terms of one's
ability to control (reduce) perceived susceptibility to the particular disease and its
consequential severity. The combined levels of susceptibility severity provide the
force to act, and the perception of benefits provides a preferred path of action.
Motivation is assumed as a necessary condition for action. It is
operationalized in the model's dual dimensions: the psychological state of
readiness to take specific action and the extent to which a particular action is
believed to be beneficial in reducing the threat (Rosenstock 1990). This concept
has its roots in Lewin's level-of-aspiration theory (one of the first decision-making
theories) wherein the individual desires to achieve success or to avoid failure.
Due to its adaptability, the HBM has inspired voluminous health-related
research. The theory itself has been the subject of considerable direct study, and it
has served as the foundation of many subsequent health behavior models.
2.2.1.3 The HBM, Health and Health Behavior
The HBM was originally formulated to explain preventive health behavior.
Health behavior, as defined by Kasl and Cobb (Kasl & Cobb 1966), is an activity
undertaken by a person who believes himself to be healthy, for the purpose of
preventing disease or detecting disease in its asymptomatic stage. Conversely, the
illness behavior is an activity undertaken by a person who feels ill, for the purpose
of defining the state of his health and of discovering a suitable remedy. Sick-role
33


behavior is the activity undertaken by individuals who consider themselves ill, for
the purpose of getting well. The three modes of behavior are not discontinuous.
Hardly anyone can be found who, upon intensive questioning, would report himself
free of all symptoms. Similarly, the boundary between illness behavior and sick-role
behavior is blurred.
Historians of medicine have labeled as ontological the view that diseases
are specific entities that unfold in characteristic ways in the typical person
(Aronowitz 1998). In this framework, diseases exist in some platonic sense outside
their manifestations in a particular individual. The other compelling account of
illness, the physiological or holistic, stresses the individual and his or her
adaptation, both psychological and physical, to a changing environment. In this
framework, illness exists only in individuals.
These notions have been in a state of dynamic tension since antiquity.
Temkin (1977) cites related distinctions drawn by other scholars, such as Platonic
versus Hippocratic, realist versus nominalist, rationalist versus empirical, and
conventional versus natural. In more recent years, these distinctions have been
used as part of an effort to critique or expand the traditional biomedical model of
disease and doctor-patient encounter, particularly in cross-cultural settings.
Perhaps the most influential formulation has been the promotion of a distinction
between illness and disease by social science critics of medicine, notably Arthur
Kleinman and Leon Eisenberg (Kleinman et al. 1978). According to Kleinman
(1988), illness refers to the innately human experience of symptoms of suffering
(p.3). Disease, on the other hand, is what the practitioner creates in the recasting
of illness in terms of theories of disorder (p. 5). These distinctions reflect often
discrepant models of health held by patients and physicians, particularly so in a
multi-cultural setting (Eisenberg & Kleinman 1981). They capture the difference
between the lived experience of the patient versus the system of knowledge
developed by biomedicine. Further, they refer to mental states of patients, which
may account for behaviors that may not be concordant with those suggested by the
biomedical model (Pachter 1994).
34


These distinctions were considered in my research as it involves subjects
drawn from a diverse population of asymptomatic women who, after obtaining their
first mammogram, are told to follow-up because the mammogram turned out not to
be normal. While the outcomes and economic implications of screening
mammography have received some attention (Moskowitz 1987, Eddy et al. 1988),
the psychological and behavioral consequences of abnormal mammograms have
been largely neglected. Research in screening for other diseases suggests that the
psychosocial effect of receiving abnormal results may be substantial (MacDonald et
al. 1984, Reelick et al. 1984). The psychological sequelae of an abnormal BC
screening result in the absence of symptoms could potentially adversely affect
subsequent participation in screening (Bellock & Breslow 1972, Rimer et ai. 1989,
Vernon et al. 1993), and adherence to a follow-up (Lerman et al. 1991 [b], Shaw de
Parades 1994). Heightened perceived vulnerability to BC has been related to
reduction in the practice of breast self-examination among high-risk women
(Anagna et al. 1987). Anxiety and worry about cancer has been associated with
reduced participation in repeat mammography screening (Lerman et al. 1990) as
well as delay in seeking medical attention for possible cancer symptoms
(Greenwald et al. 1978) or symptoms of any potentially serious disease (Sarafino
1997, Mirotznik et al. 1998). These factors seem likely to play a role in adherence
behavior; they were given attention in my research.
Health decision-making, including health behavior, illness behavior, and
sick-role behavior, is a process in which the individual moves through a series of
phases or stages. In each of them, he/she interacts with other individuals and
events. The nature of the interactions at any one of these stages may increase or
decrease the probability that a particular subsequent response will take place (Zola
1964, Zola 1973). The HBM presumes that an individual's relevant health beliefs
will serve as a setting for the subsequent responses at other stages in the decision
process. For example, individuals who accept their susceptibility to a particular
condition and are aware of actions that might be beneficial in reducing their
susceptibility may very well be the same persons who exhibit, what Freidson (1961)
35


terms cosmopolitan rather than parochial orientation toward health services. These
individuals may be more prone to leam about and seek out professional diagnosis
rather than using the lay referral system. In such cases, the initial set of beliefs
would determine the choices in the decision-making process. What the HBM terms
here as cues to action, Zola (1964) refers to as critical incidents.
Again, relative to health decision-making influences of and interactions of
study subjects with other individuals (significant others and health professionals in
particular) and a subject's life events, are as important to this inquiry as are the
women's particular health beliefs; attempts were made to address them. An
abnormal mammogram result may serve as the critical incident or a cue to action.
The meaning of abnormality to the person receiving such a result is likely to have
an impact on perceived susceptibility to BC and severity of the condition. This
meaning may be shaped by the woman's selection of BC information sources
(Meischke & Johnson 1995) as well as by society's norms, expectations, and
culturally shared rules of interpretation (Jones & Moon 1987) and thereby influence
adherence to and timeliness of follow-up.
2.2.2 Health and Health Behavior Models
2.2.2.1 Antonovsky's Models of Health
and Health Behavior
It appears that Antonovsky and Kats (1970) independently developed an
integrated model of the determinants of health behavior, which was in many ways
similar to the HBM but offered additional insights. The model successfully explained
much of the variation in behavior related to obtaining preventive dental care in a
sample of about 500 employees of the Hadassah Medical Organization in
Jerusalem. The authors took exception to at least two of the HBM concepts, one of
which is of considerable significance.
The first criticism regards cue to action. For Antonovsky and Kats (1970),
this is a superfluous concept the individual prepared to take action will create his
36


own cues. There appears to be no more than a semantic difference here. As
suggested by Zola (1964), the concept of cues, identical to his critical incidents that
serve as triggers to initiate a train of events, fits well with the psychological value
expectancy theory.
Of perhaps greater importance is the criticism that the early HBM failed to
include a motivational component. It is true, as discussed earlier, that a
motivational variable was not included in the initial model and was added later.
There is, however, a more fundamental difference in viewing the concept of
motivation. For Antonovsky & Kats, motivation is a goal-oriented behavior, and the
relevant goal is that of maintaining health. Other relevant goals, not considered by
these two scholars, may include achieving approval by significant others or
achievement of self-approval.
The HBM underpinnings cannot accommodate the view that perceptions of
severity and susceptibility play the principal role in the increase of the salience of
the motive (Becker et al. 1974). Although motivation is considered necessary for
action to take place, the HBM implies that cognitive factors may be playing a
somewhat independent role in influencing behavior.
The concept of the role of significant others has been explored by other
researchers. Kasl (1983), for instance, uses it as a neutral label that encompasses
social networks, social support, and social isolation. Belief in one's ability to
perform adequately (with approval), and congruence between one's beliefs and
societal expectations are basic tenets of this concept. They are likely to influence
health behavior, and perhaps health outcomes.
The role of significant others in the decision-making process and in carrying
out the decision to obtain mammographic screening and to adhere (or not to
adhere) to follow-up recommendations were addressed in the qualitative part of this
research.
In their original model, Antonovsky and Kats (1970) introduced the concept
of threshold level, which suggests that the relationship between a given variable
and preventive health behavior is not linear. This would mean that the motive and
37


the perceptions of susceptibility, severity, benefits, and costs may have cut-off
points for a given individual below which the variable will have no effect on the
behavioral outcome. Aaron Antonovsky further explores this concept in his studies
involving concentration camp survivors that culminate with the formulation of
salutogenic model of health (Antonovsky 1979). Health outcomes are generally
unpredictable as the life stressors and, although variable, are omnipresent.
Confronting a stressor results in a state of tension with which one needs to
deal. Whether the outcome will be pathological, neutral, or salutary, Antonovsky
asserts, depends on the adequacy of tension management. Salutogenic orientation
focuses on positive health outcomes attainable despite countless and repeated
stressors that one is bombarded with, the phenomenon of good deviant cases. In
order to effectively combat stressors, one needs general resistance resources
(GRRs), which include money, ego, strength, cultural stability, and social supports.
In addition to GRRs, the model proposes the existence of a strong sense of
coherence (SOC) as necessary for salutogenic outcomes. These concepts were
very useful in my inquiry as it involved diverse women with quite different socio-
demographic characteristics or GRRs.
2.2.2.2 Andersen's Model of Health Services' Use
How people deal with health-related stressors may affect the ways in which
they utilize health services. Numerous models offer insights into the multiple factors
that influence such behaviors. One of the first behavioral models of health services'
use was developed by Ronald Andersen (Andersen & Newman 1973) as a part of
his doctoral dissertation at Purdue University in the late 1960s (Fig. 2.2).
Its initial purpose was to assist in the understanding of why families use
health services; it attempted to integrate a number of ideas about the how and why
of health services' use. The original focus of the model on the family as the unit of
analysis grew out of the perspective that the medical care an individual receives is
a function of the demographic, social, and economic characteristics of the family as
a unit. Subsequently, the focus shifted to the individual as the unit of analysis due
38


PREDISPOSING ^ ENABLING ^ NEED 1 ^ USE OF
CHARACTERISTICS I RESOURCES I I HEALTH SERVICES
I Demographic I I Personal/famiiy I I Perceived I
I Social Structure I Community I (Evaluated)
Health Beliefs
Fig 2.2 The initial version of Andersen's Behavioral Model of 1960s (Andersen & Newman
1973)
to difficulties of developing measures at the family level that would take into
account the potential heterogeneity of family members, i.e., a summary measure of
family health status. This model suggests that people's use of health services is a
function of their predisposition to such an activity, factors that enable or impede
use, and their need for care. While there had been some question of whether this
model was meant to predict or explain use (Mechanic 1979, Rundall 1981), the
model, it seems, could be applied in both inquiries. On the one hand, each
component could be conceived as making an independent contribution to predicting
use. On the other, the model suggests an explanatory process or causal ordering
where the predisposing factors might be exogenous (especially the demographic
factors and social structure), where some enabling resources are necessary but not
sufficient conditions for use, and where some form of need must be defined for use
to actually take place.
Among the predisposing characteristics, demographic factors such as age
and gender represent biological states predictive of the likelihood that people will
need health services (Hulka & Wheat 1985). Social structure is measured by a
broad array of factors that determine the status of a person in the community, his or
her ability to cope with presenting problems and command resources to deal with
39


these problems, and how healthy or unhealthy the physical environment is likely to
be. Traditional measures used to assess social structure include education,
occupation, and race/ethnicity. Although the model has been criticized for not
paying enough attention to social networks, social interactions, and culture (Bass &
Noelker 1987, Portes et al. 1992), it would seem that measures of these concepts
adequately fit into the social structure component.
Health beliefs are attitudes, values, and knowledge that people have about
health and health care these are likely to influence their perceptions of need and
use of health services. They provide one means of explaining how social structure
might influence enabling resources, perceived need, and subsequent use. Social
psychologists have been concerned that health beliefs have not been appropriately
conceptualized and measured in much work employing the behavioral model
(Mechanic 1979, Becker & Maiman 1983). A possible consequence is that health
beliefs have not appeared to be as important as they really are in predicting and
understanding use of health services. While significant efforts have been made to
integrate elements of the behavioral model with the HBM (Green at al. 1980), it has
been argued that more effort needs to be expended to show stronger and more
meaningful relationships between beliefs and use of health care by refining
measurements of beliefs and needs, as well as types of use (Tanner et al 1983). If
we examine beliefs about a particular disease, measure need associated with that
disease, and observe the services received to deal specifically with that disease,
the relationships will probably be much stronger than if we try to relate general
health beliefs to global measures of need and a summary measure of all services
received in a given period of time.
The initial version of Andersen's behavioral model has also been criticized
as having a built-in bias that the increased use of health services is always better
and should be sought (Chen 1978). One must reflect on the health care setting at
the time of development of this model; increased utilization was a major policy goal
and cost was not quite the concern it is today. The model seems, nevertheless,
quite useful for an inquiry regarding conditions that either facilitate or impede
utilization of services.
40


This model has been further developed by adding components to
predisposing characteristics, to enabling resources, and need for services. It would
seem that with the explosive development of gene mapping, genetic counseling,
and the possibilities of gene therapy, genetic factors may represent a potentially
viable, important and definable predisposing component (Rosnau 1994). Ethical
issues involved in such considerations have so far prevented their pursuit. Another
possible component, conceptually distinct from those present in Andersen's model,
is psychological characteristics.
Psychological characteristics considered by others as predisposing
variables have included self-efficacy (Bandura 1977), cognitive impairment (Bass et
al. 1992), and autonomy (Davanzo 1994). A comprehensive effort to model
utilization of health services must consider how people view their own general
health, what they do to protect it, how they experience symptoms of illness, pain
and worries, and whether or not they judge their problems to be of sufficient
importance to seek medical care.
Additionally, the perceived need of a person to utilize health care cannot be
examined devoid of the social context. Both community and personal resources
must be present for use to take place. Health care facilities and personnel must be
available where people live and work. People must have the means and know-how
to obtain these services. A regular source of care, income, health insurance, and
travel and waiting time may be some of the measures of importance here. The
framework provided by Andersen's model offers limited insights into these issues.
One of the major strengths of this framework is the fact that it offers the
possibility to explore the very complex issue of access to medical care. Potential
access is defined as the presence of enabling resources. Better/more enabling
resources provide means for use of health services, and increase the likelihood that
use will take place. Realized access is the actual use of services. Equitable or
inequitable access measures are defined in terms of dominant predictors of
realized access (Andersen et al. 1975).
Within the framework of Andersen's model, equitable access would occur
when demographic and need variables account for most of the variance in
41


utilization. Inequitable access would occur when social structure (i.e.
race/ethnicity), health beliefs, and enabling resources (i.e. health insurance)
determine who receives medical care. One might argue, however, that people's
health beliefs, and consequently the use of services influenced by those beliefs or
by perceived need determined by those beliefs, should be considered equitable
access variables. Value judgments about which components of the model should
explain utilization in an equitable health care system are crucial to the definition;
equity is in the eye of the beholder (Aday et al. 1980).
2.2.2.3 Other Behavioral Models
The original version of Andersen's realized access concept, measured by
utilization and satisfaction, has been expanded in subsequent health behavior
models. Most notable is the addition of outcome measures in cancer care studies
(Mandelblatt et al. 1999). This measure is important for two reasons, particularly
when evaluating access to medical care throughout the continuum of cancer care.
First, having a quantifiable end point is necessary to determine whether access has
occurred and has had the intended effect. Secondly, defined outcomes may serve
as quality of care indicators (Epstein 1990, Brook & Cleary 1996).
In the continuum of cancer care, individuals (and populations) must gain
access to early detection services. Next, if a screening test is abnormal, diagnostic
services must be available. For those diagnosed with cancer, staging evaluation
precedes and often determines treatment. Patients surviving their disease also
need to have access to an on-going surveillance, so that recurrences can be
recognized and treated in a timely fashion. For those who will die of their disease,
having access to end-of-life care is critical. Thus, access has different dimensions
and outcomes across the spectrum of cancer care; some domains may be more
relevant for a particular aspect or phase of care. In this study, outcome
measurements occurred in two phases of the cancer care continuum: BC screening
and follow-up on abnormal mammograms.
42


Among numerous other health behavior models, one recently developed
and validated by Sally Vernon, Ronald Myers and others (Vernon et al. 1997), the
Preventive Health Model (PHM) particularly influenced my thoughts and guided me
in the development of a model appropriate for this inquiry. The PHM, depicted in
Fig. 2.3, brings together concepts from the work of Antonovsky (1984), the HBM,
the Theory of Reasoned Action (Ajzen & Fishbein 1980) and the Social Cognitive
Theory (Bandura 1986). It posits that background, cognitive and psychological
FACTORS RELATED TO PREVENTIVE PREVENTIVE
PREVENTIVE INTENTION & BEHAVIOR INTENTION BEHAVIOR
Background Factors:
Socio-demographic characteristics
Medical History
Past Preventive Behavior
Cognitive/Psvchological Factors:
Perceived Susceptibility to Disease
Worry about Having a Disease
Interest in Knowing Diagnostic Status
Belief in Disease Prevention & Curability
Belief in Salience & Coherence of
Behavior
Belief in Efficacy of Detection & Treatment
Belief in Self-Efficacy Related to Behavior
Concern about Behavior-Related
Discomfort
Social Support & Influence Factors:
Support & Influence of Family Members
& Health Care Professionals
Intention
to
Engage in Behavior
Behavior
Program Factors:
Provider Actions that Facilitate Preventive
Behavior
Fig 2.3 The Preventive Health Model (adapted from Myers et al., 1999)
43


characteristics, social support and influence, and preventive program factors are
associated with intention to take preventive action and practice preventive
behavior.
This framework has been useful in explaining intention and behavior
concerning prostate cancer screening in Black males (Myers et al. 1994, Myers et
al. 1996, Myers et al. 1999), and in colorectal cancer screening in White males
(Vernon et al. 1997, Myers et al. 1998). A number of constructs used in these
studies of screening behaviors became part of the Adherence to Follow-Up Model
developed specifically for this research (depicted in Fig 3.1 in Chapter 3, p. 50).
The review of health and health behavior models makes it clear that there
are overlaps in conceptual domains. The social influence (Ajzen & Fishbein 1980)
construct in the PHM, for example, can be viewed as a refinement of the benefits
and barriers dimensions of the HBM. Similarly, self-efficacy (Bandura 1977) may be
seen as an aspect of perceived barriers. The review of studies that utilized these
models indicates that, even when researchers use the same constructs, operational
definitions often vary from study to study. This presents difficulties if one attempts
to generalize the findings from different studies or draw comparisons.
The goal of my research was to determine predictors of adherence behavior
for diverse groups of women who were recommended to follow-up on their
screening mammograms. While the components of models suitable for studies of
screening behaviors among various groups of subjects were very useful in this
inquiry, they were not sufficient as, in addition to BC screening, this study was
concerned with another phase of cancer care continuum: follow-up on abnormal
mammograms. An abnormal mammogram, a result of the first screening, is viewed
as the Critical Event It is, on the one hand, an external factor that prompts a follow-
up. On the other hand, this event is internalized by the woman's perception
regarding its meaning. A number of factors, of which a personal threat of having BC
is most salient, come into play here. This aspect, while existent in the realm of
screening behavior inquiry, is brought to another level by the Critical Event. How a
woman perceives the Critical Event, and whether and how her perceptions can be
44


modified depends on a host of factors. These factors, with full awareness that the
list is not exhaustive, have been explored with the help of the Adherence to Follow-
Up Model.
Womens perceptions are likely to be influenced by cultural factors. As this
study involves a racially/ethnically diverse sample of population, it was important to
ground the inquiry in multicultural understanding and exercise cultural sensitivity in
all phases of this project. The Multicultural Understanding Model (Fig. 2.4) provided
necessary aid.
2.2.3 The Multicultural Understanding Model
The roles of teachers, researchers, and health care professionals have
recently been expanded to include considerations of the cultural identities of
students, clients, and patients. Professionals have a responsibility to increase their
awareness, knowledge, and skills so that their conduct recognizes the influences of
the cultural group on an individual. All individuals are, in some respects, like all
other individuals; they are members of the human race, and, as such, share
membership in our own species, Homo Sapiens. At the same time, all individuals
are, in some respects, like some other individuals, as a result of cultural group
membership. The cultural group serves as the basis for individuals to become
humanized; each individual becomes fully human through the process of
participating in a cultural group or groups. Finally, all individuals are, in some
respects, like no other individuals, in that there is some uniqueness in each
individual. Individuals differ from one another both biologically and socially; no two
individuals share the same experience in their society. To understand a person one
needs to be aware of these three identities, of the fact that each individual is
seeking a personal identity by acknowledging and identifying, to a greater or lesser
degree, with a cultural group while living in a world community.
There is no doubt that the experiences of culturally different groups differ
from one another and from the dominant culture. While it is recognized that
culturally different individuals and groups all face certain similar barriers and similar
45


issues related to acculturation, the key to understanding a particular cultural group
lies in an appreciation of the wide diversity of the individuals' experiences.
This project involves study of adherence behavior following an abnormal
mammogram, likely to have been influenced by cultural nuances in the meaning of
the experience among racially/ethnically diverse women. A Model of Multicultural
Understanding (Fig 2.4) developed by Locke (1992) offered a useful foundation for
this inquiry.
There are many elements of the culture in this model: the socio-political
factors, the culture's history of oppression, the experience of prejudice and racism,
poverty within the culture, influence of language and arts, influence of religious
practices, child-rearing practices, family structure, values and attitudes, and the
degree of opposition to acculturation. The model is similar to Sue & Sue's (1990)
model; both models stress the impact of socio-cultural forces as well as the
psychological and developmental influences on the behavioral expressions of
different racial and ethnic groups.
Lockes model provided guidance for more in-depth study of particular
cultures as part of this research. While race/ethnicity was a dominant variable in the
quantitative analyses that considered adherence of various groups, the aggregate
information obtained as a result was presented in an objective and culturally
sensitive manner. The awareness of the existence of differences within cultural
groups that may sometimes be greater than the difference between the dominant
culture and other cultures facilitated consideration of the cultural uniqueness of
individual study subjects in the qualitative study. Giving balanced attention to these
aspects of uniqueness prevented stereotyping women as members of particular
cultural group.
Thorough examination of all elements of Locke's model is beyond the scope
of this inquiry, however, among the model's elements, cultural values and attitudes,
were particularly useful concepts; they were explored in-depth in this research as
they strongly relate to health behavior.
46


GLOBAL INFLUENCES
Fig 2.4 Multicultural Understanding Model (adapted from Locke 1992)
This research was undertaken with full awareness of the fact it is in the nature of
many culturally diverse groups to limit their accessibility to outsiders and that no
direct research can be conducted without being somewhat intrusive. In this respect,
the choice of research methods is of fundamental importance.
47


3. Research Design and Methodology
This Chapter contains a description of the Adherence to Follow-up Model
specifically designed for this project, a review the paradigms that guided this
research, the research styles selected, the research design employed in pursuing
the study aims, the methodology selected, and reconciliation of the various
methods utilized in this project.
This study is in the broad category of explanatory research. It attempts to
find an answer to the following question:
What factors are associated with delay in recommended follow-up after
abnormal mammograms among women of different racial/ethnic groups?
In an effort to answer this complex question, numerous other questions were
addressed and ideas explored, utilizing both quantitative and qualitative methods.
The Adherence to Follow-up Model was designed for the purposes of identification,
description, and explanation-generation of factors that may have influenced
adherence behaviors.
3.1 Research Design
Philosophical inquiry provided guidance through the duration of this multi-
method project. This research style, frequently referred to as a thought
experiment (Miller & Crabtree 1999), assured clarity of ideas and provided the
basis for conceptual development of relationships between these ideas.
3.1.1 The Adherence to Follow-Up Model
The Adherence to Follow-Up Model developed for this project to aid in
pursuing its general aims is depicted in Fig 3.1. The model contains some of the
48


elements and constructs of models reviewed and discussed in the previous
Chapter. To avoid pitfalls of other comparative studies, the emphasis in the
development of this model was placed on retaining the uniformity of the constructs'
operational definitions for the two initial stages of the cancer care continuum
examined in this study: screening mammography and follow-up on possibly
abnormal findings.
For the Critical Event to take place, subjects had to have undergone the
screening. Thus a mammography screening (the first for each subject) is the
predisposing event. Its determinants, Predisposing Factors, can be grouped into
three categories: socio-demographic characteristics, cognitive/psychologicai
factors, and cultural/social support facilitators of screening. Socio-demographic
factors relative to screening were examined in the preliminary inquiry using
quantitative methods; a qualitative approach was utilized to assess the remaining
two categories of factors relative to participation in screening.
The Critical Event a possible abnormality was examined from a clinical
findings standpoint external to the subject, and from the subject's internal
perspective in terms of her perceptions regarding these findings. In addition to
result severity construct, the invasiveness of recommended follow-up procedure
was examined as a potential factor associated with adherence. Again, this
construct was considered as external to the subject clinical factor and from the
subject's internal perspective in terms of her understanding and perception of the
meaning regarding a follow-up procedure. External factors were examined using
quantitative methods. A qualitative approach was utilized to explore the internal
factors relative to clinical findings and follow-up recommendations as well as those
relative to the emotions women that women experienced following the Critical
Event.
Modifying Factors are grouped in three categories: socio-demographic
predictors, personal clinical factors, and cognitive/cultural factors. Socio-
demographic predictors of screening were examined for the cohort that received
follow-up recommendations and for the sub-cohorts of women who adhered to
49


Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
cn
o
PREDISPOSING FACTORS CRITICAL EVENT MODIFYING FACTORS LIKELIHOOD OF ACTION ACTION
(participation in BC screening) (possible abnormality) (intention) (behavior)
Soclo-demooraphlc factors:
Race/ethnicity,
Age,
Education,
Health insurance.
Coanltive/psvcholooical factors:
Belief in value of good health,
Belief in health maintenance,
Belief that one can have BC
without having symptoms,
Perceived susceptibility to BC,
Belief in screening efficacy &
self-efficacy,
Concerns about mammography
(i.e. cost, radiation, pain).
Cultural/social support factors:
Race/ethnicity,
Acceptance of BC screening
practice by seif and others,
Support of (significant) other(s),
Locus of control,
Influence of health professionals,
Concerns about mammography
(embarrassment,fear,discomfort)
External clinical factors: Mammography results (severity) Recommended follow-up (invasiveness) Socio-demooraphlc adherence predictors: Race/ethnicity, Age, Education, Health insurance.

Personal clinical
Internal: perceived Prior contact with BC
Understanding & (family history of BC),
meaning of: Communication with
result severity & health care
invasiveness of professionals
follow-up (F/U),
Treat (of BC & of

misdiagnosis), Coonitlve/cultural
Anxiety, Sources of knowledge
Worry, (about health and BC),
Understanding of Role of (significant)
F/U purpose others in decisions
(i.e. risk reduction about health,
benefit, efficacy), Social support & social
Control ability, networks,
Trust of health Sense of coherence
professionals. (comprehensibility,
manageability,
meaningfulness),
Belief in BC curability.
Paving general mwn?
to complete F/U:
(financial, time,
situational i.e.
transportation, etc.)
Psvcho-soclal factors:
Perceived benefits of
F/U outweigh perceived
barriers,
Congruence of beliefs
& social expectations,
Cultural stability,
Motivation (state of
readiness to take action
believed to be beneficial)
Strength to cope,
Salutoganic outlook.
Adherence:
Level (within
12 months:
overall, when
results possibly
abnormal, when
result probably
benign, when
results abnormal,
Timeliness: when
result probably
benign, when
results abnormal
Fig 3.1 The ADHERENCE TO FOLLOW-UP (FAJ)
MODEL (November 2000)


these recommendations, both timely and untimely. Quantitative methods were used
to examine these factors. Clinical factors were also assessed using quantitative
methods. Cognitive/cultural factors were explored using a qualitative approach.
Qualitative methods were used in effort to assess the Likelihood of Action.
General means/resources that are required to take the desired action and
psychosocial factors served as constructs for assessment of subjects' intentions to
adhere to recommended follow-up.
Finally, the Action taken was assessed using quantitative methods.
Outcome variables for various groups of women were: an overall rate of adherence
to recommended follow-up within 12 months, the rate of adherence for the cohort of
women whose first mammograms were possibly abnormal, adherence rate and
follow-up timeliness for the cohort with probably benign finding, and adherence rate
and timeliness of follow-up for the cohort with severe abnormalities, suspicious
abnormality findings and mammographic results highly suggestive of malignancy.
Measurements of these constructs are described along with the description of
specific methods selected for this research (Section 3.2 of this Chapter).
3.1.2 Quantitative Study Design
A large database maintained as part of the Colorado Mammography Project
(CMAP) was utilized in this research. The data set used for this study included
records of 167,232 ethnically diverse women who participated in screening
mammography during the 1990-1997 period.
3.1.2.1 The CMAP, Its Goals and This Research
The CMAP is an on-going community-based surveillance and research
project. It collects breast cancer (BC) screening data through cooperative
partnerships with mammography clinics and radiologists in the Denver Metropolitan
area. The project was initiated in 1988 at the Colorado Department of Public Health
and Environment (CDPHE). In 1997, the AMC Cancer Research Center partnered
with the CDPHE. Since 1994, as part of the National Breast Cancer Surveillance
51


Consortium, the CMAP project has been funded by a National Cancer Institute
(NCI) grant. Other members of the consortium are located in California,
Washington, New Hampshire, Vermont, New Mexico and North Carolina.
Data on more than 500,000 mammograms in Colorado have been collected
since 1989. Each year, the CMAP database, now maintained by the AMC Cancer
Research Center, increases by approximately 85,000 records. These data,
combined with data from the consortium, are used to conduct research that aims to
determine risk factors for BC and improve screening procedures. The ultimate goal
is to reduce BC mortality rates (CDPHE 2000).
Access to the CMAP data for the purpose of this study has been secured
with the support from Drs. Mark Dignan and Gary Cutter, co-investigators on CMAP
project (AMC) and from Dr. Jiilian Jacobellis, Director of EMS and Prevention
Programs Division of the CDPHE. Appropriate statements of confidentiality,
required by both institutions, have been signed prior to commencing this study.
Recently, in June 2001,1 was approached by the CMAP co-investigators, and
agreed to contribute on a regular basis to the project-community newsletter CMAP
Circular, in the form of Research Briefs. The initial brief, based on the findings of
the preliminary study, was published in August 2001 (Strzelczyk 2001).
A research protocol describing major elements of the study was submitted
to the University of Colorado at Denver Human Research Committee (UCD HRC);
it was reviewed using an expedited procedure and approved as non-exempt on
September 15, 2000. An extension was sought and the protocol was renewed for
another calendar year in August 2001. A copy of this protocol along with approval
letters is included in the Appendix.
3.1.2.2 Design of Quantitative Study
The underlying aim of the quantitative study was to examine whether socio-
demographic characteristics of screening participants, established through the
preliminary study, predict adherence to a follow-up on mammography, when such a
follow-up is recommended. Groups of interest were based on race/ethnicity, age,
52


educational level, health insurance type, and family history of BC status of subjects.
While it was anticipated that established predictors of participation in
screening might predict adherence behavior, the Exploratory Data Analysis (EDA)
philosophy guided this work. According to this philosophy, the data should be
explored without assumptions about probabilistic models, error distributions, or
relationships between variables for the purpose of discovering what they can tell us
about the phenomena we are investigating (Tukey 1977). The goal of such
quantitative detective work is to explore data in order to reveal patterns and
features that will enhance understanding or provide insights into underlying
processes. This philosophy determined the strategy that was implemented in the
analyses of data.
Comparative analyses attempted to identify possible patterns in adherence
behavior by answering specific research questions. Some of the basic research
questions were:
1) Are there differences in the adherence behavior among various groups of
women; if so, are they significant? (part of aim 1, described in Chapter 1, Section
1.1.3, pp. 3-4)
Experimental probabilities of adherence were calculated for the overall sample and
for each category within each group of interest, by dividing the number of women
who adhered by the number of women within that category who received
recommendations to follow-up. These proportions were compared with the
proportion of specified reference groups, i.e., the proportions for each racial/ethnic
group and the combined minority (Non-white) group were compared with that of
White women, the proportion of older (50 years of age and over) women was
compared with that of the younger (<50 years of age) group.
2) Are there differences in the proportions of women who receive follow-up
recommendations? (part of aim 1, pp. 3-4)
The technique described in 1) was applied to assess differential recommendations.
Proportions were derived at by dividing the number of women that received follow-
up recommendations by the number of screening participants in each category in
53


each group of interest, and compared with appropriate reference groups.
3) Is the result of mammography (severity of the finding) related to the rate
of adherence among various groups; if so, how significant are these differences?
(part of aims 2, 3 and 4, described in Chapter 1, Section 1.1.3, pp. 4-5)
The cohort of women whose mammograms were possibly abnormal was examined
by stratifying results by their severity: possibly benign, suspicious abnormality, and
highly suspicious of malignancy. Two-way (horizontal & vertical) comparisons of
proportions of adherent women in each category of abnormality, and for each
category of socio-demographic characteristic within groups of interest were done.
4) Does the invasiveness of a follow-up procedure play a role in adherence
behavior? (part of aim 2, p.4)
To assess this relationship, the data for the cohort with possibly abnormal
mammograms were stratified by grouping follow-up diagnostic procedures in two
categories based on the level of their invasiveness, i.e., an ultrasound was
considered a non-invasive procedure, a biopsy was placed in an invasive type
group. Comparisons of proportions were again performed in a two-way fashion
taking into consideration the invasiveness of follow-up, and each category of socio-
demographic characteristic within groups of interest.
5) Are there differences in the timeliness of adherence among various
groups, when the mammography finding suggests a malignancy; if so, are these
differences significant? (part of aim 4, p. 5)
To assess differences among groups, time intervals (means and medians) that
elapsed between index mammograms and respective (first) follow-up diagnostic
procedures were calculated for each category within each group of interest and
compared with reference groups.
A retrospective cohort design was employed for aims 1 through 4 of this
study and a documentary-historical style was chosen. The general purpose was to
examine (identify and describe) socio-demographic predictors of screening
[Predisposing Factors in the Adherence to Follow-up Model\, as predictors of
adherence behavior [Modifying Factors in the Adherence to Follow-up Model].
54


In addition, research aims 1 though 4 considered external clinical factors
[Critical Event in the Adherence to Follow-up Modet\, and personal clinical factors
[Modifying Factors in the Adherence to Follow-up Model] as possible predictors of
adherence behavior. Adherence rates and timeliness were the outcome measures
of behavior [Action in the Adherence to Follow-up Model]. These aims also had a
statistical explanation-generation purpose.
It was anticipated that findings of the quantitative study would suggest
characteristics of subjects intended to be explored further using qualitative
approach. It was also hoped that results of this part of the study would point to
productive areas of future research.
3.1.3 Qualitative Study Design
A qualitative approach was used in order to explore the range of
psychosocial, cognitive, and cultural factors that may facilitate or deter womens
adherence to recommended follow-ups.
Subjects were drawn from the following sources: an urban breast care clinic
collaborating with the CMAP project, breast cancer initiative activities (meetings of
help groups for family members of BC patients, breast cancer survivor meetings,
the Avon Breast Cancer Walk), and contacts made at community meetings and
health fairs offering free-of-charge clinical breast exams. While, admittedly, this was
a selective sample, all efforts were made to capture a diversity of life/event stories.
Women were asked to participate in focused semi-structured interviews.
This study had an exploratory aim. In the search of possible factors
influencing adherence behavior, the study was designed to explore the meanings,
understandings, and experiences of women as they relate to health and health
maintenance, BC screening, mammography, abnormal exam results, and follow-up
diagnostic procedures.
While the questions that study participants were asked, derived from the
Adherence to Follow-up Model, followed the general structure of the research
instrument (UCD HRC protocol), there were diversions in the order and formulation
55


of these questions. This approach was meant to make the questions more open-
ended and allow women talk freely about their experiences. These questions
served as a template or a priori codes in the semi-structured setting but the inquiry
involved the investigation of selected and well-defined spheres of activity. Two
events in women's lives received focused attention: their first mammograms (an
abnormal result being a Critical Event) and recommended follow-up exams.
Such a study design sometimes requires additional, a posteriori coding,
particularly when an interpretive evaluation of results involves a certain level of
generalization. As this project involved a question-specific research design, which
remained basically the same throughout the interviewing process, there was no
need for a posteriori coding. The descriptions of events were configured into each
woman's story constituting a narrative analysis. People stories are essentially
expressions of their self-understanding (Polkinghome 1988), life-event stories are
thus already interpretations of what the lives/events are about (Polkinghome 1995).
The qualitative study (aim 5, described in Chapter 1, Section 1.1.3 p. 6) with
an exploratory purpose was performed in a field inquiry style. Basic to the tenets of
this approach is the notion that (Bogdan & Bilken 1992)...
human beings are actively engaged in creating their world...
people act, not on the basis of predetermined responses to
predefined object, but rather as interpreting, defining, symbolic
animals whose behavior can only be understood by having the
researcher enter into the defining process (p. 36).
As Schwandt (1994) stated, central to this paradigm is the goal of understanding
the complex world of lived experience from the point of view of those that live it (p.
118).
Interviews of 14 subjects considered numerous factors, which were grouped
in the following categories: cognitive/psychological, cultural/social support,
perceptions of abnormality (internal), cognitive/cultural, having general means, and
psychosocial factors. Cognitive/psychological and cultural/social support categories
of factors were examined as possible predictors of participation in BC screening
(initial and subsequent) [Predisposing Factors in the Adherence to Follow-up].
56


Subjects' perceptions of abnormality [Critical Even(\ were the primary focus in
efforts to assess the range of meanings and understandings of both the severity of
mammography results and the invasiveness of a diagnostic follow-up procedure.
Cognitive/cultural factors were again examined as possible facilitators of or barriers
to adherence following the critical event [Modifying Factors in the Adherence to
Follow-up Model\. Having general means to complete a follow-up and psycho-
social factors were examined as those that may possibly influence an intention to
adhere to recommended follow-up [Likelihood of Action in the Adherence to Follow-
up Model].
Design decisions have to consider what one plans to study, under what
circumstances, for what duration of time, and with whom (Denzin & Lincoln 1998). It
is the question being asked [that] determines the appropriate research
architecture, strategies and tactics to be used not tradition, authority, experts,
paradigms, or schools of thought (Sackett & Wennberg 1997 p. 1636). This
research sought to answer the question: What factors are associated with delay in
recommended follow-up after abnormal mammograms among women of different
racial/ethnic groups?
3.2 Research Methodology
Regardless of theoretical orientation that a researcher holds, a sound mix of
qualitative and quantitative data is preferred in any study of human thought and
behavior (Goldenberg 1992). The concurrent use of qualitative and quantitative
methods can serve to cross-validate data collected by differing techniques.
3.2.1 Quantitative Methods
The CMAP data set used in this study consists of records of 167,232
women who participated in screening mammography and who self-reported their
race/ethnicity as: White (82.79%, Black (3.34%, Hispanic (11.04%), Asian (1.56%),
Native American (0.58%) or "Other (0.69%). The preliminary study utilized
screening data for this multiethnic sample of Colorado women.
57


3.2.1.1 Sample for Quantitative Analysis
The first mammograms (within CMAP system) of these women were
tracked in the database to identify a cohort (n=17,358) of women who received
follow-up recommendations. Further tracking of those cases allowed identification
of a cohort of women (n=14,001) who adhered to recommended follow-up within 12
months (aim 1). Inquiry about the results of first mammographic exams of these
women (aim 2) established a sub-cohort of women with possibly abnormal
mammograms (n=13,972). In this sub-cohort, 10,252 women had a probably
benign [BI-RADS category 3R] result (examined in aim 3), 3,279 had a
suspicious abnormality [BI-RADS category 4] and 441 had a result, which was
highly suggestive of malignancy [BI-RADS category 5"]. Data for women with the
severe results were examined in aim 4.
Among women with possibly abnormal mammograms, 12,053 were
recommended to have a non-invasive follow-up, and 1,665 to have an invasive type
of diagnostic procedure (examined in aim 2). Follow-up procedure type was not
specified for 139 women; these data were excluded from analyses pertinent to the
invasiveness of follow-up.
3.2.1.2 Data Collection
The raw data necessary for quantitative analyses were previously collected
as part the CMAP project. A CMAP History Form is included in the Appendix.
Before these data could be used for this research, it was necessary to assure that
the data set was clean. This was accomplished in collaboration with the AMC
Cancer Research Centers data managers.
The original data set for the 1990-1997 period consisted of records of about
218,000 women. The examination of these records revealed that information on
race/ethnicity was not available for about 40,000 of these women; a closer
examination of this group indicated that only a small portion of these records had
other socio-demographic data. As race/ethnicity is one of the dominant variables in
this study, a decision made to exclude these records as non-informative. The
58


investigation of records of women who self-reported their race/ethnicity revealed
that there were duplicate records for those who, in addition to their Hispanic origin,
reported being White, Black or other race. To assure that there was one record per
subject, the Hispanic group in this data set was comprised of women who reported
only Hispanic origin. These approaches reduced the study sample to about
170.000 unique subjects.
Many of these women obtained multiple mammograms within the CMAP
system, ranging from one to nine. This data set contains information on more than
400.000 mammograms. When data on recommendations to follow-up on these
mammograms were examined, it was noted that there were multiple follow-ups with
some of them appearing to have been occurred at the dates preceding
mammograms. A closer examination of these cases indicated that some of the
mammograms, which resulted in follow-up recommendations, were actually follow-
ups. It was not possible to determine which mammographic exam was followed-up
on. To make sense of these data, it was decided to consider only the first
mammograms and limit the scope to asymptomatic cases. This approach reduced
the sample to 167,232 unique women and 167,232 screening mammograms they
underwent. Additionally, only one, the first diagnostic follow-up procedure was
considered in the assessment of rates and the timeliness of adherence.
3.2.1.3 Quantitative Measures
3.2.1.3.1 Dependent Variables
Adherence to recommended follow-up is the dependent variable assessed
in this study. This variable was measured in two ways: as categorical, with two
possible outcomes, YES and NO, and as a continuous variable that reflected actual
time interval, in days, from index mammogram to the first diagnostic procedure.
These quantities are defined and measured as follows:
Level (rate) of adherence, /?,, is defined as the ratio (proportion) of the
number of subjects who returned for a follow-up exam, k,, and the total number of
59


subjects in the i-th group who received follow-up recommendation, n,; R, = k,n;1.
Rates of adherence were calculated for all groups of interest, separately for the
overall adherence, for the cohort with possibly abnormal mammograms, for the
sub-cohort with probably benign results, and for the sub-cohort whose
mammogram results indicated severe abnormality.
Adherence behavior (the YES category) can occur in a timely or untimely
manner. The recommended time interval between the first mammogram and the
next breast care visit depends on the category of abnormality. While there is no
uniform time frame within which a follow-up on a probably benign result (Bl-
RADS category "3") should be obtained, the procedure should generally be
obtained within 6 months and should certainly precede the next regularly scheduled
annual mammogram (Sickles 1991, Lewin 1999). Whether the result indicates a
suspicious abnormality or is highly suggestive of malignancy (BI-RADS
categories "4" and "5"), a follow-up should take place as soon as possible; in
practice, the target interval is 14 days (Kerlikowske 1996).
Timeliness of adherence is defined as the length of time (in days) that
elapsed between the index mammogram and the first diagnostic follow-up
procedure. This quantity has been assessed separately for category "3" and for
categories "4" and "5" combined.
For each group of interest, the timeliness was assessed by calculation of the
group's mean, the median, and the interquartile range (IQR). The initial approach of
calculating groups' mean times with their respective standard deviations was
expanded due to non-normal distributions of time intervals within groups. The
values of the median and the IQR allowed to observe the length of time intervals at
which 25%, 50%, and 75% of women in each group of interest obtained their
follow-up exams.
3.2.1.3.2 Independent Variables
Two groups of independent variables in this research were: socio-
demographic characteristics of subjects and clinical variables.
60


Socio-demographic variables (at the time of the first mammogram) included:
Race/ethnicity a categorical variable with six possible outcomes: White,
Black, Hispanic, Asian, Native American and Other. In some analyses, all minority
women were combined into a Non-white category.
Age the values of this continuous variable have been grouped at the time
of data collection converting this measure into a categorical variable with seven
possible outcomes : <=30, 30-39, 40-49, 50-59,60-69, 70-79, and =>80 (in years).
In most analyses, data were further grouped into two categories: younger women
(< 50 years of age), and older women (50 years of age and over).
Education values of this continuous variable were grouped at the time of
data collection. This categorical variable has 6 outcomes: <=8 grades, some High
School (HS), HS graduate, some college, college graduate, post-graduate
education.
In most analyses, data were grouped into two categories: less educated women
(HS education and less), and women with more formal education (>HS).
Health insurance status a categorical variable with four outcomes:
uninsured, assisted by Medicaid, assisted by Medicare, and group/private
insurance.
In most analyses, data were grouped into two categories: uninsured/underinsured
women and those with private/group health insurance.
Family history of BC (first degree relative) categorical variable measured
as YES or NO.
Clinical variables included:
Mammography Result American College of Radiology BI-RADS codes
(ACR 1998): 0" Need additional Imaging Evaluation, 1 Negative, 2" Benign
Finding, 3" Probably Benign Finding, 4" Suspicious Abnormality, and 5" -
Highly Suggestive of Malignancy were the outcomes of this categorical variable.
Mammograms with complete assessment codes "3", "4" or "5" comprised the
possibly abnormal category. Severity of abnormality increases with the increase in
the numerical value of the code.
61


Invasiveness of follow-up exam follow-up procedures are grouped in two
categories: INVASIVE and NON-INVASIVE. A biopsy, fine needle aspiration and
surgical consult were considered comprised an INVASIVE category. NON-
INVASIVE procedures included: ultrasound, diagnostic mammogram, short-term
follow-up (repeated mammogram), and a clinical breast exam (CBE).
3.2.1.4 Quantitative Analyses
SASr statistical package, version 8 (SAS 2000) was used to retrieve
appropriate data from the large CMAP database in the form of frequency
distributions of events in categories of interest. SASR and Microsoft Excel, version
2000 (Berk & Carey 2000) software packages were utilized in statistical analyses.
3.2.1.4.1 Descriptive Statistics
The preliminary phase of this study, the aim of which was to examine the
composition of screening cohort, involved calculation of proportions of women with
characteristics of interest, who participated in screening mammography during the
8-year study period. Only one record per participant was utilized; it included
information obtained at women's first mammographic exam. Consideration was
given to socio-demographic factors: race/ethnicity, age, level of education, health
insurance status, and family history of BC. The proportions of diverse women in
each socio-demographic category were calculated for the overall sample of
167,232 women. Additionally, these proportions were calculated for the samples in
the opening year (n=2,286), and the closing year (n=21,778) of the study.
Further examination of these data was done to investigate qualitative
interactions of race/ethnicity with each of the considered socio-demographic
characteristic. This approach is particularly useful when a directionality in main
effects (proportion of participants in the preliminary study or proportion of adherent
women in further research) is expected for that factor, racial/ethnic category in this
case, for different levels of other factors (Armitage & Berry 1994). As a directionality
in participation by women with characteristics of interest and among the adherent
62


was observed for the overall sample when race/ethnicity (in dichotomous mode),
age, education, health insurance status and FHBC were examined as unique
factors, it was anticipated that stratifying data within individual racial/ethnic groups
would provide additional insights. Proportions of older, less educated,
uninsured/undersinsured and those with family history of BC were calculated and
tabulated for each racial/ethnic group. The numerical values of these proportions
(experimental probabilities) were then compared with the proportions for specified
reference groups.
As described earlier in this Section, the multiethnic sample of women in the
adherence study consisted of those who received a follow-up recommendation at
their first screening. Proportions of women in each group of interest were calculated
to allow assessment of differences in recommendations received by various
groups. The same was done for the cohort of women who adhered to
recommended follow-up within 12 months; the rates of adherence were calculated
for each group of interest (aim 1). Additionally, proportions of older, less educated,
uninsured/underinsured, and women who reported family history of BC who did
adhere within 12 months, were calculated to facilitate assessment of qualitative
interactions of race/ethnicity with other socio-demographic characteristics among
adherent women (the approach developed when conducting the preliminary study).
The same descriptive statistics methods were applied for the sub-cohort
(aim 2) of women with possibly abnormal mammograms, as well as for two groups
in this sub-cohort: one, women who had probably benign result (examined in aim
3), and two, the combined group of those who had suspicious abnormality and
those who had a result, which was highly suggestive of malignancy (data for
women with the severe results examined in aim 4). Additionally, for these two
groups overall and for each category of socio-demographic characteristic,
timeliness of adherence was assessed by calculation of average time of response
(the mean and standard deviation) and the median time interval along with the IQR.
The sub-cohort of women with possibly abnormal mammograms consisted
of two groups with differing invasiveness of recommended follow-up (examined in
63


aim 2). The rates of adherence were calculated for women with characteristics of
interest in each of these two groups to assess the invasiveness of diagnostic
procedure as a factor that influenced follow-up.
Further quantitative analyses of these data were accomplished using inferential
statistics methods described below.
3.2.1.4.2 Inferential Statistics
The purpose of comparative analyses of proportions and rates was to test
the null hypotheses that proportions and rates observed for groups of interest do
not differ from appropriate reference groups' proportions and rates. These analyses
aimed to determine whether apparent numerical differences observed for various
groups of women are real or due to chance. In all comparisons of proportions and
rates, the normal approximation to a discrete distribution approach was utilized.
For all categories of proportions and rates calculated as described above
under the Descriptive Statistics heading, rate ratios (relative risk, RR) and their
95% confidence intervals (95% Cl) were determined utilizing Fisher exact test for
unpaired cases (Armitage & Berry 1994, Rosner 2000).
In the preliminary study, the proportion of women with a certain
characteristic (i.e. those 80 years of age and over) who participated in screening
mammography in 1990 was compared with a corresponding proportion for 1997.
The proportions of screening participants of each race/ethnicity possessing also
each of the characteristics determined as barriers to screening (older age, less
education, being uninsured/underinsured, or having family history of BC) were
compared to the proportions of White women possessing corresponding
characteristics.
In research on adherence, the analyses of rates proceeded on two tracks.
On the first track, the rates of adherence for each group of interest (experimental
probabilities of adherence by the group) were compared with the rate calculated for
an appropriate reference group, i.e., the rate for less educated (HS or less) women
was compared with that for the better-educated group. This was done separately
64


for the overall adherence (sample included all women who were recommended to
follow-up on their first mammogram), for the cohort with possibly abnormal
mammograms, for the sub-cohort with probably benign result, and for the sub-
cohort with two severe abnormality results combined. The second track of analyses
followed the approach developed in the preliminary study, which considered
interactions of race/ethnicity with other variables. The proportions of adherent
women of each race/ethnicity possessing also each of the characteristics as factors
that may have influenced obtaining first screening in this population (older age, less
education, being uninsured/underinsured, or having family history of BC) were
compared to the proportions of White women possessing corresponding
characteristics. Again, this was done for all cohorts and sub-cohorts of adherent
women. Additional analyses involved comparisons of proportions of adherent
(overall) women with proportions of screening participants in corresponding groups.
The differences in mammography results received by various groups of
women were also assessed, as were the differences in follow-up
recommendations, and differences in adherence levels depending on the type of
follow-up procedure for the first track of analyses. For the second track, the
proportions of adherent women whose mammograms were possibly abnormal were
compared with the proportions of all adherent (overall) women who received follow-
up recommendations, and proportions of adherent women with less severe
mammography results were compared with the proportions of adherent women
whose mammograms were abnormal.
In addition to adherence rates, the timeliness of response was evaluated for
women whose mammograms were in the less severe category of abnormality,
probably benign (aim 3) or in either of the two more severe categories, suspicious
abnormality or highly suggestive of malignancy (aim 4). Time intervals (days) from
index mammogram to the first follow-up diagnostic procedure were assessed for
various groups of women using descriptive statistics. Brief examination of the
means and corresponding very large standard deviations for each group of interest
revealed that time distributions were non-normal. Further examination of the
65


median times and IQRs revealed long right "tails" in the distributions (left
skewness) for all groups of interest.
Comparisons of such data are generally done using non-parametric
methods (Sheskin 1997, Rosner2000). Wilcoxon rank sum test with median option
was applied to compare time interval data distributions and median values for each
variable category, and to assess the level of significance of differences in
timeliness of follow-up observed for different groups. The results were tabulated
and served as a basis for drawing inferences and deriving explanations of observed
phenomena.
Additionally, another statistical approach was attempted for women with
severely abnormal results. A timely adherence to follow-up by these women is of
utmost importance and clinical significance. Associations between time interval and
independent categorical variables were assessed using Cox proportional hazard
models (Collett 1994, Themeau & Grambsch 2000). Time interval, the continuous
dependent variable, was expressed in days.
Separate models were developed for each independent dichotomous
variable: racial/ethnic group (Non-white and White), age (younger and older),
education (lesser and more), health insurance (uninsured/underinsured and
private/group), and family history of BC (Yes and No) as well as a multivariate
model that considered all socio-demographic variables together. Additional
multivariate models were also developed for individual racial/ethnic groups with
White women as the reference group. Results of this modeling of data were
summarized using hazard ratios (HR)s. Unlike in the previous analyses of
timeliness of follow-up by adherent women, proportional hazard models allow the
inclusion of censored cases, women who had the same socio-demographic
characteristics but who did not obtain a follow-up within 12 months. While the
inclusion of non-adherent women as censored cases removes some bias,
proportional hazard models did not appear advantageous when applied to this data
set. Comparison of statistical results obtained using these different methods is
presented in Chapter 4.
66


Substantial differences in the levels and timeliness of adherence among various
socio-demographic groups suggested that additional factors should be explored. It
was hoped that qualitative methods would complement the quantitative data and
provide insights into the less measurable predictors of adherence behavior.
3.2.2 Qualitative Methods
3.2.2.1 Sample for Qualitative Inquiry
The sample of women who agreed to participate in this study, and freely
and openly share their general health, health maintenance, and mammography-
related attitudes, consisted of 14 ethnically diverse women. Racial/ethnic
composition of the sample was as follows: four Black, four Hispanic, two Asian, one
Native American, and three White women. Their ages ranged from 40 to 74 with
equal proportions of younger (<50 years of age) and older women. The level of
education of these women at the time of their first mammogram ranged from having
HS or less (four women) to holding a post-graduate degree. There was only one
woman in the uninsured/underinsured group (assisted by Medicaid at the time of
her first mammogram); the remainder of the sample had private/group health
insurance. Due to the sources used to recruit women for this study, most had family
history of BC; three women were BC survivors, and one had a personal history of
cervical cancer.
The subjects were assured that their participation in the study was
voluntary, and that they could withdraw any time. Only two of the approached 16
women expressed their desire to withdraw in the initial stages of contact, citing lack
of time. Their wishes were honored and these women were excluded.
3.2.2.2 Data Collection
The data for the qualitative part of this research were obtained through in-
depth semi-structured interviews, with the consent of study subjects, as described
in the approved UCD HRC research protocol (Appendix). One-on-one interviews
67


were conducted with each subject at the urban mammography clinic following an
introductory conversation. To some extent, they resembled medical history-taking
interviews (Coulehan & Block). Some women preferred to respond to interview
questions in writing citing the necessity to verify the dates and details of specific
events. These women were given copies of the instrument and asked to return
them as soon as practical; all women returned the questions with their responses
(by mail, e-mail or in person). In two cases, when there was a discrepancy between
those responses and the notes taken during the initial interview, a telephone
conversation clarified discrepancies thus decreasing a recall bias. Each interview
lasted 30 to 45 minutes; the duration of additional telephone conversations was on
average 15 minutes. Some women have maintained telephone and/or e-mail
contact with me; these women were offered access to their own stories, as
constructed to fit the pattern of narrative reporting. This strengthened the validity of
collected data. Five women in this sample expressed their desire to know the
results of this research. Their wishes will be granted upon completion of this
project.
Confidentiality was maintained throughout the duration of this project.
Interview information has been stored on my personal computer and as a
conventional file, which was sealed after each use. Each woman in the study was
assigned a pseudonym.
3.2.2.3 Qualitative Approach
While the results of the quantitative part of this research identified
measurable predictors of adherence to recommended follow-up among various
groups of women, and allowed the assessment of significance of observed
differences, they did not explain them. These results, while providing a wealth of
information, and suggesting a role of qualitative interactions of race/ethnicity with
other socio-demographic as well as with clinical factors in adherence, did not give
an adequate response to the question that prompted this research.
Ethnographic methods were utilized in this phase of the study to leam about
68


meanings and understandings of health, health maintenance, and mammography-
related issues. The ultimate aim of ethnography is to describe culture in sufficient
detail so that the reader gains an understanding of what it might be like to be part of
that culture (Bernard 1995). Through hermeneutic/dialectical processes the
researcher and participants describe and interpret the reality of culture.
This study of culture concerns both process and content. Culture may be
defined from cognitive and/or behavioral perspectives. The behavioral perspective
sees it as observable patterns of behavior, customs, ways of life; the cognitive
perspective considers culture the sum of people's ideas, beliefs, and knowledge.
Fetterman (1989) believes that it is essential for ethnographers to understand both
perspectives. An ethnographic approach carries with it the following implications;
it acknowledges that we move within a social world, challenging
solipsistic or overly individualistic accounts,
it suggests that there is important knowledge, which can be gained in no
other way than picking things up though hanging around in the culture
[participant observation],
it suggests a certain openness to culture and learning from the setting,
where the ethnographer is prepared to see everything and suspend
premature judgment on what should be selected as data, and
the knowledge, accounts, worldviews, and lives of insiders, participants,
and native experts are given a high status (Denscombe 1995, p. 184).
3.2.2.4 Design Rationale
The task of ethnography is to describe and interpret culture, which requires
a holistic perspective, contextualization and emic, etic and non-judgmental views of
reality (Fetterman 1989, p. 29).
The holistic perspective of ethnography requires that the researcher spend
time within the culture of interest collecting data through multiple methods.
Achieving this perspective was accomplished through multiple methods; spending
time at an urban mammography clinic observing and informally interacting with
69


patients who come to obtain breast care, by participation in help groups and BC
initiatives, by social interactions with diverse women outside the health care setting,
and by in-depth interviewing of women.
Contextualization requires the researcher always to view collected data
through the prism of culture. No single piece of evidence can be examined in
isolation; the researcher must constantly connect data to the culture.
Contextualization of subjects' experiences was derived from the previously
mentioned activities as well as from considering the findings of quantitative
analyses.
Naturally, the researcher always brings an etic or outsider's perspective to
ethnography. This vantage point allows the researcher to examine the evidence
without being too eager [to avoid embellishing or even inventing patterns], to
remain skeptical, and not to go native (Miles & Huberman 1994, p. 216). The emic
or insider perspective, however, is the essence of ethnography. To remain true to
the goal of ethnography one must capture the emic perspective free of
preconceptions or interpretations.
Interpretive style research must be particularly concerned with the
trustworthiness of the findings. Questions about the true value, applicability, and
consistency of the findings need to be answered. The positivist paradigm utilizes
the labels validity, reliability, and objectivity to describe these criteria. Naturalistic,
field study such as this one, addresses trustworthiness of findings by means of the
criteria of credibility, transferability, dependability, and confirmability of the data and
their analysis (Lincoln & Guba, 1985, p. 290).
Credibility in this study was enhanced by the following techniques:
prolonged and culturally sensitive interactions with women in the clinic, at clinic-
sponsored help group meetings and at BC initiatives at various locations in the 6-
county area, intense informal (social) interactions with women of diverse cultures,
peer and committee member debriefing, triangulation, and negative case analysis.
The object of negative case analysis is to seek to include participants in the study
who have experiences that diverge from what has been previously documented
70


until deviant cases have been included in the analysis. In this study of women who
timely adhered to recommended follow-up, all efforts were made also to
recruit women who returned to the clinic in an untimely fashion as well as those
who did not return within a 12-month period.
Methodological triangulation (Denzin 1978) was utilized to strengthen the
findings. Triangulation is (Gall et al. 1996),
the process of using multiple data collection methods, data
sources, analysts or theories to check the validity of findings...
to eliminate biases that might result from relying exclusively
on one data collection method, source, analyst or theory (p. 574).
While in-depth interviews were the main source of data in this study, their semi-
structured format, along with the use of credibility enhancement techniques
described earlier, provided a richer ground for obtaining relevant data.
3.3.3 Reconciliation of Methods
Quantitative methods emphasize isolating the phenomenon under study
from its context. A qualitative approach, on the other hand, frequently emphasizes
the context of the phenomenon (Sanday 1983). Each approach has its own
strengths and weaknesses; methodological triangulation may take advantage of the
strength and minimize the weaknesses of each method.
Quantitative methods are usually used deductively. A theory or hypothesis
guides both testing and the interpretation of results. Qualitative methods are said to
be phenomenological. They stress searching for meaning from the study subject's
frame of reference. They are typically used inductively, in that observations may
lead to theory (Stange & Zyzanski 1989). These two paradigms, however, should
not be viewed dichotomously the actual process of understanding is more circular
than linear (Guba 1990). Observations lead to a theory, a model and hypotheses,
which lead to observations to test the hypotheses. This may lead to a modification
of the theory or model, and so on. A researcher may enter this circle at any point
and exit it when he/she has enough confidence in the observations or theory to
report the results.
71


The sequential use of quantitative and qualitative methods may draw
attention to the possibility of confounding as well as to selection or information
biases inherent in the application of the opposing methodology. Concurrently
obtained qualitative data may also serve to explain the results of a quantitative
study in a way that is meaningful to the population under study.
To some extent, quantitative and qualitative approaches to research
represent different data collection and analytic methodologies. However, as Stange
& Zyzanski (1989) see it, the two approaches can easily become paradigms or
ways of viewing the world. Researchers may find out that their research paradigm
influences not only how the research question is asked, but also what question is
asked, and how the answer to that question is followed with future studies.
Quantitative researchers, no doubt, speculate about the meaning behind their hard
facts, often basing this speculation more on their own personal frame of reference
than on knowledge grounded in the culture of their subjects. Likewise, qualitative
researchers are often tempted to "quantify" their findings and are drawn to attempt
generalization of these findings beyond the study group. The simultaneous use of
both research methods provides insights into a better understanding of meanings
along with hard facts. Such a strategy may yield higher quality results without
unreasonable additional effort. I found this strategy to be an efficient way to
advance understanding of the phenomenon under study.
72


4. Quantitative Data Analysis and Results
This chapter begins with an introductory Section 4.1, which examines socio-
demographic characteristics of the multiethnic sample of Colorado women who
participated in screening mammography within the Colorado Mammography Project
(CMAP) system from 1990-1997.
Section 4.2 contains results of tracking of the first (index) mammograms of
these women for the purpose of establishing the characteristics of the cohort that
received follow-up recommendations, and examination of the overall adherence to
follow-up within 12 months of the index mammogram. The data on
recommendations and on the overall adherence are presented side-by-side for
easy comparisons of proportions of women in categories of interest. The analyses
involve examination of factors, identified earlier as possible predictors of
participation in screening, that may determine which groups of women are more
likely to receive follow-up recommendations, and which groups are more likely to
adhere.
In Section 4.3, the attention focuses on adherence levels (rates) observed
for women whose first mammograms indicated possible abnormality. The
invasiveness of recommended follow-up procedure and its effect on adherence is
also considered. These two clinical variables, mammography result and the type of
a follow-up procedure, are examined as unique factors that may play a role in
adherence to recommendations, and in interactions with socio-demographic
characteristics of women to influence adherence.
The next section contains data and results of analyses involving cases of
the least severe category of abnormality probably benign screening
mammography result (BI-RADS category 3). In addition to adherence rates
observed for various groups, timeliness of a follow-up is considered.
73


Section 4.5 focuses on adherence to follow-up on abnormal mammograms
(BI-RADS category 4 and 5). Both aspects of adherence are examined:
differential rates and the timeliness. Comparative analyses of these data aim to
identify characteristics of women who are at the highest risk of not obtaining the
recommended follow-up in a timely fashion.
Section 4.6, the last in this chapter, contains a summary of findings and a
review of strengths and limitations of the approaches utilized in this quantitative
study.
4.1. Participation in Screening Mammography:
Socio-demographic Characteristics of the Sample
The data set for this study consists of records of 167,232 screening
participants who self-reported their race/ethnicity as: White (82.79%), Black
(3.34%), Hispanic (11.04%), Asian (1.56%), Native American (0.58%) or Other
(0.69%). This racial/ethnic composition reflects relatively well population estimates
for the state of Colorado as well as for the 6-county Denver metropolitan area
during the study period1. In addition to self-reported race/ethnicity, socio-
demographic characteristics of the participating women considered in this study
include: age, education, type of health insurance, and family (first degree relative)
history of breast cancer (FHBC) at the time of the first mammographic screening
within CMAP project. No income information was collected as part of the project -
education and health insurance status serve as estimators for socio-economic
status.
Low utilization of screening mammography by certain groups of women has
been an issue extensively explored in the literature in the past 20 years. Recent
74
1 From "Population Estimates for Counties by Race and Hispanic origin"
rwww.census.Qov/). as updated in August 2000, Colorado population composition was as
follows: 80.71% White (non-Hispanic), 4.14% Black, 12.32% Hispanic, 1.89% Asian/Pacific
Islander, and 0.95% American Indian/Alaska Native in 1990. The corresponding proportions
for 1997 were: 79.04%, 4.29%, 13.42%, 2.33%, and 0.93%. These estimates calculated for
the 6-county area were: 79.73%, 5.41%, 11.60%, 2.40%, and 0.86% for 1990; 78.05%,
5.55%, 12.63%, 2.94%, and 0.82% for 1997.


literature, which reviews 1980s screening data, indicates that older (>50 years of
age) women, those of low socio-economic status, as well as racial/ethnic minority
women were among the underserved groups (Champion 1994, Farley & Flannery
1993, Lacey 1993, Romans 1993). Preliminary work with the data set for this study
indicates that there may be some positive changes in the participation of these
vulnerable groups in the 1990s (Strzelczyk & Dignan 2000).
4.1.1 Dynamic Character of Participation
by Various Groups
Between 1990 and 1997, the opening and the ending years of the study
period, the number of women who had their first screening mammogram within the
CMAP system of clinics increased almost ten-fold (Table 4.1). A number of
changes in the composition of screening participants occurred during this period.
4.1.1.1 Race/ethnicity
While a large increase (>30 fold) occurred in the overall number of Non-
white women who obtained their first mammograms in respective years, and their
proportion among screening participants in respective grew from 10.63% to
39.65%, the changes did not occur uniformly for all racial/ethnic groups. The largest
increase in the proportion in respective years was observed for Hispanic women
(5.42% to 34.20%); the proportion of Native American women more than doubled,
0.26% to 0.64% and an increase also occurred for Asian women (1.27% to 1.76%);
these changes were statistically significant for Hispanic and Native American
women. While some of these increases, particularly for Hispanic and Asian women,
are reflective of changes in the composition of Colorado and 6-county population
(footnote 1), the significant increase in the participation of Native American women
appears to be due to other factors.
75


TABLE 4.1 Changes in the composition of screening cohort in the opening and the dosing
years of the study period. Relative Ratio of corresponding proportions, RR, and 95% C.l.
Number and percentage of partiapants RR (95% C.l.)
Partiapants characteristics 1990 1997
Overall sample 2,286 21,778
Race/ethnidty
Black 57 (2.49%) 505 (2.32%) 0.93 (0.71,1.21)
Hispanic 124 (5.42%) 7,447 (34.20%)b 6.30 (5.31,7.48)
Asian 29 (1.27%) 384 (1.76%) 1.30 (0.96,2.02)
Native American 6 (0.26%) 139 (0.64%)b 2.43 (1.07,5.50)
Other 27 (1.18%) 161 (0.74%)b 0.62 (0.42,0.94)
Overall Non-white 243 (10.93%) 8,636 (39.65%)b 3.73 (3.31,4.21)
White 2,043 (89.37%) 13,142(60.35%)b 0.68 (0.66,0.69)
Age (years of age)a
<30 11 (0.48%) 96 (0.44%) 0.91 (0.49.1.71)
30-39 385 (16.84%) 3,168 (14.57%)b 0.87 (0.78,0.95)
40-49 797 (34.86%) 8,825 (40.59%)b 1.16(1.10,1.23)
Overall <50 1,193 (52.19%) 12,089(55.60%) 1.07(1.02,1.10)
50-59 525 (22.97%) 4,775 (21.96%) 0.96 (0.88,1.03)
60-69 370 (16.19%) 2,698 (12.41%) 0.77 (0.69,0.85)
70-79 169 (7.39%) 1,706 (7.85%) 1.06 (0.91,1.24)
80 & over 29 (1.27%) 476 (2.19%) 1.73 (1.19,2.50)
Overall 50 & over 1,093 (47.81%) 9,655 (44.40%) 0.93 (0.89,0.97)
Education a
<8 grades 35 (1.54%) 515 (2.65%) 1.72(1.23,2.41)
some HS 80 (3.52%) 602 (3.10%) 0.88(0.70.1.11)
HS graduate 544 (23.92%) 4,650 (23.94%) 1.00 (0.93,1.08)
Overall HS or less 659 (28.98%) 5,767 (29.69%) 1.02(0.96,1.10)
Some college 687 (30.21%) 6,857 (35.30%) 1.17(1.09,1.25)
College graduate 781 (34.34%) 4,717 (24.28%) 0.63 (0.60,0.67)
Postgraduate education 147 (6.46%) 2,086 (10.74%) 1.66(1.41,1.95)
Overall >HS 1.615(71.02%) 13,660 (70.31%) 0.99 (0.96,1.02)
Health insurance a
None Medicaid Medicare Overall Un/Under-insured Private/group 332 (1.90%) 326 (1.87%) 1,860 (10.68%) 2,518 (14.45%) 14,913 (85.55%)
FHBC a
Yes 361 (15.79%) 2,330(11.37%) 0.72 (0.65,0.80)
No 1,925 (84.21%) 18,163(88.63%) 1.05(1.03,1.07)
" Missing data excluded from calculations of respective proportions: Age n=34 in 1997; Education n=12 in 1990
and n=2,351 in 1997; Insurance information not available for 1990, missing for n=4,347 in 1997; FHBC status
unknown for n=1,285 in 1997.
b Proportion in 1997 significantly different from the corresponding one in 1990 (Fisher exact test for unpaired
cases)
76


4.1.1.2 Age
The overall number of younger (<50 years of age) women who obtained
their first mammograms in respective years increased more than ten-fold and their
proportion among screening participants grew from 52.19% to 55.60%. The
numerical increase for women 50 years of age and older was slightly lower (8.8
fold) and their proportion of the total sample decreased from 47.81% to 44.40%,
comparing 1990 with 1997. The changes in proportions for both age groups were
statistically significant. In the younger group, the largest increase occurred for
women aged 40-49, from 34.86% to 40.59% but the proportion of women aged 30-
39 decreased significantly. In the older group, the proportions of women 50-59 and
60-69 decreased, 22.97% to 21.96% and 16.19% to 12.41%, respectively and the
decrease was significant for the latter group. An increase occurred in the proportion
of elderly participants from 7.39% to 7.85% for the group 70-79, and from 1.27% to
2.19% for those 80 years of age and over; the increase was significant for the
eldest group.
A decrease was observed in the proportion of the youngest women (<40
years of age) who obtained their first mammograms in the respective years; women
in this age group are generally not recommended to participate in mammographic
screening unless they are at an increased risk of developing breast cancer. This
decrease occurred in parallel with a substantial increase in the proportion of women
aged 40-49. There was an increase in the elderly group (over 70 years of age) of
women who obtained their first mammogram, with a parallel decrease for women
50-69. Along with demographic dynamics (aging population), other factors such as
changes and inconsistencies in screening recommendations as well as their
understanding by women (Wolosin 2000) may have influenced some of these
changes.
4.1.1.3 Education
A slight increase, from 28.98% to 29.69%, was observed in the proportion of
women with HS or less formal education, and corresponding decrease, 71.02% to
77


70.31% for women with more than HS. The proportion of women whose reported
educational level was less than grade 8 almost doubled, 1.54% to 2.65% (along
with 14.7-fold numerical increase). Large and statistically significant increases also
occurred for women with some college and those with postgraduate education,
30.30% to 35.30% and 6.46% to 10.74%, respectively, along with significant
decrease, 34.34% to 24.28%, for college graduates. A large increase in the number
of women with the least of formal education who obtained their first mammograms
offers hope that perhaps outreach efforts have been more effective in reaching
these women.
4.1.1.4 Health Insurance
Because insurance information was not available for the initial years of the
CMAP project, including 1990, no comparison of screening data for 1997 could be
made for this characteristic.
4.1.1.5 Family History of Breast Cancer (FHBC)
A decrease from 15.79% in 1990 to 11.37% in 1997 was noted in the
proportion of women reporting a FHBC among those who obtained their first
screening. This again may reflect debates that occurred in early 1990s and led to
changes in mammography recommendations (NIH 1997, Wolosin 2000), which call
for women at an increased BC risk, with FHBC being one of those risks, to have
diagnostic rather than screening mammograms.
4.1.1.6 Summary of the Preliminary Findings
The findings of the preliminary study, which considered each of the socio-
demographic characteristics as a unique factor, which may have influenced
screening behavior, are in good agreement with previous studies that suggested
lower participation of racial/ethnic minority women (Champion 1994, Farley &
Flannery 1993, Lacey 1993, Romans 1993). Non-white women comprised 17.2%
78


of this multiethnic sample of women who obtained their first mammograms. Other
barriers to screening identified in the cited studies, older age and lesser education,
were only to some extent reflected in the findings of this preliminary work. The
proportion of older women among those who obtained their first mammogram
decreased but the proportion of those with HS or less education slightly increased
comparing 1990 and 1997. A decrease was observed in the proportion of women
who reported FHBC.
4.1.2 Interactions of Race/ethnicity
with Other Characteristics
For some of these groups of women, a diagnosis of possible BC has more
severe consequences than for others. Black women in particular tend to receive
later stage diagnoses (ACS 2000), which in turn may lead to higher mortality rates
from BC than those observed for White women (Zavertnik 1993). Large differences
in BC survival continue to persist among racial/ethnic groups (Ries et al. 2000).
National statistics indicate that older women who are at the greatest risk of
BC have the lowest screening rates (Burg et al. 1990, Taylor et al. 1998). It is
possible that race/ethnicity interacts with age as well as with other socio-
demographic factors to influence participation in screening.
To examine these qualitative interactions, the data on 167,232 screening
participants were stratified by each of the socio-demographic characteristics within
each racial/ethnic group. Although health insurance data were missing for the first
years of the CMAP project, they were reliably present for about one-third of
screening participants (n=59,533). Based on the breast cancer screening literature
(Farley & Flannery 1993, Zavertnik 1993), it seemed appropriate to consider the
uninsured/underinsured group as one for which the proportion among screening
participants would be lower when compared with women who had private or group
insurance.
Results of this stratification and comparisons are presented in Table 4.2.
There were numerous significant differences in the proportions of women with
79


TABLE 4.2. Comparison of proportions (%) of screening participants possessing characteristics
of interest within radal/ethnic groups with the proportions of White women with corresponding
characteristics, RR (95% Cl).
Characteristics within radal/ethnic groups
Race/ethnicity Older Age1 Lesser Education2 Uninsured/ Underinsured 3 FHBC4
Sample
n=167,232 Black 47.85% 32.10% 19.01% 13.62%
n=5,582 48.07% 36.61% a 22.02% a 11.89% a
Hispanic 0.99 (0.97,1.02) 1.24(1.20,1.29) 1.18(1.08,1.30) 0.84 (0.78,0.90)
n=18,461 45.34% a 50.97% a 19.13% 10.93% a
Asian 0.94 (0.92,0.95) 1.72 (1.69,1.75) 1.02(0.98,1.08) 0.77 (0.74,0.81)
n=2,612 43.42% a 41.81%a 28.99% 8 7.77% a
Native Am 0.90 (0.86,0.94) 1.42(1.36,1.49) 1.56(1.42,1.72) 0.55 (0.48,0.63)
n=968 40.12% 38.50% a 23.34% a 12.31%
Other 0.83 (0.77,0.90) 1.31(1.21,1.42) 1.26(1.04,1.52) 0.87 (0.73,1.03)
n=1,150 44.91% a 41.17%a 25.89% * 12.72%
Non-white 0.93 (0.87,0.99) 1.40(1.30,1.50) 1.39(1.18,1.65) 0.90 (0.77,1.05)
n=28,773 45.51% a 46.03% a 20.74% a 10.95% a
White 0.92 (0.87,0.99) 1.56 (1.54,1.59) 1.12(1.07,1.16) 0.77 (0.75,0.80)
n=138,459 48.34% 29.44% 18.58% 14.17%
1 1 1 1
Partidpants excluded from calculations due to missing information: 1: age n=48; 2:education n=4875;
3: health insurance n=107,699; 4: FHBC status n=5068
a Proportion statistically different (p<0.05) from the proportion for White women (Fisher exact test for unpaired
cases).
characteristics of interest as compared with proportions observed for reference
groups.
4.1.2.1 Interaction of Race/ethnicity and Age
The overall proportion of younger women among screening participants in
this sample was significantly higher than the proportion of older women, 52.15%
versus 47.85%, (RR = 1.09, 95% Cl: 1.08,1.10). The proportions of older
80


participants within racial/ethnic groups ranged from 40.12% for Native American to
48.34% for White women, with 45.51% for the overall Non-white group. Except for
older Black women, whose proportion (48.07%) was close to that observed for
White women, the proportions of older Non-white women overall and in each
racial/ethnic group were significantly lower than the proportion of older White
women.
4.1.2.2 Interaction of Race/ethnicity and Education
About two-thirds of all screening participants had more than HS education,
67.90% versus 32.10% of those with HS or less education (RR = 2.11, Cl: 2.09,
2.13). The proportion of less educated participants ranged from 29.44% for White
to 50.97% for Hispanic women, with 46.03% for the Non-white group. The
proportions of less educated Non-white women overall, and in each racial/ethnic
group, were significantly higher than the proportion of White women with HS
education or less.
4.1.2.3 Interaction of Race/ethnicity and Insurance
In the subgroup of 59,533 participants for whom health insurance
information was available, 19.01% of women were either uninsured or
underinsured. This proportion ranged from 18.58% for White to 28.99% for Asian
women, with 20.74% for the overall Non-white group. Except for Hispanic women,
whose proportion in this category (19.13%) was dose to that of White women, the
proportions of uninsured/underinsured Non-white women overall, and in each
radal/ethnic group, were significantly higher than the proportion of
uninsured/underinsured White women.
4.1.2.4 Interaction of Race/ethnicity and FHBC
Overall, 13.62% of women participating in screening reported FHBC; the
proportions ranged from 7.77% for Asian to 14.17% for White women, with 10.95%
for Non-white women combined. Proportions of Non-white women with FHBC
81


overall, and for each racial/ethnic group, were significantly lower than the
proportion of White women reporting FHBC. Two exceptions were observed:
proportions of Native American and of "Other" women reporting FHBC, although
lower, were not significantly different from that of White women.
4.1.3 Section Summary and Discussion
With some exceptions noted for individual racial/ethnic groups, the
proportions of older women and those with FHBC among Non-white women who
obtained their first mammogram were significantly lower than respective
proportions for the White group. The reverse was observed for women with lesser
education and the uninsured/underinsured in this sample.
Among Non-white screening participants, Black women were the only group
whose proportion among screening participants did not increase over time. While
there was an increase in the Black population in Colorado in the 6-county area
comparing 1990 and 1997 (footnote 1), the proportion of screening participants
decreased from 2.49% to 2.32% (Table 4.1). This group was again an exception
among Non-white women in that it had the highest proportion of older participants
(48.07%), and the lowest proportion of those with HS education or less. Proportions
of Black women who were uninsured/underinsured, and who reported FHBC were
not different from those observed for the overall Non-white group (Table 4.2).
The largest increase in the proportion of women who obtained their first
mammogram occurred for Hispanic women (5.42% to 34.20%); this trend reflects to
some extent changes in the state and 6-county population. The proportion of older
Hispanic women was significantly lower than the proportion of older women in the
White group (Table 4.2) and this ethnic group had the highest proportion of the less
educated (50.97%). Proportions of uninsured/underinsured Hispanic women and
those with FHBC did not differ from those observed for the Non-white group.
An increase in the proportion of Asian women who obtained their screening
followed the direction of change in the Colorado and 6-county Asian/Pacific
Islander population during this study period. The proportions of older or less
82


educated Asian women did not differ from those observed for the overall Non-white
group. However, Asian women had the highest among all racial/ethnic groups
proportion of the uninsured/underinsured (28.99%). Their proportion of those with
FHBC was the lowest; there is a much lower incidence of BC among Asian women
(Ries et al. 2000).
The proportion of Native American women who obtained their first screening
mammograms more than doubled between 1990 and 1997 while the state and 6-
county American Indian/Alaskan Native population remained level (footnote 1). The
proportion of older Native American women was the lowest among all racial/ethnic
groups in this sample, and the proportion of those less educated was the second
lowest after Black women. The proportions of Native American participants who
were uninsured/underinsured and those who had FHBC were higher than those
observed for the overall Non-white group (Table 4.2).
The "Other" category included women that did not feel they belonged to any
of the specified racial/ethnic groups. Their participation in screening decreased
over time. No comparison with the directionality of Colorado or 6-county population
change can be made since there is no "Other" category in the census estimates
(footnote 1). The characteristics of participants in this group did not differ from the
overall Non-white group except for those uninsured/underinsured; the proportion of
"Other" women in this category was the second highest after Asian women.
4.2 Follow-up Recommendations and Overall
Adherence: Socio-Demographic Factors
Typically, 6-18% of women participating in screening mammography are
required to undergo some type of further examination (Sickles 1991). In this
sample, 10.4% of women (n=17,358) received follow-up recommendations at their
first screening within the CMAP system.
4.2.1 Differences in Follow-up Recommendations
The examination of recommendation data for White and Non-white
83


screening participants indicates that the proportions of women who received follow-
up recommendations were the same for both groups (Table 4.3).
Table 4.3. Socio-demographic characteristics of screening participants, of women who received
follow-up (F/U) recommendations, and women who adhered to recommendations within 12 months
Screened
Socio-demographic category N
Overall 167,232
Race/ethnidty Blade Hispanic Asian Native American Other Non-white 5,582 18,461 2,612 968 1,150 28,773
White 138,459
Age1 Younger 87,180
Older 80,004
Education2 Less 52,117
More 110,240
Insurance3 Uninsured/ Underinsured 11,319
Private/group FHBC* * 48,214
Yes 22,081
No 140,083
Recommended F/U N, (% of N) Adhered to F/U N2 (% of Ni)
17,358 (10.4%) 14,001 (80.7%)
567 (10.2%) 444 (78.3%)*
2,000 (10.8%) 1,467 (73.3%)*
186 (7.1%)* 126 (67.7%)
110 (11.4%) 78 (70.9%)
124 (10.8%) 92 (74.2%)
2,987 (10.4%) 2,207 (73.9%)
14,371 (10.4%) 11,791 (82.1%)
9,527 (10.9%)" 7,519 (78.9%)
7,825 (9.8%) 6,480 (82.8%)
5,658 (10.9%) 4,469 (79.0%)*
11,229 (10.2%) 9,193 (81.9%)
1,148 (10.1%) 832 (72.5%)
4,925 (10.2%) 3,635 (73.8%)
2,257 (10.2%) 1,981 (87.8%)
14,557 (10.4%) 11,684 (80.3%)
1,2,3,4: Excluded from calculations due to missing information: 1: n=48,2: n=4,875, 3: n=107,699,4: n=5,068
for screening participants; 1: n=6,2:471,3: n=11,285,4: n=544 for cohort that received follow-up
recommendations; and 1: n=2, 2: n=339, 3:9534; 4: n=336 for the cohort of adherent women.
* Proportion significantly different from the reference group at p<0.05 (Fisher exact test for unpaired cases)
84


However, these proportions differed for individual racial/ethnic groups, ranging from
7.1% for Asian to 11.4% for Native American women; the former being significantly
lower than the proportion of White women (10.4%) who were recommended to
follow-up, RR=0.68 (95% Cl: 0.60, 0.79). While receiving a follow-up
recommendation does not necessarily mean that BC is suspected, it is interesting
to see how the recommendations received by women in this study correspond to
BC incidence rates among diverse American women during this study period.
According to the latest report on cancer rates in the U.S. (Ries et al. 2000),
American Indian/Alaskan Native women experienced the largest increase in BC
incidence while Asian/Pacific Islander women continued to enjoy the lowest rates
during 1990-1997.
A significantly larger proportion of younger (<50 years of age), as compared
with older women, received follow-up recommendations, RR=1.12 (95% Cl: 1.09,
1.15). Less educated women (HS or less) also appeared slightly more likely than
their more educated counterparts to receive follow-up recommendation, RR=1.07
(95% Cl: 1.00, 1.10). The difference in the proportion of uninsured/underinsured
women versus those with private/group health insurance was not statistically
significant. Similarly, the proportions of women who reported FHBC versus those
with no FHBC did not differ.
To enhance survival following detection of BC, women who receive follow-
up recommendations need to receive appropriate diagnostic services. Their
adherence to follow-up on screening mammography is essential in efforts to identify
BC in its early generally curable stages.
4.2.2 Differences in the Overall Adherence
to Follow-up
While it seems counterintuitive, not all women who following a mammogram
are required to return for a diagnostic procedure do so. The minimal literature
available on this subject indicates that women of color have lower follow-up rates
particularly when mammography findings are abnormal or questionable (Burach &
Liang 1993, Manfredi et al. 1990) while the overall follow-up rates within 60 days
85