Citation
The influence of psychological determinants on rural health care utilization among older Hispanic and non-Hispanic whites

Material Information

Title:
The influence of psychological determinants on rural health care utilization among older Hispanic and non-Hispanic whites
Alternate title:
The San Luis Valley health and aging study
Creator:
Clement, Lacey
Language:
English
Physical Description:
1 electronic file (87 pages) : ;

Subjects

Subjects / Keywords:
Older people -- Health and hygiene ( lcsh )
Medicine and psychology ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
Andersen’s Behavioral Model of Health Services Utilization posits three broad categories of factors that influence individual determinants of health care utilization: predisposing characteristics, enabling resources, and basic need. Though the model does not detail exhaustive predictive factors or specific outcomes of utilization, it does provide a comprehensive framework to conceptualize the underlying influences of health care usage. However, psychological determinants of health care utilization, such as depression, quality of life, and social support, are not specifically included in this model but may influence each aspect of Andersen’s model in various ways. The primary purpose of the present study was to examine how psychological determinants may help explain the utilization of health care in a rural population. A secondary aim was to assess the relationships among the domains of determinants and explore if Andersen’s hypothesized factors would be similar to those found in the San Luis Valley area of Southern Colorado. Health care utilization outcomes examined in this study was frequency of visits to a clinic/ER, as well as use of prayer as a form of complementary and alternative medicine and various preventative measures such as flu shots, mammograms and pap smears for females, and prostate exams for males. Overall, psychological determinants added only minimally to the overall regression models. However, it is possible that depression, quality of life, and social support have a systemic effect on other determinants in the model. Using a factor analysis, similar factors were found in this sample to Andersen’s conceptual model of determinants, but some differences existed. It is clear that psychological variables can and do have a significant influence on decisions, both medical and otherwise, as well as perception of state. Continuing to test models of health care utilization in different populations with different facets emphasized will help to better understand the pathways of utilization in a more holistic manner. However, revising Andersen’s Model of Health Services Utilization to include psychological variables in a systems-based approach will help conceptualize health care utilization in different populations and cultures.
Thesis:
Thesis (M.A.) - University of Colorado Denver.
Bibliography:
Included bibliographic references
System Details:
System requirements: Adobe Reader.
General Note:
Department of Psychology
Statement of Responsibility:
by Lacey Clement

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:
945381216 ( OCLC )
ocn945381216
Classification:
LD1193.L645 2015m C53 ( lcc )

Downloads

This item has the following downloads:


Full Text
THE INFLUENCE OF PSYCHOLOGICAL DETERMINANTS ON RURAL HEALTH
CARE UTILIZATION AMONG OLDER HISPANIC AND NON-HISPANIC WHITES:
THE SAN LUIS VALLEY HEALTH AND AGING STUDY
by
LACEY CLEMENT
B.A., Stephen F. Austin State University, 2011
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Master of Arts
Clinical Health Psychology Program
2015


This thesis for the Master of Arts degree by
Lacey Clement
has been approved for the
Clinical Health Psychology Program
by
Kevin S. Masters, Chair
Krista Ranby
James Grisgby
November 12, 2015


Clement, Lacey R. (M.A., Clinical Psychology)
The Influence Of Psychological Determinants On Rural Health Care Utilization Among
Older Hispanic And Non-Hispanic Whites: The San Luis Valley Health And Aging Study
Thesis directed by Professor Kevin S. Masters.
ABSTRACT
Andersens Behavioral Model of Health Services Utilization posits three broad
categories of factors that influence individual determinants of health care utilization:
predisposing characteristics, enabling resources, and basic need. Though the model does
not detail exhaustive predictive factors or specific outcomes of utilization, it does provide
a comprehensive framework to conceptualize the underlying influences of health care
usage. However, psychological determinants of health care utilization, such as
depression, quality of life, and social support, are not specifically included in this model
but may influence each aspect of Andersens model in various ways. The primary
purpose of the present study was to examine how psychological determinants may help
explain the utilization of health care in a rural population. A secondary aim was to assess
the relationships among the domains of determinants and explore if Andersens
hypothesized factors would be similar to those found in the San Luis Valley area of
Southern Colorado. Health care utilization outcomes examined in this study was
frequency of visits to a clinic/ER, as well as use of prayer as a form of complementary
and alternative medicine and various preventative measures such as flu shots,
mammograms and pap smears for females, and prostate exams for males. Overall,
psychological determinants added only minimally to the overall regression models.
However, it is possible that depression, quality of life, and social support have a systemic
iii


effect on other determinants in the model. Using a factor analysis, similar factors were
found in this sample to Andersens conceptual model of determinants, but some
differences existed. It is clear that psychological variables can and do have a significant
influence on decisions, both medical and otherwise, as well as perception of state.
Continuing to test models of health care utilization in different populations with different
facets emphasized will help to better understand the pathways of utilization in a more
holistic manner. However, revising Andersens Model of Health Services Utilization to
include psychological variables in a systems-based approach will help conceptualize
health care utilization in different populations and cultures.
The form and content of this abstract are approved. I recommend its publication.
Approved: Kevin S. Masters
IV


ACKNOWLEDGEMENTS
I would like to acknowledge my family and friends for supporting and
encouraging me to pursue my dreams in life. I would also like to thank my advisor and
mentor, Kevin S. Masters, for pushing me to be better and constantly challenge myself.
Many thanks also to my thesis committee members, Krista Ranby and Jim Grigsby, for
helping me through this journey. A special thanks is in order for my labmates and cohort
for always supporting me and being my family in graduate school; Stephanie Hooker-
Showalter, Megan Grigsby, Kaile Ross, Jo Vogeli, Shiva Fekri, and Tattiana Romo: I am
honored not only to call you colleagues, but also wonderful friends.
v


TABLE OF CONTENTS
CHAPTER
I. BACKGROUND
Behavioral Model of Health Services Utilization......................1
Predisposing Determinants of Health Care Utilization.................3
Enabling Determinants of Health Care Utilization.....................5
Need Determinants of Health Care Utilization.........................6
Psychological Determinants of Health Care Utilization................8
Present Study.......................................................13
II. METHOD
Population and Sampling ............................................15
Measures ...........................................................16
Predisposing Determinants .......................................16
Enabling Determinants............................................17
Need Determinants................................................18
Psychological Determinants.......................................19
Outcomes of Health Care Utilization..............................21
vi


Data Analytic Strategies................................................21
Missing Data .......................................................22
III. RESULTS
Hypothesis 1............................................................29
Frequency of Clinic or Emergency Room Visits .......................29
Prayer as a Form of Complementary and/or Alternative Medicine ........34
Preventative Care: Flu Shots........................................37
Preventative Care for Men: Prostate Exams ..........................39
Preventative Care for Women: Mammograms and Pap smears .............42
Hypothesis 2............................................................46
IV. DISCUSSION................................................................49
Predisposing, Enabling, and Need Determinants of Health Care Utilization .49
Psychological Determinants of Health Care Utilization...................53
Analyzing Determinant Domains Within the Model .........................56
Amending the Conceptual Model of Health Services Utilization ...........58
A Study of Culture .....................................................60
Strengths and Limitations ..............................................61
Future Directions ......................................................61
vii


Conclusions
62
REFERENCES .................................................................63
APPENDIX
A. Linear Regression Predicting Frequency of Clinic/ER Visits Using MPlus .74
B. Logistic Regression Predicting Use of Prayer Using MPlus.................75
C. Logistic Regression Predicting Use of Flu Shots Using MPlus..............76
D. Logistic Regression Predicting Prostate Exams Using MPlus...............77
E. Logistic Regression Predicting Mammograms Using MPlus ...................78
F. Logistic Regression Predicting Pap smears Using MPlus...................79
viii


CHAPTERI
BACKGROUND
Individuals use health care services for a variety of reasons, including treating and
curing illnesses, preventing health problems, reducing pain, and increasing quality of life.
Individuals can utilize health services in many ways, such as by visiting a physician or
the emergency room or even engaging in various types of complementary and alternative
medical procedures. Myriad factors play a role in the process of deciding to utilize health
care and in the actual utilization of health services. Furthermore, as various health care
reform proposals are developed, understanding trends and factors influencing health care
utilization is crucial to determining which proposals are more likely to succeed,
especially in populations that are underserved.
Behavioral Model of Health Services Utilization
Health care utilization is the end result of a complex process that consists of both
individual and societal determinants, interacting with the particular structure and
resources of the health care system. As health care becomes more expensive and the rate
of chronic illness increases in the population, understanding health care utilization and
the determinants that influence usage of health services becomes vital. Utilization is now
often conceptualized as an individual behavior with many idiosyncratic factors playing a
role. With any individualized behavior, multiple factors in the surrounding context
influence what behavior occurs. To help conceptualize the factors that influence these
behaviors, Andersens Behavioral Model of Health Services Utilization was developed. It
posits three broad categories of factors that influence individual determinants of health
care utilization: predisposing characteristics, enabling resources, and basic need. This
1


model begins to explain the different factors that influence utilization as well as the way
they might interact with each other (Andersen, 1995; Andersen & Newman, 1973;
Wolinsky & Johnson, 1991). Due to technological advances in treatments and more
elaborate health care structures, such as specialized practices and holistic clinics, societal
influences may be increasingly impacting an individuals decision to seek treatment
(DCrus & Wilkinson, 2005; Andersen & Newman, 2005). Consequently, this model has
recently been used from a more economical or public health population-based perspective
to aid in understanding societal influences on utilization (Andersen & Newman, 2005).
That being said, individual factors still play large and important roles in the decision to
utilize health care services. Baxter, Bryant, Scarbro, and Shetterly (2001) noted that this
model tends to characterize and explain more of the individual factors, especially the
need component, rather than changes in medical infrastructure. Using Andersens model,
Baxter et al. (2001) found that patterns in health care resources and use in a rural, aging
population were based on more than just straightforward medical need in that various
factors, such as culture and attitudinal dispositions, played a role in the utilization of
services.
Andersen takes a more systems-based approach with this model in that multiple
factors and their interactions influence the likelihood of utilization taking place. It is
difficult to completely separate these determinants and their respective practical
influences on utilization because of the complex and interactive nature of system-based
models. Though this model does not detail exhaustive predictive factors or specific
outcomes of utilization, it does provide a comprehensive framework to conceptualize the
2


underlying influences of health care usage and guide further research (Andersen &
Newman, 1973).
Predisposing Determinants of Health Care Utilization
Predisposing characteristics of the individual are those which exist prior to the
need for health services but impact utilization nonetheless. Predisposing characteristics
include variables such as demographics, social structure, and health beliefs. Demographic
and biologic variables may predispose someone to have medical problems and hence the
need to seek care more than others without a similar biological and cultural
predisposition. They exist without the presence of a diagnosed medical condition yet may
impact the decision to seek care should the need arise. For instance, demographic
variables such as sex impact utilization. In previous studies of health care utilization, it
was found that females were less likely to be hospitalized or have outpatient surgery than
males, but males were less likely to contact a physician (Dunlop, Manheim, Song &
Chang, 2002). Similar to sex, age also is a factor in health care use. Older adults are more
likely to seek medical care than their younger counterparts due to the perceived
possibility of increased health problems and therefore increased salience of medical care
(Benjamins & Brown, 2004).
Besides age, ethnicity is one of the most predictive predisposing determinants of
health care utilization (Benjamins & Brown, 2004). Ethnic differences in education and
income are also more pronounced in older populations (Dunlop et al., 2002), thus ethnic
disparities play a significant role in health care utilization, especially in older adults.
Ethnicity, as well as education and culture, falls within the social structure paradigm of
predisposing characteristics. The environment and context of ones life influences that
3


persons health and by extension, the utilization of health care. Nguyen, Ugarte, Fuller,
Haas, and Portenoy (2005) found that Hispanics reported fewer years of formal education
and lower income than Whites or African Americans. Lack of education and decreased
income put individuals at a disadvantage to accessing needed health care. Latino
populations generally have poorer overall health and less access to health care services
(Rogers, 2010). According to Nguyen et al. (2005), Hispanics also were less likely to
seek medical consultations than non-Hispanic whites or African Americans. Specifically,
Hispanics with the same diagnosis as Non-Hispanic Whites were less likely to have an
invasive procedure (Carlisle, Leake & Shapiro, 1997), and ethnic minorities engage less
frequently in preventative care services (Coffield et al., 2001; Benjamins, 2005). Cultural
differences also exist in different ethnic groups, and culture plays a vital role in health
care utilization, as it impacts health beliefs, context of living, and perhaps access to care
(Rogers, 2010; Helman, 1994).
Different aspects of culture influence health behaviors, such as values, religion,
and worldviews. Religious and spiritual beliefs often are a cultural norm for Hispanic
groups, and religion may influence beliefs and perceptions of health care, especially for
rural dwellers (Arcury et al., 2005) and older individuals. Individuals who attend
religious services more frequently possess greater knowledge regarding appropriate
health care and maintenance (Apel, 1986; Benjamins, 2005) which is thought to predict
use of health services. On average, adults, 65 and older, are the most likely to attend
religious services. Benjamins (2006) found that moderate levels of church attendance
significantly predicted use of female preventative services such as mammograms. Higher
self-reported religiosity was positively correlated with use of preventative medical care
4


(Benjamins & Brown, 2004). Koenig and Larson (1998) found that individuals who
attended more religious services were less likely to be admitted into a hospital and had a
shorter duration of stay if they were admitted. However, in Hispanics no relationship was
found between church attendance and quantity of medical or hospital visits (Levin &
Markides, 1985).
Additionally, the way someone perceives the health care system and his or her
role in it is important to understanding how and why he or she utilizes it in a particular
way. The precise relationship between beliefs and usage is not yet settled and is
influenced by the type of illness, the implications of that illness, and the overall context
of the individual (Wolinsky, 1988). Beliefs about medicine and illness including attitudes
towards screening, taking medications, and trust in medical professionals can differ
across cultures and even within cultural subgroups (Le et al., 2014).Though some
predisposing and demographic factors are not modifiable and therefore unable to be the
basis of an intervention, they impact medical care usage nonetheless. It is important to
understand the role these factors play in utilization and the way they interact with other
individual determinants.
Enabling Determinants of Health Care Utilization
Enabling resources from the person and his or her family and community can
influence the way an individual views and uses health care. Income, health insurance, and
a means to travel to a physician or medical facility are vital resources for individuals,
without which utilization would likely not take place regardless of other factors. Other
variables associated with the medical facility and resources such as cost of health care,
accessibility of providers, and region of living are considered community enabling
5


resources (Andersen & Newman, 1973). In a rural area, providers might not be plentiful
and thus patients are required to wait for medical care. In environments where this is
normative, it likely impedes individuals from seeking care and a sense of helplessness
may develop. Residents of rural settings are often disparately poorer, more likely to be
older and retired, and less educated than those in urban centers (Ricketts, 1999). With
less education and lower incomes, access that others would routinely utilize becomes
limited for these special populations.
Rural health care utilization brings to the forefront obstacles that are not always
present in urban settings. Those who live in isolated areas tend to have more difficulties
accessing health care due to distance and resources. Fewer or no opportunities for public
transportation and farther distances to travel negatively impact the quality and quantity of
care (Arcury et al., 2005). Understanding factors that influence health care utilization in
rural populations as well as those of lower socioeconomic status is important to
recognizing how health care may differ depending on the setting. However, despite the
enabling resources, another factor must play a role in health care utilization, and that is
whether or not the need exists, or is perceived to exist, to access that care.
Need Determinants of Health Care Utilization
Need encompasses the biological and physical component of symptoms and
illness. The severity of an illness and the perception and evaluation of that severity will
influence whether or not one seeks medical care. Ultimately, whether an individual seeks
medical care depends on if he or she first determines the need exists for health services.
Pain is often the determinant that alerts individuals into the possibility of pathology. For
example, seeking health care for acute low back pain was associated with duration and
6


severity of pain (Carey, Garrett & Jackman, 2000). Similarly, less pain and better daily
functioning was related to less health care seeking (Cote et al., 2001). Further,
functionality in daily life greatly influences perception of illness. Days off from work,
school, or other social responsibilities due to disability are a major component of
evaluated illness (Andersen & Newman, 2005).
The existence and perceived severity of symptoms, or a medical problem,
influences if and how medical care is sought, as well as the type of medical care. For
instance, chronic pain patients who live with incurable pain that does not often diminish
may regularly seek out advice or second opinions, often going from one medical clinic
to the next on a search for some relief from their pain conditions. Furthermore, evaluation
of an illness also comprises need. Documented or diagnosed medical conditions guide
usage of health services. For example, chronic illnesses are among the most prevalent
medical problems plaguing the health care system. Moreover, the aging population, who
is more likely to have chronic illnesses, account for a large portion of health care
utilization (de Boer, Wijker & de Haes, 1997). Heart disease, cancer, stroke, and
hypertension are among those chronic illnesses that require patients to access their
medical provider more often, due to the need for frequent checkups and possible
emergency situations. Similarly, chronic illnesses have a large behavioral component.
Because of this, lack of self-management drives patients to their primary care doctor or
emergency department more often (Butterworth, Linden & McClay, 2007). Secondary
medical problems from these chronic illnesses also lead individuals to utilize services
more (e.g., diabetes mellitus often leading to myocardial infarctions) (Brownson &
Heisler, 2009).
7


Objective physical indicators such as body mass index (BMI) and blood pressure
are need determinants that predict medical use. Morbidly obese patients are more likely
to see their general practitioner, and even after controlling for chronic conditions or other
related medical diagnoses, obesity is still a significant predictor of visits to a general
practitioner (Twells, Bridger, Knight, Alaghehbandan & Barrett, 2012). Hypertension is
often concurrent with cardiovascular disorders but may exist without the presence of a
diagnosable cardiovascular problem. High blood pressure may predict other health issues
that could lead to greater utilization. Need characteristics are often found to account for
much more of the variance in health care utilization than predisposing or enabling
determinants (Hershey, Luft & Gianaris, 1975; Mechanic, 1979).
Whether or not a person actually has significant health problems and perceives
that he or she has significant health problems determines if utilization is necessary. Two
different decisions must take place. The way pain and other symptoms are experienced,
attended to, and perceived first impacts the decision of whether or not the symptom is
indicative of pathology and then whether care should be sought. However, examining
need merely from the presence of chronic illness, physical indicators, pain, or even
limited functionality undermines the complexity of need determinants.
Psychological Determinants of Health Care Utilization
Not clearly defined in Andersens model is the role of psychological factors in
health care utilization. Patients often frequent their medical care provider for symptoms
with no known organic cause. In fact, only three to 20 percent of visits actually had
known organic causes. In addition, there are consistent relationships between these
medically unexplained symptoms and psychological processes (Gask, Dowrick, Salmon,
8


Peters & Morriss, 2011; Verhaak, Meijer, Visser & Wolters, 2006; Peveler, Kilkenny &
Kinmouth, 1997; Kolk, Schagen & Hanewalk, 2004). Clearly, different factors interact
with one another to bring a person to seek care at a medical facility. Sobel (1995) argued
that despite the presence of known organic causes, emotional and mental factors play a
significant role in the onset and management of diseases. The way an individual displays
medical symptoms and functions with a disease greatly depends on his or her ability to
cope. Furthermore, patients who somaticize or have other physical symptoms, along with
a comorbid psychological disorder, are more likely to utilize health care, especially
emergency room visits (Blount et al., 2007). Kolk, Hanewald, Schagen and Gijsbers van
Wijk (2003) posited a complex model of symptom perception to understand the
evaluation of physical symptoms. In that model, emotions and affectivity influence
perception at various levels, such as input of somatic information and attention to
sensations. Moreover, negative affectivity, in particular, is associated with the presence
of physical symptoms, specifically increases of somatic information to be encoded (Kolk,
Handwald, Schagen & Gijsbers van Wijk, 2002). Those with negative affect tend to
somaticize more and generally experience more physical sensations to appraise.
Seemingly, it is not just the increased appraisal of physical sensation but also the
increased magnitude of sensations that leads to increased care utilization.
Once sensations occur, though, psychological factors such as depression, social
support, and quality of life affect perception of those symptoms and the overall decision
to utilize health care. Actual severity of illness has little to do with a decision to seek
medical care (Berkanovic, Telesky & Reeder, 1981). Those who seek medical care tend
to have worse health-related quality of life (Cote et al., 2001), although the direction
9


between these two could be and most likely is bidirectional. According to de Boer,
Wijker & de Haes (1997), depression and overall psychological distress are some of the
strongest predictors of health care use. Depression has been shown repeatedly to
significantly predict increased utilization (Keeley et al., 2008; Sullivan, Feuerstein,
Gatchel, Linton & Pransky, 2005). Patients with a diagnosed depressive disorder have
double the health care costs relative to their peers without depression (Kathol et al.,
2005). Depression can lead persons to be more attentive to symptoms and catastrophize
physical sensations, inducing them to over utilize the health care system. Even if known
organic causes exist, the added layer of psychological factors such as depression or even
overall quality of life can change the way the illness and need to seek care is appraised.
For example, interactions between pain and quality of life can be manifest in various
ways, depending on other personal and social factors. Those who attend more to painful
symptoms or catastrophize those symptoms are more likely to seek medical care.
Additionally, those with depression or negative affectivity are less likely to be compliant
and adherent with medical advice. With noncompliance and nonadherence comes
increased medical care and associated health care costs (Whooley et al., 2008; Glassman
et al., 1990; Carney et al., 1995; Ziegelstein, Bush & Fauerbach, 1998).
Older adults, for whom nonadherence may be problematic partially due to age-
related issues, tend to also have multi-morbidity. Boyd et al. (2005) reported that over
half of adults 65 and older had at least three chronic diseases. Of those, a significant
number were living with five or more chronic diseases. The combination of multiple
diagnoses as well as age-related concerns may have a compounded effect on utilization. It
is known that patients with chronic illnesses use health services more and often have
10


adjoining psychosocial sequelae. Secondary consequences of chronic illnesses such as
poor adjustment, poorer quality of life, and more intense depression all lead to greater
utilization (Yohannes, 2013). These consequences manifest uniquely in an aging
population. In particular, older adults lack of social support, or perceived lack of social
support, negatively influences their health behaviors in a number of ways, such as poor
self-efficacy and inability to appropriately self-manage their illness (Rosland et al., 2008;
McCathie, Spence & Tate, 2002). Social support is a vital component of overall quality
of life. Regardless of population, the quality of social support, rather than quantity, is
often the defining attribute that predicts better health (Yohannes, 2013). Matching
support given to support needed is often key and mismatches often result in poorer
adjustment (Yohannes, 2013). Ability to adjust and cope with an illness can influence
pathways of health care utilization. A persons overall well-being, impacted by negative
affectivity, quality of life, social support, or other related variables, is related to physical
health. Depending on the population studied, subjective well-being is understood in
different ways (Abdel-Khalek, 2014) and terminology often overlaps, but the positive
evaluation of circumstances is related to increased physical health (Abdel-Khalek, 2014).
If one evaluates his/her life more positively, the greater sense of overall well-being may
lessen negative appraisals of need.
Both positive and negative outlooks influence health evaluations; psychological
factors in general may also influence predisposing and enabling determinants of health
care utilization. The relation between utilization and gender, age, and income can all
influence, as well as be influenced by, psychosocial factors. Kolk et al. (2003) reported
that age and SES were significant indicators of physical symptom appraisal through
11


various direct and indirect pathways, such as selective attention, negative mood, and
prevalence of chronic disease. Older adults are less likely to be diagnosed with
depression (Hybels & Blazer, 2003) and community-dwelling older adults who were
independent had lower prevalence rates of depression compared to those in the general
adult population (Birrer & Vemuri, 2004). Also, depression has a known gender gap in
that women are usually twice as likely to be diagnosed with a depressive episode or
disorder but that gap lessens as age increases (Komstein, 1997).
Psychological determinants of health care utilization may influence each aspect of
this model in various ways. An individuals affective or psychological attributes could
potentially mediate the relationship among these individual determinants. In describing
his model of health care utilization, Andersen understood that psychosocial variables,
such as social networks, could influence different determinants (Andersen, 1995;
Andersen & Newman, 2005) but failed to adequately address the complexity of ways that
psychological determinants impact health care utilization. In 2012, a systematic review
by Babitsch, Gohl, and von Lengerke examined studies that implemented Andersens
model. In it, they included mental disorders in the domain of predisposing
characteristics, social/emotional support in the domain of enabling characteristics, and
health related quality of life in the domain of need characteristics. Though there is a
certain logic that follows this way of organization, it has also been shown that factors
such as depression, social support, and quality of life interact closely with each other and
influence a vast number of other determinants. Therefore, health and psychological
factors appear to have systemic influences on each other meaning bidirectional and
compound effects exist between the two.
12


Present Study
The primary purpose of the present study was to examine how psychological
determinants may help explain the utilization of health care in a rural population.
Specifically, the study set out to explore how Andersens model of health care utilization
mapped onto predictors of utilization in the rural San Luis Valley area of southern
Colorado. The primary hypothesis of this investigation was that psychological
determinants, namely depression, social support, and quality of life, would explain the
use of health services in a community sample beyond the predisposing, enabling, and
need determinants found in Andersens model. Specifically, these psychological
determinants would better predict use of health services such as clinic, hospital, and ER
visits in a 12-month period, preventative care (e.g. flu shots, prostate exams for men, and
mammograms and pap smears for women), and use of prayer for spiritual healing than
predisposing, enabling, and need determinants, as posited in Andersens model. If this
were the case, it is possible that the existing model could be construed differently to
include these psychological determinants (see figures 1, 2, and 3 for hypothesized
conceptual models).
The present study also had a secondary aim to assess the relationships among the
domains of determinants and explore if Andersens hypothesized factors would be similar
to those found in this population. Based on the theoretical model, it is presumed that in
order to create these domains of determinants, individual predictors within each domain
would be more associated to each other than to those in other domains. For instance, a
need determinant such as presence of chronic illnesses would be more related to another
need determinant, such as self-reported health, than to an enabling determinant such as
13


accessibility to health care. This study sought to examine those relationships to see if they
endured in this sample.
14


CHAPTER II
METHOD
Population and Sampling
The San Luis Valley Health and Aging Study (SLVHAS) surveyed health and
disability among Hispanic and non-Hispanic white residents of Alamosa and Conejos
counties, two mainly isolated rural southern Colorado counties, between 1993 and 1995.
Residents of the San Luis Valley live mostly in small, primarily farming, communities.
According to the 1990 population census of the region, 46 percent were Hispanic, 52
percent were non-Hispanic White, and 2 percent were other (San Luis Valley Regional
Development and Planning Commission, 1992). The majority of Hispanic residents in the
San Luis Valley report their ethnicity as Spanish or other Hispanic as opposed to
Mexican American. At the time of data collection, the population in the San Luis Valley
region had little to no new immigration occurring (Bean & Tienda, 1987). Additionally,
19 percent of residents 65 and older were below the poverty level in 1990 (Hamman et
ah, 1999). In order to participate in the study, individuals must have been: (a) 60 years of
age or older, (b) a current resident in the surveyed counties, and (c) of Hispanic or Non-
Hispanic White ethnicity.
In this sample, participants (N= 1358) were mostly Hispanic (58.4%) and female
(56.8%). The average age was 74 (SD = 7.8; range 60-99). Almost forty percent did not
complete high school, and almost half of the participants yearly income was less than
$10,000. See table 1 for complete participant characteristics.
15


Table 1
Participant Characteristics
Variable_________________________N_____________________(%)
Age M (SD) = 74(7.8)
Female Ethnicity 772 (56.8%)
Non-Hispanic White 565 (41.6%)
Hispanic Education 793 (58.4%)
Less than high school 518 (38.7%)
High school 540 (40.3%)
College 194 (14.5%)
Graduate degree 87 (6.5%)
Income
Less than $10,000 545 (45.8%)
$10,000 $19,999 358 (30.0%)
$20,000 $34,999 182 (15.2%)
$35,000 $49,999 60 (5.0%)
$50,000 $74,999 29 (2.4%)
More than $75,000 17 (1.4%)
Note. Because these data were from 1993-1995, inflation rates must be considered to understand how income rates measured in this population translate into income based on the current U.S. economy. Using the CPI calculator, (Consumer Price Index; U.S. Bureau of Labor Statistics), an inflation rate of 1.6 must be used to determine how these data convert to current income rates. For instance, income of $10,000 in this dataset would convert to $16,000, $20,000 would convert to $32,000, $75,000 would convert to
$121,000, etc.
Measures
Predisposing Determinants. Age, ethnicity, gender, and education are measures
of the predisposing characteristics. Age was measured by asking date of birth of each
participant and calculating from that to the time of data collection. Ethnicity was assessed
by asking participants Are you of Spanish/Hispanic origin or descent? Four answer
options were possible of No, Yes, Mexican, Mexican-American, Chicano, Yes,
Cuban or Puerto Rican, and Yes, other. Participants were asked if they were male or
16


female to identify gender. For assessment of education level, participants were asked to
identify the highest grade or year of school (they had) completed. Possible answer
choices ranged from 00-17+, with one-year intervals, as well as I dont know.
Enabling Determinants. Income, insurance, and accessibility of medical care are
all enabling determinations of health care usage. Income was assessed by asking
participants about how much was the total income, before taxes, of all your family
members, living in your house, from all sources last year? The following 11 options
were given to answer about income: a) less than 5,000, b) 5,000 7,499, c) 7,500 -
9,999, d) 10,000 14,999, e) 15,000 19,999, f) 20,000 24,999, g) 25,000 34,999, h)
35,000 49,999, i) 50,000 74,999, j) 75,000 or more, and dont know/refuse to
answer. Because these data were from 1993-1995, inflation rates must be considered to
understand how income rates measured in this population translate into income based on
the current U.S. economy. Using the CPI calculator, (Consumer Price Index; U.S. Bureau
of Labor Statistics), an inflation rate of 1.6 must be used to determine how these data
convert to current income rates. Each participant was also asked whether or not he/she
was covered by a health insurance plan.
Finally, accessibility of medical care was assessed by three different measures, 1)
by asking if there is a certain clinic or medical center used if needed, 2) how many miles
the clinic is from his/her home, and 3) if they had difficulties getting to their medical
center because of various factors, which was a total count of endorsement of the
following: how much it costs, didnt have a way to get there, care was not available
when needed, needed someone to take care of family member, had to wait too long
in the office or clinic, had no confidence in the staff, and staff did not speak a
17


language you were comfortable with (Hispanic Health and Nutrition Examination
Survey, 1990).
Need Determinants. Both subjective and objective determinants of need were
assessed. Self-reported health status, presence of chronic illness and pain were measured,
as were objective physical indicators of waist to hip ratio and blood pressure.
Participants were asked to rate their overall health status as excellent, good,
fair, or poor compared with other people (their) age. Chronic illnesses were assessed
by asking participants to self-report whether or not they had ever been told by a doctor
that they had cancer, heart attack, mini-stroke, major or severe stroke, angina, high blood
pressure, enlarged heart or heart failure, emphysema/chronic bronchitis/COPD, cirrhosis
of liver, kidney failure, Parkinsons disease, osteoporosis, seizure disorder, and
migraines/persistent headaches. Response options for each question were no, yes, or
dont know. Self-reported diseases tend to correlate well with documented medical
diagnoses (Edwards, Winn & Kurlantzick, 1994). The measure of chronic illness was a
total count of all illnesses reported. Pain was measured in different ways with four
separate measures included in subsequent analyses: presence of pain, severity of pain,
frequency of pain, and pain interference with activities of daily living. Presence of pain
was assessed by asking Are you ever troubled with pain? (in past year), which was
used as a dichotomous variable. Additionally, severity and frequency of pain were both
assessed ordinally by asking When your pain is at its worst, would you describe it as
mild, moderate, severe or unbearable and During the past week, how much of the time
have you been troubled with pain (all of the time, most of the time, some of the time, or
rarely/never), respectively. The interference with activities of daily living (ADLs) due to
18


pain was a continuous variable that used a total count of endorsement to five detailed
questions such as Does this pain ever..1) cause you to move around less, 2) keep
you from sleeping, 3) cause you to cut down on any of your usual activities like work,
household chores, or running errands, 4) keep you from visiting family and friends in
your own home, and 5) keep you from doing things you like to do for pleasure, like
hobbies, sports or leisure activities? The four measures of pain were used separately in
data analyses.
As well as the self-reported measures of illness or health, two objective measures
of health were taken that predict occurrence or development of cardiovascular or
metabolic disease: a) waist to hip ratio to determine adiposity, and b) an average of the
second and third blood pressure measurements. Because both low and high BMI could
signal pathology in an aging population, waist to hip ratio was determined to be the
preferred measure of adiposity. This measure is best estimate of frailty in an aging
population. Also, although three measurement of blood pressure were taken, only the last
two were averaged because the first reading was used to calibrate. The SLVHAS
employed several full-time data collectors who were long-time residents of the area.
Multiple training sessions were provided to ensure reliability of the data. Additionally, to
gain an understanding of interrater reliability, 60 participants were reinterviewed an
average of 10 days (range 3-15 days) after the initial visits for some selected items, and
researchers found no differences in reproducibility between Hispanics and Non-Hispanic
White subjects (Hamman et al., 1999).
Psychological Determinants. Three measures of psychological factors
(depression, social support, and quality of life) were used to explore participants
19


psychosocial state. Depressive symptoms were assessed as a measure of mood using the
20-item Center for Epidemiologic Studies Depression (CES-D) Scale (Radioff, 1977).
The scale asks about feelings and experiences during the last week (e.g. I felt that
everything I did was an effort and I felt depressed) with response options ranging
from rarely or none of the time to most or all of the time. The overall scale
reliability is high in the general population (a = .85). The CES-D has been shown to have
good sensitivity and specificity, with high internal consistency (Lewinsohn, Seeling,
Roberts & Allen, 1997), and it has been used with different racial and ethnic groups
(Roth et al., 2008). The total score of the CES-D will be the datum point for analyses.
Total scores range from 0 to 60, with higher scores indicating greater depressive
symptoms. Four questions were reverse scored (I felt I was just as good as other
people, I felt hopefuly about the future, I was happy, and I enjoyed life).
Social support was measured by two items assessing the social network size, 1)
How many relatives do you feel close to? That is, relatives you feel at ease with, can
talk to about private matters, and can call on for help, and 2) How many close friends
do you have? That is, people that you feel at ease with, can talk to about private matters,
and can call on for help. The two items were added together to create a singular social
support measure of both relatives and friends.
Quality of life was measured using Patrick's Perceived Quality of Life Scale
(PQoL; Patrick, Danis, Southerland, & Hong, 1988). All 20 questions were rated by the
participant on an 11-point scale from 0 to 10, where 0 was very dissatisfied and 10 was
very satisfied. Examples of questions include: how satisfied are you with... the health of
your body, the meaning and purpose in your life, and the amount of variety in your life.
20


The PQoL also asks participants to rate on the 0-10 satisfaction scale how happy are
you? A mean score of responses is calculated, and if the mean is equal or greater than
7.5, scores are interpreted as satisfied. The PQoL evaluates satisfaction with areas of
functional status in persons with varying levels of wellness and disability. It has also been
more widely used with different patient groups and specifically with this population in
previous studies (Baxter et al., 1998; Caldwell, Baxter, Mitchell, Shetterly & Hamman,
1998).
Outcomes of Health Care Utilization. Three outcomes of health care utilization
were assessed. Frequency of health care utilization in the past year was assessed by
asking participants In the past 12 months, how many times have you gone to a medical
clinic, outpatient clinic or emergency room? Participants were also asked about
preventative care by the following questions: In the past 12 months, have you had... flu
shot, prostate exam (for males only), pap smear (for females only), mammogram (for
females only)? Use of a particular complementary or alternative medicine (CAM)
approach was also used as a measure of health care utilization; participants were asked if,
in the past year, they had used prayer or spiritual healing (Eisenberg et al., 1993).
Data Analytic Strategies
Data were analyzed using IBM SPSS Statistics version 21 (SPSS Inc., 2012) and
Mplus version 7.1 (Muthen & Muthen, 2013). Descriptive statistics (e.g. means, standard
deviations, frequency distributions) were calculated to describe the sample. Continuous
variables were checked for normality. The outcome variable of frequency visits to a place
of care was transformed using a square-root transformation, and the miles from care
variable used to assess accessibility to health care was also transformed using a log
21


transformation to correct for the non-normal distributions. The distributions of all other
continuous variables were within the normal ranges (skew between 3.0). The variable
of adiposity (waist:hip ratio) was squared to create a quadratic measure to best determine
a curvilinear relationship.
Categorical demographic and predictor variables were dummy-coded so they
could be used in a linear analysis: (a) ethnicity was categorized into (0) Non-Hispanic
White and (1) Hispanic; (b) sex was categorized into (0) male and (1) female; (c)
insurance was categorized into (0) does not have insurance and (1) has insurance; (d)
place of usual care was categorized into (0) no place of usual care and (1) does have a
place of usual care; and (e) presence of pain was categorized into (0) denied pain and (1)
endorsed pain; Dichotomous outcome measures of health care utilization (flu shots,
prostate exams, mammograms, pap smears, and use of prayer) were recoded so that 0 =
denied getting service and 1 = endorsed getting service.
Scale scores were calculated according to the authors instructions (see
description of calculation in measures section). Bivariate associations among study
variables were examined. See table 2 for correlations among predictor variables and table
3 for correlations among predictors and outcomes.
Missing Data. To account for missing data, listwise deletion was used in SPSS so
only those who had complete data on the variables used were included in analyses. Table
4 illustrates the percentage of missing data for each variable. Most variables had missing
data that ranged from 2.0%-23.00%. It appears that most variables within a domain of
determinants were missing in clusters in that when one variable was missing, others
within that domain were also missing.
22


Tabic 2
Correlations Among Predictors
Variable 1 2 3 4 5 6 7 8 9 10
1. Age
2. Hispanic -.04 --
3. Female .07* -.01
4. Education -.17** -.48* .04
5. Income -.32** -.35** -.20** .49**
6. Insurance .16** -.01 .00 .03 -.03
7. Place of Care .04 .01 .02 -.02 .02 -.02
8. Distance from Care -.07* .14** -.08** -.11** -.05 -.01
9. Accessibility Difficulties .02 .06* -.00 -.10** -.12** -.06* .04 .12**
10. Self-Reported Health -.04 -.17** -.01 .25** .27** -.04 -.05 -.03 -.16**
11. Chronic Illnesses -05f -.13** .03 .02 -.02 .06* .10** -.03 .09** -.28**
12. Presence of Pain .02 -.02 .02 -.03 -.04 -.01 .07* .05 .12** -.24**
13. Severity of Pain -.06 -.01 .09 -.06 -.07 .04 -.01 -.01 .10** -.19**
14. Frequency of Pain .07* .16** .03 -.22** -.18** -.05 .06 .07* .14** -.30**
15. ADL Interference -.07* .04 .12** -.07* -.13** .03 .03 -.01 .13** -.34**
16. Waist:Hip Ratio -.03 .11** -.65** -.13** .01 -.01 .02 .08** ,06f -.08**
17. Systolic Blood Pressure .11** .12** .08** -.12** -.12** .01 -.00 -.01 .04 -.08**
18. Diastolic Blood Pressure -.14** .14** -.06* -.04 -.04 -.05 -.04 .04 .05 .01
19. Depression .03 .07* .13** -.09** -.19** -.00 .05 -.01 .25** -.33**
20. Social Support .03 -.09** -.15** .09** .11** .06* -.01 .11** -.05 .16**
21. Quality of Life .11** .17** .04 -.12** .05 -.03 -.05 .02 -.20** .36**
tp <_ .06, *p < .05, **p < .01
Note. Depression measured using Center for Epidemiologic Studies Depression Seale (CES-D). Social support is a total count
of family and friends. Quality of Life measured using Patricks Perceived Quality of Life scale (PQoL). WaistrHip Ratio was
transformed to measure quadratic trends for a curvilinear relationship.


Tabic 2
Correlations Among, Predictors (contj
Variable__________________________U_________12 13 14 IS 16 17 18 19 20
1. Age
2. Hispanic
3. Female
4. Education
5. Income
6. Insurance
7. Place of Care
8. Distance from Care
9. Accessibility Difficulties
10. Self-Reported Health
11. Chronic Illnesses
12. Presence of Pain .22**
13. Severity of Pain .13**
14. Frequency of Pain .07 .20**
15. ADL Interference .20** .59** .26**
16. Waist:Hip Ratio .01 .01 -.06 .02 -.05
17. Systolic Blood Pressure .14** .01 -.05 .06 .01 .02
18. Diastolic Blood Pressure .01 -.03 -.06 .03 -.02 .09** .58**
19. Depression .20** .18** .16 .22** .28** -.04 .08* .02
20. Social Support -.02 -.02 -.06 -.01 -.10** .04 -.02 .00 -.16** -
21. Quality of Life -.24** -.16* -.19** -.16** -.32** -.08** .02 .01 -.44** .15**
tp < .06, *p < .05, **p < .01
Note. Depression measured using Center for Epidemiologic Studies Depression Seale (CES-D). Social support is a total count
of family and friends. Quality of Life measured using Patricks Perceived Quality of Life scale (PQoL). Waist:Hip Ratio was
transformed to measure quadratic trends for a curvilinear relationship.
to
-P*.


Tabic 3
Correlations Among Predictors and Outcomes
Variable Clinic/ER Visits Prayer Flu Shots Prostate Exams Mammograms Pap smears
1. Age .12** .07* .12** -.02 -.19** -.01**
2. Hispanic .03 .01 -.08** -.14** .00 .06
3. Female .06* .17** .06
4. Education -.10** .02 .11** .15** .12** .02
5. Income -.10** -.07* .05 .09 .18** .07
6.Insurance .06* .03 .06* .07 .07 .06
7. Place of Care .18** .05 .10** .12** .09* .06
8. Distance from Care -.01 .05 -.10** .03 .01 .05
0. Accessibility Difficulties .08** .13** -.05 -.02 -.06 -.06
10. Self-Reported Health -.28** -.05 -.07* -.08 .09* .04
11. Chronic Illnesses .33** .11** .17** .13** -.01 -.04
12. Presence of Pain .19** .07** .07* .08 -.01 -.04
13. Severity of Pain .13** .07* .07* .00 -.01 -.01
14. Frequency of Pain .12** -.02 -.03 -.03 -.12** -.06
15. ADL Interference .22** .13** .07* -.01 .01 -.01
16. WaistrHip Ratio .00 -.11** -.03 .01 -.06 -.02
17. Systolic Blood Pressure .05 .02 -.03 .02 -.04 -.03
18. Diastolic Blood Pressure -.04 -.07* -.09** -.03 -.01 .04
19. Depression .19** .10** .04 .05 -.10** -.03
20. Social Support -.03 -.02 -.03 .05 .00 .01
21. Quality of Life -.20** -.08** -.07* -.08 .08* .18**
*p < .05, **p < .01
Note. Depression measured using Center for Epidemiologic Studies Depression Scale (CES-D). Social support is a total count
of family and friends. Quality of Life measured using Patricks Perceived Quality of Life scale (PQoL). WaistrHip Ratio was
transformed to measure quadratic trends for a curvilinear relationship.


Tabic 4
Percentages of Missing Data
Variable %
Age 2.0%
Hispanic 2.0%
Sex 2.0%
Education 3.4%
Income 14.1%
Insurance 3.5%
Place of Care 3.5%
Distance from Care 6.2%
Accessibility Difficulties 9.3%
Sclf-Rcportcd Health 13.7%
Chronic Illnesses 6.5%
Presence of Pain 1.9%
Severity of Pain 1.9%
Frequency of Pain 1.9%
ADL Interference 1.9%
Waist:Hip Ratio 8.1%
Systolic Blood Pressure 7.0%
Diastolic Blood Pressure 7.0%
Depression 16.8%
Social Support 16.9%
Quality of Life 23.5%
Clinic/ER Visits 3.8%
Prayer 4.4%
Flu Shots 13.4%
Prostate Exams 12.3%
Mammograms 11.8%
Pap smears 11.8%
To further explore how missing data might impact analyses, regressions were also
performed in MPlus to estimate missing data parameters. Results from analyses in both
SPSS and MPlus were similar. Because results were similar, SPSS analyses are reported
in the results section. See appendices A-F for regression tables based on MPlus analyses.
Hypothesis 1: Psychological determinants (depression, social support, and quality of
life) will explain the use of health services in a community sample beyond the
26


predisposing, enabling, and need determinants in Andersens model. Linear and
logistic regressions were used to test if psychological determinants accounted for
additional variance in health care utilization than did predisposing, enabling, and need
determinants alone. For the outcome of frequency of visits to a place of health care, a
linear regression was used. Continuous variables were centered using the grand mean
technique. Moderation analyses were used to determine how, in the presence of other
predictor variables, psychological variables are related to health care utilization at a place
of health care. The other outcomes were dichotomous in nature (either received service or
did not), therefore logistic regressions were used for those models. The models that
included psychological variables were examined to ascertain if there was a significant
increase in explained variability beyond the previous models that lacked psychological
variables. Beta-weights were examined for significant contributions to frequency of visits
to a place of health care.
Hypothesis 2: Domains of determinants in this sample will be similar to those
proposed in Andersens model of health care utilization. Predictors within a domain
of determinants will be more closely related than predictors across domains. A principal
component factor analysis was performed to check the validity of Andersens model in a
new sample.
27


CHAPTER III
RESULTS
Descriptive statistics of the psychological variables and outcomes are presented in
tables 5 and 6, respectively.
Table 5
Descriptive Statistics on Psychological Variables
Variable N M (SD) Min Max Skew Kurtosis
Depression 1153 6.84 (7.7) 0.0 57.0 2.0 5.4
Social Support 1152 7.31 (3.0) 0.0 12.0 -0.2 -0.7
Quality of Life 1060 8.25 (1.3) 2.6 10.0 -0.8 0.7
Note. Depression measured using Center for Epidemiologic Studies Depression Scale
(CES-D). Social support is a total count of family and friends. Quality of Life measured
using Patricks Perceived Quality of Life scale (PQoL).
Table 6 Descriptive Statistics on Outcomes of Health Care Utilization
Variable N M (SD) Min Max Skew Kurtosis
Clinic/ER Visits 1334 1.84 (1.23) 0.00 7.75 0.68 0.93
Variable N % Received Service
Prayer 1325 29.3%
Flu Shots 1200 64.0%
Prostate Exams 515 43.9%
Mammograms 681 36.6%
Pap smears 681 40.8%
Note. Measure of clinic/ER visits is a continuous variable that was square-root
transformed to acquire normality. Prayer, flu shots, prostate exams, mammograms, and
pap smears are all dichotomous variable
28


Hypothesis 1
To test hypothesis 1, both linear and logistic regressions were used to explore the
relationship between psychological variables and health care utilization outcomes. To
determine which variables to include in the regression models, bivariate correlations were
calculated. Predictors that did not have a significant relationship to the outcomes were
excluded from the analysis.
Frequency of clinic or emergency room visits. Before running a linear regression,
correlations between predictors and the outcome of frequency of visits to a clinic or
emergency room were analyzed to determine which variables should be included in the
analyses. Based on the correlational analyses, three predisposing determinants, four
enabling determinants, and six need determinants were significantly related to the
outcome. Within the predisposing determinant domain, age (r=.l2,p< .01), sex (r =
06, p < .05), and education (r = 10, p < .01) were significantly related to frequency of
visits. Both age and sex were positively related to frequency of clinic visits, suggesting
that older participants and females were more likely to seek treatment in a clinic or
emergency room. Education, on the other hand, was negatively related to frequency of
visits, such that those with less education were more likely to seek treatment. Within the
enabling determinant domain, income (r = -.10,p < .01), insurance (r = .06, p < .05),
place of care (r=.lS,p< .01), and accessibility difficulties (r = -.08 ,p< .01) were
significantly related to frequency of clinic visits. If a person has insurance and a place of
usual care, he or she was more likely to frequent a place of care. Income and difficulties
with accessing a place of health care were negatively correlated with visits; those with
less income and fewer difficulties were more likely to frequent clinics or emergency
29


rooms. Finally, within the need determinants, self-rated health (r = -.28,/) < .01), number
of chronic illnesses (r = .33, p < .01), presence of pain (r=.l9,p< .01), severity of pain
(r = .13,/><.01), frequency of pain (r = .12, p < .01), and interference of ADLs because
of pain (r = .22, p < .01) were all significantly related to frequency of clinic visits.
Comparing ones health to others was the only negatively correlated predictor with
frequency of visits. The lower one rated ones health, the more likely it was that he or she
would visit a clinic or emergency room. Greater pain and pain interference, along with
number of chronic illnesses, all indicated a greater likelihood of seeking treatment. There
was no significant relationship between visits and the measure of adiposity (waisfhip
ratio), (r = .0l,p = .85).
All three psychological variables were put into the regression model, although
only perceived quality of life (r = -.20, p < .01) and depression {r = .\9,p< .01) were
significantly related to frequency of clinic visits. Because social support (r = -.03,p =
.34) is a core component of the hypothesized model, it was also included to determine
how this construct might be interacting with other variables.
A regression model that included all these variables was calculated. See table 7
for the hierarchical linear regression analyses. The final model that included all three
psychological variables was significant F (15, 605) = 7.77,/) < .001, R2 = .14, and
accounted for 14% of the variance in frequency of clinic emergency room visits.
However, including psychological variables into the model only added 0.01 explained
variability to the model, and the change in R2 was not significant, F (3, 605) = 0.69,/) =
.56. With just the predisposing variables in the model, 2% of the variability was
explained and the addition of enabling variables explained 4% of the variability of
30


seeking treatment. However, when need determinants were added into the model, 14%
was explained, which was a significant change ini?2, F (5, 608)= 14.93,p< .001. Need
determinants added the most unexplained variability to the overall model, whereas
psychological determinants only added a minimal amount of variability.
To determine if multicollinearity among the psychological variables existed, each
psychological variable was entered individually in the final model with all other variables
previously entered. Adding social support into the model was not significant, R2A = 0.00,
F (1, 653) = 0.24,/> = .63. Similarly, adding depression into the model alone was not
significant, R2A = 0.00, F (1, 654) = 0.15,p = .70, nor was perceived quality of life, R2A
= 0.02, A (1, 607)= 1.73,/? = .19.
Moderation analyses were examined to determine if any interactions that included
the psychological variables were significant. There was a significant interaction between
depression and chronic illnesses, b = -.01, SE= 0.00, f$= -.01,p < .05, as depicted in
figure 4. The graph of the interaction with the data points overlaid was examined and
verified that the interaction effect was within the bounds of observed data. It appears that
as number of chronic illnesses increase, number of visits to a clinic or ER also increases.
However, depression seems to play a larger role in this relationship if a person has fewer
chronic illnesses. This moderation was also examined by plotting two slopes of clinically
significant depression (> 16 on CES-D) or non-significant depression (< 16 on CES-D) in
Figure 5. The slope was marginally significant for those who scored in the depressed
range on the CES-D, 6 = .156,/? = .06. For those were not depressed, the slope was
significant, b = .321 ,P< .001. Most of the data points fell within the range of fewer
chronic illnesses. The effect of number of chronic illnesses on frequency of visits to a
31


health care clinic depended on the presence of clinically significant depression in that
when individuals were depressed, they utilized health care more than non-depressed
individuals when they were healthier (fewer chronic illnesses). However, as number of
chronic illnesses increased, the frequency of clinic visits increased more for those who
were not depressed compared to their depressed counterparts. Other moderation analyses
were also examined between depression, quality of life, and social support among
predisposing, enabling, and need determinants, but none were significant.
- Low [>rpr*Mlon
....Medium tSapmahod
----Dfprwiion
01

to 0 0 .'0 4 0 0 SO
Naml>*r* kmk Mlwtm (mlrml)
Figure 4. Moderation of Depression and Chronic Illnesses on Frequency of Clinic/ER Visits.
32


Table 7
Linear Regression Predicting Frequency of Clinic/ER Visits (N = 605).
Model 1
Model 2
Model 3
Model 4
Variables B SEB P B SEB P B SEB P B SEB P
Age 0.02 0.01 0.11** 0.02 0.01 0.10* 0.02 0.01 0.12** 0.02 0.01 0.12**
Sex 0.14 0.10 0.05 0.14 0.10 0.05 0.08 0.10 0.03 0.10 0.10 0.04
Education -0.03 0.01 0.08* -0.03 0.02 -0.07 -0.02 0.02 -0.06 -0.03 0.02 -0.08
Income 0.00 0.03 0.00 0.04 0.03 0.06 0.04 0.03 0.07
Insurance 0.83 0.42 0.08* 0.42 0.41 0.04 0.38 0.41 0.04
Place of Care 1.40 0.38 0.15*** 1.15 0.36 0.12** 1.13 0.36 0.12**
Accessibility Difficulties 0.14 0.07 0.09* 0.07 0.06 0.04 0.05 0.06 0.03
Self-Reported Health -0.23 0.07 -0.16*** -0.26 0.07 -0.15**
Chronic Illnesses 0.16 0.04 0.18*** 0.15 0.04 0.18***
Pain Severity 0.00 0.06 0.00 0.00 0.06 0.00
Pain Frequency -0.01 0.05 -0.01 -0.01 0.05 -0.01
ADL Interference 0.10 0.03 0.14** 0.09 0.04 0.13**
Depression 0.00 0.01 0.00
Social Support 0.01 0.02 0.02
Quality of Life -0.06 0.05 -0.06
R* .02 .04 .14 .14
F for change in R~ 4.44** i 5.54*** 14.93*** .69
*p < .05, **p < .01, ***/> < .001
LtJ
OJ


Number ml Oirn
DumtN
MM OimwN
Figure 5. Moderation of Depression Cut-off and Chronic Illnesses on Frequency of Clinic/ER
Visits.
Prayer as a form of complementary and/or alternative medicine. Similar to the first
analysis, only predictors that were significantly related to the outcome of prayer use were
used in the regression analyses. Based on Pearson correlational analyses, sex (r = .17, p<
.001), age (r = .07, p < .05), income (r = -.07, p < .05), accessibility difficulties (r = .13,/?
< .001), chronic illnesses (r=.\\,p< .001), severity of pain (r = .07, p < .05), pain
interference with ADLs (r = .13,p< .001), waist:hip ratio (r = 11 ,P< .001), depression
(r = .10, p < .01), and quality of life (r = -.08, p < .05) all had significant associations
with prayer use. Social support (r = .02, p = .53) was not significantly related to prayer
use.
Use of prayer was a dichotomous variable (either a person engaged in prayer or
did not), therefore chi-square and logistic regressions were used to determine if and how
34


determinants might predict use. A chi square analysis of independence was performed to
examine the relationship between the categorical variables of sex and ethnicity with
prayer. There was a significant difference among sex and use of prayer, y2 (1, N = 1325)
= 36.18,/) < .001. Women were more likely to engage in prayer for spiritual healing than
were men. Ethnicity was also examined using chi square, and there was no difference
between Non-Hispanic White and Hispanic participants, y2 (1, N = 1325) = 0.21 ,/> = .65.
Using these variables, a hierarchical logistic regression analysis was performed to
assess if psychological determinants increased the correct prediction of prayer use above
and beyond the predisposing, enabling, and need determinants. With all determinants
entered into the model, there was a good model fit, y2 (12, N = 616) = 54.74,/) < .001,
Nagelkerke R2 = 12. Overall classification was impressive in that it showed that the
model correctly predicted 71.6% of those who prayed. Of the predictors in the model,
sex, f$= 1.12,/) < .001, accessibility difficulties, j3= 0.42,/) = .001, and diastolic blood
pressure, j3= -0.02,/) < .05, were significant. Interestingly, although there was overall
good model fit, adding psychological variables into the model did not significantly
increase the model fit, y2 (3, N = 616) = 4.34,/) = 0.23, indicating that the addition of
psychological determinants to the model did not reliably distinguish those who utilized
prayer as a source of health care. Adding enabling determinants (income and accessibility
difficulties) was the only addition of predictors that significantly increased model fit, y2
(2, N = 616) = 14.51,/) < .001. Table 8 illustrates the individual contributions of the
predictors to the model.
35


Table 8
Logistic Regression Predicting Use of Prayer (V = 616).
Model 1 Model 2 Model 3 Model 4
Variables B X Lxp(B) B X' Exp(B) B 1 X* Lxp(B) B X Exp(B)
Age 0.00 0.06 1.00 0.01 0.21 1.01 0.00 0.01 1.00 0.00 0.01 1.00
Sex 0.99 26.71 2.68*** 1.06 28.15 2.87*** 1.04 15.39 2.84*** 1.12 17.01 3.07***
Income 0.01 0.07 1.01 0.02 0.13 1.02 0.01 0.03 1.01
Accessibility Difficulties 0.43 14.44 1.54*** 0.43 13.57 1.53*** 0.42 12.31 1.52***
Chronic Illnesses 0.07 1.10 1.07 0.05 0.65 1.06
Pain Severity -0.05 0.14 0.95 -0.04 0.10 0.96
ADL Interference 0.07 1.14 1.07 0.06 0.68 1.06
Waist:Hip Ratio 0.16 0.04 1.17 0.16 0.04 1.18
Diastolic Blood Pressure -0.02 4.18 0.98* -0.02 4.07 0.98*
Depression -0.01 0.63 0.99
Social Support 0.05 1.80 1.05
Quality of Life -0.14 2.83 0.87
x 28.74*** 43.26*** 50.40*** 54.74***
X2 A 14.51*** 7.14 4.34
Nagelkerke R: 0.07 0.10 0.11 0.12
*p < .05, **/>< .01, ***/><.001
LtJ
On


Preventative Care: Flu Shots. Only predictors that were significantly related to flu shots
were used in analyses to determine if psychological variables added more explained
variability of health care utilization than predisposing, existing, and need determinants.
Use of flu shots was a dichotomous variable (either a person received a flu shot or did
not), therefore chi-square and logistic regression were used to determine if and how
determinants might predict this type of preventative care. Based on Pearson correlational
analyses, ethnicity (r = -.80,p < .01), education (r = 11 ,P< .001), insurance (r=.06,p<
.05), place of care (r = .10, p < .01), distance from care (r = -.10, p < .01), self-rated
health (r = -.07, p < .05), total chronic illnesses (r = .11, p< .001), presence of pain (r =
07, p < .05), severity of pain (r = .07, p < .05), pain interference with ADLs (r = .07, p <
.05), and quality of life (r = -,01,p< .05) all had significant associations with flu shots.
Depression (r = .04,p = .19) and social support (r = -.03,p = .40) were not significantly
related to flu shots.
A chi-square analysis of independence was performed to examine the relationship
between sex and ethnicity and flu shots. There was a significant difference among sex
regarding flu shots, y2 (1, N =1360) = 3.72,/? = .05. Women were more likely to receive
a flu shot than men. Ethnicity was also examined using chi square, and there was a
significant difference between Non-Hispanic White and Hispanic participants, y2 (1, N
=1360) = 7.76,/? < .01, in that Hispanics were less likely to receive a flu shot than Non-
Hispanic Whites.
Using these variables, a hierarchical logistic regression analysis was performed to
assess whether the addition of psychological determinants increased prediction of flu
shots beyond the selected predisposing, enabling, and need determinants. The model was
37


Tabic 9
Logistic Regression Predicting Use of Flu Shots (N = 654).
Model 1 Model 2 Model 3 Model 4
Variables B X' Exp(B) B X' Exp(B) B 1 X* Exp(B) B X' Exp(B)
Age 0.03 4.38 1.03* 0.02 3.00 1.02 0.02 2.83 1.02 0.02 2.96 1.02
Ethnicity -0.12 0.40 0.88 -0.07 0.11 0.94 0.09 0.19 1.10 0.11 0.26 1.11
Education 0.06 6.04 1.07** 0.06 5.12 1.06* 0.07 6.43 1.07** 0.07 6.41 1.07**
Insurance 0.42 0.33 1.52 0.09 0.02 1.10 0.13 0.03 1.14
Distance from Care -0.41 6.78 0.66** -0.39 5.81 0.68* -0.40 5.84 0.67**
Self-Reported Health -0.08 0.45 0.92 -0.10 0.58 0.91
Chronic Illnesses 0.25 11.39 1.28*** 0.26 12.01 1.29***
Pain Frequency -0.11 1.27 0.90 -0.09 0.99 0.91
Pain Severity 0.10 0.83 1.11 0.10 0.79 1.11
ADL Interference 0.06 1.00 1.06 0.07 1.22 1.07
Diastolic Blood Pressure -0.01 1.43 0.99 -0.01 1.45 0.99
Depression -0.01 1.09 0.99
Social Support -0.02 0.37 0.98
Quality of Life -0.01 0.01 0.99
x2 14.97' ** 22.12*** 45.90*** 47.30***
JCA 7.15* 23.78*** 1.40
Nagelkerke Rr 0.03 0.05 0.10 0.10
*p < .05, **p < .01, ***/> < .001
LtJ
00


a good fit, x2 (14, N = 654) = 47.30,/> < .001, Nagelkerke R2 = .10. Overall, classification
in the model that included psychological variables correctly predicted 68.0% of those
who received a flu shot. Table 9 illustrates the hierarchical contributions of the predictors
to the model. Adding all three psychological variables into the model increased the
predicted correct cases by 2.4% more than the model with just predisposing, enabling,
and need determinants. However, while adding enabling and need determinants into the
model significantly increased model fit, x2 (5, N = 654) = 7.15,/> < .05 and x2(ll,N =
654) = 23.78,p < .001, respectively, adding psychological variables into the model did
not significantly increase model fit, x2 (14, N = 654) = 1.40,/) = 0.71. Although
psychological variables added correct classification, the addition was not statistically
significant for a better overall model fit. Of the predictors in the final model, education, (>
= 0.07, p < .01, distance from care, f$= -0.40,/) < .01, and chronic illnesses, f$= 0.26, p <
.001, were significant. Age was significant in the first model, j3= 0.03,/) < .05, but
dropped out after other predictors were entered into the model.
Preventative Care for Men: Prostate Exams. Only predictors that were significantly
related to prostate exams were used in analyses to determine if psychological variables
added more explained variability of health care utilization than predisposing, existing,
and need determinants. Receiving a prostate exam was a dichotomous variable (either a
person received a prostate exam or did not), therefore chi square and logistic regression
were used to ascertain if these determinants might predict this type of preventative care
for men. A chi square analysis of independence was performed to examine the
relationship between ethnicity and prostate exams. Hispanic men were less likely to
receive a prostate exam than Non-Hispanic White men, x2 (1, N =515) = 10.17,/) < .001.
39


Based on Pearson correlational analyses, ethnicity (r = -.14,/> < .01), education (r
= .15,p < .01), place of care (r = .12, p < .01), and chronic illnesses (r = .13,p< .01)
were significantly related to prostate exams. None of the psychological variables were
significantly related to prostate exams (depression, r = .05,p = .25; social support, r =
.05 ,p= .32; and perceived quality of life, r = -.08 ,p =.10).
Using these variables, a hierarchical logistic regression analysis was performed to
assess whether the addition of psychological determinants increased prediction of
prostate exams beyond the selected predisposing, enabling, and need determinants. There
was a good model fit, %2 (7, N = 460) = 30.31, p < .001, Nagelkerke R2 = .09. Overall, the
model with psychological variables correctly predicted 59.8% of those who received a
prostate exam. Adding predisposing and enabling determinants into the model
significantly increased model fit, x2 (2, N = 460) = 17.11,/? < .001 and x2 (1, N = 460) =
10.36,p< .001, respectively, but adding need variables did not significantly increase
model fit, x2 (1, N = 460) = 1.84,p = 0.18. Furthermore, the addition of psychological
variables into the model did not significantly increase model fit, x2 (3, N = 460) = 1.00,p
= 0.80 indicating that the addition of psychological determinants to the model did not
reliably distinguish men who received prostate exams from those who did not. Of the
predictors in the final model, education, f$= 0.07,/) < .05, and having a place of usual
care, j3= 2.32, p < .05, were significant. Table 10 illustrates the hierarchical contributions
of the predictors to the model.
40


Tabic 10
Logistic Regression Predicting Prostate Exams (N = 460).
Model 1 Model 2 Model 3 Model 4
Variables B X' Exp(B) B X' Exp(B) B 1 X Exp(B) B X Exp(B)
Ethnicity -0.39 3.28 0.68 -0.41 3.56 0.67* -0.34 2.35 0.71 -0.33 2.14 0.72
Education 0.07 6.25 1.07** 0.07 6.05 1.07** 0.07 6.05 1.07** 0.07 5.73 1.07*
Place of Care 2.46 5.51 11.69** 2.35 5.01 10.51* 2.32 4.86 10.16*
Chronic Illnesses 0.10 1.82 1.11 0.10 1.56 1.10
Depression 0.00 0.05 1.00
Social Support 0.03 0.91 1.03
Quality of Life -0.02 0.06 0.98
r 17.11*** 27.47*** 29.31' * 30.31***
x2^ 10.36*** 1.84 1.00
Nagelkerke R: 0.05 0.08 0.08 0.81
*p < .05, **p < .01, ***p < .001
-P*.


Preventative Care for Women: Mammograms and pap smears. Use of preventative
care by women included both mammograms and pap smears. These preventative care
exams for women were used in analyses to determine if psychological variables added
more explained variability of health care utilization than predisposing, existing, and need
determinants. Both were dichotomous variables (either a person received the procedure or
did not), therefore chi square and logistic regression was used to determine if these
determinants might predict these types of preventative care. A chi square analysis of
independence was performed to examine the relationship between ethnicity and these
procedures. Neither was significant in that Hispanic and Non-Hispanic White women did
not significantly differ in terms of receiving either mammograms or pap smears.
Only predictors that were significantly related to mammograms were used in the
regression analyses. Age (r = -.19, p < .01), education (r = .12, p < .01), income (r = .18,
p < .01), place of care (r = .09, p < .05), self-rated health (r = .09,p < .05), frequency of
pain (r = -.12,p < .01), depression (r = -.10,p < .01), and perceived quality of life (r =
08 ,p < .05) were significantly correlated with mammograms using Pearson correlations.
Using these variables, a hierarchical logistic regression analysis was performed to assess
prediction of mammograms with the addition of all three psychological variables of
depression, social support, and perceived quality of life with the selected predisposing,
enabling, and need determinants. The variable of having a place of usual health care was
dropped because of multicollinearity. With the remaining variables, there was a good
model fit, %2 (8, N = 346) = 26.63,p< .01, Nagelkerke R2 = .10. Overall, the model with
psychological variables was somewhat impressive in that it correctly predicted 63.6% of
those who received a mammogram. The addition of need determinants into the model
42


significantly increased the model fit, %2 (5, N = 346) = 7.48,/) < .05, but adding
psychological variables into the model did not significantly increase model fit, %2 (8, N =
346) = 5.61,/) = 0.13. The odds ratio of all three psychological predictors was not
significant. Of the predictors in the final model, age, j3= -0.04,/) < .01, and frequency of
pain, f$= -0.26, p< .05, were significant. Table 11 illustrates the hierarchical
contributions of the predictors to the model for mammograms.
For the outcome of pap smears, only two predictors were significantly related to
the outcome. Only age (r = -.01 ,P< .01) and quality of life (r = .18,/) < .01) was
significant using Pearson correlations. Therefore, a regression model using all variables,
except for pain frequency and place of usual care, which were both dropped due to
multicollinearity, was performed for the pap smear outcome to better understand how all
of the predictor variables contributed to the overall model. Using all of the variables, a
hierarchical logistic regression analysis was performed to assess whether the addition of
psychological determinants increased prediction of pap smears beyond the selected
predisposing, enabling, and need determinants. The model was a good fit, %2 (18, N =
327) = 32.21,/) < .05, Nagelkerke R2 = .13. Overall, classification in the model that
included psychological variables was unimpressive in that it correctly predicted 61.5% of
those who received a pap smear. However, adding all three psychological variables into
the model increased the predicted correct cases by 1.6% more than the model with just
predisposing, enabling, and need determinants. For the outcome of pap smears, adding
psychological variables (depression, social support, and quality of life) significantly
increased overall model fit, %2 (3, N = 327) = 12.15,/) < .01, although of the three
43


Table 11
Logistic Regression Predicting Mammograms (N = 346).
Model 1 Model 2 Model 3 Model 4
Variables B X* Exp(B) B X' Exp(B) B X* Exp(B) B t X Exp(B)
Age -0.05 7.94 0.96** -0.04 6.34 0.96** -0.05 7.47 0.96** -0.04 6.79 0.96**
Education 0.07 3.74 1.07* 0.04 1.26 1.04 0.03 0.53 1.03 0.04 0.96 1.04
Income 0.08 1.63 1.08 0.06 0.76 1.06 0.05 0.49 1.05
Self-Rated Health 0.14 0.92 1.15 0.05 0.07 1.05
Pain Frequency -0.29 4.45 0.75* -0.26 3.68 0.77*
Depression -0.01 0.45 0.99
Social Support -0.05 1.75 0.95
Quality of Life 0.17 2.47 1.19
x2 11.90** 13.54** 21.02* * 26.63***
5fA - 1.64 7.48* 5.61
Nagelkerke R: 0.05 0.05 0.08 0.10
*p < .05, **p < .01, ***/> < .001
-P*.
-P*.


Table 12
Logistic Regression Predicting Pap smears (V = 327).
Model 1 Model 2 Model 3 Model 4
Variables B X' Exp(B) B X Exp(B) B X* Exp(B) B X' Exp(B)
Age -0.05 7.43 0.96** -0.05 8.15 0.95** -0.06 9.27 0.94** -0.06 8.93 0.94**
Ethnicity 0.36 1.84 1.43 0.37 1.82 1.45 0.41 2.00 1.51 0.20 0.45 1.23
Education 0.01 0.01 1.00 0.01 0.05 1.01 0.01 0.04 1.01 0.02 0.13 1.02
Income -0.01 0.05 0.99 -0.03 0.14 0.98 -0.05 0.42 0.96
Insurance 1.59 1.94 4.92 1.61 1.93 5.02 2.01 2.80 7.47
Distance from Care 0.00 0.00 1.00 0.04 0.03 1.04 0.05 0.05 1.05
Accessibility Difficulties -0.01 0.00 0.99 0.05 0.07 1.05 0.17 0.78 1.18
Self-Rated Health 0.00 0.00 1.00 -0.17 0.82 0.85
Chronic Illnesses -0.12 1.82 0.89 -0.11 1.37 0.90
Pain Severity 0.00 0.00 1.00 -0.04 0.06 0.96
Pain Frequency -0.22 2.42 0.80 -0.21 2.02 0.81
ADL Interference 0.01 0.03 1.01 0.08 0.81 1.08
Waist:Hip Ratio -0.01 0.00 0.99 0.04 0.00 1.04
Systolic Blood Pressure 0.01 0.58 1.01 0.00 0.13 1.00
Diastolic Blood Pressure -0.02 1.29 0.98 -0.01 0.91 0.99
Depression 0.00 0.07 1.00
Social Support -0.02 0.16 0.98
Quality of Life 0.40 10.60 1.49***
X 11.47** 14.04* 20.06 32.21*
X& 2.57 6.02 12.15*
Nagelkerke R: 0.05 0.06 0.80 0.13
*p<.05, **/><.01,, ***/><.001


variables, only quality of life was significant predictor in the model, /3= 0.40,p < .001.
Depression, /3 = 0.00, p = 0.80, and social support, /3= -0.02,p = .69, were not
significantly adding to overall model fit for pap smears. Quality of life appeared to drive
this overall significant effect. Interestingly, the only other predictor that was significant in
the model was age, (3 = -0.06, p < .01. Table 12 illustrates the hierarchical contributions
of the predictors to the model for pap smears.
Hypothesis 2
To test hypothesis 2, an exploratory principal component factor analysis was
performed to explore how factors based on the predictors in the sample from the San Luis
Valley region may differ from those posited in Andersens model. Oblimin rotation was
used to allow the factors to correlate since the different domains of determinants are not
independent. To determine composition of the factors, it was determined that variables
could not load onto more than one factor and the highest loading would be determined for
each variable and load only onto that factor. In an effort to simply understand how these
variables related to one another, no cutoff point was used to exclude individual variables;
even if they did not meet a .50 loading rule, they were still included in the factors. To
determine factors, the following decision rules were used: (a) eigenvalue greater than 1;
(b) examination of the scree plot; (c) items only loading on one factor; and (d)
interpretability of the factor. Using these criteria, six factors were extracted, although not
all variables loaded highly (> .50) on each. On the Psychosocial Characteristics factor,
depression (.66), quality of life (.70), social support (.29), and accessibility difficulties
(.35) loaded together. On the Physical Characteristics factor, sex (.90) and waist:hip ratio
(.73) loaded highly. On the Sociocultural Characteristics factor, ethnicity (.70), education
46


Table 13
Factor Loadings for Determinants
Variables 1 2 Factor 3 4 5 6
1. Psychosocial Characteristics*
Quality of Life -.70 -.03 .05 .00 -.03 .44
Depression .66 -.12 .20 .04 -.02 -.35
Accessibility Difficulties J5 .09 .18 .08 -.10 -.09
Social Support -29 .17 -.14 -.04 .08 .04
2 Physical Characteristics
Sex .13 -.90 -.02 -.03 .16 -.05
WaistiHip Ratio .07 .73 .17 .09 -.07 -.03
3. Sociocultural Characteristics'
Education -.14 -.11 -.71 -.15 .05 .05
Ethnicity .17 .13 .69 .16 -.24 .11
Income -.35 .14 -.62 -.11 -.21 .17
Distance from Care .03 .19 .21 .05 -.18 .02
4. Objective Health Indicators'1
Systolic Blood Pressure .04 -.03 .14 JO .13 -.16
Diastolic Blood Pressure .09 .12 .13 .78 -.22 .05
5. Age-Related Characteristics'
Age -.08 -.07 -.03 -.05 .72 -.08
Insurance -.05 .00 -.06 -.01 2\ -.17
6. Subjective Health Indicators*
Self-Reported Health -.42 -.07 -.35 -.08 .05 .60
Chronic Illnesses .16 -.04 -.09 .08 .16 -.52
ADL Interference .38 -.10 .10 -.01 -.09 -.43
Eigenvalue = 2.79; percentage of variance = 16.42. Eigenvalue = 2.07; percentage of
variance = 12.18. Eigenvalue = 1.66; percentage of variance = 9.76. Eigenvalue = 1.47;
percentage of variance = 8.66. 'Eigenvalue = 1.33; percentage of variance = 7.83.
Eigenvalue = 1.05; percentage of variance = 6.16.
(.71), income (.62), and distance from care (.21) loaded together, although distance from
care much lower than the others. On the Objective Health Indicators factor, systolic (.80)
and diastolic blood pressure (.78) loaded together. On the Age-Related Characteristics
factor, age (.72) and insurance (.21) loaded together, although insurance was much lower.
47


Finally, on the Subjective Health Indicators factor, self-reported health (.60), chronic
illnesses (.52), and pain interference with ADLs (.43) loaded well together. See Table 13
for factor loadings, eigenvalues, and percent variance accounted for by each factor.
48


CHAPTER IV
DISCUSSION
Health care utilization is an individualized behavior with many idiosyncratic
factors potentially interacting with one another. Andersens Behavioral Model of Health
Services Utilization (Andersen, 1995; Andersen & Newman, 1973) sought to create a
conceptual model explaining the factors that might influence health care use, including
factors that exist prior to a health need, factors that enable a person to seek care, and the
actual medical need of receiving services. This theory posited interactions among the
determinants but did not provide a clear categorization of psychological determinants.
Moreover, systematic analyses using this model, began to categorize and divide some
psychosocial variables such as mental diagnoses, social support, and health-related
quality of life into different domains of determinants (Babitsch, Gohl & von Lengerke,
2012), yet a clear explanation of how these psychological factors might influence the
overall model has yet to be succinctly posited. The present study aimed to explore how
psychological factors interact with the proposed determinants and conceptualize how a
rural, culturally specific populations health services utilization would fit onto
Andersens model.
Predisposing, Enabling, and Need Determinants of Health Care Utilization
It was hypothesized that Andersens proposed domains of determinants would
predict health care utilization similarly in this population to previously studied
populations (Andersen, 1995; Andersen & Newman, 1973; Wolinsky & Johnson, 1991;
Baxter, Bryant, Scarbro, & Shetterly, 2001), but psychological predictors would also play
a role in the explanation of health services utilization. Before addressing the
49


psychological variables, however, it is important to understand how the originally
proposed domains of determinants behaved with this sample. In the San Luis Valley
Health and Aging Study, participants were older and from a more isolated, rural area.
Because of this, the possibility remained that the predisposing, enabling, and need
determinants might have predicted somewhat differently in this sample. Given this, it was
interesting to observe how these types of factors fit into the overall model.
Understanding the demographic variables, alone, was important in predicting
types of health care use. Almost forty percent of individuals sampled did not finish high
school, and almost half had a yearly income of less than $10,000. This sample was
skewed in terms of demographics, and it is possible that individuals less educated and
below the poverty line utilize health care differently, possibly for a variety of reasons. It
is possible that age plays a larger role in this skewed sample. Age was significantly
negatively related to education and income so the older one is, the less likely they were to
have obtained higher levels of education and income. In all of the regression analyses for
utilization outcomes, predisposing characteristics alone in the model significantly
predicted the outcome, and age was a significant predictor in most regression analyses.
Given that the entire sample was 65 years of age and older, this is noteworthy and
possibly indicates that in the aging population, differences in age, even relatively small
ones, can influence well-being. Health-related changes tend to occur more rapidly in an
aging population compared to a middle-aged population.
Differences between men and women were also present in terms of utilization.
Women were significantly more likely to receive a flu shot than their male counterparts,
and they were also significantly more likely to pray than men. A conclusion can be made
50


that women utilize more types of health care than men. A difference in flu shots between
Non-Hispanic White and Hispanic participants was also present. Hispanics were less
likely to receive flu shots as a preventative measure, which mirrors previous research.
This finding could indicate that regardless of urban or rural location, Hispanics are
generally less likely to engage in preventative care.
Another related factor not measured in this study that could have impacted
outcomes was that of health literacy. Lower health literacy and greater obstacles, both
physical and mental, may create barriers to seeking health care. Cho, Lee, Arozullah, and
Crittenden (2008) explored how low health literacy affected health status and health
service utilization and found that low health literacy had a direct effect on poorer health
outcomes. Given this, it is interesting to note that these individuals sampled might have
had lower health literacy that was contributing to health outcomes. Also, a lower self-
efficacy to manage illnesses might also be playing a large role in their ability and
willingness to seek treatment.
Accessibility of health care did not appear to have a significant role in utilization.
In this sample, individuals did not endorse having many difficulties accessing health care.
In fact, out of six possible difficulties, the mean reported by individuals was less than 1,
giving evidence that those in this region did not feel these types of logistical or physical
barriers existed for them, but we are unaware of what other barriers may have existed,
especially emotional and cognitive. Of note is that accessibility difficulties were
negatively related to education, income, and insurance in that those who had less
education, income, and insurance perceived greater difficulties in accessing health care.
Accessibility difficulties were also positively related to self-reported health, number of
51


chronic illnesses, and pain. It appears that difficulties accessing care is due more to
poorer health and/or financial means than the physical availability of medical care clinics.
It is possible that in rural areas, access to primary care is comparable to that in more
urban areas (Hanratty, Zhang & Whitehead, 2007).
Based on previous literature regarding Andersens model (Andersen, 1995;
Andersen & Newman, 1973; Wolinsky & Johnson, 1991; Baxter, Bryant, Scarbro, &
Shetterly, 2001), need determinants tend to be most predictive of health services
utilization. That trend was seen in the present study, with self-reported health, total
number of chronic illnesses, pain, and pain interference with ADLs often contributing
significantly to models with different health care outcomes. Perhaps more surprising,
though, was that measures of adiposity (waist:hip ratio) and blood pressure were not
significant in most models. In the general population, obesity is a risk factor for myriad
health problems and subsequent visits to a medical professional. In the aging population,
one similar to this sample, the opposite can be true. Frailty in the aging population also
indicates pathology. Waisfhip ratio was negatively correlated with education, self-
reported health, and quality of life. Despite these significant relationships, this measure of
adiposity had little to no bearing on the predictive nature of utilization. Similar to
adiposity, hypertension is a risk factor but did not contribute to utilization outcomes in
this sample. It is quite possible that these need variables are so closely related that more
significant pathology or diagnosed illnesses has a greater impact than risk factors for
those illnesses.
All of these variables most likely have a differing role depending on the type of
outcome measured. In this study, three types of outcomes were measured, but it is not
52


entirely clear exactly what those mean in terms of pathology or appropriate behaviors.
Frequency of visits to a clinic or emergency room, for example, could be both negative
and positive. It is difficult to determine what an appropriate number of visits would be
given a persons medical history and presentation. Similarly, receiving preventative care
could be indicated if a person is more at risk for medical complications. In the elderly,
receiving a flu shot is perhaps more important than for a younger individual. Diseases
like the flu can have a far more detrimental effect for an elderly individual with comorbid
medical diseases than for a healthy young adult. Therefore, preventative care might
have a uniquely different connotation for a 75 year old immunocompromised male with
heart disease and COPD than for a 25 year old male with no medical problems and a
healthy immune system.
Psychological Determinants of Health Care Utilization
Psychological variables can and do have a significant influence on decisions, both
medical and otherwise, as well as perception of state. What was most notable in this
study was the seemingly lack of individual contribution by psychological determinants on
the outcomes of health care utilization. In almost every model, other factors, especially
need determinants, seemed to predict outcomes stronger than psychological components.
This observation falls in line with the Andersens originally hypothesized models.
However, psychological determinants were significantly related to many other variables
within all three domains of predisposing, enabling, and need determinants. Given this,
psychological determinants may help to explain how predisposing, enabling, and need
determinants predict health care utilization rather than predict on their own.
53


There was often a negative relationship between perceived quality of life and
utilization, both visits to a clinic and preventative care. These data suggest that as
perceived quality of life increases, visits to a clinic or emergency room decrease. Visiting
a doctor likely means that you have more medical complications, or at least more
perceived medical complications, thus indicating poorer quality of life. Moreover,
depression often had a positive relationship to outcomes so that as depression increased,
visits increased. Likely a bidirectional relationship is occurring in that as depression
increases, one is more likely to experience and/or catastrophize somatic symptoms
leading him or her to seek treatment more often. It is also possibly quite true that as
medical conditions worsen, depression worsens as a consequence. The inverse
relationship between perceived quality of life and depression could be indicative of
coping ability and styles. The moderation analyses that provided the most insight into this
was between depression and number of chronic illnesses. When a person had fewer
chronic illnesses, higher depression indicated a decreased frequency in visits for a health
treatment, which could possibly be categorized as underutilization. As the number of
chronic illnesses increased, however, higher depression resulted in a greater frequency of
visits. Though no other moderation analyses were significant with quality of life or
depression, this interaction effect sheds light onto a possible mechanism behind the
psychological variables. Perhaps as overall health declines, sickness behavior becomes
more susceptible to influences from negative mood.
Temporality is difficult to untangle given the interactive nature of these
constructs. These data would suggest that more visits to medical place of service are
indicative of pathology from a holistic point of view (depression, quality of life, medical
54


illnesses, etc.). A problem with these data, however, is the combination in the outcome of
both clinic visits and visits to an emergency department. Separating these two locations
into two different outcomes would likely provide a clearer picture of positive vs. negative
behavior, as well as the mediating and moderating factors playing a role in seeking types
of treatment.
One area of research that could provide an explanation to the mediating effects of
depression and other psychological variables on health care utilization outcomes is the
role of inflammation. The inflammatory hypothesis examines how
psychoneuroimmunological dysfunction may contribute to the manifestation of
depression, sickness behavior, and actual health outcomes (Zunszain, Hepgul & Pariante,
2013). Inflammatory cytokines have been shown to be associated with depression,
especially treatment-resistant depression (Raison & Miller, 2011). It is also well
documented that inflammation is known to contribute to poor health outcomes. In this
population, immunosenescence, or the decline of immune functioning due to aging, is
more apparent and involves the decreased capacity to fight infections (McElhaney &
Effros, 2009). Related to this, aging individuals often have overproduction of
inflammatory markers, most notably C-reactive protein (Aiello, Haan, Pierce, Simanek &
Liang, 2009). The aging are more likely to have decreased immunological functioning
and a greater number of morbidities. This prominent inflammation in these individuals
could contribute to both depression and the perception of ones ailments and therefore
should be taken into account when assessing determinants that predict health care
utilization.
55


The third psychological variable measured failed to contribute to outcomes across
the board. Social support never seemed to play a role in predictive nature in the models
tested. This construct of social support was based on number of family and friends, rather
than the qualitative nature of the support provided and perceived. Yohannes (2013) stated
that quality is much more indicative of adjustment than quantity. Given the fact that this
sample is an aging population with somewhat limited capabilities, caregiving and the
support from caregivers might be an even more important construct to measure than
social support alone.
Analyzing Determinant Domains Within the Model
To examine the second aim of the study, an exploratory factor analysis was
performed to better understand how the determinants load together in this sample and
compare them to Andersens proposed determinants. Six factors were extracted with this
sample, which is more than the three proposed by Andersen, but these six factors still did
not entirely describe the full picture. The first factor, Psychosocial Characteristics, was
comprised of the psychological determinants of depression, social support, and quality of
life. Surprisingly, accessibility difficulties also loaded onto this factor, although not
highly. It is possible that psychological factors such as depression and quality of life can
influence the way barriers to care are perceived. Social support did not have a strong
loading to any factor, which is not surprising given its limited contribution to models in
this study and measure of quantity rather than quality.
Education, ethnicity, income, and distance from care created the factor of
Sociocultural Characteristics. These variables were significantly related to one another
and might help to explain the availability of resources in this population. A separate, but
56


related, factor was found that included age and insurance. This Age-related
Characteristics factor by itself is difficult to interpret. In this population, it seems that age
acts uniquely from other demographic variables. It loaded highly on this factor (.72)
compared to the other variable of insurance, which loaded at .21. Because of the lack of
variability in this sample regarding insurance, it is unclear how much it truly impacting
outcomes. It seems as though for interpretability purposes, insurance could be dropped
altogether from the factors, thus creating a factor of just age and then another of
sociocultural characteristics influencing availability to resources.
A Physical Characteristics factor including both sex and waist:hip ratio was
determined to be separate from all other factors. Similar to the Age-related
Characteristics factor, this factor is difficult to interpret on its own. Both could be related
to other health indicators, and in this sample, health indicators were clearly divided into
subjective and objective. Systolic and diastolic blood pressure loaded together on
Objective Health Indicators, and self-reported health, chronic illnesses, and ADL
interference due to pain loaded onto Subjective Health Indicators. These two separate
factors support the idea of need being a large component of health care utilization. It
seems that indicators that are influenced by perception differ from those that do not
(blood pressure and measures of adiposity). It is possible that these aspects of need are
much more salient to an individual than the objective measures such as adiposity or blood
pressure. This also begs the question of how psychological variables might be influencing
subjective need determinants. This factor of subjective need correlated most highly with
depression and quality of life. These data give evidence that psychological factors
57


influence not just one determinant, but they can have a systemic influence of all
determinants.
If insurance is dropped and the physical characteristics factor is not included for
lack of interpretability, five factors remain: Psychosocial Characteristics, Sociocultural
Characteristics, Age, Objective Health Indicators, and Subjective Health Indicators.
Comparing these to Andersens proposed Predisposing determinants, Enabling
determinants, and Need determinates, it seems that conceptually, psychological variables
constitute a completely different domain of determinants influencing outcomes of health
care utilization. The Sociocultural factor closely resembles that of Andersens Enabling
determinants. Both are comprised of factors the influence availability of resources.
Interestingly, need determinants were broken into subjective and objective, but this
finding illustrates the importance of this factor in health care utilization. Andersens
enabling and need determinants were shown to be present in some respect in the present
study. However, in general, Andersens predisposing determinants loaded onto different
factors in this population, possibly due to the high correlations among variables. No clear
factor that resembled his predisposing determinants was seen in this study. In this
population, age clearly was a unique variable, and it is difficult to ascertain how this
variable might behave in other populations. Given these results, it appears that a separate
factor of psychological factors should be considered, at least in a conceptual model.
Amending the Conceptual Model of Health Services Utilization
Andersen originally postulated that many, if not all, of the determinants would
interact with one another. One of the strengths and weaknesses of his model is that it is
more of an abstract model rather than an exhaustive list of predictive factors. It takes
58


more of a systems-based approach and does not specify certain variables but rather gives
a general overview of the types of constructs that might influence utilization. As fields of
literature change and deepen, this model can be used in different situations and from
different viewpoints (e.g. inflammatory hypothesis in psychoneuroimmunology). This
study set out to better understand how psychological determinants influence the existing
model of health services utilization and empirically test Andersens conceptual model.
What became readily apparent, though, was the difficulty in empirically separating
determinants and their interactions with psychological variables, especially given the
limited number of determinants measured in the study. It appears that there is a great deal
of multicollinearity that occurs between these variables. Although psychological
variables seldom significantly added to the overall models, it is clear that they impact
various other determinants. Given the results and conclusions thus far, it would seem
appropriate that amending Andersens conceptual model to include psychological
variables as a domain in itself that influences the other domains would provide a more
robust conceptualization of variables that could influence overall health care utilization.
In addition, empirically, different factors are extracted, and although they somewhat
resemble Andersens factors, they did differ to some extent (e.g. need is separated into
objective and subjective factors and predisposing determinants are spread out among
various factors). However, to conceptually guide future research, it is likely that the best
way to think about domains influencing health care utilization is similar to Andersens
approach: age (or other basic demographic information), sociocultural factors, and need
determinants. Figure 6 illustrates this conceptual model.
59


A Study of Culture
A particular noteworthy piece of this study was the sample itself. Because of its
targeted sampling of ethnicity, age, and rural residency, all of which could influence
health care utilization in various ways, the San Luis Valley Health and Aging Study,
conducted in southern Colorado, provided a unique population to test Andersens model
in a comprehensive manner. Due to its uniqueness, one could posit that this particular
sample held its own cultural influences on health care utilization.
Prayer, for instance, could be much more of a culturally specific behavior rather
than a type of health care. It is debated in the literature whether prayer constitutes
complementary and alternative medicine, but in this population, it seems as though it
would be more complementary than alternative. Prayer is a normal part of Hispanic
culture, and given the sample of the study, it is plausible that regardless of ethnicity,
prayer is a major part of the culture. Similarly, the way depression and quality of life are
60


viewed can vary with cultural contexts. Integrating cultural beliefs and practices into the
model can help to provide a more holistic conceptualization of health care utilization.
Strengths and Limitations
As discussed, the sample in this study provided one of its biggest strengths, as did
the number of variables able to be measured. Within each study, it is impossible to
measure all variables that might be influencing an outcome. However, a limitation of the
study was the lack of data regarding health beliefs. These beliefs can strongly influence
all determinants, and because of the lack of measure assessing this, it is difficult to
ascertain how these may have interacted with determinants, especially the psychological
determinants. Nevertheless, this does not diminish the need for the addition of
psychological variables to the model. As mentioned, the sample, although unique in its
ability to give insight into this population, also can be a weakness in that there is
somewhat limited generalizability, although the argument for assessing culture gives
weight to this as a strength. Perhaps the greatest limitation was that the data were from a
previous study and were collected almost twenty years ago. Measures and questionnaires
were not asked in a way that best facilitated the aims of the study, and therefore it became
difficult to statistically analyze the data in a way that best answered the studys questions.
Future Directions
This line of research could go in many different future directions, some of which
have already been addressed. First, assessing health beliefs will be crucial in future
studies, especially to determine if and how they interact with psychological variables.
Also, in future studies, social support should be a measure of quality rather than quantity,
and if aging populations are used, caretakers will be an important component to assess.
61


Health literacy and self-efficacy related to health might be important routes to
consider as well. These constructs might play significant roles in utilizing health care,
especially in populations that are more rural and/or have residents with lower
socioeconomic status or education level. It might also show differences in receiving
different types of health care (emergency room visits vs. visits to a primary care clinic).
One aspect that may provide additional insight into an aging populations
utilization of health care is that of cognitive functioning. This was beyond the scope of
this study, but there is research that suggests links between cognitive functioning,
inflammation, disease, and depression. This would be an interesting way to conceptualize
health care utilization from a more psychoneuroimmunological framework. This also
might have ties to the question of adherence or compliance to medical regimens.
Investigations within this area of research could likely provide substantial insight into
both the area of health care utilization and the inflammatory hypothesis.
Conclusions
Continuing to test models of health care utilization in different populations with
different facets emphasized will help to better understand the pathways of utilization in a
more holistic manner. However, revising Andersens Model of Health Services
Utilization to include psychological variables in a systems-based approach will help
conceptualize health care utilization in different populations and cultures.
62


REFERENCES
Abdel-Khalek, A. M. (2014). Happiness, health, and religiosity: significant associations
among Lebanese adolescents. Mental Health, Religion & Culture, 77(1), 30-38.
doi:l 0.1080/13674676.2012.742047
Aiello, A. E., Haan, M. N., Pierce, C. M., Simanek, A. M., & Liang, J. (2008). Persistent
infection, inflammation, and functional impairment in older Latinos. The Journals of
Gerontology. Series A, Biological Sciences and Medical Sciences, 63(6), 610-618.
http://doi.Org/63/6/610
Apel, M. D. (1986). The attitudes and knowledge of church members and pastors related
to older adults and retirement. Journal of Religion and Aging, 293, 31-43.
Andersen, R. M. (1995). Revisiting the behavioral model and access to medical care:
does it matter? Journal of health and social behavior, 36(1), 1-10.
Andersen, R., & Newman, J. F. (1973). Societal and individual determinants of medical
care utilization in the United States. The Milbank Memorial Fund quarterly. Health and
society, 57(1), 95-124.
Andersen, R., & Newman, J. F. (2005). Societal and individual determinants of medical
care utilization in the United States. Milbank Quarterly, 53(4), 1-28.
Arcury, T. A., Gesler, W., M., Preisser, J. S., Sherman, J., Spencer, J., & Perin, J. (2005).
The effects of geography and spatial behavior on health care utilization among the
residents of a rural region. Health Services Research, 40(1), 135-156.
Babitsch, B., Gohl, D., & von Lengerke, T. (2012). Re-revisiting Andersens Behavioral
Model of Health Services: A systematic review of studies from 1998-2011.
Psychosomatic Medicine, 9, doi: 10.3205/psm000089
Baxter, J., Bryant, L. L., Scarbro, S., & Shetterly, S. M. (2001). Patterns of Rural
Hispanic and Non-Hispanic White Health Care Use: The San Luis Valley Health and
Aging Study. Research on Aging, 23(1), 37-60. doi: 10.1177/0164027501231003
63


Baxter J., Shetterly S. M., Eby C., Mason L., Cortese C. F., Hamman R. F. (1998) Social
network factors associated with Perceived Quality of Life. Journal of Aging and Health,
10(3), 287-310.
Bean, F.D., & Tienda, M. (1987). The Hispanic population of the United States. New
York, NY: Russell Sage Foundation.
Benjamins, M. R. (2005). Social Determinants of Preventive Service Utilization: How
Religion Influences the Use of Cholesterol Screening in Older Adults. Research on
Aging, 27(4), 475-497. doi: 10.1177/0164027505276048
Benjamins, M. R. (2006). Religious influences on preventive health care use in a
nationally representative sample of middle-age women. Journal of behavioral medicine,
29(1), 1-16. doi: 10.1007/sl0865-005-9035-2
Benjamins, M., & Brown, C. (2004). Religion and preventative health care utilization
among the elderly. Social Science & Medicine, 55(1), 109-118. doi:10.1016/S0277
9536(03)00152-7
Berkanovic, E., Telesky, C., & Reeder, S. (1981). Structural and social psychological
factors in the decision to seek medical care for symptoms. Medical Care, 19, 693-709.
Birrer, R. B., & Vemuri, S. P. (2004). Depression in later life: a diagnostic and
therapeutic challenge. American Family Physician, 69(10), 2375-2382.
Blount, A., Schoenbaum, M., Kathol, R., Rollman, B., Thomas, M., ODonohue, W., &
Peek, C. J. (2007). The economics of behavioral health services in medical settings: A
summary of the evidence. Professional Psychology: Research and Practice, 35(3), 290-
297.
Boyd, C. M., Darer, J., Boult, C., Fried, L. P., Boult, L., & Wu, A. W. (2005). Clinical
practice guidelines and quality of care for older patients with multiple comorbid diseases:
Implications for pay for performance. JAMA, 294(6), 716-724.
Brownson, C. A., & Heisler, M. (2009). The role of peer support in diabetes care and
self-management. Patient, 2(1), 5-17.
64


Bureau of Labor Statistics, United States Department of Labor. (2014). CPI (Consumer
Price Index) Calculator. Available from http://www.bls.gov/data/inflation_calculator.htm
Butterworth, S. W., Linden, A., & McClay, W., (2007). Health coaching as an
intervention in health management programs. Disease Management & Health Outcomes,
75(5), 299-307.
Caldwell E. M., Baxter J., Mitchell C. M., Shetterly S. M., Hamman R. F. (1998) The
association of non-insulin-dependent diabetes mellitus with perceived quality of life in a
biethnic population: The San Luis Valley Diabetes Study. American Journal of Public
Health. 88 (8), 1225-1229.
Carey, T. S., Garrett, J. M., & Jackman, A. M. (2000). Beyond the good prognosis.
Examination of an inception cohort of patients with chronic low back pain. Spine, 25(1),
115-120.
Carlisle, D. M., Leake, B. D., & Shapiro, M. F. (1997). Racial and ethnic disparities in
the use of cardiovascular procedures: Associations with type of health insurance.
American Journal of Public Health, 87(2), 263-267.
Carney, R. M., Freedland, K. E., Eisen, S. A., Rich, M. W., & Jaffe, A. S. (1995). Major
depression and medication adherence in elderly patients with coronary artery disease.
Health Psychology, 14, 88-90.
Cho, Y. I., Lee, S.-Y. D., Arozullah, A. M., & Crittenden, K. S. (2008). Effects of health
literacy on health status and health service utilization amongst the elderly. Social Science
& Medicine, 66(8), 1809-1816. http://doi.Org/10.1016/j.socscimed.2008.01.003
Coffield, A. B., Maciosek, M. V., McGinnis, J. M., Harris, J. R., Caldwell, M. B.,
Teutsch, S. M., Atkins, D., Richland, J. H., Haddix, A. (2001). Priorities among
recommended clinical preventative services. American Journal of Preventative Medicine,
27(1), 1-9.
Cote, P., Cassidy, J. D., & Carroll, L. (2001). The treatment of neck and low back pain:
who seeks care? who goes where? Medical care, 39(9), 956-67.
65


DCrus, A. & Wilkinson, J. M. (2005). Reasons for choosing and complying with
complementary health care: An in-house study on a south Australian clinic. The journal
of alternative and complementary medicine, 11, 6, 1107-1112.
de Boer, A., Wijker, W., & de Haes, H. (1997). Predictors of health care utilization in the
chronically ill: A review of the literature. Health Policy, 42, 101-115.
Dunlop, D. D., Manheim, L. M., Song, J., & Chang, R. W. (2002). Gender and
ethnic/racial disparities in health care utilization among older adults. The journals of
gerontology. Series B, Psychological sciences and social sciences, 57(4), S221-33.
Eisenberg, D. M., Kessler, R. C., Foster, C., Norlock, F. E., Calkins, D. R., & Delbanco,
T. L. (1993). Unconventional medicine in the United States, New England Journal of
Medicine, 328, 246-252.
Edwards, W. S., Winn, D. M., & Kurlantzick, V. (1994). Evaluation of National Health
Interview Survey Diagnostic Reporting. National Center for Health Statistics. Vital and
health statistics, 2(120), 1-116.
Gask, L., Dowrick, C., Salmon, P., Peters, S., & Morriss, R. (2011). Reattribution
reconsidered: Narrative review and reflections on an educational intervention for
medically unexplained symptoms in primary care settings. Journal of psychosomatic
research, 71, 325-334.
Glassman, A. H., Helzer, J. E., Covey, L. S., Cottier, L. B., Stetner, F., Tipp, J. E., &
Johnson, J., (1990). Smoking, smoking cessation, and major depression. JAMA, 300,
1546-1549.
Gornick, M. E., Eggers, P. W., Reilly, T. W., Mentnech, R. M., Fitterman, L. K., Kucken,
L. E., & Vladeck, B. C. (1996). Effects of race and income on mortality and use of
services among Medicare beneficiaries. New England Journal of Medicine, 335, 791-799.
Hamman, R. F., Mulgrew, C. L., Baxter, J., Shetterly, S. M., Swenson, C., &
Morgenstem, N. E. (1999). Methods and prevalence of ADL limitations in Hispanic and
non-Hispanic white subjects in rural Colorado: the San Luis Valley Health and Aging
Study. Annals of epidemiology, 9(4), 225-35.
66


Hanratty, B., Zhang, T., & Whitehead, M. (2007). How close have universal health
systems come to achieving equity in use of curative services? A systematic review.
International journal of health services, 37(1), 89-109.
Helman, C. G. (1994). Culture, Health and Illness: An Introduction for Health
Professionals. Bristol, UK: Butterworth-Heinmann.
Hershey, J. C., Luft, H. S., & Gianaris, J. M. (1975). Making sense out of utilization data.
Medical Care, 73(10), 838-854.
Hispanic health and nutrition examination survey, 1982-84 (1990). Findings on health
status and health care needs. American Journal of public health, (80), 1-70.
Hybels, C. F., & Blazer, D. G. (2003). Epidemiology of late-life mental disorders. Clinics
in Geriatric Medicine, 79(4), 663-696.
Kathol, R. G., McAlpine, D., Kishi, Y., Spies, R., Meller, W, Bernhardt, T., ... Gold, W.
(2005). General medical and pharmacy claims expenditures in users of behavioral health
services. Journal of General Internal Medicine, 20, 160-167.
Keeley, P., Creed, F., Tomenson, B., Todd, C., Borglin, G., & Dickens, C. (2008).
Psychosocial predictors of health-related quality of life and health service utilisation in
people with chronic low back pain. Pain, 135, 142-150.
Koenig, H. G., & Larson, D. B (1998). Use of hospital services, religious attendance, and
religious affiliation. Southern Medical Journal, 91, 925-932.
Kolk, A. M., Hanewald, G. J. F. P., Schagen, S., & Gijsbers van Wijk, C. M. T. (2002).
Predicting medically unexplained physical symptoms and health care utilization. A
symptom-perception approach. Journal of psychosomatic research, 52(1), 35-44.
Kolk, A. M., Hanewald, G. J. F. P., Schagen, S., & Gijsbers van Wijk, C. M. T. (2003). A
symptom perception approach to common physical symptoms. Social science & medicine
(1982), 57(12), 2343-54.
67


Kolk, A. M., Schagen, S., & Hanewald, G. J. F. P. (2004). Multiple medically
unexplained physical symptoms and health care utilization: Outcome of psychological
intervention and patient-related predictors of change. Journal of Psychosomatic
Research, 57, 379-389.
Kornstein, S. G. (1997). Gender differences in depression: implications for treatment.
Journal of Clinical Psychiatry, 58, 12-8.
Le, T. D., Carney, P. a, Lee-Lin, F., Mori, M., Chen, Z., Leung, H., Lau, C., et al. (2014).
Differences in knowledge, attitudes, beliefs, and perceived risks regarding colorectal
cancer screening among Chinese, Korean, and Vietnamese sub-groups. Journal of
community health, 39(2), 248-65. doi:10.1007/sl0900-013-9776-8
Levin J. S., & Markides K. S. (1985). Religion and health in Mexican Americans.
Journal of Religion and Health, 24, 60-69.
Lewinsohn, P.M., Seeley, J.R., Roberts, R.E., & Allen, N.B. (1997). Center for
Epidemiological Studies-Depression Scale (CES-D) as a screening instrument for
depression among community-residing older adults. Psychology and Aging, 12, 277- 287.
McCathie, H. C., Spence, S. H., & Tate, R. L. (2002). Adjustment to chronic obstructive
pulmonary disease: The importance of psychological factors. European Respiratory
Journal, 79(1), 47-53.
McElhaney, J. E., & Effros, R. B. (2009). Immunosenescence: what does it mean to
health outcomes in older adults? Current Opinion in Immunology, 21(4), 418-424.
http://doi.Org/10.1016/j.coi.2009.05.023
Mechanic, D. (1979). Correlates of physician utilization: Why do multivariate studies of
physician utilization find trivial psychosocial and organizational effects? Journal of
Health and Social Behavior, 20, 387-396.
Muthen, B. & Muthen, L. (2013). Mplus version 7.1 [Computer software].
Nguyen, M., Ugarte, C., Fuller, T, Haas, G., & Portenoy, R. K. (2005). Access to care for
chronic pain: racial and ethnic differences. The journal of pain : official journal of the
American Pain Society, 6(5), 301-14. doi: 10.1016/j.jpain.2004.12.008
68


Patrick, D. L., Danis, M., Southerland, L. I., Hong, G. (1988). Quality of life following
intensive care, Journal of General Internal Medicine, 3, 218-223.
Peveler, R., Kilkenny, L., Kinmonth, A. (1997). Medically unexplained physical
symptoms in primary care: A comparison of self-report screening questionnaires and
clinical opinion. Journal of psychosomatic research, 42, 245-252.
Radloff, L. S. (1977). The CES-D scale: a self report depression scale for research in the
general population. Applied Psychological Measurement, 1, 385-401.
Raison, C. L., & Miller, A. H. (2011). Is depression an inflammatory disorder? Current
psychiatry reports, 13(6), 467-475.
Ricketts, T. C. (1999). Rural Health in the United States. New York: Oxford University
Press.
Rogers, A. T. (2010). Exploring health beliefs and care-seeking behaviors of older USA
dwelling Mexicans and Mexican-Americans. Ethnicity & health, 75(6), 581-99.
doi:l 0.1080/13 557858.2010.500018
Rosland, A. M., Kieffer, E., Israel, B., Cofield, M., Palmisano, G., Sinco, B., Spencer,
M., & Heisler, M. (2008). When is social support important? The association of family
support and professional support with specific diabetes self-management behaviors.
Journal of General Internal Medicine, 23(12), 1992-1999.
Roth, D.L., Ackerman, M. L., Okonkwo, O. C., & Burgio, L. D. (2008). The four-factor
model of depressive symptoms in dementia caregivers: A structural equation model of
ethnic differences. Psychology and Aging, 23, 567-576.
San Luis Valley Regional Development and Planning Commission (1992). San Luis
Valley Overall Economic Development Program. Alamosa, CO: San Luis Valley
Regional Development and Planning Commission.
Sobel, D. S. (1995). Rethinking medicine: Improving health outcomes with cost-effective
psychosocial interventions. Psychosomatic Medicine, 53(3), 234-244.
doi: 10.1097/00006842-199505000-00005
69


SPSS Inc. (2012). Statistical package for the social sciences v. 21 [Computer software].
Sullivan, M. J., Feuerstein, M., Gatchel, R., Linton, S. J., & Pransky, G. (2005).
Integrating psychosocial and behavioral interventions to achieve optimal rehabilitation
outcomes. Journal of Occupational Rehabilitation, 75(4),475-489. DOI: 10.1007/sl0926-
005-8029-9
Twells, L. K., Bridger, T., Knight, J. C., Alaghehbandan, R., & Barrett, B. (2012).
Obesity predicts primary health care visits: a cohort study. Population health
management, 75(1), 29-36. doi:10.1089/pop.2010.0081
Verhaak, P. F. M., Meijer, S. A., Visser, A. P., & Wolters, G. (2006). Persistent
presentation of medically unexplained symptoms in general practice. Family practice,
23(4), 414-420.
Whooley, M. A., de Jonge, P., Vittinghoff, E., Otte, C., Moos, R., Carney, R. M.,
Browner, W. S. (2008). Depressive symptoms, health behaviors, and risk of
cardiovascular events in patients with coronary heart disease. JAMA, 300, 2379-2388.
Wolinsky, F. (1988). Seeking and Using Health Services. In The Sociology of Health
(117-144). Belmont, CA: Wadsworth.
Wolinsky, F. D., and Johnson, R. J. (1991). The use of health services by older adults.
Journal of Gerontology, 46, 345-357.
Yohannes, A. M. (2013). Is it quality or quantity of social support needed for patients
with chronic medical illness? Journal of psychosomatic research, 74(2), 87-8.
doi: 10.1016/j .j psy chores.2012.11.009
Ziegelstein, R. C., Bush, D. E., & Fauerbach, J. A. (1998). Depression, adherence
behavior, and coronary disease outcomes. Archives of internal medicine, 158, 808-809.
Zunszain, P. A., Hepgul, N., & Pariante, C. M. (2013). Inflammation and depression.
Current topics in behavioral neurosciences, 14, 135-151
70


Predisposing Determinants
Figure 1. Proposed model with psychological determinants as moderators of clinic visits as outcome of health care utilization.


Predisposing Determinants
Figure 2. Proposed model with psychological determinants as moderators of preventative care as outcome of health care utilization.


Prrdiipoiing Determinant!
Figure 3. Proposed model with psychological determinants as moderators of CAM use as outcome of health care utilization.
--j
UJ


Appendix A
Linear Regression Predicting Frequency of Clinic/ER Visits using MPlus (N = 621).
Model 1 Model 2 Model 3 Model 4
Variables B (SE) P B (SE) P B (SE) P B (SE) P
Age 0.11 0.03 0.00*** 0.09 0.03 0.00** 0.11 0.04 0.00** 0.12 0.04 0.00**
Sex 0.06 0.03 0.04* 0.04 0.03 0.16 0.04 0.04 0.28 0.04 0.04 0.33
Education -0.09 0.03 0.00** -0.06 0.03 0.07 -0.05 0.04 0.22 -0.07 0.04 0.09
Income -0.02 0.04 0.57 0.06 0.04 0.19 0.07 0.05 0.14
Insurance 0.05 0.03 0.08 0.05 0.04 0.15 0.04 0.04 0.35
Place of Care 0.17 0.03 0.00*** 0.10 0.04 0.00** 0.12 0.04 0.00**
Accessibility Difficulties 0.08 0.03 0.00** 0.02 0.04 0.61 0.03 0.04 0.43
Self-Reported Health -0.16 0.04 0.00* ** -0.15 0.05 0.00* **
Chronic Illnesses 0.18 0.04 0.00* ** 0.17 0.04 0.00* **
Pain Severity -0.01 0.04 0.85 0.00 0.04 0.96
Pain Frequency -0.01 0.04 0.97 -0.01 0.05 0.83
ADL Interference 0.14 0.05 0.00** 0.13 0.05 0.00**
Depression 0.00 0.05 0.95
Social Support 0.02 0.04 0.57
Quality of Life -0.06 0.05 0.18
R: .03 .06 .16 .16
*p< .05, **p< .01, ***/>< .001


Appendix B
Logistic Regression Predicting Use of Prayer using MPius (N = 616).
Model 1 Model 2 Model 3 Model 4
Variables B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B)
Age 0.07 0.03 1.02 0.07 0.04 1.02 0.01 0.05 1.00 0.00 0.05 1.00
Sex 0.20 0.03 2.1 !** 0.22 0.04 2.33*** 0.28 0.06 2.98*** 0.29 0.07 3.15***
Income -0.01 0.04 0.99 -0.01 0.05 1.00 0.01 0.05 1.01
Accessibility Difficulties 0.14 0.03 1.44*** 0.17 0.04 1.50*** 0.17 0.05 1.52***
Chronic Illnesses 0.04 0.05 1.06 0.04 0.05 1.06
Pain Severity -0.02 0.05 0.96 -0.02 0.06 0.96
ADL Interference 0.08 0.05 1.10 0.05 0.06 1.06
WaistiHip Ratio 0.45 0.41 0.00 0.38 0.41 0.00
Diastolic Blood Pressure -0.13 0.05 0.98** -0.10 0.05 0.98
Depression -0.04 0.06 0.99
Social Support 0.07 0.05 1.04
Quality of Life -0.09 0.06 0.87
R-' .05 .08 .14 .13
*p < .05, **p < .01, ** V < 001


Appendix C
Logistic Regression Predicting Use of Flu Shots using MPlus (N = 654).
Model 1 Model 2 Model 3 Model 4
Variables B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B)
Age 0.15 0.03 1.04*** 0.15 0.04 1.04*** 0.08 0.05 1.02 0.08 0.05 1.02
Ethnicity -0.02 0.04 0.94 -0.02 0.04 0.92 0.02 0.05 1.09 0.03 0.06 1.11
Education 0.13 0.04 1.07** 0.11 0.04 1.06** 0.16 0.05 1.09*** 0.14 0.05 1.07**
Insurance 0.05 0.03 1.96 0.03 0.04 1.56 0.01 0.04 1.14
Distance from Care -0.08 0.03 0.76* -0.10 0.04 0.70* -0.11 0.05 0.67**
Self-Reported Health -0.07 0.05 0.86 -0.04 0.06 0.91
Chronic Illnesses 0.16 0.05 1.23*** 0.19 0.05 1.29***
Pain Frequency -0.05 0.05 0.91 -0.05 0.05 0.91
Pain Severity 0.48 0.05 1.10 0.05 0.05 1.11
ADL Interference 0.03 0.05 1.04 0.06 0.06 1.07
Diastolic Blood Pressure -0.06 0.04 0.99 -0.06 0.05 0.99
Depression -0.06 0.05 0.99
Social Support -0.03 0.05 0.98
Quality of Life -0.01 0.06 0.99
R2 .04 .05 .09 .10
*p < .05, **p < .01, ***/> < .001
On


Appendix D
Logistic Regression Predicting Prostate Exams using MPius (N = 460).
Model 1 Model 2 Model 3 Model 4
Variables B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B)
Ethnicity -0.10 0.05 0.69 -0.10 0.05 0.67* -0.08 0.06 0.74 -0.09 0.06 0.72
Education 0.12 0.06 1.06* 0.12 0.05 1.06** 0.13 0.06 1.06* 0.14 0.06 1.07*
Place of Care 0.21 0.09 11.31** 0.20 0.09 9.95* 0.21 0.90 10.16*
Chronic Illnesses 0.09 0.05 1.14 0.07 0.05 1.10
Depression 0.01 0.06 1.00
Social Support 0.05 0.05 1.03
Quality of Life -0.02 0.06 0.98
R: .04 .08 .09 .10
*p<.05, **/><.01


Appendix E
Logistic Regression Predicting Mammograms using MPlus (N = 346).
Model 1 Model 2 Model 3 Model 4
Variables B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B)
Age 0.08 0.05 1.34 0.13 0.05 1.61** 0.23 0.80 2.99 0.21 0.74 2.65
Education 0.18 0.05 1.10*** 0.10 0.06 1.06 0.15 0.50 1.10 0.12 0.45 1.09
Income 0.19 0.05 1.18*** 0.13 0.43 1.15 0.10 0.35 1.12
Self-Rated Health 0.07 0.25 1.22 0.03 0.13 1.09
Pain Frequency -0.10 0.35 0.78 -0.13 0.45 0.73
Depression -0.05 0.20 0.99
Social Support -0.05 0.19 0.96
Quality of Life 0.06 0.21 1.11
R-' .03 .08 .40 .40
*p<.OS,**p<.Ol
oo


Appendix F
Logistic Regression Predicting Pap smears using MPius (N = 327).
Model 1 Model 2 Model 3 Model 4
Variables B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B) B (SE) Exp(B)
Age -0.23 0.04 0.95*** -0.24 0.05 0.94*** -0.25 0.07 0.94*** -0.22 0.07 0.94**
Ethnicity 0.06 0.05 1.24 0.05 0.05 1.96 0.07 0.07 1.29 0.05 0.08 1.22
Education 0.04 0.05 1.02 -0.01 0.06 1.00 -0.01 0.07 0.99 0.03 0.08 1.02
Income 0.01 0.06 1.01 -0.02 0.07 0.98 -0.05 0.08 0.96
Insurance 0.12 0.06 5.43 0.15 0.08 7.72 0.13 0.08 7.45
Distance from Care 0.02 0.05 1.08 0.01 0.06 1.02 0.02 0.06 1.06
Accessibility Difficulties -0.05 0.05 0.87 0.01 0.06 1.01 0.06 0.07 1.18
Self-Rated Health 0.02 0.07 1.05 -0.07 0.08 1.85
Chronic Illnesses -0.09 0.06 0.89 -0.08 0.07 0.90
Pain Severity 0.03 0.07 1.06 -0.02 0.07 0.95
Pain Frequency -0.07 0.07 0.88 -0.10 0.07 0.81
ADL Interference 0.01 0.07 1.01 0.07 0.08 1.08
Waist:Hip Ratio -0.22 0.45 0.03 -0.20 0.47 0.04
Systolic Blood Pressure 0.08 0.08 1.01 0.03 0.08 1.00
Diastolic Blood Pressure -0.07 0.08 0.99 -0.07 0.08 0.99
Depression 0.02 0.08 1.01
Social Support -0.03 0.07 0.98
Quality of Life 0.26 0.08 1.49***
R-' .06 .07 .09 .13
*p < .05, **p< .01
VO


Full Text

PAGE 1

THE INFLUENCE OF PSYCHOLOGICAL DETERMINANTS ON RURAL HEALTH CARE UTILIZATION AMONG OLDER HISPANIC AND NON HISPANIC WHITES: THE SAN LUIS VALLEY HEALTH AND AGING STUDY b y LACEY CLEMENT B.A., Stephen F. Austin State University, 2011 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Arts Clinical Health Psychology Program 2015

PAGE 2

ii This thesis for the Master of Arts degree by Lacey Clement h as been ap proved for the Clinical Health Psychology Program by Kevin S. Masters, Chair Krista Ranby James Grisgby Novemb er 12 2015

PAGE 3

iii Clement, Lacey R. (M.A., Clinical Psychology) The Influence Of Psychological Determinants On Rural Health Care Utilization Among Older Hispanic And Non Hispanic Whites: The San Luis Valley Health And Aging Study Thesis directed by Professor Kevin S. Masters. ABSTRACT Andersen's Behavioral Model of Health Services Utilization posits three broad categories of factors that influence individual determinants of health care utilization: predisposing characteristics, enabling resources, and basic need. Though the model does not detail exhaustive predictive factors or specific outcomes of utilization, it does provide a comprehensive frame work to conceptualize the underlying influences of health care usage. However, psychological determinants of health care utilization, such as depression, quality of life, and social support, are not specifically included in this model but may influence eac h aspect of Andersen's model in various ways. The primary purpose of the present study was to examine how psychological determinants may help explain the utilization of health care in a rural population. A secondary aim was to assess the relationships amon g the domains of determinants and explore if Andersen's hypothesized factors would be similar to those found in the San Luis Valley area of Southern Colorado. Health care utilization outcomes examined in this study was frequency of visits to a clinic/ER, a s well as use of prayer as a form of complementary and alternative medicine and various preventative measures such as flu shots, mammograms and pap smears for females, and prostate exams for males. Overall, psychological determinants added only minimally t o the overall regression models. However, it is possible that depression, quality of life, and social support have a systemic

PAGE 4

iv effect on other determinants in the model. Using a factor analysis, similar factors were found in this sample to Andersen's concep tual model of determinants, but some differences existed. It is clear that psychological variables can and do have a significant influence on decisions, both medical and otherwise, as well as perception of state. Continuing to test models of health care ut ilization in different populations with different facets emphasized will help to better understand the pathways of utilization in a more holistic manner. However, revising Andersen's Model of Health Services Utilization to include psychological variables i n a systems based approach will help conceptualize health care utilization in different populations and cultures. The form and content of this abstract are approved. I recommend its publication. Approved: Kevin S. Masters

PAGE 5

v ACKNOWLEDGEMENTS I would like to acknowledge my fam ily and friends for supporting and encouraging me to pursue my dreams in life. I would also like to thank my advisor and mentor, Kevin S. Masters, for pushing me to be better and constantly challenge myself. Many thanks also to my the sis committee members, Krista Ranby and Jim Grigsby for helping me through this journey A special thanks is in order for my labmates and cohort for always supporting me and being my fami ly in graduate school; Stephanie Hooker Showalter, Megan Grigsby, Ka ile Ross, Jo Vogeli, Shiva Fekri, an d Tattiana Romo: I am honored not only to call you colleagues, but also wonderful friends.

PAGE 6

vi TABLE OF CONTENTS CHAPTER I. BACKGROUND Behavioral Model of Health Services Utilization ................................ ................... 1 Predisposing Determinants of Health Care Utilization ................................ ............ 3 Enabling Determinants of Health Care Utilization ................................ .................. 5 Need Determinants of Health Care Utilization ................................ ....................... 6 Psychological Determinants of Health Care Util ization ................................ .......... 8 Present Study ................................ ................................ ................................ ......... 13 II. METHOD Population and Sampling ................................ ................................ ..................... 15 Measures ................................ ................................ ................................ .............. 16 Predisposing Determinants ................................ ................................ ............ 16 Enabling Determinants ................................ ................................ ................... 17 Need Determinants ................................ ................................ ......................... 18 Psychological Determinants ................................ ................................ ........... 19 Outcomes of Health Care Utilization ................................ ............................. 21

PAGE 7

vii Data Analytic Strategies ................................ ................................ ....................... 21 Missing Data ................................ ................................ ................................ .. 22 III. RESULTS Hypothesis 1 ................................ ................................ ................................ .......... 29 Frequ ency of Clinic or Emergency Room Visits ................................ ........... 29 Prayer as a Form of Complementary and/or Alternative Medicine ............... 34 Preventative Care: Flu Shots ................................ ................................ .......... 37 Preventative Care for Men: Prostate Exams ................................ .................. 39 Preventative Care for Women: Mammograms and Pap smears .................... 42 Hypothesis 2 ................................ ................................ ................................ .......... 4 6 IV. DISCUSSION ................................ ................................ ................................ ........... 49 Predisposing, Enabling, and Need Determinants of Health Care Utilization ....... 49 Psychological Determinants of Health Care Utilization ................................ ....... 53 Analyzing Determinant Domains Within the Model ................................ ............ 5 6 Amending the Conceptual Model of Health Services Utilization ........................ 58 A Study of Culture ................................ ................................ ................................ 6 0 Strengths and Limitations ................................ ................................ ..................... 61 Future Directions ................................ ................................ ................................ .. 6 1

PAGE 8

viii Conclusions ................................ ................................ ................................ ........... 62 REFERENCES ................................ ................................ ................................ ................. 63 APPENDIX A. Linear Regression Predicting Frequency of Clinic/ER Visits Using MPlus ............... 74 B. Logistic Regression Predicting Use of Prayer Using MPlus ................................ ........ 7 5 C. Logistic Regression Predicting Use of Flu S hots Using MPlus ................................ ... 7 6 D. Logistic Regression Predicting Prostate Exams Using MPlus ................................ ..... 7 7 E. Logistic Regression Predicting Mammograms Using MPlus ................................ ...... 7 8 F. Logistic Regression Predicting Pap smears Using MPlus ................................ ............ 79

PAGE 9

1 CHAPTER I BACKGROUND Individuals use health ca re services for a variety of reasons, including treating and curing illnesses, preventing health problems, reducing pain, and increasing quality of life. Individuals can utilize health services in many ways, such as by visiting a physician or the emergency room or even engaging in various types of complementary and alternative medical procedures. Myriad factors play a role in the process of deciding to utilize health care and in the actual utilization of health services. Furthermore, as various health care reform proposals are developed, understanding trends and factors influencing health care utilization is crucial to determining which proposals are more likely to succeed, especially in populations that are underserved. Behavioral Model of Health Services U tilization Health care utilization is the end result of a complex process that consists of both individual and societal determinants, interacting with the particular structure and resources of the health care system. As health care becomes more expensive a nd the rate of chronic illness increases in the population, understanding health care utilization and the determinants that influence usage of health services becomes vital. Utilization is now often conceptualized as an individual behavior with many idiosy ncratic factors playing a role. With any individualized behavior, multiple factors in the surrounding context influence what behavior occurs. To help conceptualize the factors that influence these behaviors, Andersen's Behavioral Model of Health Services U tilization was developed. It posits three broad categories of factors that influence individual determinants of health care utilization: predisposing characteristics, enabling resources, and basic need. This

PAGE 10

2 model begins to explain the different factors th at influence utilization as well as the way they might interact with each other ( Andersen, 1995; Andersen & Newman, 1973; Wolinsky & Johnson, 1991). Due to technological advances in treatments and more elaborate health care structures, such as specialized practices and holistic clinics, societal influences may be increasingly impacting an individual's decision to seek treatment (D'Crus & Wilkinson, 2005; Andersen & Newman, 2005). Consequently, this model has recently been used from a more economical or publ ic health population based perspective to aid in understanding societal influences on utilization (Andersen & Newman, 2005). That being said, individual factors still play large and important roles in the decision to utilize health care services. Baxter, B ryant, Scarbro, and Shetterly (2001) noted that this model tends to characterize and explain more of the individual factors, especially the need component, rather than changes in medical infrastructure. Using Andersen's model, Baxter et al. (2001) found th at patterns in health care resources and use in a rural, aging population were based on more than just straightforward medical need in that various factors, such as culture and attitudinal dispositions, played a role in the utilization of services. Anderse n takes a more systems based approach with this model in that multiple factors and their interactions influence the likelihood of utilization taking place. It is difficult to completely separate these determinants and their respective practical influences on utilization because of the complex and interactive nature of system based models. Though this model does not detail exhaustive predictive factors or specific outcomes of utilization, it does provide a comprehensive framework to conceptualize the

PAGE 11

3 underly ing influences of health care usage and guide further research (Andersen & Newman, 1973). Predisposing Determinants of Health Care Utilization Predisposing characteristics of the individual are those which exist prior to the need for health services but im pact utilization nonetheless. Predisposing characteristics include variables such as demographics, social structure, and health beliefs. Demographic and biologic variables may predispose someone to have medical problems and hence the need to seek care more than others without a similar biological and cultural predisposition. They exist without the presence of a diagnosed medical condition yet may impact the decision to seek care should the need arise. For instance, demographic variables such as sex impact u tilization. In previous studies of health care utilization, it was found that females were less likely to be hospitalized or have outpatient surgery than males, but males were less likely to contact a physician (Dunlop, Manheim, Song & Chang, 2002). Simila r to sex, age also is a factor in health care use. Older adults are more likely to seek medical care than their younger counterparts due to the perceived possibility of increased health problems and therefore increased salience of medical care (Benjamins & Brown, 2004). Besides age, ethnicity is one of the most predictive predisposing determinants of health care utilization (Benjamins & Brown, 2004). Ethnic differences in education and income are also more pronounced in older populations (Dunlop et al., 20 02), thus ethnic disparities play a significant role in health care utilization, especially in older adults. Ethnicity, as well as education and culture, falls within the social structure paradigm of predisposing characteristics. The environment and contex t of one's life influences that

PAGE 12

4 person's health and by extension, the utilization of health care. Nguyen, Ugarte, Fuller, Haas, and Portenoy (2005) found that Hispanics reported fewer years of formal education and lower income than Whites or African Americ ans. Lack of education and decreased income put individuals at a disadvantage to accessing needed health care. Latino populations generally have poorer overall health and less access to health care services (Rogers, 2010). According to Nguyen et al. (2005) Hispanics also were less likely to seek medical consultations than non Hispanic whites or African Americans. Specifically, Hispanics with the same diagnosis as Non Hispanic Whites were less likely to have an invasive procedure (Carlisle, Leake & Shapiro, 1997), and ethnic minorities engage less frequently in preventative care services (Coffield et al., 2001; Benjamins, 2005). Cultural differences also exist in different ethnic groups, and culture plays a vital role in health care utilization, as it impact s health beliefs, context of living, and perhaps access to care (Rogers, 2010; Helman, 1994). Different aspects of culture influence health behaviors, such as values, religion, and worldviews. Religious and spiritual beliefs often are a cultural norm for H ispanic groups, and religion may influence beliefs and perceptions of health care, especially for rural dwellers (Arcury et al., 2005) and older individuals. Individuals who attend religious services more frequently possess greater knowledge regarding appr opriate health care and maintenance (Apel, 1986; Benjamins, 2005) which is thought to predict use of health services. On average, adults, 65 and older, are the most likely to attend religious services. Benjamins (2006) found that moderate levels of church attendance significantly predicted use of female preventative services such as mammograms. Higher self reported religiosity was positively correlated with use of preventative medical care

PAGE 13

5 (Benjamins & Brown, 2004). Koenig and Larson (1998) found that indiv iduals who attended more religious services were less likely to be admitted into a hospital and had a shorter duration of stay if they were admitted. However, in Hispanics no relationship was found between church attendance and quantity of medical or hospi tal visits (Levin & Markides, 1985). Additionally, the way someone perceives the health care system and his or her role in it is important to understanding how and why he or she utilizes it in a particular way. The precise relationship between beliefs and usage is not yet settled and is influenced by the type of illness, the implications of that illness, and the overall context of the individual (Wolinsky, 1988). Beliefs about medicine and illness including attitudes towards screening, taking medications, and trust in medical professionals can differ across cultures and even within cultural subgroups (Le et al., 2014).Though some predisposing and demographic factors are not modifiable and therefore unable to be the basis of an intervention, they impact medi cal care usage nonetheless. It is important to understand the role these factors play in utilization and the way they interact with other individual determinants. Enabling Determinants of Health Care Utilization Enabling resources from the person and his o r her family and community can influence the way an individual views and uses health care. Income, health insurance, and a means to travel to a physician or medical facility are vital resources for individuals, without which utilization would likely not ta ke place regardless of other factors. Other variables associated with the medical facility and resources such as cost of health care, accessibility of providers, and region of living are considered community enabling

PAGE 14

6 resources (Andersen & Newman, 1973). In a rural area, providers might not be plentiful and thus patients are required to wait for medical care. In environments where this is normative, it likely impedes individuals from seeking care and a sense of helplessness may develop. Residents of rural se ttings are often disparately poorer, more likely to be older and retired, and less educated than those in urban centers (Ricketts, 1999). With less education and lower incomes, access that others would routinely utilize becomes limited for these special po pulations. Rural health care utilization brings to the forefront obstacles that are not always present in urban settings. Those who live in isolated areas tend to have more difficulties accessing health care due to distance and resources. Fewer or no oppo rtunities for public transportation and farther distances to travel negatively impact the quality and quantity of care (Arcury et al., 2005). Understanding factors that influence health care utilization in rural populations as well as those of lower socioe conomic status is important to recognizing how health care may differ depending on the setting. However, despite the enabling resources, another factor must play a role in health care utilization, and that is whether or not the need exists, or is perceived to exist, to access that care. Need Determinants of Health Care Utilization Need encompasses the biological and physical component of symptoms and illness. The severity of an illness and the perception and evaluation of that severity will influence wheth er or not one seeks medical care. Ultimately, whether an individual seeks medical care depends on if he or she first determines the need exists for health services. Pain is often the determinant that alerts individuals into the possibility of pathology. Fo r example, seeking health care for acute low back pain was associated with duration and

PAGE 15

7 severity of pain (Carey, Garrett & Jackman, 2000). Similarly, less pain and better daily functioning was related to less health care seeking (C™tŽ et al., 2001). Furthe r, functionality in daily life greatly influences perception of illness. Days off from work, school, or other social responsibilities due to disability are a major component of evaluated illness (Andersen & Newman, 2005). The existence and perceived sever ity of symptoms, or a medical problem, influences if and how medical care is sought, as well as the type of medical care. For instance, chronic pain patients who live with incurable pain that does not often diminish may regularly seek out advice or "second opinions", often going from one medical clinic to the next on a search for some relief from their pain conditions. Furthermore, evaluation of an illness also comprises need. Documented or diagnosed medical conditions guide usage of health services. For ex ample, chronic illnesses are among the most prevalent medical problems plaguing the health care system. Moreover, the aging population, who is more likely to have chronic illnesses, account for a large portion of health care utilization (de Boer, Wijker & de Haes, 1997). Heart disease, cancer, stroke, and hypertension are among those chronic illnesses that require patients to access their medical provider more often, due to the need for frequent checkups and possible emergency situations. Similarly, chronic illnesses have a large behavioral component. Because of this, lack of self management drives patients to their primary care doctor or emergency department more often (Butterworth, Linden & McClay, 2007). Secondary medical problems from these chronic illn esses also lead individuals to utilize services more (e.g., diabetes mellitus often leading to myocardial infarctions) (Brownson & Heisler, 2009).

PAGE 16

8 Objective physical indicators such as body mass index (BMI) and blood pressure are need determinants that p redict medical use. Morbidly obese patients are more likely to see their general practitioner, and even after controlling for chronic conditions or other related medical diagnoses, obesity is still a significant predictor of visits to a general practitione r (Twells, Bridger, Knight, Alaghehbandan & Barrett, 2012). Hypertension is often concurrent with cardiovascular disorders but may exist without the presence of a diagnosable cardiovascular problem. High blood pressure may predict other health issues that could lead to greater utilization. Need characteristics are often found to account for much more of the variance in health care utilization than predisposing or enabling determinants (Hershey, Luft & Gianaris, 1975; Mechanic, 1979). Whether or not a perso n actually has significant health problems and perceives that he or she has significant health problems determines if utilization is necessary. Two different decisions must take place. The way pain and other symptoms are experienced, attended to, and perce ived first impacts the decision of whether or not the symptom is indicative of pathology and then whether care should be sought. However, examining need merely from the presence of chronic illness, physical indicators, pain, or even limited functionality u ndermines the complexity of need determinants. Psychological Determinants of Health Care Utilization Not clearly defined in Andersen's model is the role of psychological factors in health care utilization. Patients often frequent their medical care provid er for symptoms with no known o rganic cause. In fact, only three to 20 percent of visits actually had known organic causes. In addition, there are consistent relationships between these medically unexplained symptoms and psychological processes (Gask, Dowr ick, Salmon,

PAGE 17

9 Peters & Morriss, 2011; Verhaak, Meijer, Visser & Wolters, 2006; Peveler, Kilkenny & Kinmouth, 1997; Kolk, Schagen & Hanewalk, 2004). Clearly, different factors interact with one another to bring a person to seek care at a medical facility. So bel (1995) argued that despite the presence of known organic causes, emotional and mental factors play a significant role in the onset and management of diseases. The way an individual displays medical symptoms and functions with a disease greatly depends on his or her ability to cope. Furthermore, patients who somaticize or have other physical symptoms, along with a comorbid psychological disorder, are more likely to utilize health care, especially emergency room visits (Blount et al., 2007). Kolk, Hanewal d, Schagen and Gijsbers van Wijk (2003) posited a complex model of symptom perception to understand the evaluation of physical symptoms. In that model, emotions and affectivity influence perception at various levels, such as input of somatic information an d attention to sensations. Moreover, negative affectivity, in particular, is associated with the presence of physical symptoms, specifically increases of somatic information to be encoded (Kolk, Handwald, Schagen & Gijsbers van Wijk, 2002). Those with nega tive affect tend to somaticize more and generally experience more physical sensations to appraise. Seemingly, it is not just the increased appraisal of physical sensation but also the increased magnitude of sensations that leads to increased care utilizati on. Once sensations occur, though, psychological factors such as depression, social support, and quality of life affect perception of those symptoms and the overall decision to utilize health care. Actual severity of illness has little to do with a decisio n to seek medical care (Berkanovic, Telesky & Reeder, 1981). Those who seek medical care tend to have worse health related quality of life (C™tŽ et al., 2001), although the direction

PAGE 18

10 between these two could be and most likely is bidirectional. According to de Boer, Wijker & de Haes (1997), depression and overall psychological distress are some of the strongest predictors of health care use. Depression has been shown repeatedly to significantly predict increased utilization (Keeley et al., 2008; Sullivan, Fe uerstein, Gatchel, Linton & Pransky, 2005). Patients with a diagnosed depressive disorder have double the health care costs relative to their peers without depression (Kathol et al., 2005). Depression can lead persons to be more attentive to symptoms and c atastrophize physical sensations, inducing them to over utilize the health care system. Even if known organic causes exist, the added layer of psychological factors such as depression or even overall quality of life can change the way the illness and need to seek care is appraised. For example, interactions between pain and quality of life can be manifest in various ways, depending on other personal and social factors. Those who attend more to painful symptoms or catastrophize those symptoms are more likely to seek medical care. Additionally, those with depression or negative affectivity are less likely to be compliant and adherent with medical advice. With noncompliance and nonadherence comes increased medical care and associated health care costs (Whooley et al., 2008; Glassman et al., 1990; Carney et al., 1995; Ziegelstein, Bush & Fauerbach, 1998). Older adults, for whom nonadherence may be problematic partially due to age related issues, tend to also have multi morbidity. Boyd et al. (2005) reported that over half of adults 65 and older had at least three chronic diseases. Of those, a significant number were living with five or more chronic diseases. The combination of multiple diagnoses as well as age related concerns may have a compounded effect on util ization. It is known that patients with chronic illnesses use health services more and often have

PAGE 19

11 adjoining psychosocial sequelae. Secondary consequences of chronic illnesses such as poor adjustment, poorer quality of life, and more intense depression all lead to greater utilization (Yohannes, 2013). These consequences manifest uniquely in an aging population. In particular, older adults' lack of social support, or perceived lack of social support, negatively influences their health behaviors in a number of ways, such as poor self efficacy and inability to appropriately self manage their illness (Rosland et al., 2008; McCathie, Spence & Tate, 2002). Social support is a vital component of overall quality of life. Regardless of population, the quality of socia l support, rather than quantity, is often the defining attribute that predicts better health (Yohannes, 2013). Matching support given to support needed is often key and mismatches often result in poorer adjustment (Yohannes, 2013). Ability to adjust and co pe with an illness can influence pathways of health care utilization. A person's overall well being, impacted by negative affectivity, quality of life, social support, or other related variables, is related to physical health. Depending on the population s tudied, subjective well being is understood in different ways (Abdel Khalek, 2014) and terminology often overlaps, but the positive evaluation of circumstances is related to increased physical health (Abdel Khalek, 2014). If one evaluates his/her life more positively, the greater sense of overall well being may lessen negative appraisals of need. Both positive and negative outlooks influence health evaluations; psychological factors in general may also influence predisposing and enabling determinants of he alth care utilization. The relation between utilization and gender, age, and income can all influence, as well as be influenced by, psychosocial factors. Kolk et al. (2003) reported that age and SES were significant indicators of physical symptom appraisal through

PAGE 20

12 various direct and indirect pathways, such as selective attention, negative mood, and prevalence of chronic disease. Older adults are less likely to be diagnosed with depression (Hybels & Blazer, 2003) and community dwelling older adults who were independent had lower prevalence rates of depression compared to those in the general adult population (Birrer & Vemuri, 2004). Also, depression has a known "gender gap" in that women are usually twice as likely to be diagnosed with a depressive episode or disorder but that gap lessens as age increases (Kornstein, 1997). Psychological determinants of health care utilization may influence each aspect of this model in various ways. An individual's affective or psychological attributes could potentially mediat e the relationship among these individual determinants. In describing his model of health care utilization, Andersen understood that psychosocial variables, such as social networks, could influence different determinants (Andersen, 1995; Andersen & Newman, 2005) but failed to adequately address the complexity of ways that psychological determinants impact health care utilization. In 2012, a systematic review by Babitsch, Gohl, and von Lengerke examined studies that implemented Andersen's model. In it, they included "mental disorders" in the domain of predisposing characteristics, "social/emotional support" in the domain of enabling characteristics, and "health related quality of life" in the domain of need characteristics. Though there is a certain logic tha t follows this way of organization, it has also been shown that factors such as depression, social support, and quality of life interact closely with each other and influence a vast number of other determinants. Therefore, health and psychological factors appear to have systemic influences on each other meaning bidirectional and compound effects exist between the two.

PAGE 21

13 Present Study The primary purpose of the present study was to examine how psychological determinants may help explain the utilization of heal th care in a rural population. Specifically, the study set out to explore how Andersen's model of health care utilization mapped onto predictors of utilization in the rural San Luis Valley area of southern Colorado. The primary hypothesis of this investiga tion was that psychological determinants, namely depression, social support, and quality of life, would explain the use of health services in a community sample beyond the predisposing, enabling, and need determinants found in Andersen's model Specificall y, t hese psychological determinants would better predict use of health services such as clinic, hospital, and ER visits in a 12 month period, preventative care (e.g. flu shots, prostate exams for men, and mammograms and pap smears for women), and use of pr ayer for spiritual healing than predisposing, enabling, and need determinants, as posited in Andersen's model. If this were the case, it is possible that the existing model could be construed differently to include these psychological determinants (see fig ures 1 2, and 3 for hypothesized conceptual models). The present study also had a secondary aim to assess the relationships among the domains of determinants and explore if Andersen's hypothesized factors would be similar to those found in this populat ion Based on the theoretical model, it is presumed that in order to create these domains of determinants, individual predictors within each domain would be more associated to each other than to those in other domains For instance, a need determinant such as presence of chronic illnesses would be more related to another need determinant, such as self reported health, than to an enabling determinant such as

PAGE 22

14 accessibility to health care. This study sought to examine those relationships to see if they endured in this sample.

PAGE 23

15 CHAPTER I I METHOD P opulation and Sampling The San Luis Valley Health and Aging Study (SLVHAS) surveyed health and disability among Hispanic and non Hispanic white residents of Alamosa and Conejos counties, two mainly isolated rural southern Colorado counties between 1993 and 1995 Residents of t he San Luis Valley live mostly in small, primarily farming, communities. According to the 1990 population census of the region, 46 percent were Hispanic, 52 percent were non H ispanic White, and 2 percent were other (San Luis Valley Regional Development and Planning Commission, 1992). The majority of Hispanic residents in the San Luis Valley report their ethnicity as "Spanish" or "other Hispanic" as opposed to Mexican American. At the time of data collection, t he population in the San Luis Valley region ha d little to no new immigration occurring (Bean & Tienda, 1987). Additionally, 19 percent of residents 65 and older were below the poverty level in 1990 (Hamman et al. 1999). In order to participate in the st udy, individuals must have been: (a) 60 years of age or older, (b) a current resident in the surveyed counties, and (c) of Hispanic or Non Hispanic White ethnicity. In this sample, participants ( N = 1358) were mostly Hispanic (58.4%) and female (56.8%). The average age was 74 ( SD = 7.8; range 60 99). Almost forty percent did not complete high school, and almost half of the participants' yearly income was less than $10,000. See table 1 for complete participant characteristics.

PAGE 24

16 Table 1 Participant Characteristics Note Because these data were from 1993 1995, inflation rates must be considered to understand how income rates measured in this population translate into income based on the current U.S. economy. Using the CPI calculator, (Consumer Price Index; U.S. Bureau of Labor Statistics), an inflation rate of 1.6 must be used to determine how these data convert to current income rates. For instance, income of $10,000 in this dataset would convert to $16,000, $20,000 would convert to $32,000, $75,000 would convert to $121,000, etc. Measures Predisposing Determinants Age ethnicity, gender, and education are measures of the predisposing characteristics. Age was measured by asking date of birth of each participant and calculating from that to the time of data collection. Ethnicity was assessed by asking participants "Are you of Spanish/Hispanic origin or descent?" Four answer options were possible of "No," "Yes, Mexican, Mexican American, Chicano," "Yes, Cuban or Puerto Rican," and "Ye s, other." P articipants were asked if they were male or Variable N (%) Age M ( SD ) = 74 (7.8) Female Ethnicity 772 (56.8%) Non Hispanic White 565 (41.6%) Hispanic 793 (58.4%) Education Less than high school 518 (38.7%) High school 540 (40.3%) College 194 (14.5%) Graduate degree 87 (6.5%) Income Less than $10,000 545 (45.8%) $10,000 $19,999 358 (30.0%) $20,000 $34,999 182 (15.2%) $35,000 $49,999 60 (5.0%) $50,000 $74,999 29 (2.4%) More than $75,000 17 (1.4%)

PAGE 25

17 female to identify gender For assessment of education level pa rticipants were asked to identify the "highest grade or year of school (they had) completed." Possible answer choices ranged from 00 17+ with one year intervals, as well as "I don't know." Enabling Determinants Income insurance, and accessibility of medical care are all enabling determinations of health care usage Income was assessed by asking participants "about how much was the tota l income, before taxes, of all your family members, living in your house, from all sources last year?" The following 11 options were given to answer about income: a) less than 5,000, b) 5,000 7,499, c) 7,500 9,999, d) 10,000 14, 999, e) 15,000 19,99 9, f) 20,000 24,999, g) 25,000 34,999, h) 35,000 49,999, i) 50,000 74,999, j) 75,000 or more, and "don't know/refuse to answer." Because these data were from 1993 1995, inflation rates must be considered to understand how income rates measured in t his population translate into income based on the current U.S. economy. Using the CPI calculator, (Consumer Price Index; U.S. Bureau of Labor Statistics), an inflation rate of 1.6 must be used to determine how these data convert to current income rates. Ea ch participant was also a sked whether or not he/she was covered by a health insurance plan. Finally, a ccessibility of medical care was assessed by three different measures, 1) by asking if there is a certain clinic or medical center used if needed, 2) ho w many miles the clinic is from his/her home and 3) if they had difficulties getting to their medical ce nter because of various factors, which was a total count of endorsement of the following: "how much it costs," "didn't have a way to get there," "care wa s not available when needed," "needed someone to take care of family member," "had to wait too long in the office or clinic," "had no confidence in the staff," and "staff did not speak a

PAGE 26

18 language you were comfortable with" (Hispanic Health and Nutrition Ex amination Survey, 1990). Need Determinants Both subjective and objective determinants of need were assessed. Self reported health stat us, presence of chronic illness and pain were measured, as w ere objective physical indicators of waist to hip ratio and blood pressure. Participants were asked to rate their overall health status as "excellent," "good," "fair," or "poor" compared with other people (their) age. C hronic illnesses were assessed by asking participants to self report whether or not they had "e ver been told by a doctor" that they had cancer, heart attack, mini stroke, major or severe stroke, angina, high blood pressure, enlarged heart or heart failure, emphysema/chronic bronchitis/COPD, cirrhosis of liver, kidney failure, Parkinson's disease, os teoporosis, seizure disorder, and migraines/persistent headaches Response options for each question were "no," "yes," or "don't know." Self reported diseases tend to correlate well with documented medical diagnoses ( Edwards, Winn & Kurlantzick, 1994). The measure of chronic illness was a total count of all illnesses reported. Pain was measured in different ways with four separate measures included in subsequent analyses : presence of pain, severity of pain, frequency of pain, a nd pain interference with acti vities of daily living. Presence of pain was assessed by asking Are you ever troub led with pain? ( in past year)," which was used as a dichotomous variable. Additionally, se verity and frequency of pain were both assessed ordinally by asking "When your pain is at its worst, would you describe it as mild, moderate, severe or unbearable" and "During the past week, how much of the time have you been troubled with pain" (all of the time, most of the time, some of the time, or rarely/never), respectively. The int erference with activities of daily living (ADLs) due to

PAGE 27

19 pain was a continuous variable that used a total count of endorsement to five detailed questions such as "Does this pain ever" 1) "cause you to move around less," 2) "keep you from sleeping," 3) "cau se you to cut down on any of your usual activities like work, household chores, or running errands," 4) "keep you from visiting family and friends in your own home," and 5) "keep you from doing things you like to do for pleasure, like hobbies, sports or le isure activities?" The four measures of pain were used separately in data analyses. As well as the self reported measures of illness or health, two objective measures of health were taken that predict occurrence or development of cardiovascular or metabol ic disease : a) waist to hip ratio to determine adiposity and b) an average of the second and third blood pressure measurements Because both low and high BMI could signal pathology in an aging population, waist to hip ratio was determined to be the prefer red measure of adiposity. This measure is best estimate of frailty in an aging population. Also, although three measurement of blood pressure were taken, only the last two were averaged because the first reading was used to calibrate. The SLVHAS employed s everal full time data collectors who were long time residents of the area. Multiple training sessions were provided to ensure reliability of the data. Additionally, to gain an understanding of interrater reliability, 60 participants were reinterviewed an a verage of 10 days (range 3 15 days) after the initial visits for some selected items, and researchers found no differences in reproducibility between Hispanics and Non Hispanic White subjects (Hamman et al., 1999). Psychological Determinants. Three measure s of psychological factors (depression, social support, and quality of life) were used to explore participant's

PAGE 28

20 psychosocial state. Depressive symptoms were assessed as a measure of mood using the 20 item Center for Epidemiologic Studies Depression (CES D) Scale (Radloff, 1977). The scale asks about feelings and experiences during the last week (e.g. "I felt that everything I did was an effort" and "I felt depressed") with response options ranging from "rarely or none of the time" to "most or all of the ti me." The overall scale reliability is high in the general population (! = .85). The CES D has been shown to have good sensitivity and specificity, with high internal consistency (Lewinsohn, Seeling, Roberts & Allen, 1997), and it has been used with differ ent racial and ethnic groups (Roth et al., 2008). The total score of the CES D will be the dat um point for analyses. Total scores range from 0 to 60, with higher scores indicating greater depressive symptoms. Four questions were reverse scored ("I felt I w as just as good as other people," "I felt hopefuly about the future," I was happy," and "I enjoyed life"). S ocial support was measured by two items assessing the social network size, 1) "How many relatives do you feel close to? That is, relatives you feel at ease with, can talk to about private matters, and can call on for help", and 2) "How many close friends do you have? That is, people that you feel at ease with, can talk to about private matters, and can call on for help." The two items were added toget her to create a singular social support measure of both relatives and friends. Quality of life was measured using Patrick's Perceived Quality of Life Scale ( PQoL; Patrick, Danis, Southerland, & Hong, 1988 ). All 20 questions were rated by the participant on a n 11 point scale from 0 to 10, where 0 was very dissatis fied and 10 was very satisfied. Examples of questions include: "how satisfied are you with the health of your bod y, the meaning and purpose in your life, and the amount of variety in your life."

PAGE 29

21 The PQoL also asks participants to rate on the 0 10 satisfaction scale "how happy are you?" A mean score of responses is calculated, and if the mean is equal or greater than 7.5, scores are interpreted as "satisfied." The PQoL evaluates satisfaction with areas of functional status in persons with varying levels of wellness and disability. It has also been more widely used with different patient groups and specifically with thi s population in previous studies (Baxter et al., 1998; Caldwell, Baxter, Mitchell, Shetterly & Hamman, 1998). Outcomes of Health Care Utilization Three outcomes of health care utilization were assessed. F requency of health care utilization in the past yea r was assessed by asking participants "In the past 12 months, how many times have you gone to a medical clinic, outpatient clinic or emergency room?" Participants were also asked about preventative care by the following questions: In the past 12 months, h ave you had flu shot prostate exam (for males only) pap smear (for females only) mammogram (for females only) ?" Use of a particular complementary or alternative medicine (CAM) approach was also used as a measure of health care utilization ; p articipants were asked if, in the past year, they had used prayer or spiritual healing (Eisenberg et al., 1993) Data Analytic Strategies Data were analyzed using IBM SPSS Statistics version 21 (SPSS Inc., 2012) and Mplus version 7.1 (MuthÂŽn & MuthÂŽn, 2013 ). Descrip tive statistics (e.g. means, standard deviations, frequency distributions) were calculated to describe the sample. Continuous variables were checked for normality. The outcome variable of frequency visits to a place of care was transformed using a square r oot transformation and t he miles from care variable used to assess accessibility to health care was also transformed using a log

PAGE 30

22 transformation to correct for the non normal distributions The distributions of all other continuous variables were within th e normal ranges (skew between 3.0). The variable of adiposity (waist:hip ratio) was squared to create a quadratic measure to best determine a curvilinear relationship. Categorical demographic and predictor variables were dummy coded so they could be use d in a linear analysis: ( a ) ethnicity was categorized into (0) Non Hispanic White and (1) Hispanic; ( b ) sex was categorized into (0) male and (1) female; ( c ) insurance was categorized into (0) does not have insurance and (1) has insurance; ( d ) place of usu al care was categorized into (0) no place of usual care and (1) does have a place of usual care; and ( e ) presence of pain was categorized into (0) denied pain and (1) endorsed pain; Dichotomous outcome measures of health care utilization (flu shots, prost ate exams, mammograms, pap smears, and use of prayer) were recoded so that 0 = denied getting service and 1 = endorsed getting service. Scale scores were calculated according to the author's instructions (see description of calculation in measures section ) Bivariate associations among study variables were examined. See table 2 for correlations among predictor variables and table 3 for correlations among predictors and outcomes. Missing Data. To account for missing data, listwise deletion was used in SPSS so only those who had complete data on the variables used were included in analyses. Table 4 illustrates the percentage of missing data for each variable. Most variables had missing data that ranged from 2.0% 23.00%. It appears that most variables within a domain of determinants were missing in clusters in that when one variable was missing, others within that domain were also missing.

PAGE 31

23

PAGE 32

24

PAGE 33

25

PAGE 34

26 To further explore how missing data might impact analyses, regressions were also performed in MPl us to estimate missing data parameters. Results from analyses in both SPSS and MPlus were similar. Because results were similar, SPSS analyses are reported in the results section. See appendices A F for regression tables based on MPlus analyses. Hypothesis 1: Psychological determinants (depression, social support, and quality of life) will explain the use of health services in a community sample beyond the

PAGE 35

27 predisposing, enabling, and need determinants in Andersen's model. Linear and logistic regressions wer e used to test if psychological determinants accounted for additional variance in health care utilization than did predisposing, enabling, and need determinants alone. For the outcome of frequency of visits to a place of health care, a linear regression wa s used. Continuous variables were centered using the grand mean technique. Moderation anal yses were used to determine how, in the presence of other predictor variables, psychological variables are related to health care utilization at a place of health car e. The other outcomes were dichotomous in nature (either received service or did not), therefore logistic regressions were used for those models. The models that included psychological variables were examined to ascertain if there was a significant increas e in explained variability beyond the previous models that lacked psychological variables Beta weights were examined for significant contributions to frequency of visits to a place of health care. Hypothesis 2 : Domains of determinants in this sample will be similar to those proposed in Andersen's model of health care utilization. P redictors within a domain of determinants will be more closely related than predictors across domains A principal component factor analysis was performed to check the validity o f An d ersen's model in a new sample.

PAGE 36

28 CHAPTER I II RESULTS D escriptive statistics of the psychological variables and outcomes are presented in t able s 5 and 6 respectively Table 5 Descriptive Statistics on Psychological Variables Variable N M ( SD ) M in Max Skew Kurtosis Depression 1153 6.84 (7.7) 0.0 57.0 2.0 5.4 Social Support 1152 7.31 (3.0) 0.0 12.0 0.2 0.7 Quality of Life 1060 8.25 (1.3) 2.6 10.0 0.8 0.7 Note. Depression measured using Center for Epidemiologic Studies Depression Scale (CES D). Social support is a total count of family and friends. Quality of Life measured using Patrick's Perceived Quality of Life scale (PQoL). Table 6 Descriptive Statistics on Outcomes of Health Care Utilization Variable N M ( SD ) Min Max Sk ew Kurtosis Clinic/ER Visits 1334 1.84 ( 1.23 ) 0.00 7.75 0.68 0.93 Variable N % Received Service Prayer 1325 29.3% 64.0% 43.9% 36.6% 40.8% Flu Shots 1200 Prostate Exams 515 Mammograms 681 Pap smears 681 Note Measure of clinic/ER visits is a continuous variable that was square root transformed to acquire normality. Prayer, flu shots, prostate exams, mammograms, and pap smears are all dichotomous variable

PAGE 37

29 Hypothesis 1 To test hypothesis 1, both linear and log istic regressions were used to explore the relationship between psychological variables and health care utilization outcomes. To determine which variables to include in the regression models, bivaria te correlations were calculated. P redictors that did not have a significant relationship to the outcomes were excluded from the analysis. Frequency of clinic or emergency room visits Before running a linear regression, correlations between predictors and the outcome of frequency of visits to a clinic or emerge ncy room were analyzed to determine which variables should be included in the analyses. Based on the correlational analyses, three predisposing determinants, four enabling determinants, and six need determinants were signifi cantly related to the outcome. W ithin the predisposing determinant domain, age ( r = 12 p < .01), sex ( r = 06 p < .0 5 ), and education ( r = .10 p < .01) w ere significantly related to frequency of visits. Both age and sex were positively related to frequency of clinic visits, suggesti ng that older participants and females were more likely to seek treatment in a clinic or emergency room. Education, on the other hand, was negatively related to frequency of visits, such that those with less education were more likely to seek treatment. Wi thin the enabling determinant domain, income ( r = .10 p < .0 1), insurance ( r = .06 p < .05 ), place of care ( r = 18 p < .0 1), and accessibility difficulties ( r = .08 p < .01) were significantly related to frequency of clinic visits. If a person has i nsurance and a place of usual care, he or she was more likely to frequent a place of care. Income and difficulties with accessing a place of health care were negatively correlated with visits; those with less income and fewer difficulties were more likely to frequent clinics or emergency

PAGE 38

30 rooms. Finally, within the need determinants, self rated health ( r = .28 p < .0 1), number of chronic illnesses ( r = 33 p < .0 1), presence of pain ( r = 19 p < .0 1), severity of pain ( r = 13 p < .0 1), frequency of pai n ( r = 12 p < .0 1), and interference of ADLs because of pain ( r = 22 p < .0 1) were all significantly related to frequency of clinic visits. Comparing one 's health to others was the only negatively correlated predictor with frequency of visits. The lowe r one rated one's health, the more likely it was that he or she would visit a clinic or emergency room. Greater pain and pain interference, along with number of chronic illnesses, all indicated a greater likelihood of seeking treatment. There was no signif icant relationship between visits and the measure of adiposity (waist:hip ratio), ( r = .01 p = 85 ). All three psychological variables were put into the regression model, although only perceived quality of life ( r = .20 p < .01) and depression ( r = 19 p < .0 1) were significantly related to frequency of clinic visits. Because social support ( r = .03 p = 34 ) is a core component of the hypothesized model, it was also included to determine how this construct might be interacting with other variables. A regression model that included all these variables was calculated See table 7 for the hierarchical linear regression analyses. T he final model that included all three psychological variables was significant F ( 15 605 ) = 7.77 p < .001, R 2 = 1 4 and a ccounted for 1 4 % of the variance in frequency of clinic emergency room visits However, including psychological variables into the model only added 0. 0 1 explained variability to the model, and the change in R 2 was not significant, F (3 605 ) = 0.69 p = 5 6 With just the predisposing variables in the model, 2 % of the variability was explained and the addition of enabling variables explained 4 % of the variability of

PAGE 39

31 seeking treatment. However, when need determinan ts were added into the model, 14 % was explai ned, which was a significant change in R 2 F ( 5, 608 ) = 14.93 p < .001. Need determinants added the most unexplained variability to the overall model, whereas psychological determinants only added a minimal amount of variability. To determine if multicol linearity among the psychological variables existed, each psychological variable was entered individually in the final model with all other variables previously entered. Adding social support into the model was not significant, R 2 = 0.00, F (1, 653) = 0.2 4, p = .63. Similarly, adding depression into the model alone was not significant, R 2 = 0.00, F (1, 654) = 0.15, p = .70, nor was perceived quality of life, R 2 = 0.02, F (1, 607) = 1.73, p = .19. Moderation analyses were examined to determine if any inte ractions that included the psychological variables were significant. There was a significant interaction between depression and chronic illnesses, b = .01, SE = 0.00, = .07, p < .05, as depicted in figure 4 The graph of the interaction with the data p oints overlaid was examined and verified that the interaction effect was within the bounds of observed data. It appears that as number of chronic illnesses increase, number of visits to a clinic or ER also increases. However, depression seems to play a lar ger role in this relationship if a person has fewer chronic illnesses. This moderation was also examined by plotting two slopes of clinically significant depression (# 16 on CES D) or non significant depression (< 16 on CES D) in Figure 5 The slope was ma rginally significant for those who scored in the depressed range on the CES D, b = .156, p = .06. F or those were not depressed, the slope was significant, b = .321, p < .001. Most of the data points fell within the range of fewer chronic illnesses. The eff ect of number of chronic illnesses on frequency of visits to a

PAGE 40

32 health care clinic depended on the presence of clinically significant depression in that when individuals were depressed, they utilized health care more than non depressed individuals when they were healthier (fewer chronic illnesses). However, as number of chronic illnesses increased, the frequency of clinic visits increased more f or those who were not depressed compared to their depressed counterparts. Other moderation analyses were also exami ned between depression, quality of life, and social support among predisposing, enabling, and need determinants, but none were significant.

PAGE 41

33

PAGE 42

34 Prayer as a form of complementary and/or alternative medicine Similar to the fir st analysis, only predictors that were significantly related to the outcome of prayer use were used in the regression analyses. Based on Pearson correlational analyses, sex ( r = .17, p < .001), age ( r = .07, p < .05), income ( r = .07, p < .05), accessibil ity difficulties ( r = .13, p < .001), chronic illnesses ( r = .11, p < .001), severity of pain ( r = .07, p < .05), pain interference with ADLs ( r = .13, p < .001), waist:hip ratio ( r = .11, p < .001), depression ( r = .10, p < .01), and quality of life ( r = .08, p < .05) all had significant associations with prayer use. Social support ( r = .02, p = .53) was not significantly related to prayer use. Use of prayer was a dichotomous variable (either a person engaged in pr ayer or did not), therefore chi square and logistic regressions were used to determine if and how

PAGE 43

35 determinants might predict use. A chi square analysis of independence was performed to examine the relationship between the categorical variables of sex and ethnicity with prayer. T here was a signi ficant difference among sex and use of prayer, 2 (1, N = 1325) = 36.18, p < .001. Women were more likely to engage in prayer for spiritual healing than were men. Ethnicity was also examined using chi square, and there was no difference between Non Hispani c White and Hispanic participants, 2 (1, N = 1325) = 0.21, p = .65. Using these variables, a hierarchical logistic regression analysis was performed to assess if psychological determinants increased the correct prediction of prayer use above and beyond th e predisposing, enabling, and need determinants With all determinants entered into the model, t here was a good model fit, 2 (12, N = 616 ) = 54.74 p < .001, Nagelkerke R 2 = .12 Overall classification was impressive in that it showed that the model corre ctly predicted 71.6 % of those who prayed. Of the predictors in the model, sex, = 1. 12 p < .001, accessibility difficulties, = 0.42 p = .001, and diastolic blood pressure, = 0.02, p < .05, were significant. Interestingly, although there was overall good model fit, adding psychological variables into the model did not significantly increase the model fit, 2 (3, N = 616) = 4.34, p = 0.23 indicating that the addition of psychological determinants to the model did not reliably distinguish those who ut ilized prayer as a source of health care. Adding enabling determinants (income and accessibility difficulties) was the only addition of predictors that significantly increased model fit 2 (2 N = 616) = 14.51, p < .001. Table 8 illustrates the individual contributions of the predictors to the model.

PAGE 44

36

PAGE 45

37 Preventative Care : Flu Shots Only predictors that were significantly related to flu shots were used in analyses to determine if psychological variables added more explained va riabilit y of health care uti lization than predisposing, existing, and need determinants Use of flu shots was a dichotomous variable (ei ther a person received a flu shot or did not ), therefore chi square and logistic regression were used to determine if and how de terminants might predict this type of preventative care. Based on Pearson correlational analyses, ethnicity ( r = .80, p < .01), education ( r = .11, p < .001), insurance ( r = .06, p < .05), place of care ( r = .10, p < .01), distance from care ( r = .10, p < .01), self rated health ( r = .07, p < .05), total chronic illnesses ( r = .17, p < .001), presence of pain ( r = .07, p < .05), severity of pain ( r = .07, p < .05), pain interference with ADLs ( r = .07, p < .05), and quality of life ( r = .07, p < .05) a ll had significant associations with flu shots. Depression ( r = .04, p = .19) and social support ( r = .03, p = .40) were not significantly related to flu shots. A chi square analysis of independence was performed to examine the relationship between sex a nd ethnicity and flu shots There was a significan t difference among sex regarding flu shots 2 (1, N =13 60 ) = 3.72 p = 05 Women were more likely to receive a flu shot than men Ethnicity was also examined using chi square, and there was a significant difference between Non Hispanic White and Hispanic participants, 2 (1, N =13 60 ) = 7.76 p < .01 in that Hispanics were less likely to receive a flu shot than Non Hispanic Whites Using these variables, a hierarchical logistic regression analysis was perf ormed to assess whether the addition of psychological determinants increased prediction of flu shots beyond the selected predisposing, enabling, and need determinants. The model was

PAGE 46

38

PAGE 47

39 a good fit, 2 (14 N = 654 ) = 47.30 p < .001 N agelkerke R 2 = .1 0 Overall, classification in the model that included psychological variables correctly predicted 68.0 % of those who received a flu shot Table 9 illustrates the hierarchical contributions of the predictors to the model. Adding all three p sychological variables into the model increased the predicted correct cases by 2.4 % more than the model with just predisposing, enabling, and need determinants However, w hile adding enabling and need determinants into the model significantly increased mod el fit, 2 (5 N = 654) = 7.15 p < .05 and 2 (11 N = 654) = 23.78 p < .001, respectively, adding psychological variables into the model did not significantly increase model fit, 2 (14, N = 654) = 1.40 p = 0.71 Although psychological variables added correct classification, the addition was not statistically significant for a better overall model fit. Of the predictors in the final model, education, = 0.07, p < .0 1, distance from care = 0.40 p < .0 1, and chronic illnesses = 0.26 p < .001 we re significant. Age was significant in the first model, = 0.03, p < .05, but dropped out after other predictors were entered into the model. Preventative Care for Men: Prostate Exams Only predictors that were significantly related to prostate exams wer e used in analyses to determine if psychological variables added more explained variability of health care utilization than predisposing, existing, and need determinants. Receiving a prostate exam was a dichotomous variable (either a person received a pros tate exam or did not), therefore chi square and logistic r egression were used to ascertain if these determinants might predict this type of preventative care for men A chi square analysis of independence was performed to examine the relationship between e thnicity and prostate exams Hispanic men were less likely to receive a prostate exam than Non Hispanic White men 2 (1, N = 515 ) = 10.17 p < .0 0 1.

PAGE 48

40 Based on Pearson correlational analyses, ethnicity ( r = .14, p < .01), education ( r = .15, p < .01), place of care ( r = .12, p < .01), and chronic illnesses ( r = .13, p < .01) were significantly related to prostate exams. None of the psychological variables were significantly related to prostate exams (depression, r = .05, p = .25; social support, r = .05, p = .32; and perceived quality of life, r = .08, p =.10). Using these variables, a hierarchical logistic regression analysis was performed to assess whether the addition of psychological determinants increased prediction of prostate exams beyond the selecte d predisposing, enabling, and need determinants There was a good model fit, 2 (7 N = 460 ) = 30.31 p < .0 01 Nagelkerke R 2 = 09 Overall the model with psychological variables correctly predicted 59.8 % of those who received a prostate exam Adding pre disposing and enabling determinants into the model significantly increased model fit, 2 (2 N = 460 ) = 17.11 p < .001 and 2 (1 N = 460 ) = 10.36 p < .001, respectively, but adding need variables did not significantly increase model fit, 2 (1, N = 460) = 1.84, p = 0.18. Furthermore the addition of psychological variables into the model did not significantly increase model fit, 2 (3 N = 460 ) = 1.0 0, p = 0.80 indicating that the addition of psychological determinants to the model did not reliably disti nguish men who received prostate exams from those who did not. Of the predictors in the final model, education, = 0.07, p < .05 and having a place of usual care, = 2.32 p < .05, were significant. Table 10 illustrates the hierarchical contributions of the predictors to the model.

PAGE 49

41

PAGE 50

42 Preventative Care for Women: Mammograms and pap smears Use of preventative care by women included both mammograms and pap smears. T hese preventative care exams for women were used in analyse s to de termine if psychological v ariables added more explained variability of health care utilization than predisposing, existing, and need determinants. Both were dichotomous variable s (either a person received the procedure or did not), therefore chi square and logistic regression was used to determine if these determinants might predict th ese type s of preventative care. A chi square analysis of independence was performed to examine the relationship between ethnicity and these procedures Neither was significant in that Hispanic and Non Hispanic White women did not significantly differ in terms of receiving either mammograms or pap smears Only predictors that were significantly related to mammograms were used in the regression analyses. A ge ( r = .19, p < .01) education ( r = .12, p < .01), income ( r = .18, p < .01), place of c are ( r = .09, p < .05) self rated health ( r = .09, p < .05), frequency of pain ( r = .12, p < .01), depression ( r = .10, p < .01), and perceived quality of life ( r = .08, p < .05) w ere s ignificantly correlated with mammograms using Pearson correlations Using these variables, a hierarchical logistic regression analysis was performed to assess prediction of mammograms with the addition of all three psychological variables of depression, so cial support, and perceived quality of life with the selected predisposing, enabling, and need determinants. The variable of having a place of usual health care was dropped because of multicollinearity. With the remaining variables, there was a good model fit, 2 (8, N = 346) = 26.63, p < .01, Nagelkerke R 2 = .10. Overall, the model with psychological variables was somewhat impressive in that it correctly predicted 63.6% of those who received a mammogram. The addition of need determinants into the model

PAGE 51

43 sig nificantly increased the model fit, 2 (5, N = 346) = 7.48, p < .05, but adding psychological variables into the model did not significantly increase model fit, 2 (8, N = 346) = 5.61, p = 0.13. The odds ratio of all three psychological predictors was not significant. Of the predict ors in the final model, age = 0.04 p < .01, and frequency of pain = 0.26 p < .05, were significant. Table 11 illustrates the hierarchical contributions of the predictors to the model for mammograms. For the outcome of p ap smears, only two predictors were significantly related to the outcome. Only age ( r = .01, p < .01) and quality of life ( r = .18, p < .01) was significant using Pearson correlations. Therefore, a regression model using all variables except for pain fre quency and place of usual care, which were both dropped due to multicollinearity, was performed for the pap smear outcome to better understand how all of the predictor variables contributed to the overall model. Using all of the variables, a hierarchical l ogistic regression analysis was performed to assess whether the addition of psychological determinants increased prediction of pap smears beyond the selected predisposing, enabling, and need determinants. The model was a good fit, 2 (18 N = 327 ) = 32.21 p < .05 Nagelkerke R 2 = .13 Overall, classification in the model that included psychological variables was unimpressive in that it correctly predicted 61.5 % of those who received a pap smear However, a dding all three psychological variables into the mo del increased th e predicted correct cases by 1.6 % more than the model with just predisposing, enabling, and need determinants. For the outcome of pap smears, adding psychological variables (depression, social support, and quality of life) significantly inc reased overall model fit, 2 (3, N = 327) = 12.15, p < .01, although of the three

PAGE 52

44

PAGE 53

45

PAGE 54

46 variables, only quality of life was significant predictor in the model, = 0.40, p < .001. Depression, = 0.00, p = 0.80, a nd social support, = 0.02, p = .69, were not significantly adding to overall model fit for pap smears Quality of life appeared to drive this overall significant effect. Interestingly, the only other predictor that was significant in the model was age, = 0.06, p < .01. Table 12 illustrates the hierarchical contributions of the predictors to the model for pap smears Hypothesis 2 To test hypothesis 2, an exploratory principal component factor analysis was performed to explore how factors based on the predictors in the sample from the San Luis Valley region may differ from those posited in Andersen's model. Oblimin rotation was used to allow the factors to correlate since the different domains of determinants are not independent. To determine compositi on of the factors, it was determined that variables could not load onto more than one factor and the highest loading would be determined for each variable and load only onto that factor. In an effort to simply understand how these variables related to one another, no cutoff point was used to exclude individual variables; even if they did not meet a .50 loading rule, they were still included in the factors. To determine factors, the following decision rules were used: (a) eigenvalue greater than 1; (b) exami nation of the scree plot; (c) items only loading on one factor; and (d) interpretability of the factor. Using these criteria, six factors were extracted, although not all variables loaded highly (> .50) on each. On the Psychosocial Characteristics factor, depression (.66), quality of life (.70), social support (.29), and accessibility difficulties (.35) loaded together. On the Physical Characteristics factor, sex (.90) and waist:hip ratio (.73) loaded highly. On the Sociocultural Characteristics factor, eth nicity (.70), education

PAGE 55

47 Table 13 (.71), income (.62), and distance from care (.21) loaded together, although distance from care much lower than the others. On the Objective Health Indicators factor, systolic (.80) and diastolic blood pressure (.78) load ed together. On the Age Related Characteristics factor, age (.72) and insurance (.21) loaded together, although insurance was much lower.

PAGE 56

48 Finally, on the Subjective Health Indicators factor, self reported health (.60), chronic illnesses (.52), and pain int erference with ADLs (.43) loaded well together. See Table 13 for factor loadings, eigenvalues, and percent variance accounted for by each factor.

PAGE 57

49 CHAPTER I V DISCUSSION Health care utilization is an individualized behavior with many idi osyncratic factors potentially interacting with one another. Andersen's Behavioral Model of Health Services Utilization ( Andersen, 1995; Andersen & Newman, 1973) sought to create a conceptual model explaining the factors that might influence health care us e including factors that exist prior to a health need, factors that enable a person to seek care, and the actual medical need of receiving services This theory posited interactions among the determinants but did not provide a clear categorization of psyc hological determinants. Moreover, systematic analyses using this model, began to categorize and divide some psychosocial variables such as mental diagnoses, social support, and health related quality of life into different domains of determinants (Babitsch Gohl & von Lengerke, 2012), yet a clear explanation of how these psychological factors might influence the overall model has yet to be succinctly posited. T he present study aimed to explore how psychological factors interact with the proposed determinant s and conceptualize how a rural, culturally specific populati on's health services utilization would fit onto Andersen's model. Predisposing, Enabling, and Need Determinants of Health Care Utilization It was hypothesized that Andersen's proposed domains o f determinants would predict health care utilization similarly in this population to previously studied populations ( Andersen, 1995; Andersen & Newman, 1973; Wolinsky & Johnson, 1991; Baxter, Bryant, Scarbro, & Shetterly, 2001), but psychological predictor s would also play a role in the explanation of health services utilization. Before addressing the

PAGE 58

50 psychological variables, however, it is important to understand how the originally proposed domains of determinants behaved with this sample. In the San Luis Valley Health and Aging Study, participants were older and from a more isolated, rural area. Because of this, the possibility remained that the predisposing, enabling, and need determinants might have predict ed somewhat differently in this sample. Given th is it was interesting to observe how these types of factors fit into the overall model. Understanding the demographic variables, alone, was important in predicting types of health care use. Almost forty percent of individuals sampled did not finish high school, and almost half had a yearly income of less than $10,000. This sample was skewed in terms of demographics, and it is possible that individuals less educated and below the poverty line utilize health care differently, possibly for a variety of reas ons. It is possible that age plays a larger role in this skewed sample. Age was significantly negatively related to education and income so the older one is, the less l ikely they were to have obtained higher levels of education and income. In all of the re gression analyses for utilization outcomes, predisposing characteristics alone in the model signi ficantly predicted the outcome, and a ge was a significant predictor in most regression analyses Given that the entire sample was 65 years of age and older, th is is noteworthy and possibly indicates that in the aging population, differences in age, even relatively small ones, can influence well being. Health related c hanges tend to occur more rapidly in an aging population compared to a middle aged population. Differences between men and women were also present in terms of utilization. Women were significantly more likely to receive a flu shot than their male counterparts, and they were also significantly more likely to pray than men. A conclusion can be made

PAGE 59

51 th at women utilize more types of health care than men. A difference in flu shots between Non Hispanic W hite and Hispanic participants was also present. Hispanics were less likely to receive flu shots as a preventative measure, which mirrors previous research This finding could indicate that regardless of urban or rural location, Hispanics are generally less likely to engage in preventative care. Another related factor not measured in this study that could have impacted outcomes was that of health literacy. Lower health literacy and greater obstacles, both physical and mental, may create barriers to seeking health care. Cho, Lee, Arozullah, and Crittenden (2008) explored how low health literacy affected health status and health service utilization and found t hat low health literacy had a direct ef fect on poorer health outcomes. Given this, it is interesting to note that these individuals sampled might have had lower health literacy that was contributing to health outcomes. Also, a lower self efficacy to manage illnesses might also be playing a large role in their ability and willingness to seek treatment. Accessibility of health care did not appear to have a significant role in utilization. In this sample, individuals did not endorse having many difficulties accessing health care. In fact, out of six possible difficulties, the mean reported by individuals was less than 1, giving evidence that those in this region did not feel these types of logistical or physical barriers existed for them, but we are unaware o f what other barriers may have existed, especially emotional and cogn i tive. Of note is tha t accessibility difficulties were negatively related to education, income, and insurance in that those who had less education, income, and insurance perceived greater difficulties in accessing health care Accessibility difficulties were also positively related to self reported health, number of

PAGE 60

52 chronic illnesses, and pain. It appears that difficulties accessing care is due more to poorer health and/or financial means than the physical availability of medical care clinics It is possible that in rural areas, access to primary care is comparable to that in more urban areas ( Hanratty, Zhang & Whitehead, 2007). Based on previous literature regarding Andersen's model ( And ersen, 1995; Andersen & Newman, 1973; Wolinsky & Johnson, 1991; Baxter, Bryant, Scarbro, & Shetterly, 2001 ), need determinants tend to be most predictive of health services utilization. That trend was seen in the present study, with self reported health, t otal number of chronic illnesses, pain, and pain interference with ADLs often contributing significantly to models with different health care outcomes. Perhaps more surprising, though, was that measure s of adiposity (waist:hip ratio) and blood pressure wer e not significant in most models. In the general population, obesity is a risk factor for myriad health problems and subsequent visits to a medical professional. In the aging population, one similar to this sample, the opposite can be true. Frailty in the aging population also indicates pathology. Waist:hip ratio was negatively correlated with education, self reported health, and quality of life. Despite these significant relationships, this measure of adiposity had little to no bearing on the predictive na ture of utilization. Similar to adiposity hypertension is a risk factor but did not contribute to utilization outcomes in this sample. It is quite possible that these need variables are so closely related that more significant pathology or diagnosed illne sses has a greater impact than risk factors for those illnesses. All of these variables most likely have a differing role depending on the type of outcome measured. In this study, three types of outcomes were measured, but it is not

PAGE 61

53 entirely clear exactl y what those mean in terms of pathology or appropriate behaviors. Frequency of visits to a clinic or emergency room, for example, could be both negative and positive. It is difficult to determine what an appropriate number of visits would be given a person 's medical history and presentation. Similar l y, receiving preventative care could be indicated if a person is more at risk for medical complications. In the elderly, receiving a flu shot is perhaps more important than for a younger individual. Diseases lik e the flu can have a far more detrimental effect for an elderly individual with comorbid medical diseases than for a health y young adult. Therefore, preventative care' might have a uniquely different connotation for a 75 year old immunocompromised male wi th heart disease and COPD than for a 25 year old male with no medical problems and a healthy immune system Psychological Determinants of Health Care Utilization Psychological variables can and do have a significant influence on decisions both medical an d otherwise, as well as perception of state. What was most notable in this study was the seemingly lack of individual contribution by psychological determinants on the outcomes of health care utilization. In almost every model, other factors, especially ne ed determinants, seemed to predict outcomes stronger than psychological components. This observation falls in line with the Andersen's originally hypothesized models. However, psychological determinants were significantly related to many other variables wi thin all three domains of predisposing, enabling, and need determinants. Given this, psychological determinants may help to explain how predisposing, enabling, and need determinants predict health care utilization rather than predict on their own.

PAGE 62

54 There was often a negative relationship between perceived quality of life and utilization, both visits to a clinic and preventative care These data suggest that as perceived quality of life increases, visits to a clinic or emergency room decrease. Visiting a do ctor likely means that you have more medical complications, or at least more perceived medical complications, thus indicating poorer quality of life. Moreover, d epression often had a positive relationship to outcomes so that as depression increased, visits increased. Likely a bidirectional relationship is occurring in that as depression increases, one is more likely to experience and/or catastrophize somatic symptoms leading him or her to seek treatment more often. It is also possibly quite true that as med ical conditions worsen, depression worsens as a consequence. The inverse relationship between perceived quality of life and depression could be indicative of coping ability and styles. The moderation analyses that provided the most insight into this was be tween depression and number of chronic illnesses. When a person had fewer chronic illnesses, higher depression indicated a decreased frequency in visits for a health treatment, which could possibly be categorized as "underutilization." As the number of chr onic illnesses increased, however, higher depression resulted in a greater frequency of visits. Though no other moderation analyses were significant with quality of life or depression, this interaction effect sheds light onto a possible mechanism behind th e psychological variables. Perhaps as overall health declines, sickness behavior becomes more susceptible to influences from negative mood. Temporality is difficult to untangle given the interactive nature of these constructs. These data would suggest tha t more visits to medical place of service are indicative of pathology from a holistic point of view (depression, quality of life, medical

PAGE 63

55 illnesses, etc.). A problem with these data, however, is the combination in the outcome of both clinic visits and visi ts to an emergency department. Separating these two locations into two different outcomes would likely provide a clearer picture of positive vs. negative behavior, as well as the mediating and moderating factors playing a role in seeking types of treatment One area of research that could provide an explanation to the mediating effects of depression and other psychological variables on health care utilization outcomes is the role of inflammation. The inflammatory hypothesis examines how psychoneuroimmunolo gical dysfunction may contribute to the manifestation of depression, sickness behavior, and actual health outcomes (Zunszain, Hepgul & Pariante, 2013) Inflammatory cytokines have been shown to be associated with depression, especially treatment resistant depression (Raison & Miller 2011 ). It is also well documented that inflammation is known to contribute to poor health outcomes. In this population immunosenescence, or the decline of immune functioning due to aging, is more apparent and involves the decr eased capacity to fight infections (McElhaney & Effros, 2009). Related to this, aging individuals often have overproduction of inflammatory markers, most notably C reactive protein (Aiello, Haan, Pierce, Simanek & Liang, 2009). The aging are more likely to have decreased immunological functioning and a greater number of morbidities. This prominent inflammation in these individuals could contribute to both depression and t he perception of one's ailments and therefore should be taken into account when assessi ng determinants that predict health care utilization.

PAGE 64

56 The third psychological variable measured failed to contribute to outcomes across the board. Social support never seemed to play a role in predictive nature in the models tested. This construct of soci al support was based on number of family and friends, rather than the qualitative nature of the support provided and perceived. Yohannes (2013) stated that quality is much more indicative of adjustment than quantity. Given the fact that this sample is an a ging population with somewhat limited capabilities, caregiving and the support from caregivers might be an even more important construct to measure than social support alone Analyzing Determinant Domains Within the Model To examine the second aim of the study, an exploratory factor analysis was performed to better understand how the determinants load together in this sample and compare them to Andersen's proposed determinants. Six factors were extracted with this sample, which is more than the three propo sed by Andersen, but these six factors still did not entirely describe the full picture. The first factor, Psychosocial Characteristics, was comprised of the psychological determinants of depression, social support, and quality of life. Surprisingly, acces sibility difficulties also loaded onto this factor, although not highly. It is possible that psychological factors such as depression and quality of life can influence the way barriers to care are perceived. Social support did not have a strong loading to any factor, which is not surprising given its limited contribution to models in this study and measure of quantity rather than quality. Education, ethnicity, income, and distance from care created the factor of Sociocultural Characteristics. These variabl es were significantly related to one another and might help to explain the availability of resources in this population. A separate, b u t

PAGE 65

57 related, factor was found that included age and insurance. This Age related Characteristics factor by itself is difficu lt to interpret. In this population, it seems that age acts uniquely from other demographic variables. It loaded highly on this factor (.72) compared to the other variable of insurance, which loaded at .21. Because of the lack of variability in this sample regarding insurance, it is unclear how much it truly impacting outcomes. It seems as though for interpretability purposes, insurance could be dropped altogether from the factors, thus creating a factor of just age and then another of sociocultural charact eristics influencing availability to resources. A Physical Characteristics factor including both sex and waist:hip ratio was determined to be separate from all other factors. Similar to the Age related Characteristics factor, this factor is difficult to i nterpret on its own. Both could be related to other health indicators, and in this sample, health indicators were clearly divided into subjective and objective. Systolic and diastolic blood pressure loaded together on Objective Health Indicators, and self reported health, chronic illnesses, and ADL interference due to pain loaded onto Subjective Health Indicators. The se two separate factors support the idea of need being a large component of health care utilization. It seems that indicators that are influen ced by perception differ from those that do not (blood pressure and measures of adiposity). It is possible that these aspects of need are much more salient to an individual than the objective measures such as adiposity or blood pressure. This also begs the question of how psychological variables might be influencing subjective need determinants. This factor of subjective need correlated most highly with depression and quality of life. These data give evidence that psychological factors

PAGE 66

58 influence not just on e determinant, but they can have a systemic influence of all determinants. If insurance is dropped and the physical characteristics factor is not included for lack of interpretability, five factors remain: Psychosocial Characteristics, Sociocultural Charac teristics, Age, Objective Health Indicators, and Subjective Health Indicators. Comparing these to Andersen's proposed Predisposing determinants, Enabling determinants, and Need determinates, it seems that conceptually, psychological variables constitute a completely different domain of determinants influencing outcomes of health care utilization. The Sociocultural factor closely resembles that of Andersen's Enabling determinants. Both are comprised of factors the influence availability of resources. Interes tingly, need determinants were broken into subjective and objective, but this finding illustrates the importance of this factor in health care utilization. Andersen's enabling and need determinants were shown to be present in some respect in the present st udy. However, i n general, Andersen's predisposing determinants loaded onto different factors in this population, possibly due to the high correlations among variables. No clear factor that resembled his predisposing determinants was seen in this study. In this population, age clearly was a unique variable, and it is difficult to ascertain how this variable might behave in other populations. Given these results, it appears that a separate factor of psychological factors should be considered, at least in a co nceptual model. Amending the Conceptual Model of Health Services Utilization Andersen originally postulated that many, if not all, of the determinants would interact with one another. One of the strengths and weaknesses of his model is that it is more of an abstract model rather than an exhaustive list of predictive factors. It takes

PAGE 67

59 more of a systems based approach and does not specify certain variables but rather gives a general overview of the types of constructs that might influence utilization. As fie lds of literature change and deepen, this model can be used in different situations and from different viewpoints (e.g. inflammatory hypothesis in psychoneuroimmunology) This study set out to better understand how psychological determinants influence the existing model of health services utilization and empirically test Andersen's conceptual model What became readily apparent, though, was the difficulty in empirically separating determinants and their interactions with psychological variables especially given the limited number of determinants measured in the study. It appears that there is a great deal of multicollinearity that occurs between these variables. Although psychological variables seldom significantly added to the overall models, it is clear t hat they impact various other determinants. G iven the results and conclusions thus far, it would seem appropriate t hat amend ing Andersen's conceptual model to include psychological variables as a domain in itself that influences the other domains would pro vide a more robust conceptualization of variable s that could influence overall health care utilization. In addition, empirically, different factors are extracted, and although they somewhat resemble Andersen's factors, they did differ to some extent (e.g. need is separated into objective and subjective factors and predisposing determinants are spread out among various factors). However, to conceptually guide future research, it is likely that the best way to think about domains influencing health care utili zation is similar to Andersen's approach: age (or other basic demographic information), sociocultural factors, and ne ed determinants. Figure 6 illustrates this conceptual model.

PAGE 68

60 Figure 6. Proposed amendment to Andersen's conceptual model of health care u tilization A Study of Culture A particular noteworthy piece of this study was the sample itself. Because of its targeted sampling of ethnicity, age, and rural residency, all of which could influence health care utilization in various ways, the San Luis Valley Health and Aging Study, conducted in southern Colorado, provided a unique population to test Andersen's model in a comprehensive manner. Due to its uniqueness, one could posit that this particular sample held its own cultural influences on health c are utilization. Prayer, for instance, could be much more of a culturally specific behavior rather than a type of health care. It is debated in the literature whether prayer constitutes complementary and alternative medicine, but in this population, it s eems as though it would be more complementary than alternative. Prayer is a normal part of Hispanic culture, and given the sample of the study, it is plausible that regardless of ethnicity, prayer is a major part of the culture. Similarly, the way depressi on and quality of life are

PAGE 69

61 viewed can vary with cultural contexts. Integrating cultural beliefs and practices into the model can help to provide a more holistic conceptualization of health care utilization. Strengths and Limitations As discussed, the samp le in this study provided one of its biggest strengths, as did the number of variables able to be measured. Within each study, it is impossible to measure all variables that might be influencing an outcome. However, a limitation of the study was the lack o f data regarding health beliefs. These beliefs can strongly influence all determinants, and because of the lack of measure assessing this, it is difficult to ascertain how these may have interacted with determinants, especially the psychological determinan ts. Nevertheless, this does not diminish the need for the addition of psychological variables to the model. As mentioned, the sample, although unique in its ability to give insight into this population, also can be a weakness in that there is somewhat limi ted generalizability, although the argument for assessing culture gives weight to this as a strength. Perhaps the greatest limitation was that t he data were from a previous study and were collected almost twenty y ears ago. Measures and questionnaires were not asked in a way that best facilitated the aims of the study, and therefore it became difficult to statistically analyze the data in a way that best answered the study's questions. Future Directions This line of research could go in many different future directions, some of which have already been addressed. First, assessing health beliefs will be crucial in future studies, especially to determine if a nd how they interact with psychological variables. Also, in future studies, social support should be a me asure of quality rather than quantity, and if aging populations are used, caretakers will be an important component to assess.

PAGE 70

62 Health literacy and self efficacy related to health might be important routes to consider as well. These construct s might play si gnificant role s in utilizing health care, especially in populations that are more rural and/or have residents with lower socioeconomic status or education level. It might also show differences in receiving different types of health care (emergency room vis its vs. visits to a primary care clinic). One aspect that may provide additional insight into an aging population's utilization of health care is that of cognitive functioning. This was beyond the scope of this study, but there is research that suggests l inks between cognitive functioning, inflammation, disease, and depression. This would be an interesting way to conceptualize health care utilization from a more psychoneuroimmunological framework. This also might have ties to the question of adherence or c ompliance to medical regimens. Investigations within this area of research could likely provide substantial insight into both the area of health care utilization and the inflammatory hypothesis. Conclusions Continuing to test models of health care utili zation in different populations with different facets emphasized will help to better understand the pathways of utilization in a more holistic manner. However, revising Andersen's Model of Health Services Utilization to include psychological variables in a systems based approach will help conceptualize health care utilization in different populations and cultures.

PAGE 71

63 R EFERENCES Abdel Khalek, A. M. (2014). Happiness, health, and religi osity: significant associations among Lebanese adolescents. Mental Health Religion & Culture 17 (1), 30 38. doi:10.1080/13674676.2012.742047 Aiello, A. E., Haan, M. N., Pierce, C. M., Simanek, A. M., & Liang, J. (2008). Persistent infection, inflammation, and functional impairment in older Latinos. The Journals of Gerontology Series A, Biological Sciences and Medical Sciences 63 (6), 610 61 8. http://doi.org/63/6/610 Apel, M. D. (1986). The a ttitudes and k nowledge of c hurch m embers and p astors r elated to o lder a dults and r etirement. Journal of Religion and Aging, 293, 31 43 Andersen, R. M. (1995). Revisiting the behavioral mo del and access to medical care: does it matter? Journal of health and social behavior 36 (1), 1 10. Andersen, R., & Newman, J. F. (1973). Societal and ind ividual determinants of medical care utiliz ation in the United States. The Milbank Memo rial Fund quarterly. Health and society 51 (1), 95 124. Andersen, R., & Newman, J. F. (2005). Societal and ind ividual determinants of medical care utilization in the United States. Milbank Quarterly, 83 (4), 1 28. Arcury, T. A., Gesler, W., M., Preisser, J. S., Sherman, J., S pencer, J., & Perin, J. (2005). The effects of geography and spatial behavior on health care utilization among the residents of a rural region. Health Services Research, 40 (1), 135 156. Babitsch, B., Gohl, D., & von Lengerke, T. (2012). Re r evisiting Andersen's Behavioral Model of Health Services: A systematic re view of studies from 1998 2011. Psychosomatic Medicine, 9, doi: 10.3205/psm000089 Baxter, J., Bryant, L. L., Scarbro, S., & S hetterly, S. M. (2001). Patterns of Rural Hispanic and Non Hispanic White Health Care Use: The San Luis Valley Health and Aging Study. Research on Aging 23 (1), 37 60. doi:10.1177/0164027501231003

PAGE 72

64 Baxter J Shetterly S M Eby C Mason L Cortese C F Hamman R F. (1998) Social network factors associated with Perceived Quality of Life. Journal of Aging and Health, 10 (3) 287 310. Bean, F.D., & Tienda, M. (1987). The Hispanic population of the United States New York, NY: Russell Sage Foundation. Benjamins, M. R. (2005). Social Determinants of Prev entive Service Utilization: How Religion Influences the Use of Cholesterol Screening in Older Adults. Research on Aging 27 (4), 475 497. doi:10.1177/0164027505276048 Benjamins, M. R. (2006). Religiou s influences on preventive health care use in a nationally representative sample of middle age women. Journal of behavioral medicine 29 (1), 1 16. doi:10.1007/s10865 005 9035 2 Benjamins, M., & Brown, C. (2004). Religion and preve ntative health care util ization among the elderly. Social Science & Medicine 58 (1), 109 118. doi:10.1016/S0277 9536(03)00152 7 Berkanovic, E., Telesky, C., & Reeder, S. (1981). Structural and social psychological factors in the decision to seek medical care for symptoms. Medic al Care, 19 693 709. Birrer, R. B., & Vemuri, S. P. (2004). Depression in later life: a diagnostic and therapeutic challenge. American Family Physician, 69 (10), 2375 2382. Blount, A., Schoenbaum, M., Kathol, R., Rollman, B ., Thomas, M., O'Donohue, W., & Peek, C. J. (2007) The economics of behavioral health services in medical settings: A summary of the evidence. Professional Psychology: Research and Practice, 38 (3), 290 297. Boyd, C. M., Darer, J., Boult, C., Fried, L. P., Boult, L ., & Wu, A. W. (2 005). Clinical practice guidelines and quality of care for older patients w ith multiple comorbid diseases: Implications for pay for performance. JAMA, 294 (6), 716 724. Brownson, C. A., & Heisler, M. (2009). The role of pe er support in diabetes care and s elf management. Patient, 2 (1), 5 17.

PAGE 73

65 Bureau of Labor Statistics United States Department of Labor (2014). CPI (Consumer Price Index) Calculator. Available from http://www.bls.gov/data/inflation_calculator.htm Butterworth, S. W., Linden, A., & McClay, W. (2007). Health coaching as an intervention in health management programs. Diseas e Management & Health Outcomes, 15 (5), 299 307. Caldwell E. M., Baxter J., Mitchell C. M., Shetterly S. M., Hamman R. F. (1998) The association of non insulin dependent d iabetes mellitus with perceived quality of life in a biethnic population: The San Luis Valley Diabetes Study. American Journal of Public Health. 88 (8), 1225 1229. Carey, T. S., Garrett, J. M., & Jackman, A. M. (20 00). Beyond the good prognosis. Examinat ion of an inception cohort of patients with chronic low back pain. Spine, 25 (1), 115 120. Carlisle, D. M., Leake, B. D., & Shapiro, M. F. (1997). Racial a nd ethnic disparities in the use of cardiovascular procedures: Associations with type of health insu rance. American Journal of Public Health, 87 (2), 263 267. Carney, R. M., Freedland, K. E., Eisen, S. A., Rich, M. W ., & Jaffe, A. S. (1995). Major depression and medication adherence in elderly patient s with coronary artery disease. Health Psychology, 14 88 90. Cho, Y. I., Lee, S. Y. D., Arozullah, A. M., & Crittenden, K. S. (2008). Effects of health literacy on health status and health service utilization amongst the elderly. Social Science & Medicine 66 (8), 1809 1816. http://doi.org/10.1016/j.socsc imed.2008.01.003 Coffield, A. B., Maciosek, M. V., McGinnis, J. M., Harris, J. R., Caldwell, M. B., Teutsch, S. M., Atkins, D., Richland, J. H., Haddix, A. (200 1). Priorities among recommended clinical preventative services. American Jo urnal of Preventat ive Medicine, 21 (1), 1 9. C™tŽ, P., Cassidy, J. D., & Carroll, L. (2001). The treat ment of neck and low back pain: who seeks care? who goes where? Medical care 39 (9), 956 67.

PAGE 74

66 D'Crus, A. & Wilkinson, J. M. (2005). Reasons for choosing and complying w ith complementary health care: An in house study on a south Australian clinic. The journal of alternative and complementary medicine, 11, 6, 1107 1112. de Boer, A., Wijker, W., & de Haes, H. (1997). Predictors of health care utilization in the chronicall y ill: A review of the literature. Health Policy, 42 101 115. Dunlop, D. D., Manheim, L. M., Song, J., & Chang, R. W. (2002). Gender and ethnic/racial disparities in health care utilization among older adults. The journals of gerontology. Series B, Psyc hological sciences and social sciences 57 (4), S221 33. Eisenberg, D. M., Kessler, R. C., Foster, C., Norlock, F. E., Calkins, D. R., & Delbanco, T. L. (1993). Unconventional medicine in the United States, New England Journal of Medicine, 328 246 252. Edwards, W. S., Winn, D. M., & Kurlantzick, V. (1994). Evaluation of National Health Interview Survey Diagnostic R eporting. National Center for Health Statistics. Vital and health statistics, 2 ( 120 ) 1 116. Gask, L., Dowrick, C., Salmon, P., Peters, S., & Mo rriss, R. (2011). Reattribution reconsidered: Narrative review and reflections on an educational interven tion for medically unexplained symptoms in primary care settings. Journal of psychosomatic research, 71, 325 334. Glassman, A. H., Helzer, J E., Covey, L. S., Cottler, L. B., Stetner, F., Tipp, J. E., & Johnson, J., (1990). Smoking, smoking cessation, and major depression. JAMA, 300, 1546 1549. Gornick, M. E., Eggers, P. W., Reilly, T. W., Mentnech, R M., Fitterman, L. K., Kucken, L. E., & Vladeck, B. C. (1996). Effects of race and income on mortality and use of services among Medicare beneficiaries. New England Journal of Medicine, 335, 791 799. Hamman, R. F., Mulgrew, C. L., Baxter, J., S hetterly, S. M., Swenson, C., & Morgenstern, N. E (1999). Methods and prevalence of ADL limitations in Hispanic and non Hispanic white subjects in rural Colorado: the S an Luis Valley Health and Aging Study. Annals of epidemiology 9 (4), 225 35.

PAGE 75

67 Hanratty, B., Zhang, T., & Whitehead, M. (2007). How clos e have universal health systems come to achieving equity in use of curative services? A systematic review. International journal of health services, 37 (1), 89 109. Helman, C. G. (1994). Culture, Health and Illness: An Introduction for Health Professiona ls. Bristol, UK: Butterworth Heinmann. Hershey, J. C., Luft, H. S., & Gianaris, J. M. (1975). Making sense out of utilization data. Medical Care, 13 (10), 838 854. Hispanic health and nutrition examination survey, 1982 84 (1990). Findings on health sta tus and health care needs. American Journal of public health, (80), 1 70. Hybels, C. F., & Blazer, D. G. (2003). Epidemiology of late life mental disorders. Clinics in Geriatric Medicine, 19 (4), 663 696. Kathol, R. G., McAlpine, D., Kishi, Y., Spies, R., Meller, W, Bernhardt, T., Gold, W. (2005). General medical and pharmacy claims expenditures in users of behavioral health services. Journal of General Internal Medicine, 20, 160 167. Keeley, P., Creed, F., Tomenson, B., Todd, C., Bor glin, G., & Di ckens, C. (2008). Psychosocial predictors of health related quality of life and health service utili s ation in people with chronic low back pain. Pain, 135 142 150. Koenig, H. G., & Larson, D. B (1998). Use of hospital serv ices, religious attendance, and religious affiliation. Southern Medical Journal, 91, 925 932. Kolk, A. M., Hanewald, G. J. F. P., Schagen, S., & Gijsbers van Wi jk, C. M. T. (2002). Predicting medically unexplained physical symptoms and health care utilization. A symptom perception app roach. Journal of psychosomatic research 52 (1), 35 44. Kolk, A. M., Hanewald, G. J. F. P., Schagen, S., & Gijsbers van Wijk, C. M T. (2003). A symptom perception approach to common physical symptoms. Social science & medicine (1982) 57 (12), 2343 54.

PAGE 76

68 Kolk, A. M., Schagen, S., & Hanewald, G. J. F. P. (2004). Multiple medically unexplained physical symptoms and health care utilization: Outcome of psychological intervention and patient related predictors of change. Journal of Psychosomatic Research, 57 379 389. Kornstein, S. G. (1997). Gender differences in depression: implications for treatment. Journal of Clinical Psychiatry, 58, 12 8 Le, T. D., Carney, P. a, Lee Lin, F., Mori, M., Chen, Z., Leu ng, H., Lau, C., et al. (2014). Differences in know ledge, attitudes, beliefs, and perce ived risks regarding colorectal cancer screening among Chinese, Korean, and Vietnamese sub groups. Jo urnal of community health 39 (2), 248 65. doi:10.1007/s10900 013 9776 8 Levin J. S., & Markides K. S. (1985). Religio n and health in Mexican Americans. Journal of Religion and Health, 24 60 69. Lewinsohn, P.M., Seeley, J.R., Roberts, R.E., & Allen, N.B. ( 1997). Center for Epidemiological Studies Depression Scale (CES D ) as a screening instrument for depression among c ommunity residing older adults. Psychology and Aging, 12, 277 287. McCathie, H. C., Spence, S. H., & Tate, R. L. (2002). Adjustment to chron ic obstructive pulmonary disease: The importance of psychological factors. European Respiratory Journal, 19 (1), 4 7 53. McElhaney, J. E., & Effros, R. B. (2009). Immunos enescence: what does it mean to health outcomes in older adults? Current Opinion in Immunology 21 (4), 418 424. http://doi.org/10.1016/j.coi.2009.05.023 Mechanic, D. (1979). Correlates of physician utilization: Why do multivariate studies of physician utilization find trivial psychosocial and organizational effects? Journal of Health and Social Behavior, 20, 387 396. MuthÂŽn, B. & MuthÂŽn, L. (2013). Mplus version 7.1 [Computer software]. Nguyen, M., Ugarte, C., Fuller, I., Haas, G., & Portenoy, R K. (2005). Access to care for chronic pain: racial and ethnic differences. The journal of pain!: o fficial journal of the American Pain Society 6 (5), 301 14. doi:10.1016/j.jpain.2004.12.008

PAGE 77

69 Patrick, D. L ., Danis, M., Southerland, L. I., Hong, G. (1988). Quality of life following intensive care, Journal of General Internal Medicine, 3, 218 223. Peveler, R., Kilkenny, L., Kinmonth, A. (1997). Medically unexplained physical symptoms in primary care: A com parison of self report screening questionnaires and clinical opinion. Journal of psychosomatic research, 42, 245 252. Radloff, L. S. (1977). The CES D scale: a self report depression scale for research in the general population. Applied Psychological Mea surement 1 385 401. Raison, C. L., & Miller, A. H. (2011). Is depression an inflammatory disorder? Current psychiatry reports, 13 (6), 467 475. Ricketts, T. C. (1999). Rural Health in the United States New York: Oxford University Press. Rogers, A. T. (2010). Exploring health beliefs and care seeking behaviors of older USA dwelling Mexicans and Mexican Americans. Ethnicity & health 15 (6), 581 99. doi:10.1080/13557858.2010.500018 Rosland, A. M., Kieffer, E., Israel, B., Cofield, M., Pal misano, G., Sinco, B., Spencer, M., & Heisler, M. (2008). When is social support important? The association of family support and professional support with specific diabetes self management behaviors. Journal of General Internal Medicine, 23 (12), 1992 1999. Roth, D .L., Ackerman, M. L., Okonkwo, O. C., & Burgio L. D. (2008). The four factor model of depressive symptoms in dementia caregivers: A structural equation model of ethnic differences. Psychology and Aging, 23, 567 576. San Luis Valley Regional Development and Planning Commission (1992). San Luis Valley Overall Economic Development Program. Alamosa, CO: San Luis Valley Regional Development and Planning Commission. Sobel D S. (1995). Rethinking medicine: Improving health outc omes with cost effective psyc hosocial interventions. Psychosomatic Medicine 53 (3), 234 244. d oi:10.1097/00006842 199505000 00005

PAGE 78

70 SPSS Inc. (2012). Statistical package for the social sciences v. 21 [Computer software]. Sullivan M. J., Feuerstein, M., Gatchel, R., Linton, S. J., & Pransky, G. (2005). Integrating psychosocial and behavioral interventions to achieve optimal rehabilitation outcomes. Journal of Occupational Rehabilitation, 15 (4),475 489. DOI: 10.1007/s10926 005 8029 9 Twells, L. K., Bridger, T., Knight, J. C., Alagheh ba ndan, R., & Barrett, B. (2012). Obesity predicts primary health care visits: a cohort study. Population health management 15 (1), 29 36. doi:10.1089/pop.2010.0081 Verhaak, P. F. M., Meijer, S. A., Visser, A. P., & Wolters, G. (20 06). Persistent present ation of medically unexplained symptoms in general practice. Family practice, 23 (4), 414 420. Whooley, M. A., de Jonge, P., Vittinghoff, E., Otte, C., Moos, R., Carney, R. M., Browner, W. S. (2008). Depressive symptoms, health behaviors, and risk of car diovascular events in patients with coronary heart disease. JAMA, 300, 2379 2388. Wolinsky, F. (1988). Seeking and Using Health Services. In The Sociology of Health (117 144). Belmont, CA: Wadsworth. Wolinsky, F. D., and Johnson, R. J. (1991). The use of health services by older adults. Journal of Gerontology, 46, 345 357. Yohannes, A. M. (2013). Is it quality or quantity of soc ial support needed for patients with chronic medical illness? Journal of psychosomatic research 74 (2), 87 8 doi:10.1016/j.jpsychores.2012.11.009 Ziegelstein, R. C., Bush, D. E., & Fauerbach, J. A (1998). Depression, adherence behavior, and coronary disease outcomes. Archives of i nternal m edicine, 158, 808 809. Zunszain, P. A., Hepgul, N., & Pariante, C. M. (2013 ). Inflammation and depression. Current topics in behavioral neurosciences, 14, 135 151

PAGE 79

71

PAGE 80

72

PAGE 81

73

PAGE 82

74

PAGE 83

75

PAGE 84

76

PAGE 85

77

PAGE 86

78

PAGE 87

79