USF Libraries
USF Digital Collections

The role of social structural and social contextual factors in shaping chronic disease and chronic disease risk behavior

MISSING IMAGE

Material Information

Title:
The role of social structural and social contextual factors in shaping chronic disease and chronic disease risk behavior a multilevel study of hypertension, general health status, and mental distress
Physical Description:
Book
Language:
English
Creator:
McKay, Caroline Mae
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Social determinants of health
Income inequality
Poverty
Social capital
Health behavior
Cardiovascular disease
Self-reported health
Mental health
Multilevel models
Dissertations, Academic -- Public Health -- Doctoral -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: At present there is a reliance on behavioral interventions that have been limited in their effectiveness to reduce the public health burden of chronic disease, partly because the effects of social context on the initiation and maintenance of health behaviors is not incorporated into public health policy and practice. Yet current research indicates that there are macro-level structural and contextual influences on population health that cannot be reduced to individual or compositional effects. This study investigated the associations between social structural factors, community social context, individual characteristics, and self-reported correlates of disease. Distal influences included social structural inequalities such as income inequality and absolute deprivation or poverty. Pertinent mechanisms through which these influences might have operated on disease included social contextual factors, such as social capital. Both political economy and the ecosocial perspect ive were selected to inform this study and to provide the theoretical framework from which hypotheses were derived.The design was a multilevel, retrospective, nonexperimental study using secondary data. The study linked three data sources (2001 Behavioral Risk Factor Surveillance System, Social Capital Community Benchmark Study, and U.S. Census) by Federal Information Processing Standards codes in order for individuals to be placed in their community or state contexts. Results provided mixed evidence of the direct role of structural and contextual inequalities on self-rated health. Any direct effects of social structural inequalities on the health outcomes disappeared once individual factors were included in the models. Findings demonstrated that one dimension of social capital, organizational activism, retained its significant direct effect on general health status, once individual characteristics were considered. Conclusions suggested indirect associations whereby the negative i nfluence of social structural inequalities on health was mediated by the erosion of social trust, which in turn was associated with engaging in risk behavior, thus increasing the odds of reporting hypertension, fair/poor general health, and mental distress. Although results were inconsistent, this study contributed to advancing Healthy People 2010 goals of increasing quality of life and reducing health disparities by advancing understanding of the multilevel nature of perceived health and the chronic diseases they predict.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2006.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Caroline Mae McKay.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 303 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001787940
oclc - 126871295
usfldc doi - E14-SFE0001434
usfldc handle - e14.1434
System ID:
SFS0025754:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001787940
003 fts
005 20070514144150.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 070514s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001434
040
FHM
c FHM
035
(OCoLC)126871295
049
FHMM
090
RA425 (ONLINE)
1 100
McKay, Caroline Mae.
4 245
The role of social structural and social contextual factors in shaping chronic disease and chronic disease risk behavior :
b a multilevel study of hypertension, general health status, and mental distress
h [electronic resource] /
by Caroline Mae McKay.
260
[Tampa, Fla] :
University of South Florida,
2006.
3 520
ABSTRACT: At present there is a reliance on behavioral interventions that have been limited in their effectiveness to reduce the public health burden of chronic disease, partly because the effects of social context on the initiation and maintenance of health behaviors is not incorporated into public health policy and practice. Yet current research indicates that there are macro-level structural and contextual influences on population health that cannot be reduced to individual or compositional effects. This study investigated the associations between social structural factors, community social context, individual characteristics, and self-reported correlates of disease. Distal influences included social structural inequalities such as income inequality and absolute deprivation or poverty. Pertinent mechanisms through which these influences might have operated on disease included social contextual factors, such as social capital. Both political economy and the ecosocial perspect ive were selected to inform this study and to provide the theoretical framework from which hypotheses were derived.The design was a multilevel, retrospective, nonexperimental study using secondary data. The study linked three data sources (2001 Behavioral Risk Factor Surveillance System, Social Capital Community Benchmark Study, and U.S. Census) by Federal Information Processing Standards codes in order for individuals to be placed in their community or state contexts. Results provided mixed evidence of the direct role of structural and contextual inequalities on self-rated health. Any direct effects of social structural inequalities on the health outcomes disappeared once individual factors were included in the models. Findings demonstrated that one dimension of social capital, organizational activism, retained its significant direct effect on general health status, once individual characteristics were considered. Conclusions suggested indirect associations whereby the negative i nfluence of social structural inequalities on health was mediated by the erosion of social trust, which in turn was associated with engaging in risk behavior, thus increasing the odds of reporting hypertension, fair/poor general health, and mental distress. Although results were inconsistent, this study contributed to advancing Healthy People 2010 goals of increasing quality of life and reducing health disparities by advancing understanding of the multilevel nature of perceived health and the chronic diseases they predict.
502
Dissertation (Ph.D.)--University of South Florida, 2006.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 303 pages.
590
Adviser: Melinda S. Forthofer, Ph.D.
653
Social determinants of health.
Income inequality.
Poverty.
Social capital.
Health behavior.
Cardiovascular disease.
Self-reported health.
Mental health.
Multilevel models.
690
Dissertations, Academic
z USF
x Public Health
Doctoral.
773
t USF Electronic Theses and Dissertations.
0 856
u http://digital.lib.usf.edu/?e14.1434



PAGE 1

The Role of Social Structural and Soci al Contextual Factors in Shaping Chronic Disease and Chronic Disease Risk Behavior: A Multilevel Study of Hypertension, General Health Status, and Mental Distress by Caroline Mae McKay, M.S.W. A dissertation submitted in partial fulfillment of the requirement s for the degree of Doctor of Philosophy Department of Community & Family Health College of Public Health University of South Florida Major Professor: Meli nda S. Forthofer, Ph.D. Jeannine Coreil, Ph.D. Robert McDermott, Ph.D. C. Hendricks Brown, Ph.D. Ichiro Kawachi, M.D., Ph.D. Date of Approval: March 31, 2006 Keywords: social determinants of health, income inequality, poverty, social capital, health behavior, cardiovascular disease, self-reported health, mental health, multilevel models Copyright 2006, Carolin e Mae McKay, M.S.W.

PAGE 2

Dedication This work is dedicated to two very important people in my life, my husband, Beau McKay, and my Mom, Shirley Frishman. Without them, this work may never have been completed. To Beau, your faith in me carried me along when I could not carry the load alone, y our irreverence helped me to laugh at myself and keep needed perspective, your c onstant curiosity reminded me there was a whole world outside my desk just waiting to be discovered and studied, and most of all, your love constantly inspires me to be the person I see when I look in your eyes. To Mom, your neve r-wavering belief that I could accomplish this has helped me through this process. I want you to know that I envision this work and what it represents as a part of a long line of women who believed in tomorrow – Ida to Clara to you to me.

PAGE 3

Acknowledgments I would like to thank the following people for their support throughout this journey: my major professo r, Dr. Lyndie Forthofer, for instilling in me a drive to always strive for excellence, to apprecia te great dinners and conversation, and to remember the lighter side of serious pursuits; Dr. Jeannine Coreil, for making me a better critical thinker and constantly challenging me to challenge myself; Dr. Ichiro Kawachi, for inspiring me and making scholarship a vibrant, exciting process; and Bethany Bell Ellison, for being a shoulder to lean on, an ear to vent to, and a confidante with an amazing abilit y to debate and clarify at the same time.

PAGE 4

i Table of Contents List of Tables v List of Figures viii Abstract ix Chapter One: Background and Significance 1 Introduction 1 Purpose of Study and Study Significance 4 Study Rationale 5 Selected Explanations for Inequalities 6 Limitations of Exis ting Knowledge Base 10 Preliminary Hypotheses 10 Overview of Design 11 Data Sources 12 Implications for Public Health 13 Delimitations 15 Limitations 16 Definitions 17 Chapter Two: Conceptual Fram ework and Review of Literature 21 Theoretical Framework 21 Literature Review 24 Social Structural Factors and Health 27 Relative Deprivation: Income Inequality 27 Absolute Deprivation: Poverty 33 Social Structural Factors and Health Behaviors 38 Social Structural Factors and Social Context 42 Social Context and Health 46 Social Context and Health Behaviors 54 Health Behaviors and Hypertension 60 Physical Inactivity 61 Obesity 64 Smoking 66 Health Behaviors and General Health Status 68 Physical Inactivity 69 Obesity 70 Smoking 71

PAGE 5

ii Health Behaviors and Mental Distress 72 Physical Inactivity 73 Obesity 74 Smoking 75 Hypotheses 79 Chapter 3: Methodology 87 Study Design 87 Sampling 89 Behavioral Risk Factor Surv eillance System (BRFSS) 90 Social Capital Community Benchmark Survey (SCCBS) 93 Census 94 Variable Measures 95 Analysis Procedures 108 Chapter 4: Results 114 Univariate Analysis 114 Sociodemographic Factors 114 Health Factors: Behavioral Variables 116 Health Factors: Outcome Variables 122 Social Contextual Factors 126 Social Structural Factors 132 Bivariate Analysis 136 Summary of Significant Biva riate Associations among Key Study Constructs 149 Correlates of Hypertension 151 Correlates of General Health Status 151 Correlates of Mental Distress 152 Associations between Soci al Contextual Factors and Health Behavior 159 Associations between Social Structural Factors and Health Behavior 162 Associations between Social Structural and Social Contextual Factors 162 Multilevel Analysis 164 Cluster 1: Behavioral variabl es only partially mediate social structure and disease 165 Hypertension 166 General Health Status 169 Mental Distress 170 Health Behavior 172 Cluster 2: Social context parti ally mediates social structure and disease 178 Social Capital and Disease 178

PAGE 6

iii Hypertension 178 General Health Status 181 Mental Distress 186 Social Structure and Social Capital 187 Cluster 3: Behavior only parti ally mediates social context and disease 190 Health Behavior and Health Outcomes 191 Health Behavior and Social Capital 192 Social Capital, Health Behavior, and General Health Status 200 Summary of Findings 203 Chapter 5: Discussion and Implications 207 Summary of Findings 207 Limitations of Study 211 Data Sources and Sampling 212 Variable Selecti on and Measurement 213 Design Issues 217 Contributions of Study and Implic ations for Public Health 219 Recommendations for Future Research 228 Conclusion 231 References 234 Appendices 276 Appendix A1: Sample Distribution Comparison C.F. of Greater Birmingham 276 Appendix A2: Sample Distribution Comparison Arizona Community Foundation 277 Appendix A3: Sample Distribution Comparison California C.F. 278 Appendix A4: Sample Distribution Comparison The San Diego Foundation 279 Appendix A5: Sample Distribution Comparison The Walter and Elise Haas Fund 280 Appendix A6: Sample Distribution Comparison (C.F. Serving Boulder County) 281 Appendix A7: Sample Distribution Comparison Denver Foundation/Rose C.F./Piton Foundation 282 Appendix A8: Sample Distribution Comparison Delaware Division of State Service Centers/ Delaware C.F. 283 Appendix A9: Sample Distribution Comparison C.F. for Greater Atlanta 284

PAGE 7

iv Appendix A10: Sample Distribution Comparison Indiana Grantmakers Alliance 285 Appendix A11: Sample Distribution Comparison Forum 35/Baton Rouge Area Foundation 286 Appendix A12: Sample Distribution Comparison Kalamazoo C.F. 287 Appendix A13: Sample Distribution Comparison C.F. for Southeastern Michigan 288 Appendix A14: Sample Distribution Comparison The St. Paul Foundation 289 Appendix A15: Sample Distribution Comparison Montana C.F. 290 Appendix A16: Sample Distribution Comparison New Hampshire Charitable Foundation 291 Appendix A17: Sample Distribution Comparison Central New York C.F. 292 Appendix A18: Sample Distribution Comparison Rochester Area C.F. 293 Appendix A19: Sample Distribution Comparison Winston-Salem Foundation 294 Appendix A20: Sample Distribution Comparison C.F. of Greater Greensboro 295 Appendix A21: Sample Distribution Comparison Cleveland Foundation 296 Appendix A22: Sample Distribution Comparison Greater Cincinnati Foundation 297 Appendix A23: Sample Distribution Comparison Northwest Area Foundation 298 Appendix A24: Sample Distribution Comparison York Foundation 299 Appendix A25: Sample Distribution Comparison Greater Houston C.F. 300 Appendix A26: Sample Distribution Comparison Northwest Area Foundation 301 Appendix A27: Sample Distribution Comparison Greater Kanawha Valley Foundation 302 About the Author End Page

PAGE 8

v List of Tables Table 2.1 Hypotheses 79 Table 3.1 Behavioral Risk Factor Su rveillance System Reliability & Validity 92 Table 3.2 Social Capital Community Benchmark Survey Sample 96 Table 3.3 Variable Sources & Definitions 97 Table 3.4 Social Capital Communi ty Benchmark Survey Social Trust Index 100 Table 3.5 Social Capital Community Benchmark Survey Informal Social Engagement Index 101 Table 3.6 Social Capital Community Benchmark Survey Formal Social Participation/Organi zational Activism Index 102 Table 3.7 Social Capital Community Benchmark Survey Mutual Aid Index 104 Table 3.8 Behavioral Risk Factor Surveillance System Health Behavior Items 105 Table 4.1 Pooled Sample So ciodemographic Factors 115 Table 4.2 Frequencies of So ciodemographic Factors 118 Table 4.3 Pooled Sample Health Behavior and Outcome Factors 120 Table 4.4 Comparison of Health Behavior Frequencies 121 Table 4.5 Comparison of Hypertension Outcome Frequency 123 Table 4.6 Comparison of Descriptive Statistics of General Health Status 124

PAGE 9

vi Table 4.7 Comparison of Descriptive Statistics of Days of Mental Distress per Month 125 Table 4.8 Comparison of Community Ranking on Outcome Variables 128 Table 4.9 Comparison of Social C apital Descriptive Statistics 129 Table 4.10 Community Ranking on Social Capital Variables 131 Table 4.11 Comparison of Soci al Structural Statistics 133 Table 4.12 Select Sociodemogr aphic by Sociodemographic Frequencies 138 Table 4.13 Sociodemographic Fa ctors by Health Behavior Frequencies 141 Table 4.14 Health Behavior by Health Behavior Frequencies 142 Table 4.15 Sociodemographic Fact ors by Hypertension Frequencies 143 Table 4.16 Health Behaviors by Hypertension Frequencies 143 Table 4.17 Sociodemog raphic Factors by Gener al Health Status Frequencies 146 Table 4.18 Health Behavior by Gener al Health Status Frequencies 146 Table 4.19 Sociodemographic Fa ctors by Mental Distress Frequencies 147 Table 4.20 Health Behavior by Mental Distress Frequencies 147 Table 4.21 CVD by General Heal th Status and Mental Distress Frequencies 148 Table 4.22 General Health Status by Mental Distress Frequencies 148 Table 4.23 Summary of Significance Outcomes 150 Table 4.24 Summary of Significance Predictors 150 Table 4.25 Outcome Bivariate Associations: Hypertension 153

PAGE 10

vii Table 4.26 Outcome Bivariate Associ ations: General Health Status 155 Table 4.27 Outcome Bivariate Associations: Mental Distress 157 Table 4.28 Social Capital Bivari ates with Health Behavior 161 Table 4.29 Social Structural Biva riates with Health Behavior 161 Table 4.30 Social Capital Bivariates with Social Structural Factors 163 Table 4.31 Community Social Stru ctural Influences on Individual Hypertension 168 Table 4.32 Community Social Stru ctural Influences on Individual General Health Status 171 Table 4.33 Community Social Stru ctural Influences on Individual Mental Distress 175 Table 4.34 Community Income In equality Influences on Individual Health Behavior 176 Table 4.35 Community Poverty In fluences on Individual Health Behavior 177 Table 4.36 Community Social C apital Influences on Individual Hypertension 179 Table 4.37 Community Social C apital Influences on Individual General Health Status 184 Table 4.38 Community Social C apital Influences on Individual Mental Distress 188 Table 4.39 Associations Between So cial Structural Factors and Social Capital Indicators 190 Table 4.40 Influence of Risk Beha vior on Self-Reported Health 193 Table 4.41 Community Social C apital Influence on Individual Physical Activity 194

PAGE 11

viii Table 4.42 Community Social Capi tal Influence on Individual Body Mass Index 196 Table 4.43 Community Social C apital Influence on Individual Smoking 198 Table 4.44 Community Organizational Activism Influence on General Health Status, With and Without Mediators Added 201 Table 4.45 Community Social Tr ust Influence on General Health Status, With and Wit hout Mediators Added 202 Table 4.46 Summary of Results 204

PAGE 12

ix List of Figures Figure 2.1 Conceptual Model 25 Figure 2.2 Strength of Evidence 78 Figure 4.1 Gini Coefficient Distribution 134 Figure 4.2 Poverty Below 200% FPL Distribution 135

PAGE 13

x The Role of Social Structural and Soci al Contextual Factors in Shaping Chronic Disease and Chronic Disease Risk Behavior: A Multilevel Study of Hypertension, General Health Status, and Mental Distress Caroline Mae McKay ABSTRACT At present there is a reliance on behavioral interventions that have been limited in their effectiveness to redu ce the public health burden of chronic disease, partly because the effects of social context on the initiation and maintenance of health behaviors is not incorporated into public health policy and practice. Yet current research indicates that there are macrolevel structural and contextual influences on population health that cannot be reduced to individual or compositional effects. This study inve stigated the associations between social structural factors, community social context, individual characteristics, and selfreported correlates of disease. Distal in fluences included social structural inequalities such as income inequality and absolute deprivation or poverty. Pertinent mechanisms through which t hese influences might have operated on

PAGE 14

xi disease included social contextual factors, su ch as social capital. Both political economy and the ecosocial perspective were selected to inform this study and to provide the theoretical framework from which hypotheses were derived. The design was a multilevel, retros pective, nonexperimental study using secondary data. The study linked th ree data sources (2001 Behavioral Risk Factor Surveillance System, Social Ca pital Community Benchmark Study, and U.S. Census) by Federal Information Pr ocessing Standards codes in order for individuals to be placed in their community or state contexts. Results provided mixed evidence of the direct role of structural and cont extual inequalities on selfrated health. Any direct effects of soci al structural inequalities on the health outcomes disappeared once individual fa ctors were included in the models. Findings demonstrated that one dimension of social capital, organizational activism, retained its significant direct effect on general health status, once individual characteristics were consi dered. Conclusions suggested indirect associations whereby the negative influence of social structural inequalities on health was mediated by the erosion of soci al trust, which in turn was associated with engaging in risk behavior, thus in creasing the odds of reporting hypertension, fair/poor general health, and mental distress. Although results were inconsistent, this study contribut ed to advancing Healthy People 2010 goals of increasing quality of life and reduci ng health disparities by advancing understanding of the mult ilevel nature of perceiv ed health and the chronic diseases they predict.

PAGE 15

1 CHAPTER 1: BACKGROUND AND SIGNIFICANCE Introduction Cardiovascular diseases (CVD) are the leading cause of death for most population groups in the Unit ed States. CVD comprise a cluster of diseases, including coronary heart disease and stroke which together explain over 40% of all deaths annually; almost one million individuals die fr om CVD each year, with greater than half among women (Americ an Heart Association, 2001). The burden that CVD place on the population is not just in terms of loss of life. For 2005, the costs of CVD are estimat ed to be $393 billion (American Heart Association, 2004). Approximately one-quarte r of the population lives with CVD. Although the human and economic costs of CVD are widespread, the diseases are not equally distributed throughout the population. There are welldocumented disparities related to age, gender, race/ethnicity, SES and geographical location. The preval ence for the entire population was 354.1/100,000 in 1999, and subgroup-specific rates illustrate the disproportionate burden of CVD (American Heart Associati on, 2001). For exam ple, rates among white males are 411.5 as compared to 526.0 among black males and 402.1 among black females, with the smallest rates for white women at 295.0 (American Heart Association, 2001).

PAGE 16

2 Approaches to the prev ention of CVD address the health problem at the individual level. The purpose of these st rategies is to focus resources on those who suffer or are most likely to suffer from CVD – that is, persons who have a history of coronary heart disease, hypert ension, smoking, inactivity, obesity or elevated cholesterol. Generally, indi viduals at greatest ri sk are targeted with resources to reduce the rates of diseas e in the small porti on of the population who currently suffer the greatest burden of disease the tail of the distribution (Rose, 1992). For example, the current National Heart, Lung, and Blood Institute (NHLBI) strategic plan is a high-risk prevent ion strategy in which the focus is on individual-level risk factors (behavioral and genetic) and a suggested intervention is medical management through better phar maceuticals (National Heart, Lung, and Blood Institute, 2002). T he primary limitation of this approach is that it does not reduce the overall burden of disease. This position is consistent with a biomedical perspective: it views risk in isol ation from other factors; it focuses on behavioral or genetic contributors; it rel egates responsibility for disease (here – CVD) squarely on the shoulders of the individual; and by labeling and focusing on volition to the exclusion of more fundament al factors, it implicitly reflects a “blame the victi m” perspective. Another implication of the biomedical approach is that it perpetuates a false Cartesian dualism, whereby the physical dimension of disease and the mental aspect are envisioned as represent ing two separate and distinct systems, only marginally related. Prevention of disease, therefore, is approached as

PAGE 17

3 prevention of either physi cal or mental symptomot ology and morbidity, not reflecting one individual’s complete experience of pathology or illness. This bifurcated view has influenced primary, secondary, and tertiary prevention efforts. The preponderance of current prevention as well as treatment approaches are focused on the intraand/or inter-indivi dual factors, to the exclusion of contextual or structural cons iderations as fundamental causes. This perspective has served not to reduce the bur den of disease, but rather to worsen it. For example, mentally and physically healthy days have decreased from 53% in 1997 to 48% in 2001, with those reporting greater than half of each month unhealthy increasing from 15% in 1997 to 18% in 2001 (Zahran et al., 2005). The burden is economic as well, with co sts of poor mental health reaching $150 billion annually (Williams Chapman, & Lando, 2005). In addition, rates reflect widening disparities in perceived poor health status by ethnicity, gender, socioeconomic status, and geographical location (J ia, Muennig, Lubetkin, & Gold, 2004; Sehili, Elbasha, Moriarty, & Zack, 2005; Zahran et al., 2005). The increase in prevalence of reporting poor general and mental health, along with the chronic diseases they often precede, reflects the ineffectiveness of these risk-factor prevention strategi es that focus on proximal causes of disease, rather than directing attention to the persistent inequalities in resources that are the result of broader determinants of health (Link & Phelan, 1995; Link & Phelan, 1996).

PAGE 18

4 Purpose of Study and Study Significance The purpose of this study was to invest igate the associations between social structural factors (e.g., income inequality) community social co ntext (e.g., social capital), and individual characteristics (e.g., risk behavioral factors) and selfreported correlates of disease (hypertens ion, general health status, and mental distress). This study ex amined the extent to whic h upstream structural and contextual factors indirectly affect disease through their influence on risk behavior and the degree to which social structure and context independently influence self-reported disease. The possibility that behavior only partially mediates relationship between context and disease has profound implicatio ns. Currently there is a reliance on behavioral interventions that have been limit ed in their effectiveness to reduce the public health burden of chronic diseas e, partly because the effects of social context on the initiation and maintenance of health behaviors is not incorporated into public health policy and practice. By adopting a society-and-health lens (Walsh, Sorensen, & Leonard, 1995), this study had the potential to extend our understanding of the multile vel nature of health disparities and the need for multilevel interventions to reduce them.

PAGE 19

5 Study Rationale Rose (1992) suggests that populationlevel prevention strategies that acknowledge the social determinants of health are much more effective in eradicating disease than high-risk, individually-based, acontextualized approaches. However, prim ary prevention is not just a question of selecting a level of analysis to target resources. Population-based strategies may be limited if they rely on high-risk population behavioral change, to the exclusion of contextual influence. There are macro-le vel structural and c ontextual influences on population health that cannot be reduced to individual or compositional effects – in essence, that which places people “at risk of risks” (Link & Phelan, 1995). Overall, macro-level features of so ciety are posited to shape health through meso, or intermediary, factors which t hen differentially expose individuals to physical, social, and psychological contexts in which health promoting or health damaging behavior occurs (Berkman & Glass, 2000). Distal influences include social structural inequalities, such as income inequality, discrimination, and absolute deprivation or pov erty. Pertinent mechanisms through which these influences may operate on disease include so cial contextual factors, such as social capital. The relative effect s of these inequalities on risk behaviors and rates of self-reported health are not know n. The extent to which these broader

PAGE 20

6 social determinants shape disparate rates of multiple indicators of health and well-being, such as hypertension, general health status, or mental distress, is also unclear. As self-rated health is a st rong predictor of future morbidity and mortality, it is critical to consider per ceived health and rates of chronic disease as reflecting assessments of both physical and mental states of well-being (Williams et al., 2005). When studying the social determinants of health, examining not just the differences between individuals within a popu lation, but also differences between populations themselves, is critical. In essence, this view posits that health disparities reflect differential exposu re and differential resources shaped by society at multiple levels. There is no single theory that encompasses all the underlying assumptions in this area of study. As no one theory satisfactorily explains the relationships that will be inve stigated, a hybrid of two approaches is selected to provide the theoretical framew ork from which hypotheses are derived. Both the political economy and the ecos ocial perspective have informed this study. Selected Explanations for Inequalities Disparities in health may result from many mechanisms. Socially related origins of these differences include social structural and contextual influences that differentially affect certain gr oups. There are a variety of current

PAGE 21

7 justifications for the dis parate burden of disease. Theoretical and empirical evidence indicates that ther e is a range of explanations for social inequalities in self-reported health. Three of the most prominent rationales inform this study – those that attribute inequalit ies to material conditions, psychosocial factors, and health-related behaviors that confer biol ogical risk (Marmot, Bobak, & Smith, 1995). Material conditions refer to both relative and absolute deprivation. Income inequality can be consider ed a fundamental cause of the disparate burden of disease in that it reflects the economic and political institut ions that generate and perpetuate inequality (Krieger 2001). In essence, the effects of income inequality on health may be viewed as t he physical consequences of structural power differentials and the resultant unequal distribution of resources (Doyal, 1995), as health and disease are socially produced (Turshen, 1989). Studies have shown that there are both direct and indirect effects of income inequality on health (Wilkinson, 1992; Kaplan, Pa muk, Lynch, Cohen, & Balfour, 1996; Kawachi, Kennedy, Lochner, & Prothrow-S tith, 1997; Kennedy, Kawachi, Glass, & Prothrow-Stith, 1998; Lynch et al., 1998) An established pattern in public health evidence is the association between area-level poverty and health. In r egards to material deprivation, one of the most consistent associations is that between health outcomes and growing up and/or living in poverty (Lynch & K aplan, 2000). Although evidence supports a gradient effect (Adler et al., 1994), t here is little debate that those living under

PAGE 22

8 an absolute level of poverty have poorer health in general. The explanation for this disparity has both material and psyc hosocial aspects. Certain groups are usually dominant in the allocation of scarce resources and this structured inequality has a major impact on the heal th of less powerful groups (Doyal, 1995). Findings show that some effects of deprivation are co ntextual in nature above and beyond composition of the area (Macintyre, Maciver, & Sooman, 1993; Jones & Duncan, 1995). Poor health results from subordi nation through social, political, and economic space to which some populations are relegated (Doyal, 1995). Evidence of this form of di scrimination is seen in the health effects of residential segregation. Studies have f ound that residential segre gation influences health through multiple pathways, includi ng concentrated deprivation and the physical/social quality of the community (LaVeist, 1993; Williams, 1997). There have been few studies that have examined the role of physical and social isolation of a group with specific CVD-re lated health outcomes. However, there is substantial evidence t hat residential segregation neg atively impacts mortality rates for some groups (Acevedo-Garc ia, Lochner, Osypuk, & Subramanian, 2003). There are collective community characte ristics that may partly or fully mediate the relationship between social structural variables and risk behaviors and disease partly through their effects on psychosocial processes. Studies indicate that one mechanism through which social structural variables affect

PAGE 23

9 health is social capital (Kaw achi et al., 1997; Kawachi & Kennedy, 1999). Social capital refers to features of social organization, such as participation in associations and civic engagement, interpers onal trust, and norms of reciprocity, which act as resources and facilitate collective action (Putnam, 2000). Another potential influence of social context is racial trust. One health im pact of the social experience of racism ma y be eroded trust in other s. The experience of institutionalized discrimi nation may erode one’s sense of connection and place in the community. Health effects of this may be engaging in risk behaviors related to stress, such as smoking and overeating. Ultimately, the social patterning of heal th disparities reflects in part the social patterning of health behaviors as we ll as the social patterning of emotions (Emmons, 2000; Kubzansky & Kawachi, 200 0). Evidence suggests that risk behaviors cluster (e.g., those who smoke often drink, those who have healthy dietary practices also tend to exercise ). One can see how contextual forces differentially place certain groups “at ri sks of risks” (Link & Phelan, 1995). For example, there is evidence that lower socioeconomic status is associated with negative emotions and distress (Kubzan sky & Kawachi, 2000) and there are substantial findings that those who ar e distressed tend to engage in risk behavior (e.g., smoke, overeat, do not exercise) that negativ ely impacts heart heath. Because of the complex contribution of social structure and context to the ecology of most health outcomes, includi ng CVD, there is some evidence that programs do not completely eliminate behavioral risk when it is focused upon in

PAGE 24

10 isolation; there is a commensurate need to concentrate on contextual effects in addition to compositional effects. Limitations of Existi ng Knowledge Base At this time, there are three major gaps in the existing knowledge related to the role of the social environment in health. One major limitation in current empirical data is a lack of studies on the influence of social structure on CVDspecific outcomes. Most of what we know comes from studies on morbidity/mortality rates, life expectan cy, and general health. Another deficiency is insufficient research on hypertension, general health status, or mental distress outcomes using data that is expressly coll ected to study the effects of social capital. Finally, there is a dearth of k nowledge on the structur al and contextual influences on risk behavior. Preliminary Hypotheses 1. Behavior only partially mediates asso ciations between social structure and hypertension, general health st atus, or mental distress. 2. Social context partially mediates associations between social structure and hypertension, general health status, or mental distress.

PAGE 25

11 3. Behavior only partially mediates a ssociations between social context and hypertension, general health st atus, or mental distress. Overview of Design The design is a multilevel, retrospe ctive, nonexperimental study using secondary data. Until recently there has been a reliance on ecological studies in the study of macro-level social determinant s of health. Primar ily, cross-sectional ecological designs have been used. Multile vel modeling is utilized in this study for two reasons. First, a multilevel desig n is selected because of the nature of questions that will be investigated and the dat a that will be utilized. Different data sources representing differing levels of anal ysis are critical to this study as no one data set has specific multilevel data re lated to any outcomes in this study. This design is chosen also because it has multiple benefits over ecological approaches, including limits problems related to fallacies (ecologic, atomistic) and allows for the unique variance of contex tual and compositional levels (e.g., to test whether income inequality and social capital effects on hypertension are significant while adjusting for individual-le vel factors, such as SES and individual health behaviors) (Subramanian, Jones, & Duncan, 2003). In addition to having unique methodological features, this study advances our substantive understanding of inequalities and health. Few studies exist that explicitly address the effect s of social capital and rela ted variables on specific

PAGE 26

12 health outcomes. At present, there are no st udies that link individual-level health outcomes to a rich source of contextual data. The study links the BRFSS with the Social Capital Benchmark Study by F ederal Information Processing Standards (FIPS) codes that are present in both – in this way, individuals can be placed in their respective community or state contexts. Data Sources The study employs three data source s: the 2001 Behavioral Risk Factor Surveillance System (BRFSS), the Soci al Capital Benchmark Study (SCCBS), and the U.S. Census. The BRFSS is an annual telephone survey administered by the CDC. The study’s purpose is to collect information on the lifestyle and health behaviors of adults in the U.S t hat can be used to inform prevention policy and public health practice. Other m easures include: sociodemographic variables, risk behaviors (BMI/diet, physi cal activity, smoking), and self-reported outcome variables (hypertension, general health status, mental distress). Data on social contextual factors (i .e., social capital) have been obtained from the Social Capital Community B enchmark Study (SCCBS) The SCCBS is the first nationwide effort to measure soci al capital and its correlates (i.e., social trust and civic participation). The SCCBS was conducted in 2000 using both a nationwide and community-specific sample of adults. Variables of interest are social trust, formal and informal soci al engagement, and mutual aid, each of

PAGE 27

13 which is measured through a structured survey administered by telephone via random digit dialing. Only 27 out of the 40 communities in the original study are used. Thirteen are omitted because t he geography of the community could not be matched with Census data, did not have FIPS codes assigned, and therefore could not be linked to the BRFSS data or did not have BRFSS data for year 2001. Geographical units that are ut ilized include both counties and lightly populated states. The number of subjects in each community range from 89 to 4068. Finally, data also are extracted fr om the 2000 Census. Specifically, measures of social structur al inequalities (i.e., absolut e and relative deprivation) are employed. Income inequality is calcul ated using the Gini Coefficient (Rogers, 1979; Kennedy, Kawachi, & Prothrow -Stith, 1996), which estimates the proportion of income above the mean t hat needs to be redistributed to approximate an equal distri bution of incomes (Kawachi & Kennedy, 1997a). Area measures of poverty include the perc ent of families living at or below 200% of the Federal Povert y Level (FPL). Implications for Public Health Future implications of this study for public health include contributions to both research and practice. In a more general sense, evidence may demonstrate that interventions focusing on disparities in multiple health

PAGE 28

14 outcomes should simultaneously address soci al and behavioral factors to inform service delivery and health policy. In regar ds to its influence on future studies, evidence from this work may promote t he needed restructuring of large national surveillance systems to include contextual data. By broadening influential public health surveillance systems, the knowl edge base from which interventions are developed and conducted for diverse popul ations regarding common behavioral risk factors for chronic disease (e.g., smoki ng, physical activity, dietary practices, substance use) will be expanded. This study will provide a significa nt contribution to understanding the relationships between social structural contextual, and behavioral aspects of self-reported health. If the social cont ext within which behavior occurs is not considered, interventions targeting behavior change as a prevention strategy will have limited effectiveness. For example, future efforts would not be exclusively expended on changing proximal factors (i.e., individual health behaviors such as inactivity or smoking), but rather attention would be gi ven to implementing social and structural changes. To reduce the disparate burden of CVD, intervention targets would include: instituting regul atory changes in polit ical and economic policy which currently shape market influences which produce and perpetuate social inequalities (Kaplan & Lynch, 1999; Terris, 1999); strengthening social capital within communities (Kawachi 1999) or perhaps directing prevention efforts towards developing community capacity (Elliott, Taylor, Cameron, & Schabas, 1998); developing models that are aimed at shaping local public

PAGE 29

15 agendas to include community-level CVD pr evention (Schmid, Pratt, & Howze, 1995; Finnegan, Viswanath, & Hertog, 1999). In addition, findings from this study may inform policies focusing on improving individual and community-level general and mental health. Policy initiatives shaped by these results would target multiple levels of the social world, in order to reduce the disparate burden of poor mental and physical health status and its impact on perpetuating disparities in chronic disease. This study advances practice by informing surveill ance and interventions focusing on the interrelated physical and mental dimensio ns of chronic disease prevention and health promotion. By advancing know ledge regarding perceived health and the chronic diseases they predict and by adding to the growing evidence base indicating health disparities reflect diffe rential exposure and resources shaped by society, this study contributes to advancing Healthy People 2010 goals of increasing quality of life and reduc ing health disparities. Delimitations The following are delimitations of th is study imposed by the researcher: 1. The study is limited to counties and lightly popula ted states – no locations are included where FIPS codes could not be assigned (e.g. cities, suburban or rural areas not identified by their county-specific locations).

PAGE 30

16 2. The study includes only those count ies that are represented in the 2001 BRFSS. 3. The study is limited to individua ls from the 2001 BRFSS sample who reside in communities represented in the SCCBS. 4. The operationalization of CVD is restricted to self-reports of one critical form of cardiovascular disease, hypertension. 5. The operationalization of general health status and mental health are constrained to self-report measures. 6. Health risk includes a small selection of discrete behaviors and is restricted to poor physical activity, being overweight or obese, and engaging in smoking behavior. 7. The operationalization of health ri sk behavior is restricted to self-report measures. Limitations 1. Individuals residing in communi ties in 1999 may be different that individuals residing in those communities in 2001. 2. The transience of residents between 1999 and 2001 may influence the social structure and social context of the community. 3. The investigation is restricted to the questionnaires, items, and survey techniques utilized in the BRFS S, SCCBS, and the Census.

PAGE 31

17 4. The three levels of sampling used may result in somewhat different comparable sociodemographic characterist ics of each community sample. 5. All prominent CVD risk behavior not included. 6. Although referred to as mental di stress, no specific psychopathology is indicated. Definitions Absolute Deprivation – an area characteristic indi cating the quantitative level of poverty; area-level SES. Cardiovascular Disease (CVD) – a cluster of diseases, including coronary heart disease, hypertension, myocardial infarcti on, stroke, peripheral arterial disease, aortic aneurysm, and deep vein thrombosis. Ecological – collective or group-level variables. Ecosocial Perspective – an inherently multilevel t heoretical framework in which the social production of disease view and biological and ecological perspectives are integrated; core constructs include embodiment and pathways of embodiment (Krieger, 1994, 2001). Federal Information Processing Standards (FIPS) codes – federally designated unique numbers assigned to each county in each state within the United States. Fundamental causes – persistent pathogenic social conditions that place individuals “at risk of risks,” are linked to various diseases through multiple

PAGE 32

18 mechanisms, and are responsible for main taining and perpetuating inequalities in health (Link & Phelan, 1995). General Health Status – a self-reported asse ssment of one’s perceived qualitative level of overall health, wit h a range from excellent to poor. Health-Related Behaviors – actions and activities of the individual that have health-associated consequences. Income Inequality – qualitative and quantitative de scription of the dispersion or distribution or range of income in a population. Informal Social Engagement a dimension of social capital referring to a collective level of participation in familiar or casual relationships (e.g., with neighbors, co-workers, friends). Macro-Level Factor – characteristics of the distal or broad social and/or physical environment. Material Conditions – physical features of the environment, such as housing, assets, available services. Mental Distress – a global estimate of perceiv ed mental status (e.g., feeling depressed, anxious or stressed); not associated with Mental Status Exam. Meso-Level Factor (aka Mezzo) – characteristi cs of the intermediary social and/or physical environment through which macro factors may influence individual health. Mutual Aid a dimension of social capital referring to a collective sense of shared or common assistance (e.g., volunt eering, charitable contributions).

PAGE 33

19 Obesity – excess body weight meeting U. S. Department of Health and Human Services (USDHHS) cutoffs (BMI > 30). Organizational Activism a dimension of social capital referring to a collective level of engagement in formal social relationships (e.g., participation in community groups, associati ons, or organizations). Overweight – excess body weight meeti ng USDHHS cutoffs (25 < BMI > 29.9). Psychosocial Factors or Processes – interand intrapersonal mechanisms through which the social environment “ gets under one’s skin” (Taylor, Repetti, & Seeman, 1997). Physical Activity – any non-work related activity that can be considered exercise. Political Economy – a theory that pos its that health and disease are socially produced; economic and politic al institutions and decisio ns that create, enforce, and perpetuate economic and social priv ilege and inequality are root or fundamental causes (Turshen, 1989; Doya l, 1995); the political economy (the national economy in interacti on with governmental policies) influences a nation’s health through the mechanisms of produc tion, distribution, and consumption (Brenner, 1995). Poverty – absolute standard of ar ea-level deprivation. Relative Deprivation – an area-level characteri stic indicating qualitative (comparative) differences in income within a population. Residential Segregation – a multidimensional c onstruct representing the differentiation and spatial distribution of tw o or more groups within a population of

PAGE 34

20 an area (Massey & Denton, 1988; Masse y, White, & Phua, 1996; AcevedoGarcia & Lochner, 2002). Risk Behavior – individual behaviors known to influence the pathogenesis of chronic disease, such as eating patte rns, physical inactivity, and smoking. Smoking – active tobacco smoking. Social Contextual Factors – a characteristic of the collective; feat ures of the social environment, such as neighborhood or community factors, which may link distal influences to indivi dual health outcomes. Social Capital – features of social organizati on, such as participation in associations and civic engagement, interpers onal trust, and norms of reciprocity, which act as resources and facilitate collective action (Putnam, 2000). Social Determinant – elements of the social envi ronment that influence (e.g., health). Social Patterning – the way in which factors are distributed and arranged by sociodemographic groups. Social Structural Factors – broad conditions of a soci ety, such as inequality, discrimination, and poverty that shape t he nature of intermediate and proximal factors. Social Trust – a dimension of social capital referring to a collective sense of faith or confidence in bonds with others (e.g ., with neighbors, co workers, police).

PAGE 35

21 CHAPTER 2: CONCEPTUAL FRAMEWOR K AND REVIEW OF LITERATURE Review of Literature Theoretical Framework The two orientations that form the broad theoretical fr amework of this study are the social structural and ecologic al perspectives. Specifically, political economy and the ecosocial approach are select ed. Each provides a unique view as well as forms a complementary perspective informing this study, which reflects the author’s epistemological stanc e of the multilevel nat ure of the social world. In general, sociological theories in heal th focus on the social rates of disease, which reflect a group or populati on-based perspective. Social structural perspectives consider the structural ba rriers that restrict people from living healthy lives. Political economy theor y examines the physical consequences of structural power differentials and the re sultant unequal distribu tion of resources (Doyal, 1995). Health and disease are so cially produced; disease/health is a function of the relative power of diffe rent groups (Turshen, 1989). Fundamental causes of the disparate burden of dis ease stem from economic and political institutions that generate and perpetuate social and econom ic inequality (Krieger, 1994, 2001). The political economy (the nat ional economy in interaction with governmental policies) infl uences a nation’s health th rough the mechanisms of

PAGE 36

22 production, distribution, and consumption (Brenner, 1995). The focus of disease shifts from the host or indivi dual to social classes defined in relation to production and the way production is organized; stress is largely a response to capitalist social relations (Turshen, 1989). Certai n groups are usually dominant in the allocation of scarce resources and this st ructured inequality has a major impact on the health of less powerful groups; poor health results from subordination through social, political, and economic space to which some populations are relegated (Doyal, 1995). The applicab ility and appropriatene ss of using this theory for this work is that it focuse s on the social production of disease, an implicit assumption in studies of social i nequalities in health. However, inherent weaknesses in using this type of perspecti ve are that it does not provide for the role of agency (e.g., the role of individual behavior) and that it may focus more on materialist explanations for social inequalities in health rather than the effects of relative deprivation. Some of the limitations of polit ical economy theory are reduced by including an ecological orientation, whic h does not restrict its view to broad features and is more inclusive of indivi dual-level factors. Specifically, the ecosocial approach (Krieger, 1994), although not a theory per se, is considered an ecological theory-in-development as it does have a coherent set of (complex) propositions linking the social production of disease with biology in a dynamic ecological analysis. The ecosocial approach is appropriate for this study in that it is explicitly multilevel in its focus on “current and c hanging patterns of [social

PAGE 37

23 inequalities in] health …in relation to each level of biologi cal, ecological and social organization (cell, organ, or ganism/individual, family, community, population, society, ecosystem)” (Kri eger, 2001). Also, the ecosocial approach incorporates the notion of embodiment ( how the social world gets under our skin to influence our biology) with the pathways of that process (similar to the political economy perspective). There are several strengths and weak nesses of this perspective. For example, its comprehensiveness is a limitation and an advantage – on one hand, one may ask what does this theoretical position not explain, but future efforts at improving the formalization wi ll increase its specificity. On the other hand, the comprehensiveness of the ecosocial appr oach is a plus, as it more adequately recognizes and seeks to explain the nested or multidimensional order of social reality, which due to methodological advan ces (e.g., multilevel modeling), can be rigorously tested. A considerable advantage of the approach is its applicability to a wide range of data from public health to education to environmental and political science. Currently, this theoret ical approach is being applied in social science and epidemiology and is in a per iod of reflection, revision, and refinement (i.e., scientific self-regulati on). Support for utilizing this view, specifically in regards to this study, includes that the perspective comprehensively explains the multileveled nature of the world. In addition, the approach incorporates social production of di sease with biological expressions of inequality (i.e., there is a place for individual agency in this perspective).

PAGE 38

24 Constraints of this view in clude that it is not a coher ent theory. Moreover, the ecosocial approach has been lim ited in its use in empirical studies, primarily due to lacking in precision and insufficient evidence of its predictive power. Although either the ec osocial perspective or political economy might individually be suitable for this study, t he most appropriate use is an integration of the two, which utilizes the strengths of each while reducing the limitations of both. These two theoretical frameworks di ffer in their emphasis on features of society and biology and how they in corporate the two (Krieger, 2001). Essentially, utilizing both fits best and is most consistent with the questions asked in this study. Literature Review This research is framed by the theoretically-driven conceptual model (Figure 2.1), which is based in ecosocial and political economy perspectives. In essence, this model proposes that disease and risk behaviors are not shaped solely by proximal individual-level factors. Rather, individual agency exists in a social context, which itself is driven by broader structural determinants of health. In following this model, the literature re view presents evidence linking social structural inequalities, such as income inequality and poverty, specifically to CVD, general health status, mental distre ss, risk behavior, and the social context

PAGE 39

25 Figure 2.1 Conceptual Model 2000 CENSUS 2001 BRFSS 2001 BRFSS 2000 SCCBS Social Structural Inequalities (income inequality, 200% FPL) Disease (self-reported hypertension/general health status/mental distress) Social Contextual Factors (social capital indicators e.g., social trust, informal social engagment, organizational activism/formal social participation, mutual aid) Health Behaviors (BMI/diet, physical activity, smoking)

PAGE 40

26 within which it occurs. Evidence is pres ented that social context is directly associated with disease as well as indi rectly through its shaping of behaviors (such as physical activity, obesity, and smoking), which have been causally linked to CVD. On the whole, very little of the cu rrent literature establishes causal relationships between the variables. Some of this may be due to the multilevel nature of the relationships and the met hodological challenges that imposes. Ecological studies demons trate the associations between broader factors and health, but only in the past few years have methodological advances enabled researchers to examine these relationshi ps with causality in mind. Multilevel modeling has permitted ecological and atomis tic fallacy to be less of a threat to validity and theref ore has allowed inferences to be made and alternative explanations to be ruled out in working with individual and cont extual levels of analysis simultaneously. Nevertheless, these advances have not allowed this literature to approximate caus ality, due to the dearth of longitudinal designs and studies of the effect of time. Temporal ordering of effects has not been systematically studied in most of the literature pertinent to this study. Timing has been considered only in regards to the relationship between health behavior and CVD and it is in that literature only that causalit y has been established. To remedy this limitation, a new stream of work encompassing lifesta ge and timing of transitions has begun. In sum, the association between macro factors and health is established and competing explanations for resu lts found have been ruled out due to

PAGE 41

27 methodological advances; however, few, if any, studies have looked at the effects of temporal ordering. Social Structural Factors and Health Relative Deprivation: Income Inequality There is a vast and growing literat ure on the influences and consequences of the positive association between in come inequality and heal th inequalities, although very little of it has been able to definitively establish temporal precedence of relative deprivation, as many designs have been cross-sectional in nature. Even so, most of the current literat ure implicates the role of social structural factors in creating envir onments that are pathogenic. Although not conclusive, there is substantial evi dence from both ecologi cal and multilevel studies of the relationship bet ween income inequality and health. Some studies have found mixed evi dence for the relative deprivation hypothesis that suggests it is the distribution of inco me across a population, not just the absolute level of deprivation, which accounts for individual health and health inequalities (Blakely, Atkinson, & O'Dea, 2003; Hou & Chen, 2003; Osler et al., 2003). However, many studies have addressed the issue of whether relative deprivation impacts health over and above absolute or individual deprivation by demonstrating associations even after accounting for area-level poverty as well as SES in their model s (Kennedy, Kawachi, Glass, & Prothrow-

PAGE 42

28 Stith, 1998). In doing so, these investigat ions have served to rule out possible alternative explanations. Consistent results demonstr ate that this social st ructural factor has both direct and indirect effect s for a host of health-related issues, such as CVD (Kennedy et al., 1996; Waitzman & Smith, 1998b; Diez-Roux, Link, & Northridge, 2000; Cooper, 2001; Mellor & Milyo, 2003), all-cause a nd premature mortality (Ben-Shlomo, White, & Marmot, 1996; Kapl an, Pamuk, et al., 1996; Kennedy et al., 1996; Kawachi & Kennedy, 1997a; Lynch et al., 1998; Cooper et al., 2001; Lochner, Pamuk, Makuc, Kennedy, & Ka wachi, 2001; Lobmayer & Wilkinson, 2002; McLaughlin & Stokes, 2002; Sanmarti n et al., 2003), self-rated health (Kennedy, Kawachi, Glass et al., 1998; Soobader & LeClere, 1999; Fiscella & Franks, 2000; Kahn, Wise, Kennedy, & Ka wachi, 2000; Subramanian, Kawachi, & Kennedy, 2001; Blakely, Lochner, & Kawa chi, 2002; Weich, Lewis, & Jenkins, 2002; Subramanian & Kawach i, 2003; Lopez, 2004), mental distress (Fiscella & Franks, 2000; Kahn et al., 2000; Weich, Le wis, & Jenkins, 2001; Shi, Starfield, Politzer, & Regan, 2002; Muramatsu, 2003), life expectancy (Wilkinson, 1992), STI/AIDS (Holtgrave & Crosby, 2003), crime (Kawachi, Kennedy, & Wilkinson, 1999), and teen birth rates (Gold, K ennedy, Connell, & Kawachi, 2002). Although both individual-level and aggregated-level outcomes have been examined, the majority of outcomes hav e been group-level due to the methods with which these studies were conducted. The reliance on findings from ecological designs has resulted in debates regarding both conceptual as well as empirical limitations in the literature.

PAGE 43

29 Current reviews of literature identif y several possible issues that may influence the detection of an associati on between income inequality and health, including concerns for lag effects, sca le, design, and confounding (Lynch et al., 2004; Subramanian & Kawachi, 2004) In regards to matters of scale, strength of associations between income inequality and various outcomes may be dependent upon the geographic level in which income inequality is assessed. Overall, in regards to U.S. studies, t here is stronger evidence of an association when income inequality is measured at the state level and more debate regarding the relationship at a smaller scale (i.e., counties, tracts, block groups) (Lynch et al., 2004; Subram anian & Kawachi, 2004). Issues related to design and the possible role of confounders are connected. There is a serious gap in t he literature related to confounding. Several individual and ecologic confounder s may influence the detection of the association between income inequality and health (Subramanian & Kawachi, 2004). Pertaining to design issues, as t he majority of studies are ecological, cross-sectional investigations, there hav e been concerns regarding the validity of interpreting associations between broader, structural factors as reflecting real influences for individual health. This i ssue has resulted in much discussion over whether the effects of income inequality are a statistical arti fact representing results that are ecologically fallacious by ascribing the effect s found to alternative sources of variance (Gravelle, 1998). Some have countered this argument wit h evidence that the relationship between income inequality and rates of mortality do not re flect a fundamental

PAGE 44

30 relationship between individual-level char acteristics and mortality and therefore are not explained by artifa ct (Wolfson, Kaplan, Lynch, Ross, & Backlund, 1999). In point of fact, the curr ent use of multilevel modelin g has added to the empirical evidence supporting the indep endent contextual effect s of income inequality on health by adjusting for cross-level confounding. In essence, the emergence of multile vel studies has served to expand the previous methodological restrictions in making conclusions regarding income inequality. Results from these works hav e allowed the contextual influence of income inequality, as com pared to the compositional in fluence of SES, to be firmly established as a critical considerat ion of social structural influences on health.1 Although findings have been consist ent from these types of studies, there is still a gap in the lit erature related to multile vel studies of specific diseases. One example is the dearth of mu ltilevel investigations of CVD. Only one study has investigated specific CVD risk and its relationship with income inequality in a multilevel study (DiezRoux, Link, & Northridge, 2000). Although there have been in vestigations into the association between income inequality and CVD-related mortalit y, there are far fewer studies of CVDrelated health behavior (K aplan et al, 1996). At this time, no studies have examined the effects of income inequa lity on both CVD risk behavior and rates of CVD simultaneously. Generally, most mu ltilevel studies of income inequality examine its association with mort ality or self-reported health. 1 Subramanian, Kawachi, & Kennedy (2001) provide the most accessibl e explanation of these terms: “contextual (the difference a place makes) and the compositional (what’s in a place)” p.10

PAGE 45

31 Although there are stud ies which have found no association between income inequality and heath status (Mellor & Milyo, 2002, 2003), there is by far more literature from multilevel studies supporting this relationship. Findings suggest that those individuals residing in areas with highest inequality are up to 30% more likely to state they have fair /poor health (Kennedy, Kawachi, Glass et al, 1998; Blakely, Lochner, & Kawachi, 2002) with a growing income inequality increasing those reporting fair/poor health by up to 39% (Subramanian & Kawachi, 2003). More recent research i ndicates that this association may be stronger, with evidence that a one point in crease in the Gini Coefficient (on a 100 percent scale) corresponds to a 4% (1.6% – 6.5%) increase in reporting fair or poor health (Lopez, 2004), with income ineq uality becoming a progressively more important influence as self-rated health deteriorates (Shi & Starfield, 2000; Lopez, 2004). Moreover, the influence may be differential based on gender (Kahn, Wise, Kennedy, & Kawachi, 2000) individual SES (Subramanian, Kawachi, & Kennedy, 2001; Weich, Le wis, & Jenkins, 2002), and geographic scale at which inequality is measured. Multilevel studies have demonstrated a strong association between self-rated heal th and income inequality at various levels of aggregation, including stat e (Kawachi, Kennedy, Lochner, & ProthrowStith, 1997; Kennedy, Kawachi, Glass et al., 1998; Subramanian et al., 2001), region (Weich et al., 2002), metropolit an area (Blakely, Lochner, & Kawachi, 2002; Lopez, 2004) and county (Soobader & LeClere, 1999). Income inequality has been shown to hav e a detrimental impact on other forms of self-reported health. However, t he body of literature as a whole is far

PAGE 46

32 smaller for mental health/ distress and demonstrates mo re mixed evidence than studies examining general health alone. For example, while some findings support a harmful influence of income i nequality on mental distress (Kahn et al, 2000; Weich, Lewis, & Jenkins, 2001; Mu ramatsu, 2003), others either find no such association, or the association disappears after other contextual or individual-level factors are added to the model (Weich, Twigg, Holt, Lewis, & Jones, 2003; Muntaner et al, 2004). Few studi es find no evidence of a significant independent association between income inequality and mental health outcomes (Sturm & Gresenz, 2002; Henderson, Liu, Diez Roux, Link, & Hasin, 2004). In spite of these results, slightly mo re studies report a positive relationship, with those living in higher income i nequality areas having up to a 70% excess risk of suffering with depressive sympt omotology (Fiscella & Franks, 2000; Kahn et al., 2000; Weich, Lewis, & Jenkins, 2001) Amongst studies which find a health damaging impact of living in hi gh income inequality region, there is evidence of an additional moderating influen ce of absolute poverty in shaping the strength and directionality of the relationship. Alt hough results are significant, they have substantively contrasting interp retations. For exam ple, some suggest that the negative influence of living in a state with high le vels of income inequality on mental distress is more profound if one has higher income (Weich et al., 2001), while other studies have indicated that it is lowe r incomes which confer a worse impact on mental distress (Kahn et al., 2000). In addition to the evidence base regarding the direct, independent association between income inequality and health outcomes, several studies

PAGE 47

33 have found the relationship to be mediated by social factors, such as the social context (Kawachi et al, 1997; Kennedy, Kawachi, Prothrow-Stith, Lochner, & Gupta, 1998; Fiscella & Fr anks, 2000; Gold et al., 2002; Veenstra, 2002) and neighborhood processes (Soobader & LeClere 1999) as well as access to health resources, such as primary care (Shi, Starfield, Politzer, & Regan, 2002). In addition, there is limited evidence of in teractional association between income inequality and minority concentration (Mc Laughlin & Stokes, 2002) as well as residential segregation (Cooper et al., 2001) However, substantial evidence finds that income inequality exerts significant independent effects on health (Kaplan, Pamuk, Lynch, Cohen, & Balfour 1996; Kennedy, Kawachi, & ProthrowStith, 1996; Soobader & LeClere, 1999; Go ld, Kawachi, Kennedy, Lynch, & Connell, 2001; Lochner, Pamuk, Makuc, Kennedy, & Kawachi, 2001). Absolute Deprivation: Poverty The association between area-level mate rial or absolute deprivation and morbidity and mortality is well established, with evidence of a growing gradient effect (Singh & Siahpush, 2002). As with relative deprivation, causality has not been established as few, if any, studi es have been able to suggest temporal ordering. Both qualitative (Cattell, 2001) and quantitative studies have found significant positive associations betw een area-level poverty and a variety of health outcomes, such as CVD (Diez-Roux et al., 1997; Davey Smith, Hart, Watt, Hole, & Hawthorne, 1998; LeClere, Rogers, & Peters, 1998; Cubbin, Hadden, &

PAGE 48

34 Winkleby, 2001; Huff & Gray, 2001; Davey Smith & Hart, 2002; Singh & Siahpush, 2002; Cohen, Farley, & Mason, 2003), mortality (Ben-Shlomo, White & Marmot, 1996; Waitzman & Smith, 1998a, 1998b; Yen & Kaplan, 1999a; Singh & Siahpush, 2002; Cohen et al., 2003), self -rated health or quality of life (Robert, 1998; Malmstrom, Sundquist, & Johansson, 1999; Cattell, 2001; Ross & Mirowsky, 2001; Steptoe & Feldman, 2001; Drukker & van Os, 2003), mental distress (Aneshensel & Sucoff, 1996; Yen & Kaplan, 1999b; Elliott, 2000; Schulz, Williams et al., 2000; Belle & Doucet, 2003; Leventhal & Brooks-Gunn, 2003; Ferrer & Palmer, 2004), teen birth rate (Gol d et al., 2002), residential instability and neighborhood violence (Sampson, Raudenbush, & Earls, 1997) and STI/AIDS (Holtgrave & Crosby, 2003). Although all these associational investigations provide empirical support fo r the contextual influence of poverty, there are two pressing issues current ly debated, both of which impact future studies of possible causality. The first one involves methodological approaches (i.e., use of multilevel methods). The se cond relates to conc eptual issues (e.g., differential influence of poverty as contex tual rather than re flecting individuallevel SES effects and the di fferential influence of pover ty and income inequality). A primary issue of both methodologica l and conceptual concern has been how and to what extent poverty at the contextual level is important above and beyond the effects of individual SES alone. Multilevel methods have allowed this issue to be investigated empirically. Overall, results demonstrate that the negative effects of deprivation are not a pr oxy for individual characteristics and, in fact, reflect a true contex tual feature of the social structural environment (Diex-

PAGE 49

35 Roux et al., 1997; Robert, 1998; Waitzman & Smith, 1998a; Ross & Mirowsky, 2001; Sundquist, Lindstrom, Malmstro m, Johansson, & Sundquist, 2004). In regards to study designs, alt hough many studies have not used multilevel designs explicitly, they hav e statistically adjusted for area-level measures in an attempt to predict i ndividual risk through proportional hazards models and therefore reduce the risk of alternative explanations. In doing so, these studies come closest to suggesting t hat there may be a causal relationship. Results from these studies are consist ent and indicate a contextual pathogenic effect of living in poverty on CVD (Davey Smith, Hart, Watt, Hole, & Hawthorne, 1998; Waitzman & Smith, 1998a, 1998b; Huff & Gray, 2001; Davey Smith & Hart, 2002). However, the adoption of multilevel st udies in this area of research has produced a growing literature to support t he already established finding that the socioeconomic environment one lives in has both direct and indirect effects on mortality. Specifically in regards to CVD risk, some have found an independent influence of absolute deprivat ion of an area (Davey Smit h et al., 1998; Robert, 1999; Cubbin, Hadden, & Winkleby, 2001; S undquist et al., 2004). Additionally, research specifically examin ing the area effects of pover ty provide evidence that risk of MI is partly explained by contex t and therefore only somewhat accounted for by composition of an area (Stjarne et al, 2002). However, it must be noted that as timing of effect has not been t horoughly investigated, causality has not been established. Findings suggest both independent and mix ed effects. For instance, a direct association is found for CHD (Diez-Ro ux et al., 1997) as well as mortality

PAGE 50

36 Waitzman & Smith, 1998a, 1998 b; Yen & Kaplan, 1999a, Bosma, van de Mheen, Borsboom, & Mackenbach, 2001). Indirect asso ciations exist as well, but there is a debate in the literature as to what ext ent individual factors, such as SES, moderate (Jones & Duncan, 1995; O'Cam po, Xue, Wang, & Caughy, 1997; Yen & Kaplan, 1999a) or mediate (Drukker & va n Os, 2003) the relationship. As of yet, there is insufficient evidence to conf irm either type of a ssociation over the other. Overall, however, research indica tes that there would be a 20% reduction in mortality rates in the U. S. if all groups had the same rates as those living in the highest SES areas, controlling for a host of individual factors (Winkleby & Cubbin, 2003). Evidence supports a direct associat ion between poverty and self-reported health status. International as well as national multilevel findings provide support for the adverse health effe cts of deprivation on self-ra ted health. There are both direct and indirect effects of living in a deprived area on self-rated health (Ross & Mirowsky, 2001), with some findings i ndicating a moderating relationship between contextual and indi vidual levels of deprivation (Ferrer & Palmer, 2004). Moreover, in regards to mediating e ffects, both qualitat ive and quantitative evidence suggests that the influence of pov erty on health status may be partially through its impact on social context (C attell, 2001; Ross & Mirowsky, 2001; Drukker & van Os, 2003). Overall, evidence of the negative impac t of living in poverty on health status is strong. Findings come from a variety of studies, with the majority of recent evidence from multileve l investigations. For exam ple, from a longitudinal

PAGE 51

37 cohort study, there is evi dence of a higher risk of reporting fair/poor health (OR 3.30, 95% CI 2.32 – 4.71) up to nine years later a ssociated with living in poverty (Yen & Kaplan, 1999a). In addition, result s indicate that living in absolute deprivation increases the odds by a range of 70% – 200% of reporting fair/poor health (Malmstrom, Sundquist, & Johanss on, 1999; Steptoe & Feldman, 2001). A significant positive associati on exists between residing in an impoverished area and general m ental distress (Schulz, Williams, et al., 2000; Leventhal & Brooks-Gunn, 2003) as well as specific disorders, such as depression and anxiety, in a variety of demographic subgroups. There is some evidence that women (Belle & Douc et, 2003) and adolescents (Aneshensel & Sucoff, 1996) appear to be at greater risk. Studies have found that poverty increases the risk of experiencing mental distress, with residing in deprivation conferring over two-times (OR 2.14, 95% CI 1.49-3.06) the risk of developing poor mental health nine years later (Y en & Kaplan, 1999b). Some suggest it is the exposure to stressful, disadvantaged conditions (Schulz et al., 2000; Steptoe & Feldman, 2001), restricted protective resources (Elliott, 2000) and neurochemical responses which result in excess morbidity for those who live in poverty (Pearlin, 1989; McEwen, 1998; Ferrer & Palmer, 2004). Although this literature is limited, it is developing a strong evidence base, as provided by both multilevel (Steptoe & Feldman, 2001) and RCT (Leventhal & Brooks-Gunn, 2003) studies. The other conceptual issue considers the differential influence of poverty compared to the impact of other structural factors. On e study found that poverty

PAGE 52

38 acts as a mediator through which re sidential segregation influences health (LeClere, Rogers, & Peters, 1998). O ne review suggests that there are independent effects of comm unity deprivation on other st ructural or contextual factors such as the social, service, and physical environment (Robert, 1999). In regards to income inequality, some studi es that have explicitly examined the combined effects of absolute and rela tive income on mortality (Ben-Shlomo, White, & Marmot, 1996; Kaplan, Pamuk, Lynch, Cohen, & Balfour, 1996; Kennedy, Kawachi, & Prothrow-Stith, 1996) Nevertheless, few have studied the impact on specific health outcomes (Gol d et al., 2002; Holtgrave & Crosby, 2003). At this time, simultaneous examinati ons of the relative effects of poverty and income inequality on the social cont ext of risk behavior and CVD, general health status, or mental distress hav e not been conducted. This gap in the literature results in com peting explanations between co ntextual influences, thus inhibiting causality to be c onfirmed or denied. Social Structural Factors and Health Behaviors Currently, the evidence base to support direct associations between various structural factors and health behaviors is sparse, with no studies explicitly examining the temporal relationships between macro factors and individual behavior. However, the small literature that exists indica tes that the relationship between social structural inequalities and disease may be partly mediated by health behavior. As contrasted with inco me inequality, the vast majority of

PAGE 53

39 studies have examined the role of poverty To date, no studi es have investigated both relative and absolute deprivation and health behavior, and therefore alternative explanations cannot be ruled out. This gap may be due, in part, to little evidence on the impact of income i nequality on behavior, as some studies have controlled for behavioral and so ciodemographic factors (Kennedy et al, 1998). However, recently, stronger evidence has accumulated. For example, in a multilevel study, Diez-Roux et al ( 2000) have found significant positive associations between income inequality and three prominent CVD risk factors BMI, hypertension, and sedentarism, adjus ting for individual SES. In addition, these authors found moderating effect s of gender and abs olute income levels. To specify, there was stronger ev idence of the relationships for all risk behaviors for women as well as a more robust positive association between income inequality and risk behavior for those living at lower income levels. Other studies have not examined a cluster of risk behavior, but rather only one or two behaviors at a time. Findings illustrate that increases in weight are positively associated with income inequalit y at the state level, espe cially for men, (OR 1.12, 95% CI 1.03 1.22) (Kahn, Tatham, Pa muk, & Heath, 1998) In addition, smoking and physical activity (Kaplan, Strawbridge, Cohen, & Hungerford, 1996) as well as sexual risk activity (Thomas & Thomas, 1999) are positively related to relative deprivation. Compared to the limited literatur e on relative deprivation and risk behavior, there are, on the whole, many more studies on the effects of poverty on

PAGE 54

40 health behavior. There is a large scientif ic base on the effects of individual SES on risk behavior, but very little on area-le vel economy or poverty and behavior. In general, there are a small number of studies of the association between poverty and singular risk behavior. Most l ook at a cluster of risk, unlike the income inequality literature, and virtually none explore caus al relationships. Both ecological and multilevel studies have found strong evidence supporting the negative impact of area-level poverty on behavior, while adjusting for individual sociodemographic factors. In controlli ng for extraneous factors, these findings support the structural influence of material deprivation above and beyond individual status on such CVD risk fact ors as physical activity, BMI/obesity, and smoking (Cubbin, Hadden, & Wi nkleby, 2001). Results indicate that not only is t here a significant negative association between poverty and physical activity, but also that there may be a doseresponse effect between area factors and activity (Parks, Housemann, & Brownson, 2003). Although this specific finding has not been replicated, the inverse relationship between absolute depriv ation and physical activity cannot be questioned. The majority of support has come from cross-sectional studies (Wister, 1996; Lantz et al., 1998; Ross, 2000b; Cubbin et al., 2001; Lee & Cubbin, 2002), although there have been multilevel investigations as well (Ross & Mirowsky, 2001). To date, the only mult ilevel studies examining the relationship of poverty on combined effects of physica l activity, BMI/obesity, smoking, or excessive alcohol use have been internat ional. Results from these studies demonstrate an increased risk of engaging in smoking, being sedentary, and

PAGE 55

41 obese for those who reside in impover ished areas (Sundquist, Malmstrom, & Johansson, 1999; Bosma, van de Mheen, Borsboom, & Mackenbach, 2001). The increased risk of obes ity for those who live in poverty has been wellestablished (Davey Smith, Hart, Watt, Hole, & Hawthorne, 1998; Cubbin et al., 2001). In most cross-sectional studies the associations found are both for independent (Lee & Cubbin, 200 2) as well as mediated effects (Lantz et al., 1998). There are two similarities bet ween the physical activity and obesity literature regarding the impact of poverty. Firstly, just as with physical activity, a dose-response relationship has been found between poverty and obesity, with the former measured by environ mental proxy. Specific ally, (Reidpath, Burns, Garrard, Mahoney, & Townsend, 2002) fi nd that those who reside in impoverished areas are exposed to 2.5 times the density of fast food outlets than those who live in more affluent areas. The second commonality is that the only multilevel studies of obesity as one of a cluster of risk fact ors are from other nations (Reijneveld, 1998; Sundquist, Malmstrom, & Johansson, 1999). By and large, the majority of studies examining the associations between poverty and risk behaviors pertain to the negative effects of smoking, where evidence has come from qualitative (Banc roft, Wiltshire, Parry, & Amos, 2003), cross-sectional (Krieger, 1992), and mult ilevel findings (Sundquist et al., 1999; Bosma et al., 2001). Risk related to area deprivation and smoking has been found in both the U.S. (Lantz et al., 1998; Cubbi n et al., 2001) and Europe (Davey Smith & Hart, 2002). On the whol e, results reveal a strong relationship between poverty and smoking, with living in disadvantaged areas conferring a

PAGE 56

42 50% to 110% increase in odds of smoki ng, controlling for individual SES (DiezRoux et al, 2003). An essential aspect of this literature is separating out area versus individual influences that hinder or facilitate smoking. In regards to this health-damaging behavior, evidence exis ts for the detrimental effects of neighborhood deprivation accounting for resi dential segregation, in dividual status (Kleinschmidt, Hills, & Elliott, 1995) as well t he moderating effect of race/ethnicity (Diez-Roux et al., 1997; Cubbin et al., 2001). Overall, as one can see from the abov e discussion, the role of social structural inequalities on CVD risk behavior is strong, even tho ugh the study of area-level absolute and relative income on risk behaviors is quite new. However, one caveat is in order in inte rpreting the findings as the majority of empirical evidence is from studies from other nations. Even so, one issue cannot be questioned – the distribution of income is reflected in the distribution of cardiovascular risk (Diez-Roux, 2003). Social Structural Factors and Social Context While there is support of a direct a ssociation between social structural inequalities and health, no studies exami ne a possible causal relationship. Nonetheless, evidence exists that some of the effect occurs through the social context. There have been several key pathw ays elucidated in current literature that make the conceptual and methodol ogical connection between social structure and social context in regards to pubic health. Generally, two schools of

PAGE 57

43 thought have emerged in the literature regarding the mechanisms through which income affects health. One group proposes that the pathway is essentially materialist in nature, primarily reflecting political changes, cla ss or economic relations, and material conditions or resources that often acco mpany poverty (Kaplan, Pamuk et al., 1996; Lynch et al., 1998; Lync h, Smith, Kaplan, & Hous e, 2000). The other view posits that although social structure’s relationship with health is related to material deprivation, these conditions are not sufficient; the association is also due to the deleterious effects on social context through relative deprivation (Wilkinson, 1992; Wilkinson, 1996) and so cial organizational processes (Marmot & Wilkinson, 2001). Fundament ally, the expansive effect of income inequality serves to corrode the social fabric of communities through such factors as increased violence, reduced civic parti cipation, and reduced productivity (Kawachi & Kennedy, 1997). It is this pers pective that informs this study. Generally, social structure is pos ited to affect heath through social contextual pathways such as disinvestm ents in forms of human and social capital as well as psychosocial mechanisms (Kawachi, 2000). Other mechanisms linking macro to meso ecological factors include physical and social characteristics of residence (MacIntyre & E llaway, 2000). The specific aspects of the social context that have been addre ssed in the relative and absolute deprivation literature are, by and large, social capital, collective efficacy, and the social environment overall.

PAGE 58

44 One potential pathway by which social structural variables, such as income inequality, shapes health is through its effects on social capital (Kawachi et al., 1997; Kawachi & Berkman, 2000; McCu lloch, 2003). To note, this pathway does not reflect the health benefits of indi vidual social networks. Rather social capital is a contextual construct, impac ting health through characteristics of the collective, such as levels of trust, reci procity, civic participation (Putnam, 2000). A seminal work by Kawachi et al (1997) has found that relative deprivation at the state-level may lead to disinvestments in m any forms of capital, including social, with associations as strong 0.46 to 0 76. Some studies demonstrate that the group experience of income inequality se rves to erode relational resources critical to health, such as mutual tr ust and civic participation (Daly, Duncan, Kaplan, & Lynch, 1998). Most studies have been ecological, cr oss-sectional investigations of associations, although there have been a few multilevel investi gations, primarily in other countries, which have adjusted for some alternative explanations. Currently, the majority of the public health literature identifies social capital as a mediator in the relationshi p between broader, st ructural factors and health. For example, there are cross-sectional studi es of the mediating effects of this variable on a variety of health outcomes, including mortality (Kawachi et al, 1997), violent crime (Kennedy Kawachi, Prothrow-Stith, Lochner, & Gupta, 1998), and teen birth rate (Gold et al., 2002). There have been far fewer studies on the associations between constructs related to social capital and macro fact ors affecting public health. To date,

PAGE 59

45 income inequality and social contex t have not been rigorously studied with respect to chronic disease. Just as with income inequality, one consistent observation in the social structural literat ure is that there is little evidence due to few empirical studies explicitly examin ing the associations between poverty and contextual factors such as collective effi cacy, social cohesion, social environment and chronic disease and risk behavior in adults, even though the association between residing in poverty and poor rates of social cohesion has been established (Coleman, 1988). There is more evidence on the direct effects of poverty on health as compared to income inequality effects (G old et al., 2002), where much of the literature has found indirect effects m ediated through social context. Although there is more evidence regarding pov erty’s association mediated through collective efficacy, social cohesion, or social environment, recently some support has been found specifically on the mediatin g effects of social capital on the relationship between absolute deprivati on and health (Cattell, 2001; Steptoe & Feldman, 2001; Gold et al., 2002; Holtgrav e & Crosby, 2003). There is, however, a substantial literature on indirect effe cts from both multile vel and ecological studies on poverty and health, mediated by collective efficacy (Sampson et al., 1997; Cohen, Farley, & Mason, 2003), social cohesion (Drukker et al, 2003), and social environment (Ross & Mirowsky, 2001). In all these studies, attempts have been made to control for competi ng explanations for findings. Overall, living in poverty result s in non-random exposure to pathogenic environments and restriction of salubrio us resources (Lynch & Kaplan, 2000).

PAGE 60

46 Consensus from multilevel studies has demonstrated a mediating effect whereby exposure to physically and socially deprived areas influence health above and beyond individual sociodemograp hic factors. Even as both income inequality and poverty reflect structural factor s, the relationship between each and the social context as a mediating force is quite different. In regards to absolute deprivation, the mechanism of influence may be more related to the experience of poverty in regards to social processes, such as stress and health behaviors (Elliott, 2000; Bosman et al., 2001; Pi ckett & Pearl, 2001; Ross & Mirowsky, 2001; Steptoe & Feldman, 2001) or material deprivation (Kaplan, Pamuk et al., 1996; Lynch et al., 1998). At this time, more ev idence of the relative effects of structural and contextual characteristics on disease and its processes is needed. Despite these advances in studying how the broader social structure shapes the social environmental context in which healt h or disease occurs, there is a gap in the literature investigating associati ons between these characteristics and specifically CVD (Diez Ro ux, 2003), especially in re lation to the temporal ordering of effects. Social Context and Health As most of what is known regarding social capital and its correlates comes predominantly from cross-sectional st udies, there is no evidence supporting causality as yet. There are, nonetheless, a number of investigations pertaining to the association of health and other context ual aspects of the social world, such

PAGE 61

47 as social cohesion, social stress and the social environment (Turner, Wheaton, & Lloyd, 1995; Schulz, Israel et al., 2000; Ellaway, Macintyre, & Kearns, 2001; Steptoe & Feldman, 2001; Drukker & v an Os, 2003). Although the number of studies investigating the health effects of social capital is relatively small, evidence is mounting of the impact of this aspect of the social environment. Consistent with its contextual nature, regional differences in indicators of social capital have been found (Cattell, 2001; Stept oe & Feldman, 2001). Associational relationships have been observed, with several studies controlling for extraneous factors. For example, so cial capital has been associated with a wide variety of outcomes in public health, including CVD mo rtality (Kawachi et al., 1997; Franzini & Spears, 2003; Lochner, Kawachi, Brennan & Buka, 2003), overall mortality (Kawachi et al., 1997; Franzini & Spear s, 2003; Lochner et al., 2003; Skrabski, Kopp, & Kawachi, 2003), se lf-rated health (Macintyre, Maciver, & Sooman, 1993; Kawachi, Kennedy, & Glass, 1999; Blakely, Kennedy, & Kawachi, 2001; Cattell, 2001; Subramanian et al., 2001; Subramani an, Kim, & Kawachi, 2002; Greiner, Li, Kawachi, Hunt, & Ahluwalia, 2004), m ental distress (McCulloch, 2001; Mitchell & LaGory, 2002; Cam pbell, Cornish, & McLean, 2004; Greiner et al., 2004; Ziersch, Baum, Macdougall, & Putland, 2005), violence (Kennedy, Kawachi, Prothrow-Stith et al., 1998; Hemenw ay, Kennedy, Kawachi, & Putnam, 2001; Galea, Karpati, & Kennedy, 2002), STI/AI DS (Holtgrave & Crosby, 2003), quality of life (Raphael et al., 2001), and teen birth rates (Gold et al., 2002). Investigations have shown that variation in area social capital do not solely reflect

PAGE 62

48 differences in individual factors, but rat her support social capital as a contextual, rather than individual-level, cons truct (Franzini & Spears, 2003). Currently, there are few studies t hat have examined the relationship between social capital and cardiovascular disease (CVD). Seminal studies of social capital have been cross-sectional an d ecological in natur e investigating the mediating influence of social capital on all-cause and coronary heart disease (CHD) mortality at the statelevel (Kawachi et al., 1997) Indicators of social capital, such as group membership and so cial trust, have independent contextual associations with CVD mortality, adjusting for other contextual factors, such as income inequality and poverty or materi al deprivation (Kawachi et al., 1997; Lochner, Kawachi, Brennan, & Buka, 2003). Recently, multilevel studies have provided evidence of the unique contextual influence of social capita l on CVD mortality while controlling for extraneous factors at differing levels of analysis. Lochner, Kawachi, Brennan, & Buka (2003) replicated earlier findings of the significant a ssociation between social capital and CVD mort ality, and extended our underst anding of indicators of social capital by measuring aspects of so cial cohesion at sma ller, substantively meaningful levels of analysis (i.e., nei ghborhoods). Additionally, research indicates that the association between soci al capital and CVD mortality occur at multiple levels of the social envir onment, such as neighborhood and county levels, after the effects of individual characteristics are taken into account (Franzini & Spears, 2003).

PAGE 63

49 In regards to the health effects of i ndicators of social capital on individual measures of health, such as self-rated health (SRH), evidence suggests that the influence is contextual rat her than compositional in nat ure (Macintyre et al., 1993; Kawachi, Kennedy, & Glass, 1999; Blakel y et al., 2001; Subramanian et al., 2001). Results supporting these conclu sions have been found in both ecological, or macro-level, as well as multilevel studi es. Therefore, criticisms regarding the validity of these findings (e.g., ecological fallacy) for individual-level outcomes such as SRH are rare. The two most pre ssing issues at this time are in what manner should social capital be conc eptualized and measured and to what extent the influence of social capital on health is moderated by individual level factors. Although social capital has been envis ioned as an aspect of the individual (Portes, 1998), the majority of social, political, and health scientists concur with its inherent contextual essenc e. What is less clear and less consistent is the way in which social capital is viewed as a characteristic of the collective having multiple dimensions (e.g., structural vs. cognitive aspects, bridging vs. bonding, participatory vs. perception). At this ti me, there are few studies of self-rated health in which multiple aspects of this construct (participation in voluntary organizations, social tr ust, mutual aid or reciprocity) are examined simultaneously (Kawachi, Kennedy, & Glass, 1999; Greiner et al., 2004). Typically, studies address only one dimensio n, usually social trust, as a single proxy for this multidimensional construct. Given this limitati on in the literature, there is a significant association between social capital and general health status,

PAGE 64

50 with up to 40% greater odds of reporting fair/poor health if one resides in a community with low forms of social capita l after adjusting for a variety of possible confounders (Kawachi, Kennedy, & Glass, 1999; Blakely et al., 2001; Jun, Subramanian, Gortmaker, & Kawachi, 2004; Rohrer, Arif, Pierce, & Blackburn, 2004). The other issue concerns the possible moderators of the social capital and self-rated health association. Paralle l to the level of analysis debate in the poverty-health literature, a debate has evol ved regarding if and to what extent contextual-level effects of social c apital are moderated by individual–level correlates (e.g., social networks, so cial support). Because of the abovementioned restriction in the current evidence base, few multilevel studies have examined the probable role of cross-level interact ions between individual and contextual characteristics. For example, in examining the possible moderating influence of individual-level correlates of social capital on the contextual level influence of social capi tal, one study found that social capital may be associated with good self-reported h ealth only for those individuals who are trusting themselves (Subramanian, Kim, & Kawachi, 2002). In general, more studies are needed on conceptual clarity and multilevel interactions to reach a consensus regarding the relationship bet ween social capital and general health status. As with other specific health outco mes, there have been few empirical studies on the association between soci al capital and mental distress. Less specifically, there is evi dence that aspects of the social environment, such as

PAGE 65

51 inadequate social cohesion (Aneshensel & Sucoff, 1996; Ellawa y, Macintyre, & Kearns, 2001) and related social proc esses (Dressler & Badger, 1985; Ross, 2000a; Schulz, Williams et al., 2000; Leventhal & Brooks-Gunn, 2003), are deleterious contextual influences on ment al health outcomes. By and large the literature on this area as well as, more expr essly, social capital is conceptual by nature (Kawachi & Berkman, 2001; McKenzie Whitley, & Weich, 2002; Sartorius, 2003). Many debates abound regarding social capital and mental distress, some of which center around validity issues. Just as in the self-rated health literature, the most pressing matters at this time include conceptualizatio n, definition, and measurement concerns. In this area, social capital has been characterized as having multiple dimensions: structur al and cognitive, bridging and bonding. Although social capital is more frequently envisioned as a characteristic of the collective, there are some who maintain that this construct reflects both individual and community-level dimensions of social relatedness. There is evidence to support both, although some question whether using individual reports of social networks and soci al support really is social capital at all, but rather testing individual res ources only. Given these concerns, both qualitative (Campbell, Cornish, & Mc Lean, 2004; Ziersch, Baum, Macdougall, & Putland, 2005) and quantitative ev idence still exists that living in areas with little social connectedness, trust and engagement confers excess risk of mental distress (Mitchell & LaGory, 2002; Greiner et al., 2004; Ziersch et al., 2005), with almost a doubling of the odds of suffering from psychiatric morbidity (OR 1.96,

PAGE 66

52 95% CI 1.39-2.75) for those living in areas with low social capital (McCulloch, 2001). As discussed above, there is a lim ited, yet strong body of evidence regarding the health effects of social capita l. However, as social capital can be seen as an aspect of social cohesion (Kawachi & Berkman, 2000), the findings regarding the association between corre lates of the social environment and health will be reviewed to lend additional su pport for the relationship. In current literature, conceptually relat ed constructs are collective efficacy, social cohesion, and the social environment. Research on the health effects of co llective efficacy includes substantial evidence demonstrating the mediating and moderating effects of this social construct on a range of health outcomes, such as CVD mortality (Cohen, Farley, & Mason, 2003), SRH (Browning & Cagney, 2002), neighborhood violence and residential instability (Sampson et al., 1997). Studies of area differences in collective efficacy report improved indi cators of health in communities or neighborhoods where this aspect of social cohesion is strongest. Consistent with evidence from other studies, in a recent review, Sampson (Sampson, 2003) encourages using social context as a crit ical unit of analysis in studies of the health effects of the social environm ent, such as collective efficacy. In contrast to the preponderance of quantitative studies, Fullilove (1998) used data from a qualitative study to ex amine the relationship between social cohesion and health in four communities. Findings were consistent with other studies, which indicated that promotion of social cohesion improves community

PAGE 67

53 health. Evidence pertaining to general health suggests, just as with most social capital studies, that social cohesion mediates the relationship between broader structural environment (e .g., area deprivation) and i ndividual perceptions of health (Macintyre et al, 2000; Drukker et al, 2003). Most studies investigating the cons equences of social location on health have studied a wide variety of aspects of the social environment. For this study, pertinent literature includes those studi es that have focused on CVD-related outcomes. Area influences include such factors as urban/rural setting (Barnett & Halverson, 2000; Ewing, Schmid, Kil lingsworth, Zlot, & Raudenbush, 2003) racial/ethnic influence or composition (Barnett & Halverson, 2000; Franzini & Spears, 2003; Reidpath, 2003), and female c oncentration (LeClere et al., 1998). The majority of these studies were multilevel, where individual level sociodemographic influences have been contro lled. This provides additional support for area or contextual nature of the social world and its effects on CVD morbidity and mortality. What is less known is the direct and indirect effects of a particular aspect of the social world, soci al capital, on specific CVD-related health and behavior. From the above review, one can see that there is a need for additional evidence of the health effects of social capital from multilevel studies, investigating its potential independent and cross-level associations, from data that is expressly intended to measur e and investigate social capital.

PAGE 68

54 Social Context and Health Behaviors Overall, there is a very small literat ure on the influences of social context on health behavior, with most focusing on school environment in childhood and adolescence. By and large, this earl y phase in the literat ure of this area examines the relationship between soci al context and health behavior, without systematic study of alternat ive explanations or temporal ordering as yet. This status is due, in part, to the decontex tualized manner in which most health promotion/behavior change studies and inte rventions are created, with little concern of the feat ures and influence of place (Ma cIntyre & Ellaway, 2000). The reliance on intraindividual determinants of behavior to the exclusion of broader societal influences, which construct or shape risk behavior (Backett & Davison, 1995), is seen in most studies of chronic disease including CVD. Just as the physical environment either inhibits or facilitates health behavior, so does the social or collective characteristics of the community affect individual action (Sorensen et al., 2003) The environment pr ovides opportunities or barriers for agency. A primary way in which the environment “gets under one’s skin” and influences the physiology and pathology of CVD is through its effect on risk behaviors, such as physical activity, di et/obesity, and smoking (Taylor et al., 1997; Kawachi & Berkman, 2000). In regards to exercise, social cont exts characterized by weak social cohesion are associated with reduced ph ysical activity. Support for this

PAGE 69

55 relationship comes from literature exam ining both the social environment in general and social capital in particular. The vast majority of research of environmental influences on physical activity focuses upon the physical context in which activity occurs and in this wa y does address timing of effect. Overall, these studies find that the most pr essing factors include accessibility, opportunity, weather, safety, and aest hetics (Humpel, Owen, & Leslie, 2002). Comparatively, the influence of the social environment on physical activity is less well understood, primarily because behav ior has been viewed in a myopic manner – completely under volitional cont rol of the individual – without concern for the way in which the social contex t impacts opportunity, agency, and choice. The limitations of this lit erature demonstrate the critical role of the social environment on physical activity. Although most studies are quantitative, there has been one qualitative study that explored the differential effects of the social environment on activity (Burton, Turrell, & Oldenburg, 2003). Generally, the majority of studies examine both physical and social aspects simultaneously (Brownson, Ba ker, Housemann, Brennanm & Bacak, 2001; Giles-Corti & Donovan, 2002; Lee & C ubbin, 2002; Ewing et al., 2003). By doing this, these studies were able to c ontrol for some alternative explanations from both environmental and indi vidual influences. Findings suggest that there is a positive association bet ween socially cohesive environments and physical activity with the physical locale necessary but not sufficient in shaping physical activity (Giles-Corti & Donovan, 2002). Other studies found that it is not just the personal factors such as enjoyment and preferences that shape activity levels,

PAGE 70

56 but the attributes of the collective, such as social disengagement, which negatively influence health behavior (Brown son, Baker, Housemann, Brennan, & Bacak, 2001; Salmon, Owen, Crawford, Bauman, & Sallis, 2003) Although there is limited knowledge regarding the effect of social context on behavior, the emerging consensus is that there are multiple dimens ions of the environment that exert influence on physical activity (Macintyre, 2000; Sallis Kraft, & Linton, 2002). At this time, there is no clear agreem ent or debate regardi ng the effects of social capital on physical activity, primar ily because there are so few studies that have examined these relationships. In a Swedish prospective cohort study, Linsdtrom, Hanson, & Ostergren (2001) ex amine the psychosocial conditions that may help to explain group differences in physical activity. The authors found that social capital, defined as social partici pation or engagement in social, civic, or political activities including formal and in formal associations, predict behavior. Specifically, those living in socially disengaged areas had over twice the odds (OR 2.2, 95% CI 1.8 – 2.7) of low physical activity, with little differential effect by individual factors. Alt hough in a follow-up study the authors find some contrary evidence of a contextual effect (Lindstr om et al, 2003), these conclusions may be called into question because the way in which social capital has been measured is not consistent with curr ent definition and usage. Ther efore, the re sults of the original study that inadequate social capita l (possibly through provision of social norms) is a mechanism explaining group di fferences in physical activity are a more robust finding.

PAGE 71

57 The only other study investigating social capital and activity was also from a European nation. An Irish study examines the associations between neighborhood design and social capital ( Leyden, 2003). Positive associations are found between social capital and physi cal activity. However, the author conceptualizes the relationship differently in terms of directional ity. He concludes that there are higher levels of social capital in walkable neighborhoods – essentially that physical activity improves social cohesion. Because there are no other studies examining this associati on, the causal pathway between these two variables cannot be presumed with any certai nty. However, the association in general cannot be questioned. As compared to the liter ature on exercise, there are an even more limited number of investigations of area effects on obesity. Ess entially, a deprived social environment is associated with poor dietary habits, specifically obesity. Another similarity between the physical activity and obesity literature is that most studies have looked at the impact of both the physical and social environment on risk behavior in order to adjust fo r alternative explanations. One commonality of the work in this area is how recent the studies are, with most empirical investigations published only in the last two years. In general, there is a positive association between area of residence and obesity (Ella way, Anderson, & Macintyre, 1997; Ecob & Macintyre, 2000; Morland, Wing, & Diez Roux, 2002), with explicit evidence of the impact of a weakened soci al context and poor diet and obesity (Lee & Cubbin, 2002; Giles-Corti & Donov an, 2003; Vandegrift & Yoked, 2004). Specifically, there is an excess odds (OR 1.6, 95% CI 1.1-2.3) of being

PAGE 72

58 overweight or obese for those residing in socially disorganized environments, controlling for individual demographic and ot her health behaviors (Caitlin et al, 2003). Other studies find that urban sp rawl (including social disengagement) is positively associated with obesity (Ewing et al., 2003; Vandegrift & Yoked, 2004), partly through its effects on physical activity. Although there is a growing literature on the social contextual influence of place on obesity, there is little exami nation of the relationship specifically between social capital and associated risk behavior. Currently, there is only one study of the role of social capital in obes ity. This European work comes from the CVD literature and posits that social ca pital should be an essential consideration of an obesity prevention strategy aim ed at reducing the prevalence of hypertension (Worsely, 2001). Other health behaviors implicated in t he etiology of CVD include smoking. Just as with the other risk factors, this area of study is quite new, with this specific literature the smallest of all. Findings indicated that living in a poor social environment is positively associated with sm oking, although the effects of timing and extraneous variables have not been well studied. Recent qualitative studies found that the daily social environm ent one resides in assists or impedes smoking through contextual influences (Poland, 2000; Bancroft et al., 2003). The few quantitative studies that have been c onducted provided findings consistent with these conclusions (Ecob & Macint yre, 2000; Diez Roux, Merkin, Hannan, Jacobs, & Kiefe, 2003).

PAGE 73

59 Also consistent with the physical acti vity and obesity literature discussed, the evidence of impact of social capital on smoking is recent, limited, and predominantly comes from international studies. Findings from these investigations demonstrated a significant positive association of weak social capital on increased smoking, with st rongest evidence on daily smoking (Lindstrom & Ostergren, 2001; Lindstr om, 2003). One such study found a positive relationship between correlates of so cial capital, such as voting behavior – an indicator of social engagement and smoking (Kelleher, Timoney, Friel, & McKeown, 2002). Although the literature re garding social capital effects on CVDrelated health behavior is incomplete, ther e are studies of its impact on other forms of behavior, such as sexual behavior and STD/HIV (Thomas & Thomas, 1999; Crosby, Holtgrave, DiClemente, Wingood, & Gayle, 2003), environmental risk (Wakefield, Elliott, Cole, & Eyle s, 2001), and alcohol use (Weitzman & Kawachi, 2000). Overall, there are three commonalit ies amongst this literature: preponderance of international investigat ions, clustering of risk behaviors, and the studying of the relati onship between risk and social context is still in its infancy and therefore unable to establish if a causal relationship exists. When studying the initiation and maintenance of ri sk behavior, it is critical to not just focus on the proximate causes of such activity, but the more contextual influences that shape the env ironment within which risk occurs. Ultimately, the environment confers opportunity or barri ers to engage in behavior. Although most studies examining the broader environmental context have looked at the impact

PAGE 74

60 of the physical environmen t, there is evidence that a commensurate (Salmon et al., 2003; Vandegrift & Yoked, 2004) need exists to address related social environmental factors as well (Ford, Ahluwalia, & Galuska, 2000) when investigating social context on CVD risk fa ctors such as physical activity, obesity, and smoking. Health Behaviors and Hypertension In general, there is a large know ledge base on the associations between health behavior and CVD. In point of fact most of the information regarding the etiology of CVD comes from studies invest igating individual factors. The majority of public health prevention efforts hav e targeted these proximal causes or associations to combat the increase in their prevalence, which then leads to increased rates of CVD and other chronic dis eases. In regards to this study, the pertinent behavioral risk factors include physi cal inactivity, obesity, and smoking. Commonalities across the literatures linki ng these behaviors to CVD include the impact of individual-level moderators (e .g., age, gender, race/ethnicity, SES) and the role of clustering of risk. There are di fferences in the literatures as well. For example, some of the earliest work on behavioral risk is in regards to the deleterious effects of smoking on CVD, whereas obesity and physical activity studies are comparatively more recent. Ov erall, evidence for all three factors as casually related to CVD is robust; t here is a large literature based on cohort, case-control, and randomized control studies from which to draw conclusions.

PAGE 75

61 Physical Inactivity Physical activity is associated with improved health. Initial studies suggest that vigorous exercise was the only way in which to obtain cardiorespiratory fitness. It has been subsequently ascertaine d that, at this time, the benefits of any form of regular exercise include a reduction in risk for a wide variety of chronic diseases, including CVD. Overa ll, these benefits cons ist of lowering the risk of: heart disease, development or worsening of hypert ension, and premature mortality from CVD (USDHHS, 1996). Although the type (i.e., vigorous, moderate, light) and amount (i.e., minutes per week) are still being studied, the general recommendation for adults is 20 30 minutes or more of moderate activity at least 5 days a week to obtai n optimum health benefit s. Nevertheless, findings demonstrate that as little as 10 or more minutes a day has significant reduction in premature CHD mortality (Leon, Myers, & Connett, 1997). Although this recommendation is widely known, t he majority of adults are physically inactive, with women, lower SES, Af rican-American and Hispanic, and older individuals having higher rates of inactivity (USDHHS, 1996). Up until relatively recently, the majori ty of findings on physical activity and CVD have been based on white men (USDHHS 1996). Currently evidence on the impact of exercise on CV D in women has essentially paralleled the benefits to men (Oguma, Sesso, Paffenbarger, & Lee, 2002). Findings suggest reduced CVD (HRR 0.64, 95% CI 0.42-0.97) is asso ciated with increasing activity levels

PAGE 76

62 for older women, even adjusted for other ri sk factors such as smoking, BMI, and comorbidities (Gregg et al., 2003). These general conclusions have been replicated on middle-aged women as we ll (Owens, Matthews, Raikkonen, & Kuller, 2003). Additional evidence of the relati onship between physical inactivity and CVD come predominately from cohort studies in which a gradient effect has been consistently demonstrated (USDHHS, 1996). On the whole, the effects of physical inactivity on this form of chr onic disease are commonly divided into two classes – one on the effects on CVD in ge neral and the other group of studies on CHD in particular, with a more limited numbe r of investigations specifically on hypertension (USDHHS, 1996). In regards to CVD risk as a whole, evidence indicates an inverse dose-response effect with physical activity (Kannel & Sorlie, 1979; Kannel, Belanger, D’Agostino, & Is rael, 1986; LaCroix, Leveille, Hecht, Grothaus, & Wagner, 1996; USDHHA 1996). Kaplan et al (1996) finds that there is a significant protective effect of ph ysical activity on CVD mortality (RR 0.81, 95% CI 0.71-0.93). Others have found a similar independent relationship, even after accounting for age, sex, BMI, and smoking (LaCroix, Leveille, Hecht, Grothaus, & Wagner, 1996). T here are parallel findings with respect to CHD. Studies have shown that risk of CHD is in versely associated with exercise (Blair, 1994; USDHHS, 1996), with an overall RR r anging from 1.21 to 1.8 of CHD for inactivity (Paffenbarger, Wing, & Hyde, 1978; Kannel et al., 1986; Berlin & Colditz, 1990).

PAGE 77

63 In regards to the pertinent pathways of effects, regular exercise influences some of the biological mechanisms associated with CHD – hypertension, dyslipidemia, obesity, and endothelial health (McKechnie & Mosca, 2003). In concordance with these findings, most of the literature proposes the physiological pathways to CVD include physical activity ’s beneficial effects on blood pressure, atherosclerosis, ischemia, levels of pl asma fibrinogen, and plasma viscosity (Gordon & Scott, 1991; Leon & Connett, 1991; USDHHS, 1996; Lindstrom, Hanson, & Ostergren, 2001). To specify, in regards to high blood pressure, evidence posits that exercise benefits both normo tensive as well as hyper tensive adults (Whelton, Chin, Xin, & He, 2002). Although there is support of moderating effects for race/ethnicity and gender, there is comm ensurate evidence of the independent effects of physical activity on the prev alence of hypertension (Bassett, Fitzhugh, Crespo, King, & McLaughlin, 2002). In addition, the progression of atherosclerosis is attenuated by regular aerobic activity (Nordstrom, Dwyer, Merz, Shircore, & Dwyer, 2003). Further, evidence suggests that those who are least physically active have a 30% gr eater risk of developing hypertension (USDHHS, 1996). In general, patterns of findings demons trate that physical activity and CVD (overall), CHD, or hypertension reflect a ro bust inverse gradient effect. By and large, findings indicate that the burden of CVD coul d be greatly decreased by increased physical activity. For exampl e, regular exercise could reduce or prevent the 13.5 million who have CHD, t he 1.5 million who suffer heart attacks,

PAGE 78

64 and the 50 million who have hypertension in a year as well as positively impact the approximately 60 million (one-third of the U.S. popul ation) who are overweight (USDHHS, 1996). Prevention e fforts would also need to take into consideration that physical inactivity and obesity often co-occur (USDHHS, 1996; Blair & Brodney, 1999) with combined negat ive effects on CVD mortality (Fang, Wylie-Rosett, Cohen, Kaplan, & Alderman, 2003). Obesity Overweight and obesity result in ex cess risk for many forms of CVD (USDHHS, 1996, 2001; NIDDKD, 1998) in both men and women (Hu, 2003). Support for this causal relationship is found in multiple types of studies including quantitative (e.g., multiple cohort, RCT, longitudinal and cro ss-sectional; NIH, 1998), qualitative cross-cultural (Treloar et al., 1999), as well as reviews of the independent effect of obesity on more prominent forms or correlates of CVD such as hypertension, dyslipidemia, C HD, CHF (Labarthe, 1998; NIDDKD, 1998; Rashid, Fuentes, Touchon, & Wehner, 2003; Sowers, 2003) with additional evidence on increased risk even for coronar y thrombosis (Wolk, Berger, Lennon, Brilakis, & Somers, 2003). The effects of obesity-related morbidity and mortality from CVD are well known, however only recently have studies demonstrated that there are effects not just for absolut e weight, but weight gain as well.

PAGE 79

65 There is support for the effect of excess weight in increasing CVD mortality, with exponential effects as we ight increases (Meyer, Sogaard, Tverdal, & Selmer, 2002; Rogers, Hummer, & Krueger, 2003). For example, in a longitudinal cohort study, in vestigators find negative effe cts of both weight and weight gain on CVD risk for young adults (Norman, Bild, Lewis, Liu, & West, 2003). There is supplementary evidence on the deleterious health effects of weight gain (Willett et al, 1995). Additi onally, Kawachi (1999), in a review, presented evidence that an increase of approximately 11-17 pounds in adulthood confers 25% excess risk of suffering from CHD, with risk increasing as weight increases (Galanis, Harris, Sharp, & Petrovitch, 1998; NIDDKD, 1998). The disparities in prevalence of CVD ma y be due, in part to the differential effects of obesity on CVD risk by gender and race/ethnicity (Patt, Yanek, Moy, & Becker, 2003) as well as age and SES (U SDHHS, 2001). Clearly, obesity is a multi-determinant risk factor with severe health consequences in and of itself, however the current literature has begun to focus more on how this feature impacts and is impacted by other forms of CVD risk behavior. Although there is a large evidence base for the direct e ffects of obesity on CVD (NIDDKD, 1998) and a growing literature on the combined effe cts of physical activity and obesity on CVD-related outcomes, currently, ther e is no clear consensus as to the directionality of effect between behaviors. What is certain is the mounting ev idence of clustering – those who are obese or overweight are less physically active by and large, with support for an inverse dose response between exercise and obesity (USDHHS, 1996, 2001).

PAGE 80

66 For example, in a study by (Kannel, W ilson, Nam, & D'Agostino, 2002), the authors have found that the clustering of risk factors associated with being overweight or obese occurs in well over half the individuals, resulting in increasing RR for CHD for men (2.07) and women (10.9). Along with physical activity, findings indicate obesity and smok ing co-occur in studies of CVD (Millen et al., 2002), with additional evidence that engaging in both behaviors compounds risk of CVD (NIDDK D, 1998). In point of fact, some suggest that morbidity due to obesity is as large as that from smoking (Sturm & Wells, 2001). Smoking Compared to the other behavioral risk fact ors (i.e., physical inactivity and obesity), smoking has by far the longes t and strongest evidence base for a causal relationship with CVD. The link between smoking and CVD has been established in the scientif ic literature since the 19th century (USDHHS, 1983). In essence, smoking has been studied from a variety of designs, which have established its causal role in t he development and progression of CVD. Empirical investigations, meta analyses, and reviews concur t hat most forms of CVD are affected including CHD, hypert ension, arteriosclerosis, aortic aneurysm, peripheral vascular disease (USDHH S, 1983, 2001; Labarthe, 1998; Burns, 2003). The knowledge base, however, is not equal amongst all groups. The majority of studies up until the past two decades have used men only in their

PAGE 81

67 samples. More recent examinations of smoking effects on CVD for women find just as robust associations. For exam ple, women smokers have a significantly greater risk of CHD (Mann, James, W ang, & Pickering, 1991; Kawachi et al., 1994) as well as hypertension (USDHHS, 1980; Mann et al, 1991). There are numerous traditional mec hanisms by which smoking confers physiological damage. Smoking causes weakening of vessel walls and exacerbates or hastens atherosclerosis and athlerosclerotic lesions (USDHHS, 1983, 2001). In addition, ni cotine and other toxins fr om smoking increase blood pressure and heart rate, resulting in an imbalance of oxygen supply to the myocardium, platelet aggr egation and function, and related pathogenic insults to the cardiovascular system such as thrombosis, hemorrhage, and vasoconstriction (USDHHS, 1983). Recent findings include studies br anching from a focus on traditional effects of smoking on pathology to more nov el factors, which are involved in CVD risk. For example, the physiological responses to smoking implicated in development of disease include increas ing cholesterol and blood pressure (USDHHS, 2001). Currently, investigat ors find a dose-response association between smoking and other, less investi gated biochemical processes such as elevated serum C-reactive protein, fibrinogen, and homocysteine levels (Bazzano, He, Muntner, Vupputuri, & Whelt on, 2003). In addition to biochemical effects, there is evidence of smoki ng’s structural damage implicated in the etiology of atherosclerosis (Pittilo, 2000; Burns, 2003).

PAGE 82

68 In essence, much is known about t he independent effects of smoking. However, there is a gap in the informa tion regarding its effect on and with other forms of risk. To fill this space in k nowledge, a growing aspect of this literature relates to the clustering e ffect of behavioral risk. It is not clear in current literature the relative and/or interactiv e effects of these behavioral risk factors (physical activity, obesity, and smoking) on the development of CVD. What is apparent is that, to some extent, all three are related in their associations with and prevention of CVD (e.g., a reduction in CHD is associated with not smoking similarly to the benefit of regular physica l activity, which is causally linked to reduced rates of obesity) (USDHHS, 1996) There are many ways in which these CVD risk factors may be linked incl uding physiological, psychological, and social factors. Health Behaviors and General Health Status Compared to the CVD liter ature, there are relative ly fewer studies that examine the influence of health behavior on self-reported general health status, even though self-rated health is a well-est ablished predictor of future morbidity and mortality. While most studies exami ne the role of individual risk behavior, such as smoking, obesity, or inactivity a few studies simultaneously examine the role of multiple risk behaviors in i ndependently predicting se lf-rated health, both

PAGE 83

69 internationally (Manderbacka, Lundberg, & Martikainen, 1999; Abu-Omar, Rutten, & Robine, 2004; Fr oom, Melamed, Triber, Ratson, & Hermoni, 2004) and nationally (Brown et al., 2003; Strine et al., 2005). Physical Inactivity The evidence of the health benefits of physical activity comes from both cross-sectional and prospective studies. T here is mounting evidence of both the direct as well as mediated influence of ex ercise on self-reported health. Findings suggested that the influence of health beha viors on self-rated health is mediated by health problems and functional lim itations (Manderbacka, Lundberg, & Martikainen, 1999). Result s demonstrated a salubrio us influence of physical activity on general health status (Ha ssan, Joshi, Madhavan, & Amonkar, 2003; Atlantis, Chow, Kirby, & Singh, 2004) with regular recommended levels of exercise conferring health benefits fr om young adulthood through to old age (Brown et al., 2003). Longitudinal inve stigations have found that infrequent exercise predicts poor self-rated health in men (OR 1.67, 95% CI 1.04-2.17) more that 7 years later (Froom, Mela med, Triber, Ratson, & Hermoni, 2004). Most of the studies of the role of ph ysical activity in general health status evaluations have treated this behavior as either indicating regular activity, insufficiently active or inactive. More recently, a study exam ining three critical aspects of exercise (intensity, frequency, and duration) finds that insufficient activity levels doubled the likelihood of reporting poor health more than half the

PAGE 84

70 month (OR 2.02, 95% CI 1. 85-2.21), controlling for a range of sociodemographic and behavioral factors (Brown et al., 2004). Obesity Body weight influences both physiological as well as psychological aspects of self. Excess weight can harm health through its role in the etiology of many chronic diseases in addition to its influence on self-esteem and selfefficacy. Although there is a vast litera ture on the impact of obesity on health, there is a relatively small group of st udies on obesity’s impact on general health status, specifically. Findings undersco red the negative impact of current weight (Ferraro & Yu, 1995; Manderbacka et al., 1999) as well as the positive influence of weight loss (Fontaine, Barofsky, Bart lett, Franckowiak, & Andersen, 2004) on self-reported general health status. Cross-sectional studies have found increased levels of BMI significantly associated with self-reported poor health (Mansson & Merlo, 2001; Kobau, Safran, Zack, Moriarty, & Chapman, 2004), with obesity predicting poor self-rated health in nationally representative studies (Ferraro & Yu, 1995). Moreover, findings from national studies indicate gradient effect in that severely obese are more likely than obese to report poor health (Hassan, Joshi, Madhavan, & Amonkar, 2003), with excess o dds increasing from 12% to 323% as weight increases (Ford, Moriar ty, Zack, Mokdad, & Chapman, 2001). Consistent with this conclusion, a non-U. S. study indicates the presence of a

PAGE 85

71 gradient; compared to normal weight indivi duals, poor health was more likely to be reported by persons overweight (OR 1.46, 95% CI 1.24-1. 71) or obese (OR 2.67, 95% CI 2.04-3.48) (Manderbacka et al., 1999) Smoking Of all three health behaviors, smoking has the least amount of literature directly examining its association with gener al health status. Only recently has a national investigation of smoking, other risk behavio rs, and self-reported health been conducted, with findings supporting the significantly negative impact of current smoking (Strine et al., 2005). Previous international studies have found the direct influence of smoking, whereby those who smoke are greater than 60% (OR 1.63, 95% CI 1.23-2.16) more likely to report wors e general health seven to ten years later (Froom et al., 2004). A dditional indirect associations indicated that those who smoke are at significantly greater odd s (OR 1.81, 95% CI 1.492.19) of reporting poor health (Manderbacka et al., 1999). Health Behaviors and Mental Distress There is a growing literature on th e impact of health behavior on mental distress. Studies have examined both type and amount of risk behavior on prevalence and treatment of mental distre ss. Specific disorders as well as general sub-syndromal moodiness (Kobau, Safran, Zack, Moriarty, & Chapman,

PAGE 86

72 2004), which may result in an increased bur den of disease, has been studied. Although the majority of studies have examined the role of individual risk behaviors, there are investigations of t he association between a cluster of risk factors and mental health/distress as well. To compare, the evidence supporting the influence of physical activity on mental health is far more rigorous, with the obesity literature less advanced. Although t he relationship between smoking and mental distress has been established, the evidence, by and large, is mixed as related to direction of influence. Physical Inactivity There is a growing literature on the benefits of physical activity for both the prevention and treatment of an array of mental disorders. There is evidence from a variety of studies, including cross-sectional, longitudinal, quasiexperimental (Dunn, Trivedi, & O'Neal 2001; Salmon, 2001), and randomized controlled investigations (Blumenthal et al ., 1999; Atlantis et al., 2004). Although the relationship between exercise and m ental health is a complex one, the protective influence of physical activity is not questioned. Findings support the benefit of acute exercise on present st ate of self-reported stress and long-term benefits in reduction of risk of psychopathology (Dunn et al., 2001). Positive influence of physical activi ty has been found with regards to specific disorders such as depression (Atlantis, Chow, Kirby, & Singh, 2004; Dunn, Trivedi, Kampert, Clark, & Cham bliss, 2005) and anxiety (Cromarty,

PAGE 87

73 Robinson, Callcott, & Freeston, 2004) as well as general stress responsivity (Scully, Kremer, Meade, Graham, & Dudgeo n, 1998; Atlantis et al., 2004), mood state, and self-esteem (S cully, Kremer, Meade, Graham & Dudgeon, 1998). Investigations on the public burden of mental disorders have found that even after controlling for sociodemographic fact ors, individuals who engage in regular exercise have a significantly lower pr evalence of many m ental disorders, including depression (OR 0.75, 95% CI 0.6-0.94), panic attacks (OR 0.73, 95% CI 0.56-0.96), social phobia (OR 0.65, 95% CI 0.53-0.8), spec ific phobia (OR 0.78, 95% CI 0. 63-0.97), and agoraphobia (OR 0. 64, 95% CI 0.43-0.94) (Goodwin, 2003). There are multiple plausible mec hanisms through which physical activity influences mental health. Biological expl anations suggest that exercise impacts neurotransmitter production and response (Scully et al., 1998). Physical activity may also act as a buffer to the body’s natural response to stress and stress hormones, thereby improvi ng individual resilience (Scully et al., 1998). Psychological explanations posit that regular exercise may lead to enhanced self-esteem and improved body image (Scully et al., 1998). Finally, some propose that the advantage stems from a dose-response relationship with respect to prevalence (Goodwin, 2003; A bu-Omar, Rutten, & Lehtinen, 2004) as well as treatment (Dunn et al., 2005) Overall, most dose-related studies have found that it is regular exercise, that is physical activity meeting the recommended level, which confers benefit (Brown et al., 2003; Brown et al., 2004).

PAGE 88

74 Obesity There is evidence of the associati on between obesity and mental distress, with findings from a variety of correlati onal and prospective studies. Early studies examining the prevalence of depression among the obese have found mixed results with some finding no asso ciation, reciprocal association, depression leading to obesit y, or obesity implicated in depression or mental distress (Friedman & Brownell, 1995). Cr oss-sectional studies have found that those who are obese are 17% to 41% more likely to report me ntal distress more than half the month (Hassan et al., 2003) More recently, findings from prospective cohort studies have indicated that being obese confers excess risk of future depression and other adverse ment al health outcomes (Roberts, Deleger, Strawbridge, & Kaplan, 2003; Hasler et al ., 2004). For example, controlling for a variety of sociodemographic and psychosocia l factors, there is a 70% to 200% excess risk of depression up to five years later for those who are obese (Roberts, Kaplan, Shema, & Strawbridge, 2000; Robe rts, Strawbridge, Deleger, & Kaplan, 2002; Roberts et al., 2003). There is an array of explanator y models and hypothesized mechanisms underlying this relationship. Some suggest t hat it is not being ov erweight that is distressing, but rather the relationship is better explained by the negative experience of dieting and related stress in volved in attempting (and often failing)

PAGE 89

75 to meet societal norms which dictate th inness as a requirement for attractiveness (Ross, 1994). Others have pr oposed that there are multiple biopsychosocial pathways accounting for the association, which include dispar ate consumption of carbohydrates and reduced neurotransmitter pr oduction resulting from inactivity (Palinkas, Wingard, & Barrett-Connor, 1996; Roberts, Deleger, Strawbridge, & Kaplan, 2003). Although more rigorous studies have est ablished the role of weight in the etiology of mental distress, the effe ct is not consistent across subgroups. Sociodemographic differences have been found based on age as well as gender. Increased body weight and depression has been found more often among women than men (Carpenter, Hasin, Allison, & Faith, 2000) with evidence of an inverse relationship in older men (Palinkas et al., 1996). Smoking The use of tobacco and nicotine is associated with mental distress, although evidence regarding the directionality of the relationship is mixed (Lasser et al., 2000). There is evidence to s upport both those suffering from mental disorders are more prone to smoke as we ll as increased prevalence of mental disorders among smokers (Williams & Zi edonis, 2004). Some hypothesize that this complex relationship is due to genetic diathesis whereby one form of

PAGE 90

76 neurobiological or endocrine response may be involved in the etiology of both outcomes (Picciotto, Brunzell, & Caldar one, 2002; Williams & Ziedonis, 2004). Additionally, findings demonstrat e that these systems may also be implicated as mechanisms under lying the smoking mental distress relationship. For example, smoking is posited to alter neurotransmitters associated with depression (Dierker, Avenevoli, Stolar, & Merikangas, 2002) as well as chronic use exacerbating more severe symptom otology associated with schizophrenia and Post-Traumatic Stress Disorder (Williams & Ziedonis, 2004). There are a variety of mental disor ders associated with previous smoking. In a study involving two national data se ts, Breslau and Klei n (1999) found that daily smoking conferred a significant in creased risk of panic attacks for both men and women. There was no evidence to support the reverse – panic attacks were not associated with initiation of subs equent smoking behavior. There was additional evidence for the link from sm oking to anxiety from a prospective longitudinal study. Controlling for a myriad of individual charac teristics, smoking one pack or more per day in adole scence assigned excess risk in early adulthood of the following diso rders: generalized anxiet y disorder (OR 5.53, 95% CI 1.84-16.66), agoraphobia (OR 6.79, 95% CI 1.53-30.17), and panic disorder (OR 15.58 95% CI 2.31 -105.14) (Johnson et al., 2000). There is also consistent evidence on the association between sm oking and depressive symptomotology (Breslau, Kilbey, & Andreski, 1991; Die rker et al., 2002; Williams & Ziedonis, 2004), with findings indicating that chronic smoking results in almost four times

PAGE 91

77 the risk of having major depression (O R 3.90, 95% CI 1. 85-8.20) (Goodman & Capitman, 2000). In sum, engaging in the risk behavio rs examined in this study are significantly associated with frequent ment al distress (Kobau et al., 2004; Strine et al., 2004). In addition, taken as a whole, there is sufficient empirical evidence to support associational relationships amon gst social structur e, social context, risk behaviors, and the three health outcome s under study, even after adjusting for various competing factor s. The strength of causal evidence, however, is mixed (Figure 2.2). For ex ample, there is moderate st rength of evidence of a causal relationship between social stru cture and both social context and health behavior, with a moderate-strong level with CVD. In regards to social contextual factors, there is moderate strength of evidence with respect to CVD, however weak support of a causal link with heal th behaviors, which may be due to the dearth of studies in that area. By far, the strongest evidence base of a causal relationship is between health behaviors and CVD, with the majority of the literature demonstrating tempor al ordering of effects. In essence, there are no studies that have attempted to examine t hese different levels of association and their possible causal effects on self-r eported CVD. Ther efore, based upon the above review of literature, this study has the potential to advance the knowledge base in this area and make a contribution to the field (Table 2.1).

PAGE 92

78 Figure 2.2: Strength of Evidence Moderate-Strong Moderate Moderate Strong Moderate Weak Weak = Correlational evidence only Moderate = Some evidence ruling Strong = Some causality established out of alternative explanations (evidence of temporal ordering) Social Structural Inequalities (income inequality, 200% FPL) Disease (self-reported hypertension, general health status, mental distress ) Social Contextual Factors (social capital e.g., social trust, informal social engagement, organizational activism/formal social participation, mutual aid) Health Behaviors (BMI/diet, physical activity, smoking)

PAGE 93

79Table 2.1: Hypotheses Overall Hypertension General Health Status Mental Distress 1 Behavioral variables only partially mediate social structure and disease 1a Social structural inequality in the community in which one resides will positively influence self-reported hypertension. 4a Social structural inequality in the community in which one resides will negatively influence self-reported general health status. 7a Social structural inequality in the community in which one resides will negatively influence selfreported mental health. 1a1 Greater income inequality in the community in which one resides will positively influence self-reported hypertension. 4a1 Greater income inequality in the community in which one resides will negatively influence self-reported general health status. 7a1 Greater income inequality in the community in which one resides will negatively influence selfreported mental health. 1a2 Greater poverty in the community in which one resides will positively influence self-reported hypertension. 4a2 Greater poverty in the community in which one resides will negatively influence self-reported general health status. 7a2 Greater poverty in the community in which one resides will negatively influence self-reported mental health. 1b The effect of social structure on selfreported hypertension is only partly mediated by known risk behaviors (BMI/diet, physical activity, smoking). 4b The effect of social structure on selfreported general health status is only partly mediated by known risk behaviors (BMI/diet, physical activity, smoking). 7b The effect of social structure on self-reported mental health is only partly mediated by known risk behaviors (BMI/diet, physical activity, smoking). 1b1 Greater income inequality in the community in which one resides will positively influence self-reported hypertension after controlling for individual risk behavior. 4b1 Greater income inequality in the community in which one resides will negatively influence self-reported general health status after controlling for individual risk behavior. 7b1 Greater income inequality in the community in which one resides will negatively influence selfreported mental health after controlling for individual risk behavior. 1b2 Greater poverty in the community in which one resides will positively influence self-reported hypertension after controlling for individual risk behavior. 4b2 Greater poverty in the community in which one resides will negatively influence self-reported general health status after controlling for individual risk behavior. 7b2 Greater poverty in the community in which one resides will negatively influence self-reported mental health after controlling for individual risk behavior. 1c Greater social structural inequalities in the community in which one resides positively influences engaging in high risk behavior.

PAGE 94

80Table 2.1: Hypotheses Overall Hypertension General Health Status Mental Distress 1c1 Greater income inequality in the community one resides in positively influences physical inactivity. 1c2 Greater income inequality in the community one resides in positively influences overweight/obesity. 1c3 Greater income inequality in the community one resides in positively influences smoking. 1c4 Greater poverty in the community one resides positively influences physical inactivity. 1c5 Greater poverty in the community one resides positively influences overweight/obesity. 1c6 Greater poverty in the community one resides positively influences smoking. 2 Social context partially mediates social structure and disease. 2a The level of social capital in the community in which one resides influences self-reported hypertension. 5a The level of social capital in the community in which one resides influences self-reported general health status. 8a The level of social capital in the community in which one resides influences self-reported mental health. 2a1 Communities characterized by less social trust will positively influence selfreported hypertension. 5a1 Communities characterized by less social trust will negatively influence self-reported general health status. 8a1 Communities characterized by less social trust will negatively influence self-reported mental health.

PAGE 95

81Table 2.1: Hypotheses Overall Hypertension General Health Status Mental Distress 2b Social structural inequalities will be negatively associated with a salubrious social context. 2b1 Income inequality will be negatively associated with social trust. 2b2 Income inequality will be negatively associated with informal social engagement. 2b3 Income inequality will be negatively associated with formal group involvements, or organizational activism. 2b4 Income inequality will be negatively associated with mutual aid. 2b5 Poverty will be negatively associated with social trust. 2b6 Poverty will be negatively associated with informal social engagement. 2b7 Poverty will be negatively associated with formal group involvements, or organizational activism. 2b8 Poverty will be negatively associated with mutual aid.

PAGE 96

82Table 2.1: Hypotheses Overall Hypertension General Health Status Mental Distress 2c Social structural inequality in the community in which one resides will positively influence self-reported hypertension after controlling for community social context. 5b Social structural inequality in the community in which one resides will negatively influence self-reported general health status after controlling for community social context. 8b Social structural inequality in the community in which one resides will negatively influence selfreported mental health after controlling for community social context. 2c1 Income inequality in the community in which one resides will positively influence self-reported hypertension after controlling for levels of social trust. 5b1 Income inequality in the community in which one resides will negatively influence self-reported general health status after controlling for levels of social trust. 8b1 Income inequality in the community in which one resides will negatively influence selfreported mental health after controlling for levels of social trust. 2c2 Income inequality in the community in which one resides will positively influence self-reported hypertension after controlling for informal social engagement. 5b2 Income inequality in the community in which one resides will negatively influence self-reported general health status after controlling for informal social engagement. 8b2 Income inequality in the community in which one resides will negatively influence selfreported mental health after controlling for informal social engagement. 2c3 Income inequality in the community in which one resides will positively influence self-reported hypertension after controlling for formal group involvements, or organizational activism. 5b3 Income inequality in the community in which one resides will negatively influence self-reported general health status after controlling for formal group involvements, or organizational activism. 8b3 Income inequality in the community in which one resides will negatively influence selfreported mental health after controlling for formal group involvements, or organizational activism. 2c4 Income inequality in the community in which one resides will positively influence self-reported hypertension after controlling for levels of mutual aid. 5b4 Income inequality in the community in which one resides will negatively influence self-reported general health status after controlling for levels of mutual aid. 8b4 Income inequality in the community in which one resides will negatively influence selfreported mental health after controlling for levels of mutual aid.

PAGE 97

83Table 2.1: Hypotheses Overall Hypertension General Health Status Mental Distress 2c5 Poverty inequality in the community in which one resides will positively influence self-reported hypertension after controlling for levels of social trust. 5b5 Poverty inequality in the community in which one resides will negatively influence self-reported general health status after controlling for levels of social trust. 8b5 Poverty inequality in the community in which one resides will negatively influence selfreported mental health after controlling for levels of social trust. 2c6 Poverty in the community in which one resides will positively influence selfreported hypertension after controlling for informal social engagement. 5b6 Poverty in the community in which one resides will negatively influence self-reported general health status after controlling for informal social engagement. 8b6 Poverty in the community in which one resides will negatively influence self-reported mental health after controlling for informal social engagement. 2c7 Poverty in the community in which one resides will positively influence selfreported hypertension after controlling for formal group involvements, or organizational activism. 5b7 Poverty in the community in which one resides will negatively influence self-reported general health status after controlling for formal group involvements, or organizational activism. 8b7 Poverty in the community in which one resides will negatively influence self-reported mental health after controlling for formal group involvements, or organizational activism. 2c8 Poverty in the community in which one resides will positively influence selfreported hypertension after controlling for levels of mutual aid. 5b8 Poverty in the community in which one resides will negatively influence self-reported general health status after controlling for levels of mutual aid. 8b8 Poverty in the community in which one resides will negatively influence self-reported mental health after controlling for levels of mutual aid. 3 Behavior only partially mediates social context and disease. 3a Engaging in risk behavior (BMI/diet, physical inactivity, smoking) is positively associated with self-reported hypertension. 6a Engaging in risk behavior (BMI/diet, physical inactivity, smoking) is negatively associated with selfreported general health status. 9a Engaging in risk behavior (BMI/diet, physical inactivity, smoking) is negatively associated with self-reported mental health. 3b Weaker social context in the community in which one resides positively influences engaging in high-risk behavior.

PAGE 98

84Table 2.1: Hypotheses Overall Hypertension General Health Status Mental Distress 3b1 Weaker social trust in the community in which one resides positively influences physical inactivity. 3b2 Less informal social engagement in the community in which one resides positively influences physical inactivity. 3b3 Less organizational activism in the community in which one resides positively influences physical inactivity. 3b4 Less mutual aid in the community in which one resides positively influences physical inactivity. 3b5 Weaker social trust in the community in which one resides positively influences overweight/obesity. 3b6 Less informal social engagement in the community in which one resides positively overweight/obesity. 3b7 Less organizational activism in the community in which one resides positively overweight/obesity.

PAGE 99

85Table 2.1: Hypotheses Overall Hypertension General Health Status Mental Distress 3b9 Weaker social trust in the community in which one resides positively smoking. 3b10 Less informal social engagement in the community in which one resides positively influences smoking. 3b11 Less organizational activism in the community in which one resides positively influences smoking. 3b12 Less mutual aid in the community in which one resides positively influences smoking. 3c Weaker social context in the community in which one resides positively influences self-reported hypertension after controlling for individual risk behavior. 6b Weaker social context in the community in which one resides negatively influences self-reported general health status after controlling for individual risk behavior. 9b Weaker social context in the community in which one resides negatively influences self-reported mental health after controlling for individual risk behavior. 3c1 Weaker social trust in the community in which one resides positively influences hypertension after controlling for individual risk behavior. 6b1 Weaker social trust in the community in which one resides negatively influences general health status after controlling for indi vidual risk behavior. 9b1 Weaker social trust in the community in which one resides positively influences mental distress after controlling for individual risk behavior. 3c2 Less informal social engagement in the community in which one resides positively influences hypertension after controlling for indivi dual risk behavior. 6b2 Less informal social engagement in the community in which one resides negatively influences general health status after controlling for individual risk behavior. 9b2 Less informal social engagement in the community in which one resides positively influences mental distress after controlling for individual risk behavior.

PAGE 100

86Table 2.1: Hypotheses Overall Hypertension General Health Status Mental Distress 3c3 Less organizational activism in the community in which one resides positively influences hypertension after controlling for indivi dual risk behavior. 6b3 Less organizational activism in the community in which one resides negatively influences general health status after controlling for individual risk behavior. 9b3 Less organizational activism in the community in which one resides positively influences mental distress after controlling for individual risk behavior. 3c4 Less mutual aid in the community in which one resides positively influences hypertension after controlling for individual risk behavior. 6b4 Less mutual aid in the community in which one resides negatively influences general health status after controlling for indi vidual risk behavior. 9b4 Less mutual aid in the community in which one resides positively influences mental distress after controlling for individual risk behavior.

PAGE 101

87 CHAPTER 3: METHODOLOGY Methods Study Design This study employed a multilevel, retrospective, nonexperimental design with longitudinal elements utilizing se condary data. Multilevel designs were utilized when data have been clustered or nested within different levels of analysis. The design was nonexperimental in that it employs data from a naturally occurring study population wit hout randomization and although did not clearly include the temporal ordering nece ssary to generate evidence concerning causal relationships, it did allow for tem poral precedence to a limited extent. This work was retrospective, in that it def ined the study population in terms of 2001 data and then linked those data to exposures in 2000 data sources. The data utilized in this study were originally co llected for different studies and purposes, but they have been judged to be suitable for addressing this study’s objectives based upon criteria outlined by McCall and Applebaum (McCall & Applebaum, 1991) and Stewart and Kamins (Stewart & Kamins, 1993). Until recently, the study of macro-le vel social determinants of health has relied on ecological studies, which have employed cross-sectional designs. The proposed study employed a multilevel design for two reasons. First, a multilevel design was selected because of the nature questions that were investigated and

PAGE 102

88 the data that was utilized. Different dat a sources representing differing levels of analysis were critical to this study as no one data set had specific multilevel data related to CVD, general health status, or mental distress. This design was chosen also because it had multiple benefits over ecological approaches, including limiting problems related to falla cies (ecologic, atomistic) and allowing for the unique variance of contextual and compositional levels (e.g., to test whether income inequality and social ca pital effects on CVD were significant while adjusting for individual-level fact ors, such as SES and individual health behaviors) (Subramanian et al., 2003). This retrospective study defined the sa mple in terms of availability of outcome data (CVD, general health stat us, mental distress) in 2001 and then examined associations with ex posure (social structural and social contextual inequalities) from other sources of aggr egation (from the 2000 Census for social structural variables and fr om the 2000 SCCBS for social contextual variables). Normally in this type of study, the concer n making valid inferences included recall bias, however as this study was not re lying on one group at one level of analysis, this threat was not applicable. This study was a nonexperimental des ign with longitudinal elements in that the data was observational in natur e and deliberately selected to address the temporal ordering of possible effects through observations gleaned from two different time points. Typically in a nonexperimental design, directionality cannot be established; however, because this design had a temporal component, exposure was speculated to precede dis ease (and not visa versa) to some

PAGE 103

89 extent. As with most none xperimentation, this study presented a more rigorous design than cross-sectional, however one still cannot infer causality due to nonrandomness of assignment to groups. Secondary data sources were used in this study for a few reasons. First of all, in order to answer the research questions proposed, multiple sources of data were needed at various levels of aggregation (i.e., individuals in communities in counties). In addition, t here was no single publicly available data set in the United States that included all the variables of interest. These three data sources were selected because they included some portion of the variables of interest in each and, as importantly, they all were linkable based on federal codes included in each. Furthermore, few studies existed that explicitly addressed the effects of social capital and related variables on specific health outcomes. At present, t here have been no studies that link the BRFSS healthrelated data to a rich source of contex tual data. The study linked the BRFSS with the Social Capital Benchmar k Study by FIPS codes that were present in both – this way individuals were placed in their respective community or state contexts. Sampling The sample included in each dataset (Census, SCCBS, and BRFSS) was intended to represent the same populati on, but the data sources reflected different levels of aggregation and different timepoints. The sample for this study

PAGE 104

90 was limited to individu als from the 2001 BRFSS sa mple who resided in communities represented in the SCCBS. BRFSS The BRFSS was an ongoing public healt h surveillance system of health behaviors of adults in the United States. Its purpose was to gather information on health practices, knowledge, and ri sk associated with major burdens of disease and disability. It s intent was to collect prevalence estimates on the lifestyle and health behaviors of adults in the U.S, which have been used to inform prevention policy and public health practice. The data reflected both national and state-spec ific trends on a variety of public health-related factors. It was an annual telephone survey administered by the CDC to a random sample of adults. Implementation of the survey was conducted by state and local health departments. The reliability of th e BRFSS has been evaluated through testretest studies. Overall, the survey ex hibited good reliability, with Kappa ranging from 0.60 (for minority parti cipants) to 0.80 (for White respondents) in regards to sociodemographic items and ranging from 0.70 to 0.80 for behavioral items (Stein, Lederman, & Shea, 1993). Other examinations of the BRFSS have found it to have both moderate (physical activity ) and high reliability and validity in most of its items (smoking, blood pressure, height, weight, and demographics), which come from the core instrument (Nelso n, Holtzman, Bolen, Stanwyck, & Mack,

PAGE 105

91 2001) (Table 3.1). Specifical ly in regards to CVD, t he BRFSS has been used in many studies to assess trends in fa ctors such as hypertension (Ayala, Greenlund, & Croft, 2002) as well as mult iple risk (Jackson, Jatulis, & Fortmann, 1992; Greenlund et al., 2004). There were three sections to the su rvey: core questions (all participate), optional modules (participation is dec ided upon by state), and state-added questions. The number of questions in any given survey ranged from 90 – 150 items. Probability sampling was used for all households with telephones in each state. The majority of participating states utilized a disproportionate stratified sample (DSS) design. Interviews were carried out using computer assisted telephone interviewing (CATI), with interv iews lasting an average of 15-20 minutes.

PAGE 106

92Table 3.1: Behavioral Risk Factor Su rveillance System Reliability and Validity1 Variable Topic Reliability2 Validity3 Consistent with Other Surveys Smoking Current Smoking Status High High Yes Obesity Height High High Yes Weight High High Yes; slightly underreported Physical Activity Level Moderate Moderate Mixed CVD Hypertension High Moderate Yes; slightly underreported General Health Status Health-Related Quality of LifeModerate High Measures only from other surveys Mental Health/Distress Health-Related Quality of LifeDifficult to determine Moderate Mixed 2where reliability High: 3where validity High: Sensitivity & Specificity > 80% or correlation coefficients >0.60 Moderate: Moderate: 60% < Sensitivity & Specificity > 79% or 0.40 < correlation coefficients >0.59 Low: Low Sensitivity & Specificity < 60% or correlation coefficients <0.40 1Nelson, Holtzman, Bolen, Stanwyck, & Mack, 2001

PAGE 107

93 SCCBS The Social Capital Community Benc hmark Survey (SCCBS) was to date the largest survey of civic engagement in the United States. The study was designed by scholars in social capi tal and measurement who attended the Saguaro Seminar at Harvard Universi ty in 1999 and was obtained for the purposes of this study from the Roper Center for Pub lic Opinion Research. The SCCBS was the first of it s kind to measure aspects and correlates of social capital across the United States conduct ed on individuals and then aggregated to community-level. Another intended use of the data was to provide the communities who participated with info rmation to support efforts targeting improvement in community connect edness. Data was aggregated from individual responses to contextual le vel constructs, as communities, not individuals themselves, made up the samp le. Communities were invited to participate during an annual meeting of Foundations in 1999. Thirty-four Foundations were selected for the range of communities they represented across the U.S. The communities consisted of counties, cities, and lightly populated states. Each Foundation selected the areas within their communities to be surveyed, with the majority using proporti onate sampling. The purpose of this work was to provide researchers and practitioners a comprehensive benchmark database to enhance current kn owledge and future initiatives. The survey interviews were approximately 26 minut es and were completed between July and November 2000 and was carried out using random-digit dialing. As certain

PAGE 108

94 Foundations represented more than one co mmunity, there were 40 communities in the final sample, each represent ed by approximately 500-1500 residents (Table 3.2). The contextual-level data from the SCCBS were linked to the individuallevel data from the BRFSS through F ederal Informational Processing Standards (FIPS) codes. FIPS codes were federa lly designated unique numbers that were assigned to every county in each stat e. Utilizing FIPS codes allowed each person in the BRFSS sample to be locat ed within his/her community represented in the SCCBS sample. However, only 27 out of the 40 communities in the original study were used. These thirt een were omitted because the geography of the community could not be matched with Census data, did not have FIPS codes assigned, and therefore could not be link ed to the BRFSS data (12) or did not have data collected for the 2001 BRFSS (1). Census Data was obtained from the 2000 Census. The Census has been conducted every ten years and was a survey of individuals and households in the United States. The Census provided statistical information regarding the population. Results have informed nat ional and local public planning and program funding as well as research. Specif ically, measures of social structural inequalities (i.e., absolute and relative deprivation) were employed. Relative deprivation was assessed by income inequalit y. Absolute or area measures of

PAGE 109

95 poverty included the percent of families li ving at or below 200% of the Federal Poverty Level (FPL). Community-level data from the SCCB S was linked to county-level data from the Census through t he use of FIPS codes as well. The 27 communities from the SCCBS that make up this study’s sample varied in the number of counties they represent. Therefore, some communities included only one county, some included several, and a few included lightly populated states. Variable Measures Variables selected were grouped into f our categories: social structural, social contextual, health behavior, and outcome (Table 3.3). The social structural inequalities were represented by relative and absolute deprivation and were measured by the 2000 Census. The social contextual factors included social capital and its correlates and were measured by the SCCBS (2000). The third and fourth groups of variables, health behavior and outcomes, were both measured by the 2001 BRFSS.

PAGE 110

96Table 3.2: Social Capital Community Benchmark Survey Sample Actual Sample Size Location (State/County) Response Rate (%) 500 Alabama/Jefferson, Shelby 31.6 501 Arizona/Maricopa 31.7 515 California/Los Angeles 24.1 504 California/San Diego 30.9 500 California/San Francisco 27.1 500 Colorado/Boulder 22.4 501 Colorado/Denver 14.9 1379 Delaware/state of 27.3 510 Georgia/DeKalb, Fulton, Cobb, Rockdale, Henry 29.8 1001 Indiana/state of 26.7 500 Louisiana/Baton Rouge 25.0 500 Michigan/Kalamazoo 27.1 501 Michigan/Wayne, Oakland, Macomb, St. Clair, Washtenaw, Monroe, Livingston 30.1 503 Minnesota/Dakota, Ramsey, Washington39.2 502 Montana/state of 44.1 711 New Hampshire/state of 32.2 541 New York/Onondaga 24.8 988 New York/Monroe, Wayne, Ontario, Livingston, Genesee, Orleans 27.1 750 North Carolina/Forsyth 34.8 750 North Carolina/Guilford 32.7 1100 Ohio/Cuyahoga 20.0 1001 Ohio/Butler, Clermont, Hamilton, Warren38.7 Kentucky/Boone, Campbell, Kenton Indiana/Dearborn 500 Oregon/Crook, Deschutes, Jefferson 34.1 500 Pennsylvania/York 28.2 500 Texas/Harris 28.7 500 Washington/Yakima 34.6 500 West Virginia/Kanawha, Putnam, Boone 27.4

PAGE 111

97Table 3.3: Variable Sources & Definitions Source Variable Defined as Census Relative Deprivation: Income Inequality Qualitative and quantitative description of the dispersion or distribution or range of income in a population Census Absolute Deprivation: PovertyArea-level socioeconomic status SCCBS Social Capital & Correlates Featur es of social organization, such as participation in associations and civic engagement, interpersonal trust, and norms of reciprocity, which act as resources and facilitat e collective action BRFSS Physical Activity Any acti vity that can be considered exercise (e.g., not related to work) BRFSS Obesity/Overweight Body Mass Index meeting USDHHS cutoffs BRFSS Smoking History of or active tobacco smoking BRFSS CVD Hypertension BRFSS General Health Status Global health assessment BRFSS Mental Distress Number of days in past 30 mental health not good

PAGE 112

98 Income inequality was a form of relati ve deprivation. It was measured by the commonly used Gini coe fficient. The Gini has bee n derived from the Lorenz curve, which was a diagram of the cu mulative proportion of income plotted against the cumulative percentage of the population (Kawachi & Kennedy, 1997b; Soobader & LeClere, 1999). T he proportion approaching 0 indicated perfect equality whereas 1 demonstrated perfe ct inequality. Area-level poverty or absolute deprivation was measured by the percent of the population in a specified area who lived at or below 200% of the Federal Poverty Line. For 2000, the FPL ranged from $8,350 (family of 1) and $11,250 (family of 2) to $17,050 (family of 4). Specif ic data related to inco me, number of households, and population economic indicators were obtained from the 2000 Census in order to calculate both measures of social structural inequalities. Social contextual variables consisted of social capital and its correlates including measures of social trus t, participation in formal and informal organizations, and mutual aid. All items measuring these constructs came from the 2000 SCCBS. Social trust described a c haracteristic of the collective (e.g., general interpersonal trust, level of trus t amongst neighbors, coworkers, etc.). There were five questions in this inde x (Table 3.4). Item s related to this construct included whether most people c an be trusted, to w hat extent one can trust the police in one’s comm unity, or to what extent one trusts people who work in the stores in which one shops. Res ponse options for each item ranged from a five to seven point Likert scale.

PAGE 113

99 Participation in formal organizations, or organizational activism (Table 3.5), included a count of eighteen various activities such as involvement in professional, neighborhood, service, c harity groups whereas informal social activity or engagement (Table 3.6) in cluded five items related to other relationship-based actions such as engaging with relatives, friends, or having people to your home. Mutual aid was measured by several items (Table 3.7). Questions were related to volunteering (e.g., for a nei ghborhood or civic group, for a school or youth program, for a place of worship) and donating. Response options were either yes/no or a five to seven point Likert scale, depending upon the specific item. For the purposes of this study, health risk was restricted to poor physical activity, being overweight or obese, and engaging in smoking behavior (Table 3.8). BRFSS items related to physical ac tivity pertained specifically to non-work related activity. Physical activity was measured as meeting current recommended levels in type (light vs. moder ate vs. vigorous), duration (minutes), and frequency (days/week or month), engaging in some activity, or not active. Activities were defined as those which caused small changes in respiration and heart rate, such as brisk walking, bi cycling, vacuuming, and gardening. In addition, activity included those that caus e significant increase in respiration and heart rate, such as running, aerobics, and heavy yard work. Reliability and validity estimates for BRFSS items relat ed to physical activity were moderate (Table 3.1).

PAGE 114

100Table 3.4: Social Capital Community Benchmark Survey Social Trust Index Variable Item Response TRUST “Whether most people can be trusted ” 1. P eople can be trusted 3.Depends 8.Don't Know 2.You can't be too careful 9.Refused TRNEI "How much you can trust people in your neighborhood" 1.Trust them a lot 3.Trust them a little 5.Does not apply 2.Trust them some 4.Trust them not at all 8.Don't Know 9.Refused TRWRK “How much you can trust people you work with” 1.Trust them a lot 3.Trust them only a little 5.Does not apply 2.Trust them some 4.Trust them not at all 8.Don't Know 9.Refused TRREL “How much you can trust people at your church or place of ” 1.Trust them a lot 3.Trust them only a little 5.Does not apply 2.Trust them some 4.Trust them not at all 8.Don't Know 9.Refused TRSHOP “How much you can trust people who work in the stores where you shop" 1.Trust them a lot 3.Trust them only a little 5.Does not apply 2.Trust them some 4.Trust them not at all 8.Don't Know 9.Refused TRCOP “How much you can trust the police in your local community” 1.Trust them a lot 3.Trust them only a little 5.Does not apply 2.Trust them some 4.Trust them not at all 8.Don't Know 9.Refused

PAGE 115

101Table 3.5 : Social Capital Community Benchmar k Survey Informal Social Engagement Index Variable Item Response CFRDVIST “In the past twelve months, how often h ad friends over to your home” Continuous CFAMVISI “In the past twelve months, how oft en visited with relatives” Continuous CJOBSOC “In the past twelve months, how often socialized with co-workers outside of work” Continuous CFRDHANG “In the past twelve months, how often hung out with friends in a public place” Continuous CCARDS “In the past twelve months, how often played cards or board games with others” Continuous

PAGE 116

102Table 3.6: Social Capital Community Benchmark Survey Formal Social Participation/Organizational Activism Index Variable Item Response GRPREL Participate in organization affiliated with religion 1. Yes 2. No 3. Don't Know 4. Refused GRPSPORT Participate in sports club, league, or outdoor activity 1. Yes 2. No 3. Don't Know 4. Refused GRPYOUTH Participate in youth organization 1. Yes 2. No 3. Don't Know 4. Refused GRPPTA Participate in parent association or other school support group 1. Yes 2. No 3. Don't Know 4. Refused GRPVET Participate in veterans group 1. Yes 2. No 3. Don't Know 4. Refused GRPNEI Participate in neighborhood association 1. Yes 2. No 3. Don't Know 4. Refused GRPELD Participate in seniors group 1. Yes 2. No 3. Don't Know 4. Refused GRPSOC Participate in charity or social welfare or ganization 1. Yes 2. No 3. Don't Know 4. Refused GRPLAB Participate in labor union 1. Ye s 2. No 3. Don't Know 4. Refused GRPPROF Participate in professional, trade, farm, or busine ss as 1. Yes 2. No 3. Don't Know 4. Refused GRPFRAT Participate in service or fraternal organi zation 1. Yes 2. No 3. Don't Know 4. Refused GRPETH Participate in ethnic, nationality, or civil ri ghts org 1. Yes 2. No 3. Don't Know 4. Refused

PAGE 117

103Table 3.6: Social Capital Community Benchmark Survey Formal Social Participation/Organizational Activism Index Variable Item Response GRPPOL Participate in political group 1. Yes 2. No 3. Don't Know 4. Refused GRPART Participate in literary, art, or musical gr oup 1. Yes 2. No 3. Don't Know 4. Refused GRPHOB Participate in hobby, investment, or garden club 1. Yes 2. No 3. Don't Know 4. Refused GRPSELF Participate in self-h elp program 1. Yes 2. No 3. Don't Know 4. Refused GRPWWW Involved in group that meets over the Inte rnet 1. Yes 2. No 3. Don't Know 4. Refused GRPOTHR Belong to other kinds of clubs or organiza tions 1. Yes 2. No 3. Don't Know 4. Refused OFFICER “Served as an officer or on a committee” 1. Yes 2. No 3. Don't Know 4. Refused CCLUBMET “In the past twelve months – How often attended a club meeting” 4 Point Likert Scale CPUBMEET "In the past twelve months How often attended public meeting discussing school or" 4 Point Likert Scale

PAGE 118

104Table 3.7: Social Capital Community Benchm ark Survey Mutual Aid Index Variable Item Response VOLARTS Volunteered for cultural or arts organizations 1. Yes 3. Don't Know 2. No 4. Refused VOLHEA Volunteered for health care or fight disease 1. Yes 3. Don't Know 2. No 4. Refused VOLHUM Volunteered to help poor or elderly 1. Yes 3. Don't Know 2. No 4. Refused VOLNEI Volunteered for neighborhood or civic group 1. Yes 3. Don't Know 2. No 4. Refused VOLREL Volunteered for a place of worship 1. Yes 3. Don't Know 2. No 4. Refused VOLYOU Volunteered for school or youth programs 1. Yes 3. Don't Know 2. No 4. Refused CVOLTIME “In the past twelve months, number of times volunteered” Continuous GIVEOTHR Dollars contributed to non-religious charit ies 0. None 3. $500 < $1000 8. Don't know 1. < $100 4. $1000 < $5000 9. Refused 2. $100 < $500 5. > $5000 GIVEREL Dollars contributed to church or religious causes 0. None 3. $500 < $1000 8. Don't know 1. < $100 4. $1000 < $5000 9. Refused 2. $100 < $500 5. > $5000

PAGE 119

105Table 3.8: Behavioral Risk Factor Surve illance System Health Behavior Items Health Behavior Item Response Physical Activity "During the past 30 days, other than your regular job, did you participate in any physical activities or exercise such as running, calisthenics, golf, gardening, or walking for exercise?" Yes or No "Now thinking about the moderate physical activities you do in a usual week, do you do moderate activities for at least 10 minutes at a time, such as brisk walking, bicycling, vacuuming, gardening or anything else that causes small increases in breathing or heart rate?" Yes or No "How many days per week do you do these moderate activities for at least 10 minutes at a time?" Days per week "On days when you do moderate activities for at least 10 minutes at a time, how much total time per day do you spend doing these activities?" Hours and minutes per day "Now thinking about the vigorous physical activities you do in a usual week, do you do vigorous activities for at least 10 minutes at a time, such as running, aerobics, heavy yard work, or anything else that causes large increases in breathing or heart rate?" Yes or No "How many days per week do you do these vigorous activities for at least 10 minutes at a time?" Days per week

PAGE 120

106Table 3.8: Behavioral Risk Factor Surve illance System Health Behavior Items Health Behavior Item Response "On days when you do vigorous activities for at least 10 minutes at a time, how much total time per day do you spend doing these activities?" Hours and minutes per day Calculated physical activity level categorized Meets recommendation/some activity/physically inactive Tobacco Use "Do you now smoke cigarettes every day, some days, or not at all?" Current smoker(every day-some days)/former smoker/never smoked Smoking status (Derived) Current &/or history of smoking/never smoked Overweight/ "About how much do you weigh without shoes?" in pounds Obesity "About how tall are you without shoes?" in feet/inches Calculated Body Mass Index categorized Not overweight or obese (BMI<25)/ Overweight (25 > BMI < 30)/ Obese (BMI > 30)

PAGE 121

107 Overweight or obesity was measured us ing the standard BMI cutoff points. The Body Mass Index has been calculat ed using height and weight, both of which were available from the BRFSS. Reliability and validit y of self-report of height and weight in t he BRFSS was high, although there have been slight underestimation of weight across populati ons (Nelson et al., 2001) (Table 3.1). Smoking status was measured by a single composite variable. The specific questions reflected both history of smoking and whet her the individual currently smoked every day, some days, or not at all. Reliability and validity of these items were high and were consis tent with other surveys of smoking behavior (Nelson et al., 2001). The outcome of interest was a restricted range of CVD, which has comprised a cluster of diseases, but for the purposes of this research study CVD indicator was limited to self-reported hypert ension. The validity of utilizing selfreport in assessments of CVD has been es tablished for all race-sex groups (Giles, Croft, Keenan, Lane, & Wheeler, 1995). Reliability and validity were high to moderate, respectively, specifically for the BRFSS hypertension item (Nelson et al., 2001). Hypertension was treated as a binary va riable (yes/no), with an affirmative response on the one item indicating hypert ension. Other forms of CVD were not being studied as there were either insu fficient data in the 2001 BRFSS (Coronary Heart Disease and Myocardial Infarction) or they were either rare in occurrence (e.g., peripheral arterial disease, aorti c aneurysm, deep vein thrombosis) or may have had a different physiological pathogen ic process (e.g., stroke).

PAGE 122

108 The original form of general health st atus was a Likert scale consisting of five response options: excellent, very good, good, fair, poor. Often due to the subjective nature of this variable, evaluat ion of reliability have been challenging. Although there are no reliability esti mates for the BRFSS item per se, international studies on self-reported heal th were moderately reliable (Nelson et al., 2001). Validity issues were sim ilar, in that although there have been no validity issues on the BRFSS item, ot her national surveillance surveys have demonstrated that self-reported health is highly valid and a strong predictor of future morbidity/mortality (Nelson et al ., 2001). Consistent with previous studies of general health status and for ease of in terpretation, the item was dichotomized in this study (0 = excellent/ver y good/good and 1 = fair/poor. Mental distress was a self-report measure assessing the number of days out of the past 30 when one’s mental health was not good (including feelings of sadness, anxiety, stress). There have been no studies of the reliability or validity of this BRFSS item. However, other scales with similar questions have found strong reliability and moderate va lidity (Nelson et al., 2001). Analysis Procedures To begin, preparation of the data fo r analysis included cleaning data and decisions regarding the treatment of missi ng data. The proce ss of cleaning the data consisted of identifying and correcting er rors in the data sets. The possible multiple sources of error that needed to be investigated in cluded respondents or

PAGE 123

109 coders mismarking responses, data entry er rors, and “not applicable” or missing coded as 0. Detecting suspicious, erroneous, or illogical values involved examining the data through descriptive stat istics, such as the range of possible values, outliers, frequencies, means, and standard deviations. Inconsistencies between related variables were explored as well. Errors that were found resulted in the variable in question being recoded, wit hout jeopardizing the integrity of the item. Once the data were cleaned, arrangements were made for missing data. Decisions regarding managing missing data included investigating the type and pattern of the missi ng information. First and foremost, it was imperative to assess whether the data was missing at random or was systematic. Every effort was made to retain or approximate t he original distribution of responses in order to maintain the integrity of the data. Following these steps in data preparat ion, the three data sources were linked and measures were taken to arr ange the data for analysis. Data was weighted based upon the respecti ve weighting schemes. The weight variable for the SCCBS was derived in a three step process whereby the initial weight (number of household adul ts/number of phone lines) was multiplied by the balancing weight derived from population distributions of variables such as gender, age, education, and race/ethnicity. The data weighting variable for the BRFSS was the product of several features of the sample and population. These factors included the probability of select ion among strata of phone numbers, the number of phone lines in a respondents hous ehold, the number of adults in the respondents household, and an adjustment fo r non-coverage and non-response.

PAGE 124

110 At this time, some variables were recoded and new variables were created (i.e., reporting health as fair/poor or not). Initially, univariate and bivariate analyses were conducted in order to see the ways in which the variables were distributed within the sample and their basic associations. Univariate statistics included mean, standard devia tion, range, and distributi on of responses on all items and composites, whether they repr esent Level-1 or Level-2 variables. Examination of the shape of responses was critical, as much of the analysis is based upon the assumption of normality. The only variables that were not anticipated to meet this cr iterion were the outcome variables (hypertension, general health status, and ment al distress), which were assumed to be positively skewed in the population. The informati on permitted me to better understand the nature of the data employed and enhanc ed the process of analysis and interpretation. The next step in examining the data was through bivariate analysis. Just as with univariate analysis, separate analyses were conducted for Level-1 and Level-2 data. As the range of variables in this study included nominal, ordinal, interval, and ratio levels of measuremen t, a variety of statistics were used, including Chi-Square, ANOVA, and Spearman and Pearson Correlations. Measures of association examined the re lationships between, for example, each sociodemographic variable (e.g., SES, race/ethnicity, gender) and individual health behaviors and self-reported hypertens ion as well as covariation between social capital indicators and social struct ural variables (i.e., income inequality and

PAGE 125

111 poverty). Bivariate analyses permitt ed the basic relationships between the variables to be elucidated and therefore informed the next step in the analysis. Hypothesis testing included identification of first-level, second-level, and cross-level relationships utilizing Hierar chical Linear Modeling. Two outcomes were treated as binary (hypertensi on and general health status) and one comprised a count (mental distress). For the two outcomes treated as dichotomous, (Yij {0,1}), the probability distribut ion of the sampling model was Bernoulli and the logit link function was us ed for transformation. As mental distress was treated as a count of days out of last 30, (Yij {0,1,…, 30}), the probability distribution of the sampling model was Po isson and the log link function was used for transformation. Basic assumptions of a two level hierarchical generalized model were main tained. These assumptions included (Raudenbush & Bryk, 2002): 1. Each Level-1 random effect cann ot be distributed normally, as it has either two discrete values or a count from 0-30, and variance 2 for each Level-1 unit within every Level-2 unit is heterogeneous. 2. There are restrictions on predicted values. 3. Level-1 predictors are independe nt of Level-1 random effects. 4. Level-2 random effects are multiv ariate normal, with a mean of 0 and variance of qq 5. Level-2 predictors are independe nt of Level-2 random effects. 6. Level-1 and Level-2 errors are independent.

PAGE 126

112 7. Predictors at each level are not correlated with random effects at the other level. The first level variables included so ciodemographic characteristics (i.e., gender, SES, race/ethnicity) and behavioral ri sk factors (i.e., physical inactivity, obesity, smoking). Control sociodemographi c factors were modeled in a cluster with each behavior then added individually to the model. As behavioral risk factors tend to cluster in the population, different combinations of risk were examined. The second level variables that were introduced in the model consisted of social contextual indicato rs (i.e., social capital and correlates) and social structural variables (i.e., income inequa lity and poverty). As with the level-1 equations, the relationship of level-2 pr edictors in explaining the outcome were investigated separately as well as together For example, two-level models were tested utilizing income inequality and pov erty (individually) and rotating social capital indicators as level-2 variable s with differing combinations of level-1 predictors to explain self-r eported hypertension, general health status, or mental distress. Although multiple comparisons were analyzed, no statistical adjustments were made. In this study, the devel opment of a priori hypotheses involved planned testing of theoretically-driven ques tions. As such, hypotheses testing of multiple comparisons was from a c onfirmatory mode, and not ad hoc, and therefore no adjustments were made. However, by not adjusting for multiple

PAGE 127

113 tests within each model, an inflated Type I error rate may have resulted. This would have produced a rejection of the null hypothesis when it was actually true – in other words, concluding a statistically significant re lationship existed when in reality it did not. Additional effort was made to adjust for confounders. At level-1, a host of individual characteristics (i.e., gender age, race/ethnicity, SES, education, marital status) were controlled for in order to isolate the relationships under study. At level-2, confounders under st udy included (for each community): median household income, percent unem ployed, and percent who have completed high school or less. Bec ause of the commonalities between the possible confounders, multic ollinearity was thoroughly examined and controlled.

PAGE 128

114 Chapter 4 The findings of this study are pres ented in the following chapter. Both univariate and bivariate statis tical results are presented, followed by the results of multilevel models testing the study’s hypot heses. For all variables and their associations, patterns of the pooled sample are described initially in order to give a global perspective of the relationshi ps, followed by discussion of selected individual patterns by community. Fina lly, a summary of support for hypotheses is presented. Univariate Analysis Sociodemographic Factors A description of the sociodemographic composition of the sample as a whole is presented in Table 4.1. Over all, the sample is 41% male and 59% female. One-third of the sample is between 45 and 64, with the next larger groups at age 20 to 34, 35 to 44, and 65 and older, respectively. In regards to race/ethnicity, the sample is 5.2% Hispan ic, 84.1% White, 9.6% Black, and 6.4% other. The majority of individuals are married (54.3%), have completed college (32.6%), and have an annual income between $20,000 and $50, 000 (42.9%). There is relatively small amount of missing data for these sociodemographic

PAGE 129

115 variables, with means across communiti es ranging from none (e.g., sex) to 13.2% (e.g., income). Table 4.1: Pooled Sample Sociodemographic Factors Gender Male 40.7% Female 59.3% Age 20-34 24.9% 35-44 22.5% 45-64 33.8% 65+ 18.9% Race/Ethnicity Hispanic 5.2% White 84.1% Black 9.6% Other 6.4% Income < $20,000 18.4% $20,000 < $50,000 42.9% $50,000 < $75,000 17.8% > $75,000 20.9% Education < 12 9.9% 12 30.7% 13-15 26.9% 16+ 32.6% Marital Status Married 54.3% Separated/Widow ed/Divorced 27.5% Never Married 18.2% The distribution of these factors for i ndividual communities is presented in Table 4.2. Comparing characteristics bet ween communities reveals that there is a wide range of incomes represented acro ss the sample, the highest proportion of residents earning $75,000/ye ar or more located in San Francisco (38.5%), Atlanta (36.3%), and St. Paul (34.4%). The three comm unities with the greatest

PAGE 130

116 proportion of disadvantaged individuals earning less than $20,000/year are Los Angeles (27.7%), Kalamazoo (27.6%), and Birmingham (26.3%). The sample is predominantly middle-aged, with the eldes t population, on average, living in Yakima, Washington (32.2% age 65 and ol der) and the youngest residing in East Baton Rouge, Louisiana (36.3% between 20 and 34 years old). Specifically, in comparing the 27 co mmunities that make up the study sample, in regards to racial/ethnic com position, the most diverse community is Georgia (59.3% White, 34.1% Black, 6.6% other) and the most homogeneous is Oregon (96.9% White, 0% Black, 3.1% other). Health Factors: Behavioral Variables The distribution of health behaviors and outcomes for the pooled sample is displayed in Table 4.3. Overall, t he sample is moderately active – over 85% report engaging in some activity, with le ss than half (44.8%) meeting current recommendations of regular exercise fo r sustaining health benefits (20 minutes of moderate activity most days). In addi tion, less than half of the pooled sample (42.1%) has a normal Body Mass Index; the ma jority of the sample is overweight or obese (58%), with only 4% missing data. Smoking is the leas t prevalent of the three risk behaviors under study, with under one-quarter (23.6%) engaging in this behavior. Approximately 26. 8% of the sample suffers from hypertension. Although the majority report their general health to be very good or excellent (56.9%), one-third (34.2%) su ffer from some form of m ental distress each month.

PAGE 131

117 In comparing the health activities of inte rest (physical activity, Body Mass Index, and smoking) between communities, ther e is a wide range in the frequency of behavior (Table 4.4). The most active community is Boul der (CO), where 65% of the sample meets recommended levels of activity. The community with the least active inhabitants is East Baton Rouge (LA), with only 34% reporting engaging in the recommended levels of physical activity. In regards to Body Mass Index, Boulder (CO) is also the healthiest in the sample with 60% of the residents reporting they are neither overweight nor obese, as opposed to Kanawha Valley (WV), where only 34% reported having a healthy Body Mass Index. In addition, the sample as a whole is largely comprised of nonsmokers (76%), with a wide span of prevalence of 8.1% in Central Oregon to 27.4% of residents in the Kanawha Valley (WV) community.

PAGE 132

118Table 4.2: Frequencies of Sociodemographic Factors* Sex Age Race/Ethnicity N M F 20-3435-4445-6465+ HispWhiteBlack Other Birmingham (AL) 496 36.7 63.326.5 20.0 33.5 20.01.0 65.4 31.8 2.8 Maricopa (AZ) 856 42.5 57.526.8 21.7 32.1 19.513.882.4 4.1 13.6 Los Angeles (CA) 1002 43.1 56.931.7 22.4 32.0 13.934.475.8 11.4 12.8 San Diego (CA) 346 41.3 58.727.7 22.6 30.1 19.622.387.8 4.1 8.1 San Francisco (CA) 95 48.4 51.634.0 18.1 33.0 14.914.774.7 7.4 17.9 Boulder (CO) 124 44.4 55.730.8 25.8 29.2 14.27.3 90.2 0.8 9.0 Denver (CO) 228 39.9 60.134.6 19.4 31.8 14.322.876.1 12.0 12.0 Delaware (DE) 3514 38.7 61.323.3 21.8 33.3 21.62.8 81.0 14.6 4.5 Atlanta (GA) 646 40.3 59.828.5 27.7 31.6 12.22.0 59.3 34.1 6.6 Indiana (IN) 3993 40.4 59.626.4 21.4 33.1 19.12.9 90.7 6.2 3.0 E. Baton Rouge (LA) 461 39.9 60.136.3 16.8 32.7 14.31.7 63.8 32.1 4.0 Kalamazoo (MI) 89 44.9 55.136.1 15.1 30.2 18.62.3 85.4 9.0 5.6 Southeast (MI) 1554 38.2 61.824.7 23.9 34.1 17.43.2 72.8 20.5 6.7 St.Paul (MN) 844 41.2 58.826.2 27.0 31.0 15.81.9 91.7 4.4 3.9 Montana (MT) 3338 42.6 57.419.8 20.1 37.6 22.52.5 86.4 0.2 13.5 New Hampshire (NH) 4068 42.5 57.522.0 25.5 35.3 17.21.6 96.0 0.4 3.6 Central (NY) 106 37.7 62.320.0 23.0 34.0 23.02.8 93.3 4.8 1.9 Rochester (NY) 164 34.2 65.926.0 24.1 32.9 17.14.3 84.7 9.2 6.1 Winston-Salem (NC) 454 38.8 61.222.3 19.8 33.5 24.51.8 70.7 26.9 2.5 Greensboro (NC) 413 35.6 64.424.6 22.6 31.8 20.92.4 68.7 26.7 4.7 Cleveland (OH) 459 37.7 62.321.7 23.9 33.6 20.83.5 73.7 22.6 3.8 Cinncinati (OH) 1038 40.9 59.128.6 21.3 30.8 19.41.7 87.5 10.5 2.1 Central (OR) 99 47.5 52.529.8 17.0 34.0 19.25.1 96.9 0.0 3.1 York (PA) 127 40.9 59.130.9 15.5 40.7 13.02.4 92.9 4.0 3.2 Houston (TX) 802 41.0 59.031.5 25.8 30.7 12.024.968.7 16.0 15.3 Yakima (WA) 119 39.5 60.521.7 18.3 27.8 32.214.393.2 0.9 5.9 Kanawha Valley (WV) 497 40.4 59.622.5 21.1 36.4 20.01.0 93.6 3.8 2.6 TOTAL SAMPLE 25932 40.7 59.3 29.4 22.5 33.8 18.9 5.2 84.1 9.6 6.4 *NOTE: All numbers reflect percentages

PAGE 133

119Table 4.2: Frequencies of Sociodemographic Factors* Education Income Marital N <12th 12 13-1516+ <20k20<50k50<75k75k+ Married Divorced/ Widowed/ SeparatedNever Birmingham (AL) 496 10.9 30.429.2 29.626.341.6 15.3 16.9 50.9 31.0 18.1 Maricopa (AZ) 856 8.9 25.533.3 32.313.245.8 17.5 23.5 54.4 27.7 17.9 Los Angeles (CA) 1002 17.0 21.426.5 35.227.731.5 16.5 24.4 46.0 28.2 25.8 San Diego (CA) 346 9.3 24.428.1 38.321.333.0 20.6 25.1 56.9 24.7 18.4 San Francisco (CA) 95 6.3 9.5 24.2 60.019.825.3 16.5 38.5 33.8 30.0 36.3 Boulder (CO) 124 3.2 12.125.8 58.914.240.7 16.8 28.3 57.4 15.7 27.0 Denver (CO) 228 17.1 18.021.1 43.924.240.6 16.4 18.8 39.5 29.4 31.2 Delaware (DE) 3514 9.3 34.525.7 30.520.140.4 17.9 21.6 53.5 27.7 18.8 Atlanta (GA) 646 7.0 17.522.5 53.011.436.3 16.1 36.3 49.5 22.2 28.3 Indiana (IN) 3993 10.4 38.724.1 26.818.747.0 18.7 15.6 56.7 28.1 15.3 E. Baton Rouge (LA) 461 7.6 24.628.1 39.720.141.7 16.6 21.6 48.1 25.2 26.7 Kalamazoo (MI) 89 12.4 23.622.5 41.627.638.2 6.6 27.6 54.7 23.3 22.1 Southeast (MI) 1554 8.5 27.730.5 33.414.839.7 18.8 26.8 50.6 27.8 21.6 St.Paul (MN) 844 5.1 21.132.0 41.88.2 36.6 20.7 34.4 55.4 25.1 19.5 Montana (MT) 3338 10.9 34.129.1 25.925.552.7 12.5 9.3 56.3 30.4 13.3 New Hampshire (NH) 4068 7.4 29.826.5 36.412.340.4 21.6 25.7 59.4 25.2 15.4 Central (NY) 106 7.7 26.027.9 38.519.834.1 22.0 24.2 51.5 30.1 18.5 Rochester (NY) 164 5.5 26.228.1 40.216.338.8 15.0 29.9 50.9 25.8 23.3 Winston-Salem (NC) 454 13.5 27.024.6 35.020.846.8 17.8 14.6 49.9 31.2 18.9 Greensboro (NC) 413 11.4 25.626.3 36.720.845.2 16.0 18.1 46.4 31.2 22.4 Cleveland (OH) 459 8.3 30.128.2 33.416.349.0 14.5 20.2 46.8 29.2 24.1 Cinncinati (OH) 1038 10.6 32.626.3 30.517.442.6 17.9 22.1 52.4 27.0 20.7 Central (OR) 99 3.0 38.434.3 24.212.455.1 16.9 15.7 62.0 22.8 15.2 York (PA) 127 8.7 40.923.6 26.816.450.9 23.3 9.5 61.8 23.6 14.6 Houston (TX) 802 15.7 22.023.7 38.719.738.5 17.2 24.6 51.7 26.7 21.6 Yakima (WA) 119 17.7 35.326.9 20.225.248.5 17.5 8.7 58.0 26.9 15.1 Kanawha Valley (WV) 497 14.5 36.926.4 22.222.946.9 16.8 13.4 57.1 27.6 15.3 TOTAL SAMPLE 25932 9.9 30.726.9 32.6 18.442.9 17.8 20.9 54.3 27.5 18.2 *NOTE: All numbers reflect percentages

PAGE 134

120 Table 4.3: Pooled Sample Health Behavior and Outcome Factors Physical Activity Inactive 15.1% Some activity 40.1% Meets recommendations 44.8% Body Mass Index Normal 42.1% Overweight 36.1% Obese 21.8% Smoking Yes 23.6% No 76.4% Hypertension Presence 26.8% Absence 73.2% General Health Excellent 22.6% Very Good 34.3% Good 28.4% Fair 10.8% Poor 3.8% Days of Mental Distress per Month0 65.8% 1 3.5% 2 6.3% 3 3.3% 4-7 7.8% 8-15 6.0% > 15 7.3%

PAGE 135

121Table 4.4: Comparison of Health Behavior Frequencies Physical Activity Body Mass Index Smoke Inactive Some Activity Meets Recommendations Normal OverweightObese Yes No Birmingham (AL) 16.21% 41.89% 41.89% 39.17% 34.79% 26.04% 22.67% 77.33% Maricopa (AZ) 10.97% 37.53% 51.50% 45.01% 37.66% 17.33% 20.84% 79.16% Los Angeles (CA) 13.10% 44.28% 42.62% 46.87% 32.72% 20.41% 16.00% 84.00% San Diego (CA) 11.47% 40.88% 47.65% 40.29% 38.24% 21.47% 14.83% 85.17% San Francisco (CA) 10.00% 36.67% 53.33% 55.91% 30.11% 13.98% 20.00% 80.00% Boulder (CO) 5.83% 29.17% 65.00% 60.33% 32.23% 7.44% 14.52% 85.48% Denver (CO) 11.42% 38.81% 49.77% 56.94% 28.71% 14.35% 24.12% 75.88% Delaware (DE) 17.55% 41.83% 40.63% 40.67% 37.48% 21.85% 24.63% 75.37% Atlanta (GA) 19.77% 37.87% 42.36% 47.66% 32.47% 19.87% 19.00% 81.00% Indiana (IN) 14.13% 41.11% 44.76% 40.25% 35.59% 24.16% 27.20% 72.80% E. Baton Rouge (LA) 28.88% 37.23% 33.89% 43.99% 35.83% 20.18% 18.70% 81.30% Kalamazoo (MI) 11.76% 48.24% 40.00% 45.35% 39.53% 15.12% 17.98% 82.02% Southeast (MI) 14.70% 41.87% 43.43% 39.76% 34.22% 26.06% 23.92% 76.08% St.Paul (MN) 7.34% 42.46% 50.20% 44.23% 36.12% 19.66% 21.26% 78.74% Montana (MT) 16.68% 35.70% 47.62% 40.66% 37.98% 21.36% 23.90% 76.10% New Hampshire (NH) 11.89% 38.11% 50.00% 43.94% 36.58% 19.48% 23.61% 76.39% Central (NY) 19.79% 35.42% 44.79% 46.08% 31.37% 22.55% 23.81% 76.19% Rochester (NY) 14.19% 42.57% 43.24% 43.95% 36.31% 19.75% 24.54% 75.46% Winston-Salem (NC) 17.73% 43.64% 38.64% 43.02% 34.42% 22.56% 25.00% 75.00% Greensboro (NC) 18.16% 45.52% 36.32% 46.67% 31.54% 21.79% 22.68% 77.32% Cleveland (OH) 15.90% 43.86% 40.24% 38.76% 36.01% 25.23% 26.36% 73.64% Cinncinati (OH) 21.34% 40.62% 38.04% 41.76% 36.23% 22.01% 26.06% 73.94% Central (OR) 7.22% 35.05% 57.73% 44.33% 41.24% 14.43% 8.08% 91.92% York (PA) 10.08% 51.26% 38.66% 35.20% 38.40% 26.40% 22.83% 77.17% Houston (TX) 15.19% 42.67% 42.14% 41.78% 35.41% 22.81% 21.13% 78.87% Yakima (WA) 6.36% 41.82% 51.82% 34.51% 35.40% 30.09% 21.85% 78.15% Kanawha Valley (WV) 18.70% 38.70% 42.61% 34.04% 39.32% 26.64% 27.42% 72.58% Pooled Sample 15.07% 40.09% 44.84% 42.12% 36.10% 21.79% 23.58% 76.42%

PAGE 136

122 Health Factors: Outcome Variables A primary outcome of interest is a CV D-related factor, hypertension. In the overall sample, an average of 26.8% of participants reported that a professional has inform ed them that they have high blood pressure; the prevalence of self-reported hypertension ranged from 16.1% (Boulder) to 35.8% (Kanawha Valley). The five communities with the highest rates of hypertension are Kanawha Valley (WV), Birmingham (A L), Yakima (WA), York (PA), and Winston-Salem (NC). Table 4.5 compares each individual community’s rate of disease. Additional outcomes of interest incl ude general health and mental distress. Approximately 14.6% of the sample rate s their health as poor or fair. The communities with the highest frequency of residents reporting fair or poor health are: Kanawha Valley, WV (24.59%), Birmi ngham, AL (21.97%), and Denver, CO (19.30%). However, most residents report their general health status to be good to excellent (mean 3.61, sd 1.07); differ ences in distributions reflect a negative skew. Slight platykurtosis is observ ed. Descriptive statistics for all 27 communities are presented in Table 4.6.

PAGE 137

123Table 4.5: Comparison of Hypertension Outcome Frequency Community Prevalence of Hypertension1 Se 95% CI Birmingham (AL) 33.7% 0.0212(29.5%, 37.9%) Maricopa (AZ) 24.1% 0.0146(21.2%, 27%) Los Angeles (CA) 23.3% 0.0134(20.1%, 26%) San Diego (CA) 26.3% 0.0237(21.7%, 30.9%) San Francisco (CA) 17.9% 0.0393(10.2%, 25.6%) Boulder (CO) 16.1% 0.0330(9.6%, 22.6%) Denver (CO) 21.9% 0.0271(16.1%, 26.7%) Delaware (DE) 29.7% 0.0077(28.2%, 31.2%) Atlanta (GA) 21.4% 0.0161(18.2%, 24.6%) Indiana (IN) 26.9% 0.0070(25.5%, 28.3%) E. Baton Rouge (LA) 27.2% 0.0207(23.1%, 31.3%) Kalamazoo (MI) 24.7% 0.0457(15.7%, 33.7%) Southeast (MI) 28.7% 0.0115(26.5%, 31%) St.Paul (MN) 20.2% 0.0138(17.5%, 22.9%) Montana (MT) 29.9% 0.0079(28.4%, 31.5%) New Hampshire (NH) 23.7% 0.0067(22.4%, 25%) Central (NY) 28.6% 0.0439(20%, 37.2%) Rochester (NY) 22.0% 0.0324(15.7%, 28.3%) Winston-Salem (NC) 31.8% 0.0219(27.5%, 36.1%) Greensboro (NC) 26.6% 0.0217(22.3%, 30.9%) Cleveland (OH) 24.8% 0.0202(20.9%, 28.8%) Cinncinati (OH) 26.4% 0.0137(23.7%, 29.1%) Central (OR) 25.3% 0.0437(16.7%, 33.9%) York (PA) 31.8% 0.0414(23.7%, 39.9%) Houston (TX) 23.6% 0.0150(20.7%, 26.5%) Yakima (WA) 32.2% 0.0428(23.8%, 40.6%) Kanawha Valley (WV) 35.8% 0.0215(31.6%, 40%) Pooled Sample 26.8% 0.0143(24.0%, 29.6%) 1where higher number reflects hi gher frequency of hypertension

PAGE 138

124Table 4.6: Comparison of Descriptive Statistics of General Health Community Excellent Very Good Good Fair Poor Mean1 sd Skew Kurtosis Birmingham (AL) 19.56% 33.06% 25.40% 15.12% 6.85% 3.43 1. 16 -0.41 -0.66 Maricopa (AZ) 24.82% 32.71% 27.06% 11.18% 4.24% 3.63 1. 10 -0.50 -0.44 Los Angeles (CA) 24.05% 31.94% 28.44% 10.98% 4.59% 3.60 1. 10 -0.47 -0.44 San Diego (CA) 30.92% 34.68% 23.70% 9.83% 0.87% 3.85 1.00 -0.50 -0.58 San Francisco (CA) 26.32% 32.63% 24.21% 12.63% 4.21% 3.64 1. 13 -0.52 -0.51 Boulder (CO) 33.06% 39.52% 24.19% 2.42% 0. 81% 4.02 0.86 -0.57 0.02 Denver (CO) 20.18% 30.26% 30.26% 13.16% 6.14% 3.45 1. 14 -0.37 -0.55 Delaware (DE) 20.23% 34.71% 29.98% 11.57% 3.51% 3.57 1. 05 -0.41 -0.39 Atlanta (GA) 29.04% 37.73% 23.76% 7.45% 2.02% 3.84 0.99 -0.64 -0.07 Indiana (IN) 19.26% 34.39% 31.95% 10.13% 4.26% 3.54 1. 05 -0.43 -0.26 E. Baton Rouge (LA) 31.89% 35.36% 21.91% 8.46% 2.39% 3.86 1.04 -0.69 -0.14 Kalamazoo (MI) 20.22% 38.20% 31.46% 6.74% 3. 37% 3.65 0.99 -0.54 0.17 Southeast (MI) 20.04% 33.38% 30.35% 12.24% 3.99% 3.53 1. 07 -0.39 -0.44 St.Paul (MN) 22.87% 40.17% 27.01% 7.58% 2. 37% 3.74 0.97 -0.57 0.01 Montana (MT) 20.28% 34.02% 29.10% 12.15% 4.44% 3.54 1. 08 -0.43 -0.43 New Hampshire (NH) 28.06% 36.63% 25.01% 7.56% 2.73% 3.80 1.02 -0.64 -0.07 Central (NY) 25.71% 34.29% 30.48% 5.71% 3. 81% 3.72 1.03 -0.59 0.08 Rochester (NY) 21.95% 40.24% 25.00% 11.59% 1.22% 3.70 0. 98 -0.44 -0.44 Winston-Salem (NC) 20.84% 31.49% 29.27% 12.42% 5.99% 3.49 1. 13 -0.42 -0.51 Greensboro (NC) 21.84% 35.68% 24.03% 13.35% 5.10% 3.56 1. 12 -0.51 -0.49 Cleveland (OH) 23.58% 32.53% 26.20% 12.88% 4.80% 3.57 1. 12 -0.46 -0.54 Cinncinati (OH) 22.57% 30.67% 31.53% 11.67% 3.57% 3.57 1. 07 -0.35 -0.51 Central (OR) 28.28% 37.37% 24.24% 7.07% 3. 03% 3.81 1.03 -0.70 0.11 York (PA) 14.17% 38.58% 29.92% 13.39% 3.94% 3.46 1. 02 -0.43 -0.27 Houston (TX) 24.13% 28.25% 29.38% 15.38% 2.88% 3.55 1. 10 -0.26 -0.80 Yakima (WA) 21.01% 27.73% 33.61% 14.29% 3.36% 3.49 1. 08 -0.21 -0.64 Kanawha Valley (WV) 17.54% 30.04% 27.82% 15.32% 9.27% 3.31 1. 20 -0.33 -0.74 Pooled Sample 22.63% 34.34% 28.44% 10.75% 3.84% 3.61 1.07 -0.47 -0.36 1where higher number indicates better self-reported health

PAGE 139

125 Table 4.7: Comparison of Descriptive Statistics of Days of Mental Distress per Month Community 0 1 2 3 4-7 8-15 >15 Mean1 sd Skew Kurtosis Birmingham (AL) 60.41% 4.08% 6.94% 4.08% 9.60% 4.90% 9.98% 4.06 8.14 2.27 4.03 Maricopa (AZ) 63.63% 3.91% 7.11% 4.62% 8.41% 6.17% 6.17% 3.28 7.19 2.76 6.94 Los Angeles (CA) 62.34% 3.60% 7.49% 4.10% 9.40% 5.90% 7.20% 3.48 7.40 2.63 6.12 San Diego (CA) 66.96% 2.61% 7.25% 3.77% 8.12% 5.80% 5.51% 3.01 6.94 2.94 8.10 San Francisco (CA) 49.47% 3.16% 3.16% 10.53% 13.69% 9.47% 10.52% 5.32 8.80 1.96 2.81 Boulder (CO) 60.16% 5.69% 9.76% 4.88% 10.57% 4.87% 4.06% 2.65 5.68 3.03 9.49 Denver (CO) 59.47% 4.85% 7.05% 3.08% 8.81% 7.92% 8.81% 4.14 8.17 2.28 4.16 Delaware (DE) 67.55% 3.72% 5.30% 3.05% 7.51% 5.21% 7.64% 3.44 7.78 2.61 5.69 Atlanta (GA) 62.75% 3.76% 6.42% 2.66% 8.45% 8.77% 7.20% 3.80 7.69 2.39 4.89 Indiana (IN) 62.05% 3.83% 6.64% 3.37% 9.05% 7.10% 8.01% 3.80 7.81 2.43 4.91 E. Baton Rouge (LA) 72.03% 2.86% 5.51% 1.98% 5.72% 5.94% 5.94% 2.93 7.09 2.83 7.19 Kalamazoo (MI) 67.05% 3.41% 7.95% 3.41% 10.23% 4.55% 6.82% 2.77 6.37 3.00 8.86 Southeast (MI) 59.22% 3.40% 7.78% 3.73% 8.89% 8.37% 8.65% 4.18 8.14 2.25 4.02 St.Paul (MN) 54.46% 7.47% 7.95% 5.18% 10.84% 7.58% 6.50% 3.64 7.04 2.52 5.81 Montana (MT) 76.14% 1.88% 4.31% 1.85% 5.52% 4.45% 5.85% 2.70 7.10 3.03 8.24 New Hampshire (NH) 66.42% 3.53% 7.03% 3.40% 7.51% 5.52% 6.64% 3.15 7.21 2.80 7.03 Central (NY) 60.38% 6.60% 10.38% 0.94% 8.49% 6.60% 6.60% 3.40 7.41 2.76 6.95 Rochester (NY) 61.96% 4.29% 7.36% 2.45% 8.59% 7.36% 7.97% 3.51 6.97 2.34 4.82 Winston-Salem (NC) 74.55% 1.79% 4.46% 3.57% 5.13% 3.12% 7.36% 2.93 7.57 2.92 7.29 Greensboro (NC) 71.22% 2.73% 4.71% 2.23% 6.70% 4.96% 7.45% 3.21 7.63 2.73 6.39 Cleveland (OH) 60.79% 4.41% 6.17% 5.51% 8.37% 7.04% 7.71% 3.71 7.60 2.48 5.25 Cinncinati (OH) 66.86% 2.63% 4.97% 3.31% 6.82% 6.63% 8.77% 3.85 8.27 2.39 4.48 Central (OR) 66.33% 5.10% 8.16% 3.06% 5.10% 3.06% 9.18% 3.53 8.45 2.64 5.57 York (PA) 53.17% 3.17% 11.90% 0.79% 12.69% 6.35% 11.90% 3.56 7.53 2.53 5.56 Houston (TX) 65.70% 4.18% 7.59% 4.30% 6.96% 5.57% 5.70% 2.95 6.83 2.94 8.08 Yakima (WA) 70.34% 2.54% 1.69% 2.54% 6.77% 8.47% 7.62% 3.68 7.86 2.38 4.80 Kanawha Valley (WV) 62.70% 1.84% 5.12% 2.87% 7.58% 6.96% 12.89% 5.09 9.53 1.86 1.98 Pooled Sample 65.77% 3.48% 6.26% 3.29 % 7.82% 6.03% 7.32% 3.45 7.59 2.60 5.79 1where higher number indicates higher mean number of days in past month self-reported mental health not good

PAGE 140

126 In addition to general health status, mental distress was evaluated. As with the other outcome variables, there is only a very small amount of data missing (less than 2%). The overall samp le reports a mean of 3.45 (range 2.65 – 5.32) days in the past m onth when their mental health was not good. The distributions of every community are positively skewed and leptokurtic (Table 4.7). Communities with the highest frequencies of residents who report their mental health is not good half of every m onth or more are: Kanawha Valley, WV (12.9%), York, PA (11.9%), and San Fr ancisco, CA (10.5%). Table 4.8 compares the relative ranking of the co mmunities in terms of each outcome. There is some consistency in the ranking of communities with respect to all three outcomes at the extreme ends of the dist ributions (which co mmunities have the worst health and which have the best). Howeve r, there is not a stable pattern of ranking the burden of disease towards the middle of the distributions of investigated outcomes. Social Contextual Factors Descriptive statistics of both pooled and community-specific individual indices are presented in Table 4.9. Ther e is very little missing data for any one index, with less than 1% for any community on every index. Sc ores for most of the indices have been standardized on nat ional norms by the original investigators. Therefore, it is cons istent that the comm unities’ pooled mean

PAGE 141

127 levels of social trust (0.02, 0.69), in formal social engagement (-0.002, 0.64), and organizational activism or formal soci al engagement (0.068, 1.036) remain around zero, with a standard deviation just under one. As the mutual aid index is more of a count-based dimension, similar standardization has not been used (pooled mean 5.243, sd 4. 312). Descriptions of the differences between individual communities are found in examinin g the type of distribution of scores. Overall, the shape of the distribution of scores for the social trust index pooled across communities is slightly negatively skewed (-0.892) and leptokurtic (0.517). Individual communities distributions range from moderate deviations (New Hampshire and Boulder, CO) to relatively normal shape (Houston, TX). The shape of pooled scores for informal social participation is s lightly positively skewed (0.87) and barely leptokurtic (0 .177). There are only very small distinctions between community-level dist ributions, with Atlanta, GA the most positively skewed (1.05) and slightly l eptokurtic (0.48). The communities’ organizational activism score distributions are very similar, with all communities both positively skewed and sharply peaked, which reflects a concentration of lower scores indicating less overall engagement in formal participation. Mutual aid is also consistent with respect to skewness, with Los Angeles the most deviant (positive skew .99). The peakedne ss of the distributions, however, are more diverse, ranging from flatter (Win ston-Salem, NC) to more peaked (Yakima, WA).

PAGE 142

128 Table 4.8: Comparison of Community Rankings on Outcome Variables Community Hypertension1 General Health2 Mental Distress3 Birmingham (AL) 2nd 2nd 5th Maricopa (AZ) 18th 16th 18th Los Angeles (CA) 21st 15th 15th San Diego (CA) 14th 25th 21st San Francisco (CA) 26th 17th 1st Boulder (CO) 27th 27th 27th Denver (CO) 23rd 3rd 4th Delaware (DE) 7th 12th 16th Atlanta (GA) 24th 24th 8th Indiana (IN) 11th 9th 7th E. Baton Rouge (LA) 10th 26th 24th Kalamazoo (MI) 17th 18th 25th Southeast (MI) 8th 7th 3rd St.Paul (MN) 25th 21st 11th Montana (MT) 6th 8th 26th New Hampshire (NH) 19th 22nd 20th Central (NY) 9th 20th 17th Rochester (NY) 22nd 19th 14th Winston-Salem (NC) 4th 6th 23rd Greensboro (NC) 12th 11th 19th Cleveland (OH) 16th 14th 9th Cinncinati (OH) 13th 13th 6th Central (OR) 15th 23rd 13th York (PA) 5th 4th 12th Houston (TX) 20th 10th 22nd Yakima (WA) 3rd 5th 10th Kanawha Valley (WV) 1st 1st 2nd 1where higher number reflects hi gher frequency of hypertension 2where higher number indicates worse self-reported health 3where higher number indicates higher mean number of da ys in past month self-reported mental health not good

PAGE 143

129Table 4.9: Comparison of Social Capital Descriptive Statistics Social Trust1 Informal Participation1 Organizational Activism1 Mutual Aid1 Mean sd Skew Kurtosis Mean sd Skew Kurtosis Meansd Skew Kurtosis Meansd Skew Kurtosis Birmingham (AL) 1.99 0.81 0.02 -1.49 1.98 0.81 0.04 -1.47 2.04 0.84 -0.08-1.57 2.08 0.80 -0.15-1.42 Maricopa (AZ) 1.99 0.83 0.02 -1.54 2.01 0.81 -0.02 -1.46 1.94 0.83 0.11 -1.55 1.97 0.82 0.05 -1.51 Los Angeles (CA) 1.74 0.80 0.50 -1.28 1.89 0.83 0.20 -1.51 1.93 0.85 0.14 -1.61 1.96 0.82 0.08 -1.51 San Diego (CA) 1.98 0.83 0.04 -1.54 1.96 0.81 0.07 -1.49 2.01 0.83 -0.03-1.56 1.97 0.80 0.05 -1.43 San Francisco (CA) 1.96 0.83 0.08 -1.54 2.04 0.82 -0.08 -1.52 2.07 0.82 -0.12-1.51 2.02 0.80 -0.04-1.42 Boulder (CO) 2.27 0.76 -0.50 -1.11 2.10 0.81 -0.18 -1.44 2.17 0.82 -0.31-1.46 2.10 0.81 -0.19-1.47 Denver (CO) 1.94 0.81 0.11 -1.46 2.00 0.81 0.00 -1.49 2.05 0.87 -0.09-1.67 2.05 0.81 -0.10-1.46 Delaware (DE) 2.06 0.82 -0.11 -1.51 2.02 0.80 -0.04 -1.45 2.07 0.85 -0.13-1.62 2.05 0.83 -0.10-1.55 Atlanta (GA) 1.82 0.83 0.34 -1.45 1.90 0.82 0.19 -1.49 2.08 0.81 -0.14-1.45 2.16 0.80 -0.29-1.39 Indiana (IN) 2.14 0.81 -0.25 -1.44 2.13 0.78 -0.24 -1.33 1.99 0.85 0.03 -1.63 2.05 0.81 -0.08-1.45 E. Baton Rouge (LA) 1.86 0.80 0.25 -1.41 2.09 0.83 -0.16 -1.51 2.17 0.81 -0.31-1.42 2.18 0.78 -0.33-1.30 Kalamazoo (MI) 2.16 0.82 -0.31 -1.45 2.19 0.80 -0.37 -1.36 2.06 0.83 -0.12-1.55 2.12 0.84 -0.22-1.56 Southeast (MI) 1.96 0.84 0.08 -1.58 2.09 0.83 -0.16 -1.53 2.02 0.84 -0.03-1.57 2.05 0.80 -0.09-1.42 St.Paul (MN) 2.30 0.77 -0.57 -1.11 2.01 0.81 -0.02 -1.45 2.04 0.82 -0.07-1.50 2.19 0.81 -0.35-1.40 Montana (MT) 2.35 0.75 -0.67 -0.92 2.09 0.79 -0.16 -1.37 2.10 0.86 -0.19-1.61 2.09 0.83 -0.17-1.53 New Hampshire (NH)2.33 0.76 -0.64 -1.00 2.06 0.80 -0.11 -1.43 2.06 0.84 -0.12-1.58 2.06 0.82 -0.12-1.51 Central (NY) 2.10 0.81 -0.18 -1.47 2.11 0.82 -0.20 -1.49 2.03 0.85 -0.06-1.61 2.10 0.80 -0.18-1.43 Rochester (NY) 2.01 0.83 -0.01 -1.56 2.03 0.80 -0.06 -1.45 1.92 0.86 0.15 -1.63 2.00 0.80 0.00 -1.45 Winston-Salem (NC) 1.98 0.83 0.04 -1.53 1.89 0.82 0.20 -1.47 2.04 0.84 -0.08-1.56 2.18 0.81 -0.34-1.40 Greensboro (NC) 1.96 0.82 0.08 -1.52 1.96 0.81 0.08 -1.49 2.06 0.85 -0.12-1.61 2.15 0.81 -0.28-1.42

PAGE 144

130Table 4.9: Comparison of Social Capital Descriptive Statistics Social Trust1 Informal Participation1 Organizational Activism1 Mutual Aid1 Mean sd Skew Kurtosis Mean Sd Skew Kurtos is Mean sd SkewKurtosis Meansd SkewKurtosis Cleveland (OH) 1.90 0.83 0.19 -1.52 2.00 0.83 0.00 -1.53 1.99 0.84 0.02 -1.58 1.94 0.82 0.12 -1.50 Cinncinati (OH) 2.08 0.80 -0.15 -1.42 2.04 0.82 -0.07 -1.49 2.01 0.84 -0.02-1.58 2.06 0.81 -0.10-1.46 Central (OR) 2.18 0.79 -0.33 -1.34 2.01 0.80 -0.03 -1.43 2.04 0.82 -0.07-1.51 1.93 0.82 0.13 -1.50 York (PA) 2.18 0.78 -0.32 -1.29 2.06 0.81 -0.11 -1.48 1.92 0.83 0.15 -1.54 2.04 0.82 -0.08-1.52 Houston (TX) 1.75 0.81 0.50 -1.30 1.87 0.83 0.24 -1.50 1.84 0.82 0.30 -1.45 1.89 0.82 0.20 -1.48 Yakima (WA) 1.99 0.82 0.02 -1.51 2.01 0.85 -0.01 -1.61 1.91 0.86 0.17 -1.64 1.92 0.84 0.15 -1.55 Kanawha Valley (WV)2.07 0.85 -0.13 -1.60 2.03 0.79 -0.06 -1.39 2.00 0.86 0.00 -1.64 1.99 0.81 0.02 -1.47 Mean 2.04 0.81 -0.07 -1.40 2.02 0.81 -0.04 -1.47 2.02 0.84 -0.04-1.56 2.05 0.81 -0.09-1.46 1where 1 = low, 2 = moderate, 3 = high

PAGE 145

131Table 4.10: Community Ranking on Social Capital Variables Informal Rank Social Trust Social Participation Organizational Activism Mutual Aid 1 Montana MT Kalamazoo MI Boulder CO Winston-Salem NC 2 New Hampshire NH Indiana IN East Baton Rouge LA Greensboro NC 3 St.Paul MN East Baton Rouge LA Greensboro NC East Baton Rouge LA 4 Boulder CO Central New York NY Montana MT St.Paul MN 5 York PA SE Michigan MI Central Oregon OR Atlanta GA 6 Central Oregon OR Boulder CO Birmingham AL Kalamazoo MI 7 Kalamazoo MI Montana MT Delaware DE Boulder CO 8 Indiana IN Yakima WA New Hampshire NH Birmingham AL 9 Cinncinati OH New Hampshire NH San Francisco CA Delaware DE 10 Central New York NY Cinncinati OH Kalamazoo MI Montana MT 11 Delaware DE Central Oregon OR Denver CO Cinncinati OH 12 Kanawha Valley WV San Francisco CA Atlanta GA New Hampshire NH 13 Winston-Salem NC Rochester NY St.Paul MN Central New York NY 14 Rochester NY Arizona AZ Central New York NY SE Michigan MI 15 Birmingham AL York PA Winston-Salem NC Denver CO 16 Arizona AZ Kanawha Valley WV Cleveland OH Indiana IN 17 SE Michigan MI Denver CO Kanawha Valley WV York PA 18 Yakima WA Delaware DE Cinncinati OH San Francisco CA 19 Greensboro NC Cleveland OH SE Michigan MI Rochester NY 20 San Diego CA St.Paul MN Indiana IN Kanawha Valley WV 21 San Francisco CA Birmingham AL San Diego CA Arizona AZ 22 Denver CO Greensboro NC Ro chester NY Los Angeles CA 23 East Baton Rouge LA San Diego CA Yakima WA Central Oregon OR 24 Cleveland OH Atlanta GA Los Angeles CA San Diego CA 25 Atlanta GA Winston-Salem NC York PA Cleveland OH 26 Houston TX Los Angeles CA Arizona AZ Yakima WA 27 Los Angeles CA Houston TX Houston TX Houston TX Comparisons of the communities on t he four indices related to social capital are displayed in Table 4.10, in which the 27 communities are ranked. The community with the most social trust is Montana, followed by New Hampshire and St.Paul (MN). Residents in Kalamaz oo (MI), Indiana, and East Baton Rouge (LA) are the most active in informal social engagement whereas organizational activism or formal social engagement is at its greatest in Boulder (CO), East

PAGE 146

132 Baton Rouge (LA), and Greens boro (NC). Those living in Winston-Salem (NC), Greensboro (NC), and East Baton Rouge (LA) report the highest amount of mutual aid within their communities. Houst on (TX) is consistently last on almost every dimension of social capital in this study. Social Structural Factors Social structural variables, including income inequality (Gini Coefficient) and absolute deprivation (200% below FPL), are reported at the state-level for each community in Table 4.11. The mean Gini across communities is .453 (range .414 .499), indicating substantively significant relative deprivation. The shape of the distribution is both posit ively skewed and platykurtic and is illustrated in Figure 4.1. The most in come inequality is observed in New York, Louisiana, and California, with the least in New Hampshire. In regards to absolute deprivation, approxim ately 30% of residents withi n the sample live at or below 200% FPL on average. Figure 4.2 displays the distribution of poverty rates. Communities range from the mo st impoverished areas, Louisiana (40.4%) and West Virginia (40.3%), to least depr ived locales, New Hampshire (19%) and Minnesota (21.6%).

PAGE 147

133 Table 4.11: Comparison of Social Structural Characteristics Gini Coefficient Percent Below 200% FPL Median HH Income % Unemployed % Completed High School or Less Birmingham (AL) 0.475 36.1 $34,135 16.9 55.1 Maricopa (AZ) 0.450 33.5 $40,558 15.2 43.3 Los Angeles (CA) 0.475 33.1 $47,493 19.5 43.3 San Diego (CA) 0.475 33.1 $47,493 19.5 43.3 San Francisco (CA) 0.475 33.1 $47,493 19.5 43.3 Boulder (CO) 0.438 24.2 $47,203 13.2 36.2 Denver (CO) 0.438 24.2 $47,203 13.2 36.2 Delaware (DE) 0.429 23.2 $47,381 15.1 48.8 Atlanta (GA) 0.461 30.5 $42,433 15.7 50.1 Indiana (IN) 0.424 25.8 $41,567 14.0 55.1 E. Baton Rouge (LA) 0.483 40.4 $32,566 18.4 57.6 Kalamazoo (MI) 0.440 25.4 $44,667 16.7 47.8 Southeast (MI) 0.440 25.4 $44,667 16.7 47.8 St.Paul (MN) 0.426 21.6 $47,111 13.7 40.9 Montana (MT) 0.436 37.1 $33,024 18.6 44.2 New Hampshire (NH) 0.414 19.0 $49,467 14.0 42.7 Central (NY) 0.499 30.5 $43,393 20.9 48.7 Rochester (NY) 0.499 30.5 $43,393 20.9 48.7 Winston-Salem (NC) 0.452 30.5 $39,184 15.9 50.2 Greensboro (NC) 0.452 30.5 $39,184 15.9 50.2 Cleveland (OH) 0.441 26.4 $40,956 14.2 53.2 Cinncinati (OH) 0.441 26.4 $40,956 14.2 53.2 Central (OR) 0.438 29.6 $40,916 18.5 41.1 York (PA) 0.452 27.4 $40,106 17.5 56.1 Houston (TX) 0.470 36.0 $39,927 16.5 49.1 Yakima (WA) 0.436 25.9 $45,776 18.3 37.8 Kanawha Valley (WV) 0.468 40.3 $29,696 18.3 64.2

PAGE 148

134 Figure 4.1: Gini Coefficient Distribution0 1 2 3 4 5 6 7 8 9 10 0.420.440.460.480.50 Gini CoefficientFrequency

PAGE 149

135 Figure 4.2: Below 200% FPL Distribution0 2 4 6 8 10 12 2025303540 Percent Below 200% FPLFrequency

PAGE 150

136 Bivariate Analysis Relationships between sociodemographi c characteristics of the sample are presented in Table 4.12. In general, the majority of associations between factors are statistically signi ficant at p<.0001 level. Specifically, men and women statistically significantly differ on almo st all other factor s (p<.001), with the exception of the gender distri bution of Hispanics. Racial/ethnic differences are also apparent with respect to age, education, income and marital status (p<.0001). There is also a consistent pattern of statistically significant associations between sociodemographic factors and t he frequencies of health behaviors (Table 4.13). Those engaging in health behaviors such as physical activity or diet differ significantly based upon personal characteristics. For example, both education and income are positively co rrelated with physical activity (rsp = .14, p < .0001 for both), with thos e who are more educated and have more income reporting more activity. A statistically significant negative association is reported for the same factors and Body Mass I ndex, with the more educated and wealthier individuals reporting lower Body Mass Index (rsp -.09, p < .0001 and rsp = -.06, p < .0001, respectively). In regards to smoking behavior, significant associations are observed with all factors, except for race/ethnicity. Generally, smokers tend to be white, male, younger, less educated, have less income, and not married. The majority of the sample engages in multiple risk behaviors, with only 21.3% both active and not overweight. In addition, slightly more smokers than

PAGE 151

137 non-smokers are overweight or obese (52.7% vs. 47.4%) and do not meet recommended physical activity levels (55.4% vs. 44.6%). Alt hough there is some evidence for the clustering effect of behav ior in this sample, the associations among risk factors are not all positive (Tabl e 4.14). Specifically, those who are inactive have higher Body Mass Index (rsp = -.12, p < .0001) however are less likely to smoke with an OR 0.95 (95% CI 0.91 0.99). Mo reover, smokers tend to have lower Body Mass Index with OR 0.82 (95% CI 0.79 0.86). The sociodemographic distribution of hypertension is reported in Table 4.15. There are no gender differences, with both men and women reporting little more than a quarter (26.8%) suffer fr om hypertension. As expected, hypertension is more prevalent in older individuals (52.6%). The racial/ethnic spread demonstrated disparities: over one-third of Blacks report hypertension whereas only about one-quarter of Whites do. Two-thirds (66.3%) of those reporting hypertension have a high school diploma or less and over one-third (36%) make $20,000 or less per year, whic h also is consistent with current literature. The distribution of risk behavior amongst those with hypertension demonstrates a trend evident in the rela tionship between hy pertension and both physical activity and Body Mass Index (T able 4.16). Indivi duals who reported having hypertension also reported that t hey were less active and had higher Body Mass Index on average than thos e who reported no hypertension.

PAGE 152

138 Table 4.12: Select Sociodemographic by Sociodemographic Frequencies Gender Age Male Female 20 3435 4445 6465+ Gender Male 25.4% 23.6% 34.7%16.4% Female 24.6% 21.7% 33.2%20.5% Age 2 = 69.38 df3 p<.0001 20 – 34 41.4%58.6% 35 – 44 42.7%57.3% 45 – 64 41.7%58.3% 65 + 35.4%64.6% Race/Ethnicity p < .0001 White 41.1%58.9% 2 = 10.6 df1 p<.0011 22.8% 22.2% 34.7%20.3% Black 33.0%67.0% 2 = 65.58 df1 p<.0001 32.9% 23.6% 30.6%12.9% Other 46.4%53.6% 2 = 23.77 df1 p<.0001 39.9% 23.6% 27.0%9.4% Hispanic 40.4%59.6% NS 47.2% 23.6% 22.8%6.4% Education 2 = 87.27 df3 p<.0001 p < .0001 >12 41.8%58.2% 20. 2% 16.2% 27.9%35.7% 12 38.9%61.0% 23.0% 22.1% 32.8%22.1% 13-15 37.7%62.3% 26.6% 22.5% 34.4%16.4% 16+ 44.5%55.5% 26.6% 24.6% 35.9%12.9% Income 2 = 340.55 df3 p<.0001 p < .0001 <$20k 30.6%69.4% 26.4% 14.9% 28.1%30.7% $20<$50k 41.5%58.5% 29.2% 22.2% 31.2%17.5% $50<$75k 45.9%54.1% 27.6% 28.4% 36.7%7.3% $75k+ 49.3%50.7% 19.4% 30.3% 43.7%6.6% Marital Status 2 = 559.36 df2 p<.0001 p < .0001 Married 44.1%55.9% 20.4% 25.5% 38.5%15.6% Separated/Divorced/Widowed 29. 0%71.0% 8.8% 17.3% 38.1%35.9% Never Married 48.2%51.8% 61.1% 20.8% 14.1%4.1% 1Hispanic vs. non-Hispanic NOTE: Rows within factors sum to 100%

PAGE 153

139 Table 4.12: Select Sociodemographic by Sociodemographic Frequencies Race/Ethnicity Education Hispanic1WhiteBlackOther <12 12 13 1516+ Gender Male 5.2% 85.0%7.8% 7.3% 10.1%29.4% 24.9%35.6% Female 5.3% 83.5%10.8%5.8% 9.7% 31.6% 28.2%30.5% Age 20 34 9.8% 77.4%12.5%10.1% 7.8% 28.0% 28.8%35.4% 35 44 5.4% 83.4%9.9% 6.6% 7.0% 29.8% 27.0%36.3% 45 64 3.5% 86.4%8.6% 5.0% 8.0% 29.4% 27.5%35.2% 65 + 1.8% 90.4%6.4% 3.1% 18.3%35.6% 23.5%22.7% Race/Ethnicity p < .0001 White 8.7% 30.8% 26.8%33.6% Black 14.6%33.4% 28.7%23.4% Other 17.7%26.2% 25.0%31.2% Hispanic 31.7%28.3% 23.4%16.6% Education >12 16.8% 74.4%14.2%11.5% 12 4.8% 84.2%10.4%5.4% 13-15 4.6% 83.9%10.2%5.9% 16+ 2.7% 87.0%6.9% 6.1% Income <$20k 10.6% 74.3%15.6%10.0% 24.8%39.5% 24.3%11.4% $20<$50k 5.2% 83.1%10.8%6.2% 8.6% 36.1% 29.6%25.6% $50<$75k 3.6% 88.5%6.4% 5.1% 2.4% 23.7% 30.0%43.9% $75k+ 2.7% 90.2%5.0% 4.9% 1.0% 14.2% 22.6%62.2% Marital Status p < .0001 Married 5.1% 88.6%5.6% 5.9% 7.7% 29.9% 25.8%36.7% Separated/Divorced/Wi dowed 3.9% 83.7%11.1%5.3% 14.4%33.6% 28.3%23.8% Never Married 6.3% 72.5%19.1%8.4% 9.2% 28.7% 28.2%34.0% 1Hispanic vs. non-Hispanic NOTE: Rows within factors sum to 100%

PAGE 154

140Table 4.12: Select Sociodemographic by Sociodemographic Frequencies Income Marital Status <$20k$20<$50k$50<$75k$75k+ Ma rried Separated/Widowed/Di vorcedNever Gender Male 13.5%42.5% 19.5% 24. 6% 58.8% 19.6% 21.6% Female 22.0%43.2% 16.6% 18.2% 51.2% 32.9% 15.9% Age 20 34 18.3%47.6% 18.7% 15. 4% 47.1% 10.2% 42.8% 35 44 11.5%40.2% 21.5% 26. 8% 62.9% 21.5% 15.6% 45 64 15.0%39.1% 19.2% 26. 7% 62.1% 31.1% 6.9% 65 + 35.5%47.5% 8.3% 8.7% 44.6% 51.9% 3.6% Race/Ethnicity p < .0001 White 16.3%42.5% 18.8% 22.4% 57.0% 27.3% 15.7% 2 = 664.07 df2 p<.0001 Black 29.7%47.7% 11.8% 10.7% 31.6% 32.0% 36.5% 2 = 746.93 df2 p<.0001 Other 28.8%41.3% 14.1% 15.9% 51.8% 23.4% 24.8% 2 = 51.22 df2 p<.0001 Hispanic 36.1%41.6% 11.8% 10.5% 55.2% 21.7% 23.1% 2 = 33.11 df2 p<.0001 Education p < .0001 >12 51.3%41.5% 4.8% 2.4% 42.6% 40.3% 17.1% 12 24.2%51.8% 14.1% 9.9% 52.9% 30.1% 17.0% 13-15 16.4%46.7% 19.6% 17. 4% 52.0% 28.9% 19.1% 16+ 6.2%32.5% 23.0% 38.3% 61.0% 20.0% 19.0% Income <$20k 23.5% 50.1% 26.4% $20<$50k 47.6% 31.3% 21.1% $50<$75k 70.7% 15.7% 13.6% $75k+ 82.1% 8.7% 9.2% Marital Status p < .0001 Married 7.9%37.4% 23.1% 31.6% Separated/Divorced/Widowed 33.7% 49.3% 10.3% 6.7% Never Married 26.5%49.6% 13.3% 10.6% 1Hispanic vs. non-Hispanic NOTE: Rows within factors sum to 100%

PAGE 155

141Table 4.13: Sociodemographic Factors by Health Behavior Frequencies Physical Activity BMI Smoking None SomeMeets NormalOverweight Obese Yes No Gender p < .0001 p < .0001 2 = 34.85 df 1 p<.0001 Male 14.4 38.1 47.5 32. 3 45.6 22.2 25.5 74.5 Female 15.6 41.5 43.0 49. 3 29.2 21.5 22.3 77.7 Age rsp = -.13, p<.0001 rsp = .08, p<.0001 OR 0.75 (0.73, 0.77) 20 – 34 10.9 38.8 50.3 50. 7 31.4 17.9 28.3 71.7 35 – 44 11.6 40.9 47.5 41. 6 36.0 22.4 29.0 71.0 45 – 64 14.8 42.4 42.8 34. 7 38.6 26.7 23.1 76.9 65 + 26.3 37.5 36.2 41.2 39.5 19.3 11.7 88.3 Race/Ethnicity p < .0001 White 13.7 39.9 46.4 43. 4 36.2 20.4 23.4 76.6 Black 23.7 42.3 34.0 29.4 37. 1 33.5 p < .0001 23.4 76.6 NS Other 19.4 39.7 40.9 43. 8 33.5 22.7 NS 26.9 73.2 OR 1.2 (1.07, 1.35) Hispanic 20.0 39.0 41.1 36. 6 37.3 26.1 p < .0001 20.6 79.4 2 = 7.15 df 1 p<.0075 Education rsp = .14, p<.0001 rsp = -.09, p<.0001 OR 0.64 (0.62, 0.66) >12 31.2 35.1 33.7 37.1 35.1 27.9 36.0 64.0 12 18.1 40.3 41.7 39.3 36.6 24.2 30.0 70.0 13-15 14.1 41.0 44.9 41. 8 35.2 23.1 24.7 75.3 16+ 8.5 40.7 50.9 46.5 36.7 16.8 12.8 87.2 Income rsp = .14, p<.0001 rsp = -.06, p<.0001 OR 0.72 (0.70, 0.74) <$20k 25.8 37.1 37.1 40. 2 32.1 27.7 32.4 67.6 $20<$50k 14.3 41.4 44.4 40. 5 36.4 23.1 27.0 73.1 $50<$75k 9.1 42.2 48.7 40. 1 38.6 21.2 21.7 78.3 $75k+ 7.0 40.7 52.3 45.1 37.6 17.4 14.9 85.1 Marital Status p < .0001 p < .0001 2 = 395.94 df 2 p<.0001 Married 13.0 41.2 45.7 39. 3 38.5 22.2 18.5 81.5 Separated/Widowed/Divorced 21.0 39. 2 39.9 42.3 35.5 22. 2 28.2 71.9 Never Married 13.0 38.8 48.2 49. 3 30.3 20.4 29.4 70.6 Note: all numbers reflect percentages

PAGE 156

142 Table 4.14: Health Behavior by Health Behavior Frequencies* Physical Activity BMI Smoking None Some Meets Normal Overweight Obese Yes No Physical Activity None 35.0% 34.6% 30.5% 26.3% 73.7% Some 38.9% 36.7% 24.4% 22.7% 77.3% Meets Recommendations 47. 0% 36.1% 17.0% 23.4% 76.6% BMI Normal 12.2% 37.1% 50.7% rsp = -.12, p < .0001 26.8% 73.2% Overweight 14.1% 40.7% 45.2% 22.6% 77.4% Obese 20.4% 44.6% 35.0% 20.1% 79.9% Smoking Yes 16.8% 38.6% 44.6% OR .95 (.91, .99) 47.4% 34.4% 18.4% OR .82 (.79, .86) No 14.5% 40.5% 45.0% 40.5% 36.7% 22.9% *rows within behaviors sum to 100%

PAGE 157

143 Table 4.15: Sociodemographic Factor s by Hypertension Frequencies Hypertension Gender Male 26.77% Female 26.81% Age 20 – 34 9.16% 35 – 44 16.14% 45 – 64 33.70% 65 + 52.62% Race/Ethnicity White 26.33% Black 34.57% Other 21.45% Hispanic 17.23% Education >12 36.36% 12 29.92% 13-15 26.58% 16+ 21.04% Income <$20k 36.04% $20<$50k 26.62% $50<$75k 21.62% $75k+ 19.05% Marital Status Married 25.80% Separated/Widowed/Divorced 37.83% Never Married 15.08% Table 4.16: Health Behavior by Hypertension Frequencies Hypertension Yes No Physical Activity None 37.1% 62.9% Some 27.0% 73.0% Meets Recommendations 22.4% 77.6% BMI Not Overweight 16.5% 83.5% Overweight 28.8% 71.2% Obese 42.7% 57.3% Smoking Yes 22.0% 78.0% No 28.3% 71.7%

PAGE 158

144 The sociodemographic distribution of gener al health status is reported in Table 4.17. In regards to racial/ethnic di vision of responses, Hispanics report fair or poor health most often (22.3%), follo wed by Blacks (20.1%), Others (19.1%), and Whites (13.6%). T he economic breakdown of responses displays a wellestablished trend – as income increas es, so does reported general health. Expected patterns in general health are r eported across health behaviors as well (Table 4.18). In regards to physical activity, almost on e-third of those who meet current recommendations for activity report excellent gener al health (whereas only just over one-tenth of non-active indi viduals do). The majo rity of those who are not overweight (65.4%) report their general health as very good or excellent while fewer obese individuals report t he same (41.4%). The distribution of sociodemogra phic factors and reported days when mental health not good is displayed in T able 4.19. Men consistently have fewer days per month of poor mental heal th, with those who are younger, less educated, and poorer having the most. For example, only slightly more than one-third as many higher income individuals (4.4%) as lowest income individuals (12.8%) report more than hal f the month with mental di stress. In regards to behavior and mental distress, a pattern emerges whereby there are similar frequencies across all health behaviors for t he first half of the month, after which differences in behavior become more evident (Table 4.20). While all levels of physical activity have similar number of no poor mental health days/month (64%66%), differences exist especially among those reporting more than half the

PAGE 159

145 month with mental distress. These dist inctions are most evident between those who engage in no activity (10.9%) or some activity (6.8%). Of the same group who report at least two wee ks of mental distress, ther e are more obese persons (10.2%) compared to normal we ight individuals (6.7%). In regards to smoking, although nonsmokers report 10% more days of no poor mental health (68.5%) compared to smokers (57%), there are over twice as many nonsmokers (12.5%) as smokers (5.8%) who report 15 or more days per month their mental health is not good. Table 4.21 displays the relative frequenc ies of all three outcomes is this study. There is a consistency where t hose who suffer from hypertension report poor general health more often than those wi thout hypertension. An interesting finding is that the frequencie s of reported days when mental health is not good is relatively similar whether one suffers from hypertension or not. Of the individuals who report excellent health approximately 75% also report no days/month of mental distress (Table 4.22). Alternativel y, of the individuals who report poor general health, approximately 31% repor t poor mental health at least two weeks/month.

PAGE 160

146Table 4.17: Sociodemographic Factors by General Health Status Frequencies General Health poor fair good very good excellent Gender Male 3.4% 9.9% 28.1% 34.6% 24.0% Female 4.1% 11.3% 28.7% 34.2% 21.7% Age 20 34 1.1% 6.3% 26.4% 38.5% 27.7% 35 44 1.8% 8.2% 26.0% 37.3% 26.7% 45 64 4.8% 11.6% 28.4% 33.0% 22.2% 65 + 8.5% 18.7% 34.1% 27.4% 11.3% Race/Ethnicity White 3.6% 10.0% 27.6% 35.6% 23.1% Black 4.7% 15.4% 33.5% 27.4% 19.1% Other 5.4% 13.7% 31.1% 28.7% 21.1% Hispanic 5.0% 17.3% 32.2% 25.0% 20.5% Education >12 12.2% 24.6% 34.4% 18.7% 10.1% 12 4.3% 13.2% 33.0% 32.5% 17.1% 13-15 3.2% 9.9% 28.8% 36.4% 21.7% 16+ 1.4% 4.9% 22.1% 39.2% 32.4% Income <$20k 10.4% 20.1% 32.7% 24.3% 12.5% $20<$50k 2.9% 10.9% 31.2% 35.1% 20.0% $50<$75k 1.1% 6.0% 25.1% 40.6% 27.1% $75k+ 0.8% 3.8% 20.5% 39.7% 35.3% Marital Status Married 2.7% 9.1% 27.3% 35.8% 25.2% Separated/Widowed/Di vorced7.4% 15.3% 30.8% 30.0% 16.6% Never Married 2.3% 8.7% 28.0% 36.3% 24.7% Table 4.18: Health Behavior by Gene ral Health Status Frequencies General Health poor Fair good very goodexcellent Physical Activity None 12.1% 19.5% 32.8% 24.0% 11.6% Some 2.8% 10.4% 30.9% 36.4% 19.6% Meets Recommendations 1.9% 7.6% 24.1% 36.6% 29.8% BMI Not Overweight 3.1% 8.3% 23.3% 35.7% 29.7% Overweight 3.2% 9. 8% 29.2% 36.4% 21.4% Obese 6.2% 16.5% 35.9% 29.5% 11.9% Smoking Yes 5.2% 13.5% 33.1% 33.5% 14.7% No 3.4% 9.9% 27.0% 34.6% 25.1%

PAGE 161

147Table 4.19: Sociodemographic Factors by Mental Distress Frequencies Days of Mental Distress per Month 0 1 2 3 4 7 8 15>15 Gender Male 72.6% 2.8% 5.2% 2.7% 6.2% 4.4% 6.1% Female 61.1% 4.0% 7. 0% 3.7% 9.0% 7.2% 8.7% Age 20 34 56.1% 4.8% 8.3% 4.4% 11.2% 8.0% 7.3% 35 44 60.1% 4.1% 7.5% 3.9% 8.9% 6.8% 8.5% 45 64 68.1% 3.2% 5.9% 2.9% 6.6% 5.3% 8.0% 65 + 82.9% 1.5% 2.5% 1.5% 3.7% 3.4% 4.6% Race/Ethnicity White 65.9% 3.5% 6.4% 3.2% 7.7% 6.0% 7.2% Black 64.3% 3.3% 5.2% 4.0% 8.0% 6.6% 8.7% Other 65.8% 3.7% 6.6% 3.3% 7.2% 6.2% 7.2% Hispanic 61.5% 3.7% 6. 8% 4.1% 8.7% 6.3% 8.8% Education >12 65.9% 1.9% 3.9% 3.2% 6.5% 7.0% 11.5% 12 67.2% 2.6% 5.4% 2.9% 7.6% 6.1% 8.3% 13-15 62.7% 3.6% 6.9% 3.2% 8.5% 7.0% 8.2% 16+ 67.0% 4.7% 7.3% 3.8% 8.0% 4.9% 4.5% Income <$20k 59.3% 2.5% 5.6% 3.1% 7.8% 8.9% 12.8% $20<$50k 64.7% 3.3% 6. 3% 3.4% 8.5% 6.6% 7.2% $50<$75k 63.7% 4.6% 7. 4% 3.4% 9.3% 5.5% 6.1% $75k+ 67.9% 4.7% 7.8% 3.8% 7.1% 4.3% 4.4% Marital Status Married 69.7% 3.5% 6.1% 3.0% 7.0% 4.8% 5.9% Separated/Widowed/Divorc ed65.6% 2.8% 5.5% 3. 0% 7.0% 6.4% 10.2% Never Married 56.5% 4.5% 7.6% 4.3% 10.5% 8.7% 8.0% Table 4.20: Health Behavior by Mental Distress Frequencies Days of Mental Distress per Month 0 1 2 3 4 7 8 15>15 Physical Activity None 66.6% 2.5% 4.6% 2.4% 6.2% 6.9% 10.9% Some 64.1% 3.8% 7.0% 3.7% 8.3% 6.2% 6.8% Meets Recommendations 65.8% 3. 6% 6.6% 3.4% 8.3% 5.8% 6.6% BMI Not Overweight 64.4% 4.2% 6.8% 3.5% 8.7% 5.8% 6.7% Overweight 69.4% 2.9% 5. 9% 3.2% 7.0% 5.4% 6.3% Obese 62.2% 3.1% 5.9% 3.1% 8.0% 7.5% 10.2% Smoking Yes 57.0% 3.1% 6.4% 3.2% 8.9% 8.9% 5.8% No 68.5% 3.6% 6. 2% 3.3% 7.5% 5.2% 12.5%

PAGE 162

148 Table 4.21: Hypertension by General Health and Mental Distress Frequencies General Health Status Days of Mental Distress per Month Excellent Very Good Good Fair Poor 0 1 2 3 4 7 8 15> 15 Hypertension No 27.3% 36.8% 26.0%7.7% 2.2%65.0%3.9%6.8%3.6% 8. 1% 6.1% 6.7% Yes 10.1% 27.7% 34.9%19.1%8.2%67.9%2.5%4.9%2. 6% 7.1% 6.0% 9.1% *rows within outcomes sum to 100% Table 4.22: General Health Status by Mental Distress Frequencies Days of Mental Distress per Month 0 1 2 3 4 7 8 – 15 > 15 General Health Status Excellent 74.3% 4.2% 6. 0% 2.6% 5.9% 3.9% 3.1% Very Good 65.7% 4.2% 7. 6% 4.0% 8.6% 5.2% 4.7% Good 64.9% 2.8% 6.1% 3.3% 8.3% 7.0% 7.7% Fair 55.9% 2.5% 4.6% 3.0% 9.0% 9.3% 15.8% Poor 48.6% 0.4% 2. 3% 1.4% 5.4% 10.9% 31.0%

PAGE 163

149 Summary of Significant Bivariat e Associations among Key Study Constructs As presented in Table 4.23, the majority of relationships between structural, contextual, i ndividual and outcome vari ables are statistically significant. Of the three outcomes in this study, general health status has the most number of significant associat ions, followed by mental distress and hypertension. There is mixed evidence of the relationship between one structural factor, income inequality, and one contextual factor, informal social engagement, and the outcomes. Results are consis tent across all health behaviors with significant findings for every outcome. Within associations between predic tor variables, there is evidence of significant relationships between both st ructural factors and most contextual factors, with one exception (T able 4.24). Organizational acti vism or formal social engagement is not significantly related to t he majority of other predictors. In regards to health behavior and broader char acteristics under study, only poverty and informal social engagement show a simila r pattern of result s with respect to physical activity, BMI, and smoking.

PAGE 164

150 Table 4.23 Summary of Significance of Bivariate Associations: Outcomes Hypertension General Health Mental Health Income inequality NS S NS Poverty S S S Social trust NS S S Informal social engagement S S NS Organizational activism S S S Mutual aid NS S S Physical Activity S S S Body Mass Index S S S Smoking S S S Table 4.24 Summary of Significance of Bivariate Associations: Predictors Income Inequality Poverty Physical Activity Body Mass Index Smoking Social trust S S S NS S Informal social engagement S S S S S Formal social participation S NS S NS NS Mutual aid S S S S NS Income inequality S S NS S Poverty S S S S

PAGE 165

151 Correlates of Hypertension There is mixed to weak evidence of t he possible role of structural factors in influencing the occurrenc e of hypertension, as seen in Table 4.25. Income inequality is non-significant while although po verty is statistically significant, living in poverty has a very small influenc e on the odds of having hypertension (OR 1.01, 95% CI 1.01-1.02). Social contextual factors fare the same, with only two dimensions of social capital significant ly associated with hypertension – informal social engagement (OR 1.60, 95% CI 1.02-2.49) and formal social engagement (OR 1.86, 95% CI 1.26-2. 74). An interesting note is that both of these associations are in the counterintuitive di rection in that the odds of a resident reporting hypertension are hi gher if s/he lives in a community characterized by higher levels of informal or formal so cial engagement. As expected, all health behaviors are significantly a ssociated with hypertension. Correlates of General Health Status Of all outcomes, findings of associat ions with general health status are most consistent. All structural, contextual, and behavioral variables are statistically significantly related to this variable (Table 4.26). Although numerically small, both inco me inequality and poverty are correlated with it at the p<.0001 level. While all social capita l indicators are associated with general

PAGE 166

152 health status, some results are in the unexpected direct ion. One such unanticipated finding is the negative associat ion between health status and living in a community with more informal so cial engagement while there is a positive association between living in a community with more organizational activism and self-reported general health. All thr ee health behaviors are significantly associated with this outcome at the p < .0001 level. Correlates of Mental Distress As with hypertension, social struct ural factors have mixed associations with self-reported mental distress (Table 4.27). Income inequality is not significantly related to the number of da ys per month for which mental health is reported as not good. Although poverty is statistically significantly associated at the p < .0001 level, directi onality of the relationship is unexpected; living in an impoverished community is negatively associ ated with reports of mental distress. A more stable pattern exists between so cial context and mental distress. Findings indicate that living in a communi ty with more social trust, Organizational activism, or mutual aid, is significant ly associated with reporting fewer days per month of mental distress. Only informa l social engagement is not statistically significantly related to mental distress. Consistent resu lts are also observed with respect to the health behaviors, with all three behaviors significantly associated in the anticipated direction with this outcome.

PAGE 167

153 Table 4.25: Outcome Bivariate Associations: Hypertension HypertensionLevel Interpretation Social Structural Factors Income inequality NS p > .25 Poverty S OR 1.01 (1.01, 1.02) Although statistically significant, living in poverty has a very small influence on the odds of reporting hypertension Social Contextual Factors Social trust NS p > .41 Informal social engagement S OR 1.6 (1.02, 2.49) Almost 60% more likely to report hypertension if live in a community with high levels of informal social engagement Organizational activism S OR 1.86 (1.26, 2.74) Almost twice as likely to report hypertension if live in a community with high levels of organizational activism Mutual aid NS p > .59 Health Risk Behavior Physical Activity S OR 0.71 (0.69, 0.74) Odds of reporting hypertension lower as activity increases Body Mass Index S OR 1.94 (1.87, 2.02) Odds of reporting hypertension almost 2x higher as BMI increases Smoking S 2 = 95.11 df 1 p<.0001 Non-smoking associated with reporting hypertension

PAGE 168

154 Table 4.25: Outcome Bivariate Associations: Hypertension Hypertension Level Interpretation Sociodemographic Factors Sex NS p > .95 Income S OR 0.75 (0.73, 0.77) Odds of having hypertension lower as income increases Race/Ethnicity Hispanic S OR 0.55 (0.48, 0.64) Odds of reporting hypertension lower if Hispanic (vs. Non-Hispanic) Black S OR 1.48 (1.35, 1.62) Odds of reporting hypertension higher for Blacks compared to Whites Other S OR 0.76 (0.68, 0.86) Odds of reporting hypertension lower for Other compared to Whites Age S OR 2.29 (2.22, 2.36) Odds of reporting hypertension higher with age Education S OR 0.79 (0.76, 0.81) Odds of reporting hypertension lower as education level increases Marital Status S 2 = 745.28 df 2 p<.0001Marital status signific antly associated with reporting hypertension

PAGE 169

155Table 4.26: Outcome Bivariate Asso ciations: General Health Status General Health Level Interpretation Social Structural Factors Income inequality S rsp = -.03, p<.0001 Living in a community with higher income inequality is significantly associated with reporting poorer general health Poverty S rsp = -.05, p<.0001 Living in a community with higher levels of poverty is significantly associated with reporting poorer general health Social Contextual Factors Social trust S rsp = .02, p=.0077 Living in a community with more social trust is significantly associated with reporting better general health Informal social engagementS rsp = -.01, p=.0434 Living in a community with more ISE is significantly associated with reporting poorer general health Organizational activism S rsp = .01, p=.0166 Living in a community with more OA is significantly associated with reporting better general health Mutual aid S rsp = .02, p=.0063 Living in a community with more mutual aid is significantly associated with reporting better general health Health Risk Behavior Physical Activity S rsp = .23, p<.0001 Physical activity is significantly associated with reporting better general health Body Mass Index S rsp = -.20, p<.0001 Higher BMI is significantly associated with reporting poorer general health Smoking S p < .0001 Nonsmoking is significantly associated with reporting better general health

PAGE 170

156 Table 4.26: Outcome Bivariate Asso ciations: General Health Status General Health Level Interpretation Sociodemographic Factors Sex S p > .0001Men report significantly better levels of general health Income S rsp = .29, p<.0001 Higher income is significantly associated with reporting better general health Race/Ethnicity Hispanic S p > .0001Hispanics report significantly poorer general health (compared to nonHispanics) Black S p > .0001Blacks report significantly p oorer general health compared to Whites Other S p > .0001Other race/ethnicities report si gnificantly poorer general health compared to Whites Age S rsp = -.21, p<.0001 Older individuals report significantly poorer general health Education S rsp = .28, p<.0001 Education level positively associated with general health Marital Status S p > .0001Marital status significantly associated with level of general health

PAGE 171

157Table 4.27: Outcome Bivariate Associations: Mental Distress Mental Health Level Interpretation Social Structural Factors Income inequality NS p > .83 Poverty S rsp = -.03, p<.0001 Living in a community with higher levels of poverty is significantly associated with reporting mental health not good fewer days/month Social Contextual Factors Social trust S rsp = -.05, p<.0001 Living in a community with more social trust is significantly associated with reporting mental health not good fewer days/month Informal social engagement NS p > 0.58 Organizational activism S rsp = -.07, p<.0001 Living in a community with more organizational activism is significantly associated with reporting mental health not good fewer days/month Mutual aid S rsp = -.04, p<.0001 Living in a community with more mutual aid is significantly associated with reporting mental health not good fewer days/month Health Risk Behavior Physical Activity S rsp = -.02, p=.02 Physical activity is significantly associated with reporting fewer days/month mental health not good Body Mass Index S rsp = .01, p=.04 Higher BMI is significantly associated with reporting more days/month mental health not good Smoking S p < .0001 Smoking is significant ly associated with reporting more days/month of mental health not good

PAGE 172

158 Table 4.27: Outcome Bivariate Associations: Mental Distress Mental Health Level Interpretation Sociodemographic Factors Sex S p < .0001 Significantly more women report more days/month of poor mental health than men Income S rsp = -.08, p<.0001 Lower income is signific antly associated with reporting more days/month mental health not good Race/Ethnicity Hispanic S p = .01 Significantly more Hispanics report more days/month of poor mental health than NonHispanics Black S p = .002 Significantly more Blacks report more days/month of poor mental health than Whites Other NS p > .60 Age S rsp = -.17, p<.0001Younger adults report significantly more days/month mental health not good compared to older adults Education S rsp = -.02, p=.0002Less educated report significantly more days/month mental health not good compared to more educated Marital Status S p < .0001 Marital status signifi cantly associated with number of days/month poor mental health reported

PAGE 173

159 Associations between Social Contex tual Factors and Health Behavior Overall, the relationship between indi cators of social capital and health behaviors is mixed. Table 4.28 demons trates that there is no pattern of association across all behaviors. Moreover, some of the significant associations are in a counterintuitive direction. In regards to the associations of the dimensions of social capital within type of health behavior, the most consistent results are with physical activity. Every as pect of social capital is statistically significantly associated with physical activity, although direction of the relationships is inconsistent across dimens ions. Results indicate that the greater the social trust (rsp = .05, p<.0001) and info rmal social engagement (rsp = .02, p=.0034) in the community, the more acti ve a resident is. Less Organizational activism (rsp = -.01, p=.0426) and less mutual aid (rsp = -.03, p<.0001) are significantly associated with inactivity within a locale. In contrast, both BMI and smoking have inconsistent patterns, with respect to both statistical significance and directiona lity of associations. Findings indicate that neither social trust nor Organizational activism is significantly associated with BMI. Informal soci al engagement is associated with th is risk factor, albeit in the positive direction (rsp = .02, p=.0019). Only mutual aid influences BMI in the anticipated manner (rsp = -.01, p=.048). Aspects of social capital either have no relationship to smoking or are positively associated.

PAGE 174

160 Viewing the results by dimension of social capital yields additional information. For example, the influence of social trust is different across health behaviors; an individual living in a communi ty with greater social trust is more likely to smoke and be more active. Although informal social engagement is significantly associated with all health beha viors, directionality is inconsistent; residing in a community with more informa l social engagement is associated with an inhabitant being more active, having hi gher BMI, and engaging in smoking. The third dimension of social capital under study, informal social participation, is the least associated with the behaviors, with a negative relationship only with physical activity. Mutual aid is also inconsistent in its influence on behaviors; individuals residing in communities with less mutual aid are both less active and have a lower BMI. These results suggest that there may be different mechanisms underlying the relationships between the dimensions of social capital and the three behaviors.

PAGE 175

161Table 4.28: Social Capital Bivariates with Health Behavior Physical Activity Body Mass Index Smoking Social trust rsp = .05, p<.0001 p > .95 OR 1.47 (1.22,1.77) The greater the social trust in the community, the more active a resident is More likely to smoke if living in a community with high social trust Informal social engagement rsp = .02, p=.0034 rsp = .02, p=.0019 OR 4.71 (2.95, 7.54) The more informal social engagement in the community, the more active a resident is The more informal social engagement in the community, the higher a resident's BMI Almost 5 times as likely to smoke if live in a community with more informal social engagement Organizational activism rsp = -.01, p=.0426 p > .14 p > .36 The more Organizational activism in the community, the less active a resident is Mutual aid rsp = -.03, p<.0001 rsp = -.01, p=.0480 p > .78 The more mutual aid in the community, the less active a resident is The more mutual aid in a community, the lower a resident's BMI Table 4.29: Social Structure Bivari ates with Health Behavior Physical Activity Body Mass Index Smoking Income inequality rsp = -.05, p<.0001 p > 0.91 OR 0.01 (0.002, 0.044) The greater the income inequality in a community, the less active a resident is Less likely to smoke if live in a community with higher income inequality Poverty rsp = -.03, p<.0001 rsp = .01, p=.0359 OR 0.993 (0.989,0.998) The greater the poverty in a community, the less active a resident is The greater the poverty on the community, the higher a resident's BMI Less likely to smoke if live in a community with more poverty

PAGE 176

162 Associations between Social Struct ural Factors and Health Behavior Both income inequality and poverty ar e significantly associated with the majority of health behaviors. An unexpec ted finding is that wh ile the greater the income inequality in a community is signi ficantly associated with less activity (rsp = -.05, p<.0001), it is not significantly related to a resident’s BMI (Table 4.29). Poverty does influence both physical activi ty and BMI similarly; the greater the poverty in a community, the less active (rsp = -.03, p<.0001) and heavier (rsp = .01, p<.0359) a resident is. As with the social contex tual factors, smoking is associated with social structure, but in the unanticipated direct ion. One is less likely to smoke if s/he lives in a comm unity with higher income inequality (OR 0.01, 95% CI 0.002-0.044) or more impoverished (OR 0.993, 95% CI 0.9890.998). Associations between Social Structur al and Social Contextual Factors The most consistent bivariate re lationships are found between social contextual and social structural factors (Table 4.30). Every aspect of social capital is negatively associated with struct ural inequalities, other than the finding that the level of poverty in a communi ty appears to be unrelated to formal social engagement. Of the two structur al factors, income inequality seems to have the

PAGE 177

163 stronger association with social capita l, with correlations ranging from -.13 (organizational activism) to -.68 (social trust). Table 4.30: Social Capital Bivariates with Social Structure Income Inequality Poverty Social Trust rsp = -.68, p<.0001 rsp = -.19, p<.0001 The greater the income inequality in a community, the less social trust The greater the poverty in a community, the less social trust Informal social engagement rsp = -.46, p<.0001 rsp = -.02, p=.0026 The greater the income inequality in a community, the less informal social engagement The greater the poverty in a community, the less informal social engagement Formal social participation rsp = -.13, p<.0001 p > .54 The greater the income inequality in a community, the less formal social participation Mutual aid rsp = -.05, p<.0001 rsp = -.12, p<.0001 The greater the income inequality in a community, the less mutual aid The greater the poverty in a community, the less mutual aid

PAGE 178

164 Multivariate Analysis For each set of hypotheses, a series of hierarchical linear models were analyzed using HLM 6 (Raudenbush & Bryk, 2004). A two-level model was used, where individual attributes and behav iors were considered level-1 factors and social contextual and social structur al variables were considered level-2 factors. A random-intercept model was one where only the level-1, or individual level, intercept was treated as random. Essentially, this form of modeling is comparable to a one-way ANOVA with random effects. Only the intercept at level-1 was modeled as a func tion of level-2 predictors. A basic model building approach was used for testing of the randomintercepts models. The sequence of st eps began with testing an unconditional model, followed by a control and then fu ll model. The unconditional model was one in which no predictor variables were ent ered at either level – it was used to test whether there were basic differenc es between communities in the outcome. The control model accounted for both level-1 sociodemographic and level-2 social structural controls, whereas the fu ll model included controls and the level-2 predictor under study. The control model at level-1 included individual sociodemographic variables c oded predominantly as dummy variables in order to facilitate interpretation of the intercept, as the intercept in multilevel modeling did not have the same interpretation as in OL S analysis. In these multilevel models,

PAGE 179

165 the intercept did not equate to the overa ll mean, but rather the mean for the referent group (where all dummy va riables representing sociodemographic characteristics of the individual at leve l 1 = 0) – here, a poor, young, white male. The selection of level-2 controls was based upon which social structural factor was included in the model. The choice of the possible level-2 control variables (median household income, percent comple ted high school or less, and percent unemployed) was specifically tailored to each social structural i ndicator to reduce the possible effects of mult icollinearity; only a subs et was used. For models testing the effects of income inequa lity, only median household income was utilized as a control in level-2. For te sts of the effects of poverty, only percent completed high school or less and percent unemployed were accounted for in the models. Results are organized by cluster of hypotheses. For each section, the overall findings are characterized, follo wed by detailed discussion of results. Cluster 1: Behavioral variables only par tially mediate social structure and disease. Only very limited evidence was f ound to support the relationships between social structural inequalities and any of the outcomes under study (hypertension, general health status, or m ental distress). Most test s of direct relationships between either income inequality or poverty and disease state were statistically

PAGE 180

166 non-significant at the p < .05 level. Because the lack of a statistically significant direct effect of social structure on dis ease may indicate full mediation by health behaviors in that the relationship between income inequality or poverty and any outcome only exists through their effects on physical activity, BMI, or smoking, additional models were examined which included possible mediators. Again, there was no support found for the relati onships between social structure and outcomes. As only limited ev idence of a direct relationship was found between social structure and hypertension, general he alth status, or mental distress, it was not surprising that indirect effects would be insignificant as well. Although the results for the three out comes were similar, some notable differences in parameter estimates were found. Hypertension The possible direct effects of a comm unity’s social structural environment on residents’ reporting of hypertension were examined. In these models, individual odds of reporti ng hypertension were regressed on community income inequality and community rates of poverty separately. The unconditional model, testing between community differences in the odds of a resident reporting hypertension, was significant at p < 0001 level (Table 4.31). The odds of a typical resident in an average community to report hypertension is OR 0.35 (95% CI 0.32 – 0.39). The control model was sign ificant as well (p < .0001), indicating

PAGE 181

167 individual sociodemographic va riation in individual odds of having hypertension in these communities. Once other vari ables were entered into the model, no statistically significant associations were found (income inequality without (p = .715) and with (p = .946) control variables ; poverty without (p = .071) and with (p = .464) control variables). As there wa s no support for a direct relationship, no evidence was found for the possible influence of behavior mediating such relationship between social structure and hy pertension. Therefore, variation in the odds of a resident reporti ng hypertension is not relat ed to the level of income inequality or poverty in the community within which s/he lived. Some of the models testing these hypotheses either di d not converge (poverty) and/or had non-significant between level2 variance, Tau. Non-convergence occurred when the model was not able to be estimated, given the type of data and/or parsimony of the model. The majority of the models having a non-significant Tau (at p < .05) indicated that there may not be significant differ ences between communities, with respect to hypertension, in t hese data once additional variables are considered.

PAGE 182

168Table 4.31 Community Social Structural Influences on Individual Hypertension Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis No predictors model Intercept, 00 0.0370.35 0.32 0.39 <.0001*** Control model for income inequality Intercept, 00 0.011 0.33 0.20 0.54 <.0001*** Model for income inequality 1a1 Intercept, 00 0.0390.25 0.04 1.70 0.15 0.012 0.31 0.06 1.75 0.176 income inequality, 01 2.14 0.03 147.55 0.715 1.12 0.04 28.90 0.946 Control model for poverty Intercept, 00 0.003 0.05 0.03 0.09 <.0001*** Model for poverty 1a2 Intercept, 00 0.0330.24 0.15 0.37 <.0001*** 0.002 0.05 0.03 0.09 <.0001*** poverty, 01 1.01 1.00 1.03 0.071 1.00 0.98 1.01 0.464 Control health behavior model for income inequality Intercept, 00 0.003 0.12 0.08 0.18 <.0001*** Full health behavior model for income inequality 1b1 Intercept, 00 0.003 0.08 0.02 0.30 0.001** income inequality, 01 2.45 0.19 31.71 0.477 Control health behavior model for poverty Did not converge 1b2 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, income, race/ethnicity, and age at level 1 and median household income (for income inequality) or percent unemployed a nd percent completed high school or less (for poverty) at level 2. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 183

169 General Health Status The potential direct effects of a comm unity’s social structural environment on residents’ ratings of general health stat us were examined. In these models, individual odds of reporting fair or poor health were regressed on income inequality and poverty s eparately. The unconditional model, testing between community differences in self-reported gener al health status, was significant at p < .0001 level (table 4.32). The odds of a typical resident in an average community to report his/her health as fa ir or poor was OR 0.17 (95% CI 0.15 – 0.20). The control model was significant as well (p < .032), indi cating individual sociodemographic variation in individual odds of reporting health as fair or poor in these communities. Once social structur al factors were enter ed individually into the model (where no controls were present), some significant associations were found. Although income inequality was not significantly associated with general health status (p = .567), significant result s were obtained (p = .047) in the model with only poverty entered as a predictor. However, as control variables were entered into the model, no statistically si gnificant associations were found for either income inequality (p = .629) or poverty (p = .4 74). As there was little support for a direct relationship betw een social structure and general health status, no evidence was found for the possi ble influence of behavior a mediating this association. Therefore, income inequality and poverty did not significantly

PAGE 184

170 explain differences in self -reported health status. Variation in the odds of a resident reporting fair or poor health was not related to the level of income inequality or poverty in the community within which s/he lived, once individual sociodemographic characteristics were taken into account. All of the models had a statistically significant Tau (at p < .05) indicating that there were still significant differences between communities, with res pect to general health status, in these data once additional variables are consider ed. These results suggest that there were other variables not considered in this model which may explain the between community differences in the odds of resi dents reporting fair or poor health. Mental Distress The possible direct effects of a comm unity’s social structural environment on residents’ reports of mental distre ss were examined. In these models, individual reports of the number of days of mental distress per month were regressed on income inequality and pover ty separately. The unconditional model, testing between community differences in mental distre ss, was significant at p < .0001 level (table 4.33) indicating that t he communities differed significantly in residents’ days of mental distress per month. The typical resident in an average community reported 3.54 (3.2 9 – 3.81) days of mental distress out of the last thirty days. The control model was significant as well (p < .002),

PAGE 185

171Table 4.32 Community Social Structural Influences on Individual General Health Status Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio95% confidence interval p value Hypothesis No predictors model intercept, 00 0.1020.17 0.15 0.20 <.0001*** Control model for income inequality Intercept, 00 0.053 0.37 0.15 0.91 0.032* Model for income inequality 4a1 Intercept, 00 0.1040.07 0.004 1.44 0.083 0.055 0.19 0.01 3.78 0.262 income inequality, 01 6.37 0.01 4534.17 0.567 3.82 0.01 1088.11 0.629 Control model for poverty Intercept, 00 0.048 0.08 0.03 0.24 <.0001*** Model for poverty 4a2 Intercept, 00 0.0850.09 0.04 0.17 <.0001*** 0.050 0.09 0.03 0.30 <.0001*** poverty, 01 1.02 1.00 1.05 0.047* 1.01 0.98 1.04 0.474 Control health behavior model for income inequality Intercept, 00 0.054 0.10 0.04 0.27 <.0001*** Full health behavior model for income inequality 4b1 Intercept, 00 0.058 0.09 0.004 2.18 0.133 income inequality, 01 1.27 0.003 483.65 0.935 Control health behavior model for poverty Intercept, 00 0.059 0.04 0.01 0.15 <.0001*** Full health behavior model for poverty 4b2 Intercept, 00 0.056 0.06 0.02 0.20 <.0001*** poverty, 01 1.02 0.99 1.05 0.24 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, income, race/ethnicity, and age at level 1 and median household income (for income inequality) or percent unemployed a nd percent completed high school or less (for poverty) at level 2. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 186

172 indicating individual sociodemographic va riation in days of mental distress reported in these communities. As such, this variability was partly explained by individual sociodemographic characteristi cs. However, once other variables were entered into the model, no statistically significant associations were found (income inequality without (p = .627) and with (p = 603) control variables; poverty without (p = .969) and with (p = .165) control variables). As there was no support for a direct relationship, no evi dence was found for the possible influence of behavior mediating such relationship between social structure and mental distress. Therefore, variation in the days a resident reported mental distress was not related to the level of income inequa lity or poverty in the community within which s/he lived. All of the models had a st atistically significant Tau (at p < .05) indicating that there were still signific ant differences between communities, with respect to mental distress, in these data even after additional variables were considered. This situation suggests that there were other variables not included in this model which may better explai n the between community differences in days of reported mental distress. Health Behavior The possible direct influence of a community’s social structural environment on residents’ health behaviors was examined. In these models,

PAGE 187

173 individual health behaviors were regr essed on income inequality and poverty separately. The unconditional models, te sting between community differences in health behaviors, were significant at p < .0001 level (Table 4.34). The odds of a typical resident in an average community to report limited or no activity is OR 1.24 (95% CI 1.12 – 1. 37), overweight or obese OR 1.31 (95% CI 1.21 – 1.43), or smoking OR 0.28 (95% CI 0. 26 – 0.31). The control m odel was significant only for one health behavior, Body Mass Index (p < .0001), demonstrating individual sociodemographic variation in individual odds of reporting being overweight or obese in these communities. Once income inequality was entered into the models, no statistically significant associ ations were found for physical activity ((income inequality without (p = .109) and with (p = .196) control variables) or Body Mass Index ((income inequality wit hout (p = .828) and with (p = .361) control variables). The results did i ndicate, though, that the odds of smoking were significantly influenced by the leve l of income inequality in one’s community (p = .032), controlling for sociodemographi c characteristics of the resident and median household income of the community. There were some differences in re sults regarding tests of the poverty’s influence on health behavior (Table 4.35). T he control model was significant only for one health behavior, physical activity (p = .008), demonstrating individual sociodemographic variation in individual odd s of reporting limited or no activity in these communities. Once poverty was ent ered into the models, no statistically significant associations were found for phy sical activity ((poverty without (p =

PAGE 188

174 .141) and with (p = .266) control variables ) or Body Mass Index ((poverty without (p = .316) and with (p = .084) control variables). The results did indicate, though, that the influence of the level of pover ty in one’s community on odds of smoking was not significant (p = .055), albeit by a relatively small degree, controlling for sociodemographic factors of the resi dent and the percent unemployed and percent completed high school or less within the community.

PAGE 189

175Table 4.33 Community Social Structural Influences on Individual Mental Distress Community Unconditional1 Conditional1 Characteristic Event Rate Ratio 95% confidence interval p value Event Rate Ratio 95% confidence interval p value Hypothesis No predictors model intercept, 00 0.033 3.54 3.29 3.81 <.0001*** Control model for income inequality Intercept, 00 0.040 3.18 1.63 6.18 0.002** Model for income inequality 7a1 intercept, 00 0.034 2.46 0.54 11.34 0.237 0.041 1.94 0.25 14.96 0.509 income inequality, 01 2.24 0.08 64.95 0.627 2.66 0.06 123.65 0.603 Control model for poverty intercept, 00 0.040 2.76 1.26 6.02 0.014* Model for poverty 7a2 intercept, 00 0.035 3.52 2.33 5.32 <.0001*** 0.038 2.51 1.15 5.44 0.023* poverty, 01 1.00 0.99 1.01 0.969 0.99 0.97 1.01 0.165 Control health behavior model for income inequality intercept, 00 0.037 1.68 0.89 3.20 0.108 Full health behavior model for income inequality 7b1 intercept, 00 0.038 0.97 0.14 6.94 0.974 income inequality, 01 3.01 0.07 121.93 0.545 Control health behavior model for poverty intercept, 00 0.039 1.85 0.85 4.01 0.114 Full health behavior model for poverty 7b2 intercept, 00 0.039 1.71 0.78 3.75 0.169 poverty, 01 0.99 0.97 1.01 0.266 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, income, race/ethnicity, and age at level 1 and median household income (for income inequality) or percent unemployed a nd percent completed high school or less (for poverty) at level 2. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 190

176 Table 4.34 Community Income Inequality Influences on Individual Health Behavior Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Physical Activity2 No predictors model intercept, 00 0.0501.24 1.12 1.37 <.0001*** Control model intercept, 00 0.0421.62 0.75 3.48 0.206 Model for income inequality 1c1 intercept, 00 0.0450.25 0.04 1.82 0.164 0.0400.37 0.03 4.05 0.402 income inequality, 01 34.05 0.43 2690.96 0.109 18.22 0.20 1631.17 0.196 Body Mass Index3 No predictors model intercept, 00 0.0301.31 1.21 1.43 <.0001*** Control model intercept, 00 0.0287.79 4.04 15.01 <.0001*** Model for income inequality 1c2 intercept, 00 0.0331.58 0.28 9.01 0.593 0.02819.15 2.35 156.10 0.008** income inequality, 01 0.66 0.01 31.25 0.828 0.17 .003 8.69 0.361 Smoking4 No predictors model intercept, 00 0.0380.28 0.26 0.31 <.0001*** Control model intercept, 00 0.0481.28 0.56 2.92 0.547 Model for income inequality 1c3 intercept, 00 0.0311.10 0.19 6.52 0.915 0.03013.88 1.41 136.75 0.026* income inequality, 01 0.05 .001 2.56 0.129 0.01 0.000 0.64 0.032* 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, inco me, race/ethnicity, and age at level 1 and median household income at level 2 *significant at p<.05 2where 0 = meets recommended levels of activity, 1 = lim ited/no activity **significant at p<.01 3where 0 = normal BMI, 1 = overweight/obese BMI ***significant at p<.0001 4where 0 = non-smoker, 1 = smoker

PAGE 191

177 Table 4.35 Community Poverty Influences on Individual Health Behavior Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Physical Activity2 No predictors model intercept, 00 0.050 1.24 1.12 1.37 <.0001*** Control model 0.023 0.33 0.15 0.72 0.008** intercept, 00 Model for poverty 1c4 intercept, 00 0.047 0.85 0.51 1.43 0.530 0.023 0.29 0.13 0.65 0.005** poverty, 01 1.01 1.00 1.03 0.141 0.99 0.97 1.01 0.266 Body Mass Index3 No predictors model intercept, 00 0.030 1.31 1.21 1.43 <.0001*** Control model intercept, 00 0.017 1.23 0.60 2.50 0.559 Model for poverty 1c5 intercept, 00 0.031 1.06 0.69 1.64 0.779 0.015 1.01 0.49 2.10 0.972 poverty, 01 1.01 0.99 1.02 0.316 0.99 0.97 1.002 0.084 Smoking4 No predictors model intercept, 00 0.038 0.28 0.26 0.31 <.0001*** Control model intercept, 00 0.022 1.11 0.49 2.51 0.786 Model for poverty 1c6 intercept, 00 0.038 0.34 0.21 0.56 <.0001*** 0.019 0.86 0.37 1.98 0.709 poverty, 01 0.99 0.98 1.01 0.404 0.98 0.96 1.00 0.055 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, income, race/ethni city, and age at level 1 and percent unemployed and percent completed high school or less at level 2. *significant at p<.05 2where 0 = meets recommended levels of activity, 1 = limited/no activity **significant at p<.01 3where 0 = normal BMI, 1 = overweight/obese BMI 4where 0 = non-smoker, 1 = smoker ***significant at p<.0001

PAGE 192

178 Cluster 2: Social context partially mediates social structure and disease. Mixed evidence was found to support the hypothesis that t hat the level of social capital in the community in whic h one resides directly influences selfreported health (e.g., hypertension, general health, mental health). Consistent with the findings for social structure, most tests of a direct relationship between indicators of social capital and disease st ate were statistically non-significant at the p < .05 level. A ll four social capital dimensions (social trust, informal social engagement, organizational activism, and mutual aid) were tested separately. Social Capital and Disease Hypertension The possible direct effects of a comm unity’s social contextual environment on residents’ reporting of hypertension were examined. In these models, individual odds of reporting hypertension were regressed on social trust, informal social engagement, organizational activism and mutual aid individually. As stated earlier, the unconditio nal model, testing between community differences in resident’s odds of hypertension, was sign ificant at p < .0001 level (Table 4.36).

PAGE 193

179 Table 4.36 Community Social Capital Influences on Individual Hypertension Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis No predictors model intercept, 00 0.0370.35 0.32 0.39 <.0001*** Level 1 control model intercept, 00 0.015 0.16 0.14 0.19 <.0001*** Model for social trust 2a1 intercept, 00 0.0390.35 0.32 0.39 <.0001*** 0.016 0.16 0.14 0.19 <.0001*** social trust, 01 1.11 0.58 2.11 0.742 0.96 0.60 1.54 0.863 Model for informal social engagement 2a2 intercept, 00 0.0390.36 0.32 0.39 <.0001*** 0.015 0.16 0.14 4.24 <.0001*** informal social engagement, 01 1.33 0.32 5.59 0.689 1.44 0.50 4.24 0.498

PAGE 194

180 Table 4.36 Community Social Capital Influences on Individual Hypertension Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Model for organizational activism 2a3 intercept, 00 0.040 0.35 0.31 0.40 <.0001*** 0.016 0.17 0.14 0.20 <.0001*** organizational activism, 01 0.98 0.32 3.05 0.978 0.89 0.37 2.16 0.795 Model for mutual aid 2a4 intercept, 00 0.040 0.38 0.13 1.13 0.079 0. 016 0.18 .080 0.43 <.0001*** mutual aid, 01 0.99 0.80 1.22 0.913 0.98 0.83 1.16 0.833 1Unconditional models do not c onsider any additional covariates in t he model at level 1; conditional models account for individual sociodemographic factors, including gender, income, race/ethnicity, and age. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 195

181 The odds of a typical resident in an average community reporting hypertension was OR 0.35 (0 .32 – 0.39). The control m odel was significant as well (p < .0001), indicating individual so ciodemographic variation in individual odds of having hypertension in these comm unities. Once other variables were entered into the model, no statistically sign ificant associations were found (social trust without (p = .742) and with (p = 863) control variables; informal social engagement without (p = .689) and with (p = .498) control variables; organizational activism without (p = .978) and with (p = .795) control variables; mutual aid without (p = .913) and with (p = .833) control variables). As there was no support for a direct relationship, no tests were performed on the possible influence of behavior mediating such rela tionship between social context and hypertension. Results suggested that vari ation in the odds of a resident reporting hypertension was not related to the leve l of social trust, informal social engagement, organizational activism, or mu tual aid in the community within which s/he lived. In sum, these findings indicated that above and beyond sociodemographic factors, social capita l did not explain a resident’s odds of reporting hypertension. General Health Status The potential direct effects of a comm unity’s social contextual environment on residents’ reports of general health stat us were examined. In these models,

PAGE 196

182 individual odds of reporting fair or poor health were regressed on social trust, informal social engagement, or ganizational activism, and mu tual aid individually. As stated earlier, the unconditional model, testing between community differences in resident’s odds of reporting fair or poor health, was significant at p < .0001 level (Table 4.37). The odds of a typical resident in an average community to report his/her health as fa ir or poor was OR 0.17 (95% CI 0.15 – 0.20). The control model was significant as well (p < .0001), indi cating individual sociodemographic variation in individual odd s of reporting fair or poor health in these communities. Once other variables were entered into the model, several statistically significant associations were found. The level of social trust in the community in which one resided decreased the odds of reporting fair or poor health (OR 0.47, 95% CI 0.23 – 1.00, p = .05), controlling for individual sociodemographic characteristics. Addition ally, living in a community with strong organizational activism also decreased t he odds of reporting fair or poor health (OR 0.19, 95% CI 0.05 – 0.69, p = .01), controlling for individual sociodemographic characteristics. T here was no evidence of a direct relationship for the other two dimensions of social capital (informal social engagement without (p = .157) and with (p = .081) control variables; mutual aid without (p = .15) and with (p = .081) contro l variables). As there was support for a direct relationship between social trus t and organizational activism and general health status, tests were performed on the possible influence of behavior mediating such relationship between social context and general h ealth status. In

PAGE 197

183 essence, results suggest ed that variation in the odds of a resident reporting his/her health as fair or poor was related to the level of organizational activism, after controlling for both sociodemographic characteristics and health behavior of the individual. However, once the sociodemographic characteristics and health behavior of the individual was considered, social trust no longer significantly predicted the odds of a resident reporting his/ her health as fair or poor. In sum, these findings indicated that above and beyond sociodemographic factors and behavior, only certain indicators of social capital explained a resident’s odds of reporting his/her health as fair or poor.

PAGE 198

184 Table 4.37 Community Social Capital Influences on Individual General Health Status Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis No predictors model intercept, 00 0.102 0.17 0.15 0.20 <.0001*** Level 1 control model intercept, 00 0.0570.18 0.15 0.23 <.0001*** Model for social trust 5a1 intercept, 00 0.096 0.17 0.15 0.20 <.0001*** 0.0480.19 0.15 0.23 <.0001*** social trust, 01 0.43 0.16 1.14 0.087 0.47 0.23 1.0020.051* Model for informal social engagement 5a2 intercept, 00 0.101 0.17 0.15 0.19 <.0001*** 0.0550.18 0.15 0.23 <.0001*** informal social 0.21 0.02 1.91 0.157 0.20 0.03 1.23 0.081 engagement, 01

PAGE 199

185 Table 4.37 Community Social Capital Influences on Individual General Health Status Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Model for organizational activism 5a3 intercept, 00 0.098 0.19 0.16 0.23 <.0001*** 0.4400.21 0.17 0.26 <.0001*** organizational 0.22 0.04 1.20 0.078 0.19 0.05 0.69 0.014* activism, 01 Model for mutual aid 5a4 intercept, 00 0.096 0.55 0.11 2.80 0.457 0.0490.59 0.16 2.23 0.422 mutual aid, 01 0.80 0.59 1.09 0.15 0.80 0.62 1.03 0.081 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, income, race/ethnicity, and age. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 200

186 Mental Distress The possible direct effects of a comm unity’s social contextual environment on residents’ reporting of mental distress were examined. In these models, an individual’s days per month of reported mental distress were regressed on social trust, informal social engagement, or ganizational activism, and mutual aid individually. As stated earlier, th e unconditional model, testing between community differences in resident’s days per month of report ed mental distress, was significant at p < .0001 le vel (Table 4.38). The typi cal resident in an average community to reported 3. 54 (3.29 – 3.81) days per month when their mental health was not good. The control model was significant as well (p < .0001), indicating individual sociodemographic vari ation in days per month of mental distress in these communities. Once other variables were entered into the model, no statistically significant associat ions were found (social trust without (p = .553) and with (p = .825) control vari ables; informal social engagement without (p = .864) and with (p = .974) control variables; organizational activism without (p = .217) and with (p = .235) control vari ables; mutual aid without (p = .255) and with (p = .318) control variables). As there was no support for a direct relationship, no tests were performed on the possible influence of behavior mediating such relationship between social context and mental distress. Results suggested that variation in the days per month a resident reported mental distress was not related to the level of social trust, informal social engagement,

PAGE 201

187 organizational activism, or mutual aid in the community within which s/he lived. In sum, these findings indicated that above and beyond sociodemographic factors, social capital did not explain a reside nt’s mental distress. Social Structure and Social Capital There was limited evidence to support the relationship between social structural factors (incom e inequality and poverty) an d social context (social capital indicators such as social trust, informal social engagem ent, organizational activism, and mutual aid). As all variabl es are ecological in nature and pertinent hypotheses were related to tests of a ssociation only, correlational analysis was performed. Only social trust was c onsistently negatively associated with both poverty and income inequality (Table 4.39). Correlations of -.402 (p = .037) and .549 (p = .003), respec tively, point to 16% to 30% of the variance in social trust explained by social structure, controlling for no other predictors. However, no support was found regarding any other soci al capital indicator and income inequality or poverty.

PAGE 202

188 Table 4.38 Community Social Capital Influences on Mental Distress Community Unconditional1 Conditional1 Characteristic Event Rate Ratio 95% confidence interval p value Event Rate Ratio 95% confidence interval p value Hypothesis No predictors model intercept, 00 0.0333.54 3.29 3.81 <.0001*** Level 1 control model intercept, 00 0.04 3.70 3.40 4.03 <.0001*** Model for social trust 8a1 intercept, 00 0.0343.56 3.30 3.84 <.0001*** 0.0403.71 3.39 4.05 <.0001*** social trust, 01 0.85 0.50 1.47 0.553 0.94 0.52 1.68 0.825 Model for informal social engagement 8a2 intercept, 00 0.0353.54 3.29 3.82 <.0001*** 0.0403.70 3.39 4.03 <.0001*** informal social engagement, 01 0.91 0.28 2.94 0.864 0.98 0.28 3.49 0.974

PAGE 203

189Table 4.38 Community Social Capital Influences on Mental Distress Community Unconditional1 Conditional1 Characteristic Event Rate Ratio 95% confidence interval p value Event Rate Ratio 95% confidence interval p value Hypothesis Model for organizational activism 8a3 intercept, 00 0.033 3.69 3.35 4.07 <.0001*** 0.0383.86 3.46 4.31 <.0001*** organizational activism, 01 0.58 0.24 1.40 0.217 0.57 0.22 1.47 0.235 Model for mutual aid 8a4 intercept, 00 0.033 5.79 2.43 13.78 <.0001*** 0.0395.89 2.30 15.11 0.001** mutual aid, 01 0.91 0.77 1.07 0.255 0.91 0.76 1.10 0.318 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, Including gender, income, race/ethnicity, and age. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 204

190Table 4.39: Associations between Social Structural Factors and Social Capital Indicators1 Indicator Income Inequality Povertyp-value Hypotheses Social Trust -.549 .003** 2b1 -.402 .037* 2b5 Informal Social Engagement -.286 .148 2b2 -.246 .215 2b6 Organizational Activism -.197 .324 2b3 -.100 .619 2b7 Mutual Aid -.113 .574 2b4 -.167 .404 2b8 1Analysis completed using correlational analysis as hierarchical linear modeling not appropriate. *significant at p<.05 **significant at p<.01 Cluster 3: Behavior only partially me diates social context and disease Tests were performed to investigate t he possible direct effects of health behaviors (physical activity, Body Mass Index, and smoking) on the three outcomes under study (hypertension, gener al health status, mental distress), direct effects of social capital on t he health behaviors, and the extent of the mediating role of these behaviors on the relationship between social context and the outcomes. The strongest evidence was f ound for the direct effect of behavior on all three outcomes. In addi tion, some findings point ed to a direct effect of social capital on all three behaviors. T here was limited evidence to support the role of health behaviors as only mediating the association between social capital and hypertension, general health status, or mental distress. Significant results

PAGE 205

191 were found for organizational activism’s direct and indirect influence on general health status. In general, although data re vealed the significant role of health behavior in shaping hypertension, general health status, or mental distress directly, findings were mix ed regarding social context ual influences on behavior. Health Behavior and Health Outcomes The testing of the effects of behav ior on hypertension, general health status, and mental distress was model ed using OLS, as these variables represented only the individual level of analysis and therefore were not appropriate for hierarchical linear modeling. As displa yed in Table 4.40, results demonstrated that the odds of reporting hypertension increased if one is inactive (OR 1.11, 95% CI 1.03 – 1. 19, p = .006) or overweight or obese (OR 2.52, 95% CI 2.34 – 2.73, p < .0001) but no significant associ ation was found for smoking (OR 0.95, 95% CI 0.87 – 1.03, p = .2063), after accounting for individual sociodemographic characteristics. In regards to general health status, one was more likely to report fair or poor health if one was inactive (OR 1.62, 95% CI 1.48 – 1.77, p < .0001), overwei ght (OR 1.47, 95% CI 1. 34 – 1.61, p < .0001), or smoked (OR 1.66, 95% CI 1.50 – 1. 83, p < .0001), after individual sociodemographic characteristics were c ontrolled. Deleterious mental health effects of engaging in risk behavior were fo und as well. Days of reported mental

PAGE 206

192 distress significantly increased with physical inactivity (parameter estimate 0.35, se 0.079, p < .0001), overwei ght/obesity (parameter esti mate 0.62, se 0.071, p < .0001) or smoking (paramet er estimate 2.12, se 0. 127, p < .0001). Standardized estimates showed that this negative in fluence was not uniform across behavior, with smoking (0.11963) comparably the mo st influential, and weight (0.06331) and inactivity (0.03246) following. Health Behavior and Social Capital Overall, very little evidence was f ound supporting the role of community social context in shaping individual health behavior. This finding was consistent across behavior, regardless of which soci al capital indicator was modeled, with and without accounting for individual so ciodemographic characteristics (Table 4.41 – 4.43). The only exc eption was the influence of social trust on physical activity. Both unconditional and conditio nal models demonstrat ed that the odds of a resident reporting limited or no activity was signific antly lower (OR 0.51, 95% CI 0.28 – 0.90, p = .023), if s/he lives in a community with higher levels of social trust (Table 4.41). Living in a community with higher levels of social capital did not significantly influence the odds of a resident being overweight or obese (pvalues ranged from .158 to .537) or smoking (p-values ranged from 269 to .646).

PAGE 207

193 Table 4.40: Influence of Risk Behavior on Self-Reported Health1 Hypertension General Health Status Mental Distress Model Odds Ratio 95% confidence interval p-value Odds Ratio 95% confidence interval p-value Parameter estimate p-value Standardized estimate Control variables: gender 0.94 0.88 1.01 0. 095 1.010.92 1.100.8605 1.37(0.11) <.0001*** age 2.24 2.16 2.33 <.0001*** 1.651.58 1.73<.0001*** -0.51(0.053) <.0001*** income 0.81 0.78 0.84 <.0001 ***0.510.48 0.54<.0001*** -0.65(0.055) <.0001*** race/ethnicity 1.20 1.13 1.28 <.0001***1.261.17 1.35<.0001*** -0.2(0.1) 0.0466* Health behaviors: physical activity 1.11 1.03 1. 19 0.006** 1.621.48 1.77<.0001*** 0.35(0.079) <.0001***0.03246 body mass index 2.52 2.34 2.73 <.0001***1.471.34 1.61<.0001*** 0.62(0.071) <.0001***0.06331 smoking 0.95 0.87 1.03 0. 2063 1.661.50 1.83<.0001*** 2.12(0.127) <.0001***0.11963 Hypothesis 3a Hypothesi s 6a Hypothesis 9a 1Analysis completed using OLS and logistic regressi on as hierarchical linear modeling not appropriate. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 208

194 Table 4.41 Community Social Capital Influence on Individual Physical Activity Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis No predictors model intercept, 00 0.500 1.24 1.12 1.37<.0001*** Level 1 control model intercept, 00 0.0411.02 0.88 1.190.777 Model for social trust 3b1 intercept, 00 0.040 1.27 1.15 1.39<.0001 ***0.0311.04 0.90 1.210.542 social trust, 01 0.50 0.27 0.960.037* 0.51 0.28 0.900.023* Model for informal social engagement 3b2 intercept, 00 0.050 1.24 1.12 1.37<.0001 ***0.0411.02 0.88 1.180.811 informal social 0.47 0.10 2.260.336 0.48 0.11 2.090.314 engagement, 01

PAGE 209

195 Table 4.41 Community Social Capital Influence on Individual Physical Activity Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Model for organizational activism 3b3 intercept, 00 0.0531.27 1.11 1.46 0.002**0 .0441.04 0.88 1.240.620 organizational 0.71 0.21 2.43 0.567 0.74 0.23 2.390.600 activism, 01 Model for mutual aid 3b4 intercept, 00 0.0470.50 0.16 1.58 0.226 0.0380.37 0.13 1.100.072 mutual aid, 01 1.19 0.96 1.48 0.114 1.21 0.99 1.490.064 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, income, race/ethnicity, and age. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 210

196 Table 4.42 Community Social Capital Influence on Individual Body Mass Index Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis No predictors model intercept, 00 0.030 1.31 1.21 1.43 <.0001*** Level 1 control model intercept, 00 0.0333.79 3.26 4.39 <.0001*** Model for social trust 3b5 intercept, 00 0.033 1.31 1.20 1.43 <.0001*** 0.0343.76 3.23 4.37 <.0001*** social trust, 01 1.11 0.62 1.98 0.724 1.25 0.69 2.29 0.446 Model for informal social engagement 3b6 intercept, 00 0.033 1.31 1.21 1.43 <.0001*** 0.0333.79 3.27 4.40 <.0001*** informal social 1.09 0.29 4.05 0.895 1.51 0.39 5.86 0.537 engagement, 01

PAGE 211

197 Table 4.42 Community Social Capital Influence on Individual Body Mass Index Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Model for organizational activism 3b7 intercept, 00 0.031 1.40 1.25 1.56 <.0001*** 0.0344.00 3.38 4.73 <.0001*** organizational 0.42 0. 16 1.15 0.088 0.47 0.16 1.37 0.158 activism, 01 Model for mutual aid 3b8 intercept, 00 0.029 2.60 1.002 6.73 0.049* 0.0336.42 2.28 18.120.001** mutual aid, 01 0.88 0.73 1.05 0.152 0.9 0.74 1.10 0.300 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, income, race/ethnicity, and age. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 212

198 Table 4.43 Community Social Capital Influence on Individual Smoking Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis No predictors model intercept, 00 0.038 0.28 0.26 0.31 <.0001*** Level 1 control model intercept, 00 0.0451.02 0.87 1.20 0.801 Model for social trust 3b9 intercept, 00 0.036 0.28 0.26 0.31 <.0001 ***0.0411.01 0.86 1.19 0.892 social trust, 01 1.34 0.72 2.50 0.345 1.44 0.74 2.82 0.277 Model for informal social engagement 3b10 intercept, 00 0.034 0.28 0.26 0.31 <.0001 ***0.0411.03 0.87 1.21 0.747 informal social 2.23 0.56 8.92 0.244 2.31 0.50 10.580.269 engagement, 01

PAGE 213

199 Table 4.43 Community Social Capital Influence on Individual Smoking Community Unconditional1 Conditional1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Model for organizational activism 3b11 intercept, 00 0.042 0.29 0.25 0.33 <.0001 ***0.0491.04 0.86 1.26 0.672 organizational 0.76 0. 24 2.44 0.628 0.75 0.21 2.68 0.646 activism, 01 Model for mutual aid 3b12 intercept, 00 0.041 0.24 0.08 0.74 0.016* 0.0480.73 0.21 2.49 0.596 mutual aid, 01 1.03 0.83 1.28 0.78 1.07 0.85 1.35 0.571 1Unconditional models do not consider any additional covariates in the model at level 1; conditional models account for individu al sociodemographic factors, including gender, income, race/ethnicity, and age. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 214

200 Social Capital, Health Behavio r, and General Health Status As evidence was found for the direct association between two correlates of social capital, organizational acti vism and social trust, and general health status, models were construc ted that included health behav iors in order to test for mediating effects. Results demonstr ated that the stro nger organizational activism in the community in which one resided was statistically significantly associated with lower odds of a resident re porting his/her health as fair or poor (Table 4.44), both when controlling for i ndividual characteristics (OR 0.19, 95% CI 0.05 – 0.69, p = .014) and accounting fo r individual health behaviors (OR 0.19, 95% CI 0.05 – 0.66, p = .012). This fi nding established that the influence of social context on health status was not co mpletely mediated by individual factors in these data. Different results were obtained for models test ing social trust. Although findings demonstrated that higher levels of social trust in the community in which one resided was statistically sign ificantly associated with lower odds of a resident reporting his/her health as fair or poor (Table 4.45), this only occurred when controlling for individual characteri stics (OR 0.47, 95% CI 0.23 – 1.00, p = .05), but was non significant after account ing for individual health behaviors (OR 0.54, 95% CI 0.25 – 1. 18, p = .117). Once a resident’s level of activity, weight, and smoking status were added to the model the level of social trust in the community no longer explained the odds of reporting fair or poor health.

PAGE 215

201 Table 4.44 Community Organizational Activism Influence on G eneral Health Status,With and Without Mediators Added Community Without Health Behaviors1 With Health Behaviors Added1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Organizational activism model 6b3 intercept, 00 0.044 0.21 0.17 0.26 <.0001*** 0.0370.07 0.06 0.09 <.0001*** physical activity, 10 1.57 1.48 1.67 <.0001*** body mass index, 20 1.43 1.35 1.51 <.0001*** smoking, 30 1.70 1.54 1.88 <.0001*** organizational activism, 01 0.19 0.05 0.69 0.014* 0.19 0.05 0.66 0.012* 1Both models account for individual sociodemographic covari ates including gender, income, race/ethnicity, and age. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 216

202 Table 4.45 Community Social Trust Influence on Gener al Health Status, With and Without Mediators Added Community Without Health Behaviors1 With Health Behaviors Added1 Characteristic Odds Ratio 95% confidence interval p value Odds Ratio 95% confidence interval p value Hypothesis Social Trust Model 6b1 intercept, 00 0.048 0.19 0.15 0.23 <.0001*** 0.0500.07 0.05 0.08 <.0001*** physical activity, 10 1.57 1.48 1.67 <.0001*** body mass index, 20 1.43 1.35 1.51 <.0001*** smoking, 30 1.70 1.54 1.88 <.0001*** social trust, 01 0.47 0.23 1.00 0.051* 0.54 0.25 1.18 0.117 1Both models account for individual sociodemographic covari ates including gender, income, race/ethnicity, and age. *significant at p<.05 **significant at p<.01 ***significant at p<.0001

PAGE 217

203 Summary of Findings Although the data did not support the majo rity of hypotheses in this study, there were several significant findings. Table 4.46 summarizes overall results. The data demonstrated some evidence of t he mediating role of health behavior on the relationships between social struct ural and social cont extual inequalities and self-reported health. Direc t effects were found between dimensions of social capital, organizational activism and soci al trust, and general health status. The direct effects of income inequality or pover ty did not explain an individual’s health status once behavioral factors were consider ed. In essence, the level of social structural inequality in the community in which one lived did not emerge as an independent influence on a resident’s self-reported health. The negative influence of income inequality or povert y in the community in which one lived appeared to negatively affect health through its erosion of social trust, which impacted a resident’s health behavior, thereby worsening self-reported health, including hypertension, general health, or mental distress.

PAGE 218

204 Table 4.46: Summary of Significance Cluster Hypothesis Support Behavioral variables only partially mediate social structure and disease 1a Social structural inequality in the community in which one resides will positively influence self-reported hypertension No 4a Social structural inequality in the community in which one resides will negatively influence self-reported general health status. Partial 7a Social structural inequality in the community in which one resides will negatively influence self-reported mental health. No 1b The effect of social structure on self-reported hypertension is only partly mediated by known risk factors (BM I, physical activity, smoking). No 4b The effect of social structure on self-reported general health status is only partly mediated by known risk factors (BMI, physical activity, smoking). No 7b The effect of social structure on self-reported mental health is only partly mediated by known risk factors (BM I, physical activity, smoking). No 1c Greater social structural inequalities in the community in which one resides positively influences engaging in high risk behavior. Partial Social context partially mediates social structure and disease

PAGE 219

205 Table 4.46: Summary of Significance Cluster Hypothesis Support 2a The level of social capital in the community in which one resides influences self-reported hypertension No 5a The level of social capital in the community in which one resides influences self-reported general health status. Partial 8a The level of social capital in the community in which one resides influences self-reported mental health. No 2b Social structural inequalit ies will be negatively asso ciated with a salubrious social context. Partial 2c Social structural inequality in the community in which one resides will positively influence self-reported hypertension after controlling for community social context. N/A* 5b Social structural inequality in the community in which one resides will negatively influence self-reported general health status after controlling for community social context. N/A* 8b Social structural inequality in the community in which one resides will negatively influence self-reported mental health after controlling for community social context. N/A* Behavior only partially mediates social context and disease

PAGE 220

206 Table 4.46: Summary of Significance Cluster Hypothesis Support 3a Engaging in risk behavior (BMI, physical activity, smoking) is positively associated with self-reported hypertension Partial 6a Engaging in risk behavior (BMI, physical activity, smoking) is negatively associated with self-reported general health status. Supported 9a Engaging in risk behavior (BMI, physical activity, smoking) is negatively associated with self-reported mental health. Partial 3b Weaker social context in the community in which one resides positively influences engaging in high risk behavior. Partial 3c Weaker social context in the community in which one resides positively influences self-reported hypertension after controlling for individual risk behavior. N/A* 6b Weaker social context in the community in which one resides negatively influences self-reported general health status after controlling for individual risk behavior. Partial 9b Weaker social context in the community in which one resides negatively influences self-reported mental health after controlling for individual risk behavior. N/A* *Tests of mediating effects no longer pertine nt as direct effects are non-significant.

PAGE 221

207 CHAPTER 5: DISCUSSION AND IMPLICATIONS Discussion Summary of Findings The purpose of this study was to in vestigate the associations between social structural factors (e.g., income inequality and poverty), community social context (e.g., social capital dimensions), and individual characteristics (e.g., risk behavioral factors) and self-reported correlate s of disease (hypertension, general health status, and mental distress). This study examined the extent to which upstream structural and contex tual factors indirectly affect disease through their influence on risk behavior and the degree to which social structure and context independently influence self -reported disease. The majority of findings supported the ro le of mediating factors in social structural and social cont extual influences on self-r eported health. In addition, little support was observed for most of the hypotheses in this study. No direct effects of social structural inequalities on any of the outcomes were found. There were only two instances of the direct effect of social capital. As the organizational activism of the communi ty decreased, the odds of a resident reporting fair/poor health increased. No ot her indicator of soci al capital exerted direct influence, once individual characteri stics were considered. In regards to

PAGE 222

208 contextual influences on health behavior, as the resident’s community’s level of social trust decreased, the odds of r eporting not meeting re commended levels of physical activity increased. Other signifi cant findings in this study confirmed those of previous investigations. S pecifically, income inequality and poverty were negatively associated with community social trust. In addition, risk behaviors were associated with indi viduals increased odds of reporting hypertension, fair/poor general health, and more days of mental distress. Once individual characteristics of the residents were included in the models, any significant direct effect of social structur al inequalities on the health outcomes disappeared. There was only mixed evidence supporting the influence of social contextual factors on self-report ed health. Results suggested that social structural and contextual inequalities di d shape certain individual behaviors. In addition, results confirmed those of numerous previous studies to indicate that engaging in risk behavior did explain self-r eported hypertension, poor/fair health status, and mental distress. In essence, it appeared that any significant negative influence of social structural inequalitie s on health was mediated by their effects on reducing social trust. Living in a commu nity with lower levels of social trust was associated with engaging in limited/no physical activity. Reduced physical activity increased the odds of reporting hypertension. Inactivity was also associated with reporting fair /poor general health as well as more days per month of mental di stress.

PAGE 223

209 Neither income inequality nor poverty in the community in which a resident lived explained reports of hypertension or mental distress. Because no direct relationships were found, tests of hypot heses that behavioral variables only partially mediate social structure and disease were not performed for these outcomes. One exception to the dear th of evidence to support associated hypotheses was the finding that living in an impoverished community did increase one’s odds of reporting fair/poor health. However, once adjustments were made for individual sociodemographi c factors, poverty no longer directly explained general healt h status. Findings indicated that most of the so cial contextual fa ctors under study did not have any direct influence on any of the three health outcomes. No significant associations were observ ed between informal social engagement or mutual aid and hypertension, general health status, or mental distress. Significant negative effects of the other tw o correlates of social capital on general health status were noted. Lower levels of either community social trust or organizational activism, which includes org anizational activism were associated with increased odds of a resident repor ting fair/poor health. However, once sociodemographic and behavioral variables were included in the model, only organizational activism retai ned its direct effects. Ther efore, this association was the only direct relationship found between a social contextual factor and a selfreported outcome. None of the social capi tal indicators significantly influenced reports of hypertension or mental distre ss directly. In re gards to support of

PAGE 224

210 indirect or mediated relationships, only soci al trust was indirectly associated with greater odds of reporting hypertension and mental distress, through its effects on health behavior. These results confi rmed prior evidence that the social environment exerts its influence primarily through more proximal individual behavior. Additional previous findings were conf irmed as well. Consistent with some previous literature, both income inequality and poverty were negatively associated with social trust, although these social structural variables were not related to any of the other indicators of so cial capital. In turn, community levels of social trust significantly predicted a re sident’s activity level, however had no influence on whether the individual was overweight or smoked. No other dimension of social capital significantly influenced individual risk. In addition, only limited direct influence of social st ructure on health behaviors was observed. Income inequality and poverty had no statistica lly significant effect on whether a resident was either inactive or ov erweight or obese, but did approach significance in explaining smoking. Again, similar to previous findings, most of the statistically significant relationsh ips were observed between individual risk behavior and self-reported health. Physicall y active individuals and those who were of normal BMI reported less hypertens ion, better general health status, and less mental distress. Smoking was not associated with reporting hypertension, although it was related to general and mental health status.

PAGE 225

211 In sum, findings from this study were not only inconsistent as a whole, but several were contrary to previous literatu re as well. In earlier studies, the direct negative influences of income inequality and poverty on general health status were found at multiple levels of aggr egation (e.g., state, metropolitan area, and county). In addition, results had supported a stronger effect of social trust on health outcomes. In comparison, soci al structural and social contextual characteristics of the community in whic h one lived had relatively little direct significant influence on either engaging in risk behavior or reporting poor health in this investigation. Tests of hy pothesized mediating roles of context and behavior were conducted only where di rect effects were found between environment and individual factors. Solely organizational activism retained its significant direct effect on general health status, once individual characteristics (sociodemographic and behavioral) were considered. These somewhat unanticipated results point to several possible ex planations for such findi ngs. Limitations of Study There were a number of po ssible limitations of this study, which may have resulted in few significant findings. Methodological issues included: use of secondary data sources and sampling, vari able selection and measurement, and design issues. The use of secondar y data and the linking of data sets constrained the use of the data and rest ricted the power with which conclusions

PAGE 226

212 could be drawn regarding the hypotheses. Variable selection and measurement were also negatively influenced by the data in that the scale of measurement of the constructs under study and prior cons truction of indices used served to introduce bias and reduce reliability and valid ity. Finally, the most limiting factor of the study was the design type. Cro ss-sectional studies such as this one cannot infer causality, but rather only suggest associations, as exposure and disease were measured at the same time. Data Sources and Sampling Both the data sources selected an d sampling procedure used may have limited the possibility of finding support fo r the hypotheses in this study. Weaknesses of this study related to t he linked nature of t he data and the small sample size at level-2. Because this was a study using secondary data, the researcher was limited to using only those communities where structural and contextual data could be linked by FIPS c odes to individual data obtained from the BRFSS. The three data sources em ployed may have resulted in somewhat different comparable sociodemographic characteristics of each community, thereby resulting in different, less than va lid or reliable comparisons. Another related difficulty was the linking of data, wh ich may not be specific to this study, but rather a growing issue in multileve l investigations. Sampling bias was introduced through the use of the Social Capital Community Benchmark Survey,

PAGE 227

213 as this survey was not a probability sample, but rather was a convenience sample obtained at an annual meeting of community foundations. In addition, several of the communities did not have FIPS codes assigned, as they were geographically diverse. The lack of av ailable FIPS codes for all communities may have biased the findings. In regards to sample size, only 27 out of a possible 41 communities were eligible fo r inclusion, thus reducing both possible variance and power. Current suggestions in the literature indicated that having a sample size of 25 – 30 at level-2 is the lo wer limit to obtain confidence in results (Kreft & de Leeuw, 1998). Conclusions draw n from the subsample employed in the study might have been quite different t han if the whole dataset were utilized the inclusion of more communities may yield different results. Variable Selection and Measurement Additionally, the selection and meas urement of the variables may have constrained the capacity of this study to support the hypotheses. Variables were selected based upon theoretical consi derations and informed by previous literature. However, thei r use was restricted by their measurement, which was not defined by the researcher. Each dat aset used introduced limitations to this study. In regards to BRFSS data, there may have been threats to the validity of multiple variables selected. For exam ple, there was some suggestion that the use of BMI as indicator of obesity may have limited validity if the muscle mass if

PAGE 228

214 the individual is not considered (Nati onal Heart, Lung, & Blood Institute, 2005). Also, several of the original sociodemogr aphic variables were not in continuous form, but rather already in ordinal-level categories (e.g., income), which limited their utility. Additionally, the outco mes (e.g., CVD or hypertension) were assessed with only single item measur es and although had demonstrated good reliability and validity, self -report measures still might have introduced bias from sources including recall, resulting in underes timation of poor heal th. In addition, as reliability and validity of the scores measuring the constructs are sample specific, conclusions drawn from this study must consider the possibility that the variables measured did not represent the sa me level of reliability and validity that have been shown in previous stud ies utilizing the same data. Measures of social capital were selected based upon availability. The selected indices were created by the aut hors of the Social Capital Community Benchmark Survey and reflected mean soci al trust, informal social engagement, organizational activism, and mutual aid. It might hav e been more suitable to use individual items or a subset of questions Unfortunately, there were inadequate data on bridging social capital, the correla te of reciprocity, and no index available assessing global social capital within a community. Currently, there have been no studies in public health literature in vestigating the influence of multiple dimensions of social capital on specific health outcomes. Future studies should include selection of cert ain items and create additional groupings/indices. This study was just an initial step in examin ing whether social capital dimensions

PAGE 229

215 previously identified in the literature had direct or indi rect effects on self-rated health outcomes. The lack of support in this study fo r a significant association between income inequality and general he alth status was in contra st to previous findings in the social epidemiology literature. There are several reasons for this discrepancy – all methodological in nature. The small sample at level 2 and the restricted range of gini coefficients (0.4 – 0.49) may have resulted in no relationship being detected. In addition, communities themselves were highly geographically heterogeneous within the sample – comparisons were made between individual counties, cluster of contiguous counties, and lightly populated states. There may have been more va riance within a community than between them. Due to the design and the need to link secondary data sets, the sample was intended to represent the same population, but the data sources did reflect different levels of aggregation. The social structural variables, inco me inequality and poverty, may have been constrained not by measurement alon e, but also by their restricted opportunity for variance. For example, due to a narrow range of Gini values, the data selected may not demonstrat e the true direct effect of income inequality on both health behaviors and health outcomes. In addition, the use of too broad of an indicator with this type of data may have limited validity and usefulness (e.g., using the community-level proportion of t hose living at or below 200% FPL may have been too broad an indicator and actual ly washed out possible effects;

PAGE 230

216 moreover, several of the communities included in the dataset were intrageographically heterogeneous). Therefore, the ex planatory value of poverty in elucidating disparities in chronic disease and self-r eported health status might have been better assessed by a more sensitive indicator. The lack of evidence regarding social structural inequalitie s influence on health might have been due to the lack of important characteristics in cluded in the models, such as residential segregation and/or political environment. In addition, Macintyre and Ellaway (2003) suggest that as aspects of place s hape individual characteristics, and visa versa, (e.g., historical shi fts in industry creates context-specific opportunities for individual occupation and hence, income – and supply of trained individuals impacts demand/local labor market), m easuring structural effects while controlling for individual level factors may result in a “partialling” fallacy whereby overcontrol of characteristics of the indivi dual may result in insignificant findings where a relationship may exist; variance is concealed by correlates (Macintyre & Ellaway, 2003). Lastly, a limitation of this study that had both measurement and design implications pertained to stability and fluidity of the composition of each community. The role of population densit y and transience might have biased the assessment of the “true” val ue of community constructs (e.g., individuals residing in communities in 1999 might have been di fferent from individuals residing in those communities in 2001). For example, gross migration ra tes (including into and out of the state) from 1995 – 2000 r anged from 111.6/1000 in Michigan to

PAGE 231

217 303.3/1000 in Colorado (Census 2000 Special Report, 2003). Generally, the Southern region witnessed more over all migration, with the Northeast experiencing the least. This limitation might have influenced both social structure and social context, thereby influencing exposure (both time and type) to the detrimental aspects of the so cial environment under study. Design Issues There were several limitations in t he design of this study. Concerns included the role of time and exposure in addition to the absenc e of moderators. The cross-sectional nature of the dat a did not permit the design to examine temporal effects on the outco mes; length of exposure to the detrimental influence of social structural and social contextu al inequalities could not be considered. The result was that it c ould not be demonstrated in this study that the level of poverty (or dose, as it were) influenci ng the outcomes was cumulative in nature or an instantaneous effect. Only a longi tudinal design would have permitted a more reliable assessment of exposure. Because there was no one nationally available dataset that includes information on multiple levels of the env ironment, data had to be linked in order for this investigation to be conducted. This issue restricted both the type of design as well as the type of inferences that can be made. Due to the limitations of linking data, the social trust, for in stance, that was being assessed in the

PAGE 232

218 community by the SCCBS may not hav e represented the same group of individuals represented in the BRFSS. Therefore, conclusions drawn regarding the influence of social trust in the communi ty in which one resided on the odds of an individual rating his/her general health as poor must be considered in light of this limitation. Threats to external validi ty of findings include those related to the lack of randomization with which t he communities in the SCCBS were selected. Selection and setting bias may be operatin g (Cook & Campbell, 1979) – whereby unmeasured attributes of those resident s, or their communities, who responded to the 2000 Social Capital survey diffe r from individuals (and areas) responding to the 2001 BRFSS questionnaires. Thes e characteristics may have impacted the generalizability of findings, alternatively known as po pulation, ecological, and temporal validity (Onwuegbuzie, 2003). Specifically, individuals residing in communities in 1999 might have been different that individuals residing in those communities in 2001. In addition, the transience of residents might have influenc ed the social structure and social context of the community, as discussed earlier, as well as the physical characteristics (e.g., land use, zoning, dev elopment or dilapidatio n) of the areas. Another limitation was that the design did not examine the possible cross-level interactions among predictors or the role of moderators. For example, the moderating influence of individual sociodem ographic characteri stics (i.e., gender) on the influence of poverty on behavior or mental distress was not included.

PAGE 233

219 Eliminating this path may have led to erro neous results. The lack of consistent findings might not be due to no relationship existing, but rather to relationships not measured in the models. Perhaps poverty’s negative influence on selfreported mental distress is significant only for young men of color or perhaps it shapes activity levels only for older wom en. This design did not allow for these considerations. Finally, the design may have reflected a fully mediated relationship by a variable not included in the model. Contributions of this Study and Implications for Public Health This study contributed to public health by adding to the knowledge base in three areas where major gaps existed in knowledge related to the role of the social environment in health. The first contribution was that it added to the current empirical data regarding studies on the influence of social structure on specific outcomes. Most of what we know has come from studies on morbidity/mortality rates, life expectan cy, and general health. Another advantage was that this study investigated hypert ension, general health status, and mental distress outcomes using data t hat was expressly collected to study the effects of social capital. Finally, this study prov ided additional knowled ge on the structural and contextual influences in which, specifically, risk behavior occurred.

PAGE 234

220 Given the theoretical rationale and pr evious empirical studies, a greater number of significant findi ngs supporting the hypotheses were anticipated. Although findings from this study were more limited and failed to conclusively demonstrate the extent to which contex t affects behavioral and self-rated health outcomes, the possibility that behavior only partially mediates the relationship between context and disease remains. This investigation pointed to the need for further examination of the effects of social context on the initiation and maintenance of health behaviors in order to broaden our understanding as well as to incorporate relevant findings in to public health policy and practice. Moreover, this investigati on added to the current literat ure by demonstrating that, despite methodological limitations, ther e is empirical evidence to support the influence of broader factors on health dispar ities. For example, poor social trust in the community in which one resi ded was associated with the resident participating in limited, if any, physical acti vity. If residents of a community do not trust their neighbors or environment, t hey may be less willing to spend time outside engaging in activity, such as walk ing. This study’s findings support the continued study of macro and meso dete rminants of the social environment and their influence on the public health burden of chronic disease. By doing so, this study extended the understanding of the multilevel natur e of health disparities and the need for multilevel interventions to reduce them (e.g., state-level policies targeting improved funding for well-lit si dewalks, community advocacy to obtain funds and implement changes to promote community cohesion and/or establish

PAGE 235

221 new norms related to physical activi ty, individual-level behavioral change strategies). Future implications of this study for public health included contributions to both research and practice. In a mo re general sense, limited evidence demonstrated that interventions focusing on disparities in multiple health outcomes should simultaneously address soci al and behavioral factors to inform service delivery and health policy. Severa l practical implications derived from this study include those related to design and data. One suggestion that has received att ention more recently is the need for quasi or experimental studies to relie ve the reliance on observational investigations (Berkman, 2004) Randomized community trials that permit causal inferences to be made compel the scient ist to clearly and precisely identify exposure (Oakes, 2004). For example, te sting contextual influence on health through evaluation of housin g policies, to which families have been randomly assigned to programs (K aufman et al., 2003). Ot her possibilities include community intervention trials aimed at health promotion (e.g., targeted zoning of affordable fruit/vegetable grocer y stores). In order to implement many of these studies, improved data is needed. In regards to its influence on future studies, evidence from this work may promote the needed restructuring of lar ge national surveillance systems to include contextual data. For exampl e, assessment of ph ysical and social characteristics of local environments through “ecometrics” incorporating

PAGE 236

222 qualitative methods of systematic soci al observation (Raudenbush & Sampson, 1999; Raudenbush, 2003) would enrich av ailable data and add to definitional clarity of constructs mo st commonly used in public health. Use of qualitative data, such as that collected vi a focus groups and open-ended interviews regarding notions of community, social tr ust, racial relations, and poverty, would augment currently available survey data. In addition to these issues, implications include the need to examine certain types of outcome data in the study of neighborhoods and health – such as measur es of variance as well as knowledge suggested by measures of associati on (Merlo et al., 2005a,b). Also, the availability of longitudinal data that includes important biomarkers (e.g., cortisol, fibrinogen levels, blood pressure, norepi nephrine and epinephrine levels) is imperative. By broadening influential public health surveillance systems to include these types of dat a, the knowledge base from which interventions are developed and conducted for diverse popul ations regarding common behavioral risk factors for chronic disease (e.g., smoki ng, physical activity, dietary practices, substance use) is expanded. Specifically in regards to hypertension, this study provided a significant contribution to understanding the relations hips between social structural, contextual, and behavioral aspects of self -reported hypertension. Results from this study provided further evidence t hat if the social context within which behavior occurs is not considered, inte rventions targeting behavior change as a prevention strategy will have limited effe ctiveness. Educating individuals

PAGE 237

223 regarding the benefits of regul ar exercise may not be as valuable in changing behavior if the individual lives in an unsafe community, with no clean areas in which to walk, and in addition, without community support to shift current practices to new, more heal thy norms of behavior. In addition, this work may inform an expansion of social and structural changes. For example, to reduce the disparate burden of chronic disease, intervention targets might include: inst ituting regulatory change s in political and economic policy which currently shape market influences which produce and perpetuate social inequalities (Kapl an & Lynch, 1999; Terris, 1999); strengthening social capital within co mmunities (Kawachi, 1999) or perhaps directing prevention efforts towards devel oping community capacity (Elliott et al., 1998); developing models that are aimed at shaping local public agendas to include community-level CVD prevention (Finnegan, Viswanath, & Hertog, 1999; Schmid, Pratt, & Howze, 1995). In light of some of the results of this study (e.g., income inequality and poverty’s negative associ ation with social trust), structural changes based on intervening at the policylevel include institution of a living wage in lieu of the inadequate “minimum wage” as currently legislated. Further studies are needed to examine if reduci ng the experience of poverty (both individually and community-wide) and/or income inequality would result in a commensurate reduction in isolation and disconnection. In a similar vein, improved affordable housing and medical ca re may reduce the disparate health burden of chronic disease current ly plaguing the less fortunate in our country.

PAGE 238

224 Moreover, this study added to the lit erature pertaining to self-reported general health and mental distress. It was unique in that it was the only investigation that exami ned the influence of broader, more fundamental social determinants on general health status and mental distress in a multilevel design, while accounting for individual attributes and behaviors. Previous research had pointed to the need for studies examining mu ltiple dimensions of Health Related Quality of Life indicators in the pursuit of Healthy People 2010 goals of improving quality of life and reducing hea lth disparities (Zack, Moriarty, Stroup, Ford, & Mokdad, 2004). By providing additi onal data assessing the burden and indicators of mental distress in a geogr aphically diverse sample in addition to expanding possible intervention targets to improve self-reported health, this study added to the advancement of knowledge in the field of population health. This study sought to contribute to filli ng some of the gaps present in the social epidemiology literatu re. This work empirically examined the role of social structure on specific health outcomes. Another innovation was the use of data that is expressly collected to study the effects of social capital. This work expanded the knowledge base of structural and contextual influences in which risk behavior occurs. Public health implications of these contributions include practice, policy, and theoretical benefits. In regards to public health practice, findings from this study indicated a need to improve service de livery (e.g., by contextualizing health education programs). There were many possible policy implications of this

PAGE 239

225 investigation. Firstly, national surveys might include assessment of social environmental factors as part of st andard surveillance procedures. This improved surveillance might in clude both the linking capacity of publicly available datasets as well as broadening the inclusi on of contextual factors in behavioral surveys. By finding some supporting evidence of the negative influence of social structural and social contextual inequal ities on health behavior and self-rated health despite methodological limitati ons, this study may inform not just assessment of public health, but also in tervention strategies by adding to the growing data on the need to target wider, more fundamental, levels of the social environment, such as local, st ate, and federal law. Recognition that public health policy begins with the economic and is driven by the politic al is a critical step in order to then envision changes in polit ical and economic policy which shape market influences which produce and perpet uate social inequalities in health (Kaplan & Lynch, 1999; Terris, 1999). At the national level, possible policy interventions might include: instituting a living wage in order to both attenuate absolute deprivation as well as narrow the range of inco me inequality; improving education funding to prevent early dropout among at-risk yout h; restructuring public housing to reduce residential segr egation, promote safety and encourage a sense of community; nati onalizing health care to provide for a more equitable distribution of benefits; acce ss to opportunities to partic ipate in the democratic process through legislative and other governmental initiatives.

PAGE 240

226 Finally, this study contributed to the need in present public health literature to explicitly use theory to inform inve stigations. This study extended present empirical testing and applied use of a newer theoretical orientat ion, the ecosocial perspective. The work also dem onstrated the benefits of combining complementary perspectives (ecosocial and political economy) to inform empirical investigations in social epidem iology. The limitations of each were reduced through employ ment of both. In regards to the application and empi rical testing of these theoretical positions, there were several strengt hs and weaknesses evident in both perspectives. Specifically in regards to political economy, the strength of its application in this study is rooted in it s concentration on material conditions and power relations in addition to the notion of nonspecific mortality – getting rid of one disease is ineffective in dealing with public health, because health effects of social inequalities are not manifested in a specific disease per se, but rather are reflected in many diseases (sick indivi duals because a sick society). The use of self-rated health is aligned with this premis e of general susceptibil ity. In addition, it is a formalized and coherent theory. It is integrated and yet relatively parsimonious. However, the application of political economy necessitated an ecological design, due to it s explicit focus on broader aspects of the social and economic environment. Whereas it does expl ain reality in one sense, it fails to account for the influence of individual di fferences and within group variation. This inherent weakness is more evident when studying the multilevel nature of

PAGE 241

227 the social determinants of health. Therefor e, it was complemented by the use of the ecosocial framework. Overall, the ecosocial approach was a better framework for this study. The social production of disease viewpo int does not include a place in its theoretical framework for neither individual agency nor intraindividual (i.e., biological or psychological) influences. The ecosocial perspective better ex plained the nest ed nature of the phenomena under study by combining the social production of disease with biology and ecology in a dynamic proce ss. Its focus includes the physiological pathogenic responses to social structur al conditions (Krieger, 1994, 2001). Through its application, however, several lim itations were found. Several tenets were difficult to operationalize and theref ore unable to be test ed empirically (e.g., pathways of embodiment and the cumula tive interplay between exposure, susceptibility, and resistance). This weak ness was not necessar ily the result of inadequate theory development. It might better reflect t he state of (inadequate) surveillance in public health in that ther e is a complex data re quirement to test this framework. Utilizing this perspecti ve necessitates improved surveillance and availability of multilevel and/ or linkable datasets, in order to include aspects from the structural world to the biological syst em. With improved data, the relevance of the ecosocial framework for pu blic health will surely expand.

PAGE 242

228 Recommendations for Future Research There are multiple recommendations fr om this study for future research, including theoretical, conceptual and methodol ogical areas. At present, research in social epidemiology does not adequately focus on expansion of theory development in the field. The development and use of cohesive, predictable, parsimonious theoretical frameworks woul d assist in advancing this area of scientific inquiry. Moreover, studies are needed which explicitly examine the relative theoretical contribution of differential explanatio ns for inequalities (e.g., material conditions versus psychosocial factor s versus genetic or biological risk). Some pertinent rationales may be mutually exclusive whereas some may provide a complementary framework from which to conceptualize future studies. At this time, it is not known. Further, potential directions of st udy include topics that have both methodological and conceptual implicat ions. Studies are needed to improve assessment of the broader social dete rminants of health. To do so, changes need to be made regarding not just what kind of data that is collected, but also how it is obtained. As biomedical bias shapes research agendas – in addition to the questions posed and studies that are funded, national surveillance systems need to be restructured (through lobbyin g and engaging of policymakers) to accommodate qualitative data on contex t, such as the aforementioned

PAGE 243

229 systematic social observation. Collection of complem entary forms of data would permit improved investigations utilizing secondary data. In addition to newer forms of data, an expanded use of design types is required. For example, t here is a need for longitudinal studies using multilevel data, which may capture the lag effects of social structural and social contextual inequalities on health as these may not be immediate. Considerations include issues related to lifespan and length of exposure at struct ural and contextual levels for child, adolescent, and adult outcome s. There may be critical periods and transitions in development across the lifespan which may buffer or exacerbate the negative influence of inequalities. Conceptual and methodological advances such as studies utilizing a longitudinal design combining developmental theory and an ecosocial approach may generate important hypotheses related to the role of the life-cycle. Longitudinal studies would also allow per tinent questions taking account of the level of variable most appropriate (s tate, community, neighborhood, block) for what type of timing, for exam ple, temporal (e.g., cohort) effects versus role of (e.g., individual) development. Probability sampling in these studies would also permit causality to be examined and perhaps established. These studies would also utilize improved meas urement and availability of characteristics of the environment (such as integrating local and national qualitative data on social context with enhanced quantitative asse ssment), perhaps using data easily linked (by common geographical identifier) to GIS or ot her federal, state, and

PAGE 244

230 local data. Moreover, there is a need to clarify and consistently refine the operationalization of the constructs we examine in the social epidemiology literature. In regards to methodological innova tion and to advance the field, crosslevel analyses would be a critical com ponent of a future study, as this would enable contextual effects to be isolated from compositional effects. It would also permit a more refined examination of t he subtleties of level-1 and level-2 interactions (e.g., testing whether individuals with a low SES living in a community characterized by high inco me inequality and low social capital reported mental distress more than i ndividuals with a low SES living in a community characterized by low income inequality and high social capital or whether living in a community characteriz ed by high poverty and disparate levels of social capital differently affect ed whether one engaged in risk behavior and reported hypertension). This analysi s process may also tease apart the differential effects of types of social stru ctural inequalities and correlates of social capital in explaining risk behavior and self-reported health for disparate populations (e.g., racial/ethnic groups, men vs. women, etc.). In regards specifically to social c ontextual variables, separate analyses might be conducted for both composite measur es as well as a global indicator of social capital. This method has been suggested, as it may provide both summary information about the relationshi ps between the variables as well as elucidate the possible distinct effect s of individual items (Putnam, 2004).

PAGE 245

231 Additional considerations regarding future studies of social capital include investigations of the com parative contribution of each dimension in explaining disparities in specific health outcomes in addition to general susceptibility of subpopulations; a follow-up study to the Social Capital Community Benchmark Study; examinations of the role of densit y and transience of the population on the social capital of a community and its possible health effects Conclusion Two overarching themes emerge from this res earch. One theme that surfaces relates to the need to include both physical and mental aspects of health when studying chronic disease. Isolating the two hemispheres of experience serves to limit our underst anding of pathology and the role of selfrated health in chronic disease. Further, harkening back to Rose (1992) and Wilkinson (1996), although the stated focus in public hea lth research is on populat ion health outcomes, there is still a dearth of evidence on the ways in which population-based multilevel studies of health disparities in chroni c disease can not just acknowledge, but rather impact broader soci al environmental influences. Understanding precedes advocacy; it is critical to make this fo rm of research germ ane through increasing the role social epidemiological research pl ays in development of health policy.

PAGE 246

232 Ultimately the value of public health res earch is in its relevance and ability to improve the public’s health. The app lication of knowle dge is a critical component of the research process w hereby findings should translate into effective policy initiatives. Knowledge of the issues is insufficient. Heymann and Fischer (2003) ask the question “Will any of this research make a difference in public policy and practice?” Sc ientific study for the sake of research does little to improve public health without generating possible polic y propositions; societal problems necessitate societal solutions (Heymann, 2000). In regards to the first matter, populat ion-based strategies may be limited if they rely on high-risk population behavioral change, to the exclusion of the effects of structural and contextual ineq ualities. Despite mounting evidence that this approach has been limited in reducing th e overall burden of disease, present public health interventions continue to focu s on individual-level risk factors. Although the lack of support for structur al and contextual influences may be interpreted as evidence for compositi onal and/or individual risk factor explanations for health dispar ities, I conceptualize a different interpretation of the significant findings (or lack thereof) of this study. I must be conceded that it is possible that there is no direct re lationship between fundamental or broader factors of the social environment and s pecific health outcomes. However, I interpret the limited results of this study as not indicative of support for the “composition” argument or perspective (focus on individuals rather than contextual), but rather point to t he need for improved measurement and data

PAGE 247

233 issues. The lack of evidence from this study to support the contextual influence of place points to several conditions of th is investigation that may have shaped conclusions found – the restricted range of variables (e.g., income inequality), sample size (e.g., only 27 communities with available data), inability to incorporate temporal effect (e.g., role of transience wi thin communities, length of exposure to environment), and additional issues related to the use of crosssectional secondary data (e.g., incompat ibility of communities between linked datasets and external validity compromises) Despite these concerns, this study has added limited support to the growing evidence base that there are macrolevel structural and contextual influenc es on population health that cannot be reduced to individual or compositional effe cts. Therefore, public health goals, such as Healthy People 2010 twin goals of increasing quality of life while decreasing health disparities, will fail to be met without due consideration to fundamental factors which serve to perpet uate and maintain disparities in health.

PAGE 248

234 References Abu-Omar, K., Rutten, A., & Lehtinen, V. (2004). Mental health and physical activity in the European Union. Soz Praventivmed, 49 (5), 301-309. Abu-Omar, K., Rutten, A., & Robine, J. M. (2004). Self-rated health and physical activity in the European Union. Soz Praventivmed, 49 (4), 235-242. Acevedo-Garcia, D., & Lochner, K. (2003). Residential segregation and health. In I. Kawachi & L. F. Berkman (Eds.), Neighborhoods and health (pp. 265287). New York: Oxford University Press. Acevedo-Garcia, D., Lochner, K. A., Osypuk T. L., & Subramanian, S. V. (2003). Future directions in residential segr egation and health research: A multilevel approach. American Journal of Public Health 93 (2), 215-221. Adler, N. E., Boyce, T., Chesney, M. A., Cohen, S., Folkman, S., Kahn, R. L., & Syme, S. L. (1994). Socioeconomic st atus and health. The challenge of the gradient. American Psychologist, 49 (1), 15-24. Aneshensel, C. S., & Sucoff, C. A. (1996). The neighborhood context of adolescent mental health. Journal of Health & Social Behavior, 37 (4), 293310. American Heart Association. (2001). 2002 Heart and Stroke Statistical Update Dallas, Texas: American Heart Association. American Heart Association. (2004). 2005 Heart and Stroke Statistical Update Dallas, Texas: American Heart Association.

PAGE 249

235 Atlantis, E., Chow, C. M., Ki rby, A., & Singh, M. F. (200 4). An effective exercisebased intervention for improving mental health and quality of life measures: a randomized controlled trial. Preventive Medicine, 39 (2), 424-434. Ayala, C., Greenlund, K. J., & Croft, J. B. (2002). State-specific trends in selfreported blood pressure screening and hi gh blood pressure United States, 1991-1999. MMWR Morbidity and Mortality Weekly Report, 51 456-460. Backett, K. C., & Davison, C. (1995). Lifecourse and lifestyle: the social and cultural location of health behaviours. Social Science & Medicine, 40 (5), 629-638. Bancroft, A., Wiltshire, S., Parry, O., & Am os, A. (2003). "It's like an addiction first thing...afterwards it's like a habi t": daily smoking behaviour among people living in areas of deprivation. Social Science & Medicine, 56 (6), 1261-1267. Barnett, E., & Halverson, J. (2000). Disparities in prem ature coronary heart disease mortality by region and urbani city among black and white adults ages 35-64, 1985-1995. Public Health Reports, 115 (1), 52-64. Bassett, D. R., Jr., Fitzhugh, E. C., Crespo, C. J., King, G. A., & McLaughlin, J. E. (2002). Physical activity and ethnic diffe rences in hypertension prevalence in the United States. Preventive Medicine, 34 (2), 179-186. Bazzano, L. A., He, J., Muntner, P., Vupput uri, S., & Whelton, P. K. (2003). Relationship between cigarette smok ing and novel risk factors for cardiovascular disease in the United States. Annals of Internal Medicine, 138 (11), 891-897.

PAGE 250

236 Belle, D., & Doucet, J. (2003). Poverty, inequality, and discrimination as sources of depression among U.S. women. Psychology of Women Quarterly, 27 101-113. Ben-Shlomo, Y., White, I. R., & Marmot, M. (1996). Does the variation in the socioeconomic characteristics of an area affect mortality? British Medical Journal, 312 (7037), 1013-1014. Berkman, L. F. (2004). Introduction: S eeing the forest and the trees – from observation to experiments in social epidemiology. Epidemiologic Reviews, 26 2-6. Berkman, L. F., & Glass, T. (2000). Social integration, social networks, social support, and health. In L. F. Berkman & I. Kawachi (Eds.), Social epidemiology (pp. 137-173). New York: Oxford University Press. Berlin, J. A., & Colditz, G. A. (1990). A meta-analysis of physical activity in the prevention of cor onary heart disease. American Journal of Epidemiology, 132 (4), 612-628. Blair, S. N., & Brodney, S. (1999). Effects of physica l inactivity and obesity on morbidity and mortality: current evidence and research issues. Medicine & Science in Sports & Exercise, 31 (11 Suppl), S646-662. Blakely, T., Atkinson, J., & O'Dea, D. (2003). No associ ation of income inequality with adult mortality within New Zealand: a multi-level study of 1.4 million 2564 year olds. Journal of Epidemiology & Community Health, 57 (4), 279-284.

PAGE 251

237 Blakely, T. A., Kennedy, B. P., & Kawachi, I. (2001). Socioeconomic inequality in voting participation and self-rated health. American Journal of Public Health, 91 (1), 99-104. Blakely, T. A., Lochner, K., & Kawach i, I. (2002). Metropolitan area income inequality and self-rated heal th--a multi-level study. Social Science & Medicine, 54 (1), 65-77. Blumenthal, J. A., Babyak, M. A., Moore, K. A., Cr aighead, W. E., Herman, S., Khatri, P., Waugh, R., N apolitano, M. A., Forman, L. M., Appelbaum, M., Doraiswamy, P. M., & Krishnan, K. R. ( 1999). Effects of exercise training on older patients with major depression. Archives of Internal Medicine, 159 (19), 2349-2356. Bosma, H., van de Mheen, H. D., Borsboom, G. J., & Ma ckenbach, J. P. (2001). Neighborhood socioeconomic stat us and all-cause mortality. American Journal of Epidemiology, 153 (4), 363-371 Brenner, M. H. (1995). Po litical economy and health. In B. C. Amick III, S. Levine, A. R. Tarlov, & D. Chapman Walsh (Eds.), Society and health (pp. 211-246). Oxford: Oxford University Press. Breslau, N., Kilbey, M., & Andreski P. (1991). Nicotine dependence, major depression, and anxiety in young adults. Archives of General Psychiatry, 48 (12), 1069-1074. Brown, D. W., Balluz, L. S., Heath, G. W., Moriarty, D. G., Ford, E. S., Giles, W. H., & Mokdad, A. H. (2 003). Associations between recommended levels of

PAGE 252

238 physical activity and health-related qua lity of life. Findings from the 2001 Behavioral Risk Factor Surveill ance System (BRFSS) survey. Preventive Medicine, 37 (5), 520-528. Brown, D. W., Brown, D. R. Heath, G. W., Balluz, L., G iles, W. H., Ford, E. S., & Mokdad, A. H. (2004). Associations between physical activity dose and health-related quality of life. Medicine & Science in Sports & Exercise, 36 (5), 890-896. Browning, C. R., & Cagney, K. A. (2002). Neighborhood structural disadvantage, collective efficacy, and self-rated ph ysical health in an urban setting. Journal of Health & Social Behavior, 43 (4), 383-399. Brownson, R. C., Baker, E. A., Housemann, R. A., Brennan, L. K. & Bacak, S. J. (2001). Environmental and policy determi nants of physical activity in the United States. American Journal of Public Health, 91 (12), 1995-2003. Burns, D. M. (2003). Epid emiology of smoking-induced cardiovascular disease. Progress in Cardiovascular Diseases, 46 (1), 11-29. Burton, N. W., Turrell, G. & Oldenburg, B. (2003). Part icipation in recreational physical activity: why do socioeconomic groups differ? Health Education & Behavior, 30 (2), 225-244. Campbell, C., Cornish, F., & McLean, C. (2004). Social capital, participation and the perpetuation of health inequalities: obstacles to African-Caribbean participation in 'partnerships' to improve mental health. Ethnicity & Health, 9 (4), 313-335.

PAGE 253

239 Carpenter, K. M., Hasin, D. S., Allison, D. B., & Faith, M. S. (2000). Relationships between obesity and DSM-IV major depressi ve disorder, suicide ideation, and suicide attempts: results fr om a general population study. American Journal of Public Health, 90 (2), 251-257. Cattell, V. (2001). Poor people, poor plac es, and poor health: the mediating role of social networks and social capital. Social Science & Medicine, 52 (10), 1501-1516 Centers for Disease Control and Prevention (CDC). (2001). Behavioral Risk Factor Surveillance System Survey Data Atlanta, Georgia: U.S. Department of Health and Human Services, Center s for Disease Control and Prevention. (retrieved May 2003). Cohen, D. A., Farley, T. A., & Mason, K. (2003). Why is poverty unhealthy? Social and physical mediators. Social Science & Medicine, 57 (9), 16311641. Cook, T. D., & Campbell, D. T. (1979). Quasi-Experimentation: Design & Analysis Issues for Field Settings Boston: Houghton Mifflin Company. Cooper, R. S. (2001). Social inequality, ethnicity and cardiovascular disease. International Journal of Epidemiology, 30 (Suppl 1), S48-52. Cooper, R. S., Kennelly, J. F., Durazo-Arviz u, R., Oh, H. J., Kaplan, G., & Lynch, J. (2001). Relationship between premature morta lity and socioeconomic factors in black and white populati ons of US metropolitan areas. Public Health Reports, 116 (5), 464-473.

PAGE 254

240 Cromarty, P., Robinson, G., Callcott, P., & Freeston, M. (2004). Cognitive therapy and exercise for panic and agorapho bia in primary care: pilot study and service development. Behavioural and Cognit ive Psychotherapy, 32 371-374. Crosby, R. A., Holtgrave, D. R., DiClemente, R. J., Wingood, G. M., & Gayle, J. A. (2003). Social capital as a predictor of adolescents' sexual risk behavior: a state-level exploratory study. AIDS and Behavior, 7 (3), 245-252. Cubbin, C., Hadden, W. C., & Winkleby, M. A. (2001). Neig hborhood context and cardiovascular disease risk factors: the contribution of material deprivation. Ethnicity & Disease, 11 (4), 687-700. Daly, M. C., Duncan, G. J., Kaplan, G. A. & Lynch, J. W. (1998). Macro-to-micro links in the relation between income inequality and mortality. Milbank Quarterly, 76 (3), 315-339, 303-314. Davey Smith, G., & Hart C. (2002). Life-course socioeconomic and behavioral influences on cardiovascular disease mortality: the collaborative study. American Journal of Public Health, 92 (8), 1295-1298. Davey Smith, G., Hart, C., Watt, G., Hole, D., & Hawthor ne, V. (1998). Individual social class, area-based deprivation, cardiovascular disease risk factors, and mortality: the Renf rew and Paisley study. Journal of Epidemiology and Community Health, 52 399-405.

PAGE 255

241 Dierker, L. C., Avenevoli, S., Stolar, M., & Merikangas, K. R. (2002). Smoking and depression: an examination of mechanisms of comorbidity. American Journal of Psychiatry, 159 (6), 947-953. Diez-Roux, A. V., Link, B. G., & Northri dge, M. E. (2000). A mu ltilevel analysis of income inequality and cardiovasc ular disease risk factors. Social Science & Medicine, 50 (5), 673-687. Diez-Roux, A. V., Nieto, F. J., Muntaner, C., Tyroler, H. A ., Comstock, G. W., Shahar, E., Cooper, L. S., Watson, R. L., & Szklo, M. (1997). Neighborhood environments and coronary heart dis ease: a multilevel analysis. American Journal of Epidemiology, 146 (1), 48-63. Diez Roux, A. V. (2003). Residential environments and cardiovascular risk. Journal of Urban Health, 80 (4), 569-589. Diez Roux, A. V., Merkin, S. S., Hannan, P., Jacobs, D. R., & Kiefe, C. I. (2003). Area characteristics, individual-level socioeconomic indicators, and smoking in young adults: the coronary artery disease risk development in young adults study. American Journal of Epidemiology, 157 (4), 315-326 Doyal, L. (1995). What makes women sick New Brunswick, New Jersey: Rutgers University Press. Dressler, W. M., & Badger, L. W. (1985). Epidemiology of depressive symptoms in black communities. A comparative analysis. Journal of Nervous & Mental Disorders, 173 (4), 212-220.

PAGE 256

242 Drukker, M., & van Os, J. (2003). M ediators of neighbour hood socioeconomic deprivation and quality of life. Social Psychiatry & Psychiatric Epidemiology, 38 (12), 698-706. Dunn, A. L., Trivedi, M. H., Kampert, J. B., Clark, C. G., & Chambliss, H. O. (2005). Exercise treatment for depression: efficacy and dose response. American Journal of Preventive Medicine, 28 (1), 1-8. Dunn, A. L., Trivedi, M. H., & O'Neal, H. A. (2001). Physical activity doseresponse effects on outcomes of depression and anxiety. Medicine & Science in Sports & Exercise, 33 (6 Suppl), S587-597; discussion 609-510. Ecob, R., & Macintyre, S. (2000). Sma ll area variations in health related behaviours; do these depend on the behaviour itself, its measurement, or on personal characteristics? Health & Place, 6 (4), 261-274. Ellaway, A., Anderson, A., & Macintyre, S. (1997). Does area of residence affect body size and shape? International Journal of Obesity Related Metabolic Disorders, 21 (4), 304-308. Ellaway, A., Macintyre, S., & Kearns, A. (2001). Perceptions of place and health in socially contrasting neighborhoods. Urban Studies, 38 (12), 2299-2316. Elliott, M. (2000). The stress process in neighborhood context. Health & Place, 6 (4), 287-299. Elliott, S. J., Taylor, S. M., Cameron, R., & Schabas, R. (1998). Assessing public health capacity to support communitybased heart health promotion: The

PAGE 257

243 Canadian Heart Health Initiati ve, Ontario Project (CHHIOP). Health Education Research,13 (4), 607-622. Emmons, K. M. (2000). Health behaviors in a social context. In L. F. Berkman & I. Kawachi (Eds.), Social epidemiology (pp. 242-266). New York: Oxford University Press. Ewing, R., Schmid, T., K illingsworth, R., Zlot, A., & Raudenbush, S. (2003). Relationship between urban sprawl and physical activity, obesity, and morbidity. American Journal of Health Promotion, 18 (1), 47-57. Fang, J., Wylie-Rosett, J., Cohen, H. W. Kaplan, R. C., & Alderman, M. H. (2003). Exercise, body mass index, ca loric intake, and cardiovascular mortality. American Journal of Pr eventive Medicine, 25 (4), 283-289. Ferraro, K. F., & Yu, Y. (1995). Body weight and self-ratings of health. Journal of Health & Social Behavior, 36 (3), 274-284. Ferrer, R. L., & Palmer, R. (2004). Variat ions in health status within and between socioeconomic strata. Journal of Epidemiology & Community Health, 58 (5), 381-387. Finnegan, J. R., Jr., Viswan ath, K., & Hertog, J. ( 1999). Mass media, secular trends, and the future of cardiovascu lar disease health promotion: an interpretive analysis. Preventive Medicine, 29 (6 Pt 2), S50-58. Fiscella, K., & Franks, P. (2000). Individual income, income inequality, health, and mortality: what are the relationships? Health Services Research, 35 (1 Pt 2), 307-318.

PAGE 258

244 Fontaine, K. R., Barofsky, I., Bartlett, S. J., Franckowiak, S. C., & Andersen, R. E. (2004). Weight loss and health-related quality of life: re sults at 1-year follow-up. Eating Behaviors, 5 (1), 85-88. Ford, E. S., Ahluwalia, I. B., & Galuska D. A. (2000). Social relationships and cardiovascular disease risk factors: findings from the third national health and nutrition examination survey. Preventive Medicine, 30 (2), 83-92. Ford, E. S., Moriarty, D. G., Zack, M. M., Mokdad, A. H., & Chapman, D. P. (2001). Self-reported body mass index and health-rela ted quality of life: findings from the Behavioral Ri sk Factor Surveillance System. Obesity Research, 9 (1), 21-31. Franzini, L., & Spears, W. (2003). Contributio ns of social context to inequalities in years of life lost to heart disease in Texas, USA. Social Science & Medicine, 57 (10), 1847-1861. Friedman, M. A., & Brownell, K. D. ( 1995). Psychological correlates of obesity: moving to the next research generation. Psychological Bulletin, 117 (1), 320. Froom, P., Melamed, S., Triber, I., Ra tson, N. Z., & Hermoni, D. (2004). Predicting self-reported health: the CORDIS study. Preventive Medicine, 39 (2), 419-423. Fullilove, M. T. (1998). Promoting social cohesion to improve health. Journal of the American Medical Wo mens Association, 53 (2), 72-76.

PAGE 259

245 Galanis, D. J., Harris, T., Sharp, D. S., & Petrovitch, H. (1998). Relative weight, weight change, and risk of coronary hear t disease in the Honolulu Heart Program. American of Journal Epidemiology, 147 (4), 379-386. Galea, S., Karpati, A., & Kennedy, B. (2002). Social capital and violence in the United States, 1974-1993. Social Science & Medicine, 55 (8), 1373-1383. Giles-Corti, B., & Donovan, R. J. (2002) The relative influence of individual, social and physical environment det erminants of physical activity. Social Science & Medicine, 54 (12), 1793-1812. Giles-Corti, B., & Donovan, R. J. (2003). Re lative influences of individual, social environmental, and physical environm ental correlates of walking. American Journal of Public Health, 93 (9), 1583-1589. Giles, W. H., Croft, J. B., Keenan, N. L ., Lane, M. J., & Wheel er, F. C. (1995). The validity of self-repor ted hypertension and correlates of hypertension awareness among blacks and whites within the stroke belt. American Journal of Preventive Medicine, 11 (3), 163-169. Gold, R., Kawachi, I., Kenne dy, B. P., Lynch, J. W. & Connell, F. A. (2001). Ecological analysis of teen birth rates: association with community income and income inequality. Maternal & Child Health Journal, 5 (3), 161-167. Gold, R., Kennedy, B., Connell, F., & Kawa chi, I. (2002). Teen births, income inequality, and social capital: devel oping an understanding of the causal pathway. Health & Place, 8 (2), 77-83.

PAGE 260

246 Goodman, E., & Capitman, J. (2000) Depressive symptoms and cigarette smoking among teens. Pediatrics, 106 (4), 748-755. Goodwin, R. D. (2003). Association bet ween physical activity and mental disorders among adults in the United States. Preventive Medicine, 36 (6), 698-703. Gordon, N. F., & Scott, C. B. (1991). The role of exercise in the primary and secondary prevention of co ronary artery disease. Clinical Sports Medicine, 10 (1), 87-103. Gravelle, H. (1998). How much of the relation between population mortality and unequal distribution of income is a statistical artefact? British Medical Journal, 316 (7128), 382-385. Greenlund, K. J., Zheng, Z. J., Keenan, N. L., Giles, W. H., Casper, M. L., Mensah, G. A., & Croft, J. B. (2004) Trends in self-reported multiple cardiovascular disease risk factors am ong adults in the United States, 19911999. Archives of Internal Medicine, 164 (2), 181-188. Gregg, E. W., Cauley, J. A., Stone, K., Thompson, T. J., Bauer, D. C., Cummings, S. R., & Ensrud, K. E. (2003). Relationship of changes in physical activity and mort ality among older women. Journal of the American Medical Association 289 (18), 2379-2386. Greiner, K. A., Li, C., Kawach i, I., Hunt, D. C., & Ahlu walia, J. S. (2004). The relationships of social participation and community ratings to health and

PAGE 261

247 health behaviors in areas wit h high and low population density. Social Science & Medicine, 59 (11), 2303-2312. Hasler, G., Pine, D. S., Gamma, A., Milos, G., Ajdacic, V., Eich, D., Rossler, W., & Angst, J. (2004). The associations between psychopathology and being overweight: a 20-year prospective study. Psychological Medicine, 34 (6), 1047-1057. Hassan, M. K., Joshi, A. V., Madhavan, S. S., & Amonkar, M. M. (2003). Obesity and health-related quality of life: a cr oss-sectional analysis of the US population. International Journal of Obesit y Related Metabolic Disorders, 27 (10), 1227-1232. Hemenway, D., Kennedy, B. P., Kawachi, I., & Putnam, R. D. (2001). Firearm prevalence and social capital. Annals of Epidemiology, 11 (7), 484-490. Henderson, C., Liu, X., Diez Roux, A. V ., Link, B. G., & Hasin, D. (2004). The effects of US state income inequalit y and alcohol policies on symptoms of depression and alcohol dependence. Social Science & Medicine, 58 (3), 565-575. Heymann, J. (2000). Health and social policy. In L. F. Berkman & I. Kawachi (Eds.), Social Epidemiology (pp. 368-382). New York: Oxford University Press. Heymann, J. & Fischer, A. (2003). Neighborhoods, health research, and its relevance to public policy. In I. Kawachi & L. F. Berkman (Eds.),

PAGE 262

248 Neighborhoods and health (pp. 335-347). New York: Oxford University Press. Holtgrave, D. R., & Crosby, R. A. (2003) Social capital, poverty, and income inequality as predictors of gonorrhoea, syphilis, chlamydia and AIDS case rates in the United States. Sexually Transmitted Infections, 79 (1), 62-64. Hou, F., & Chen, J. (2003). Neighbour hood low income, income inequality and health in Toronto. Health Reports, 14 (2), 21-34. Hu, F. B. (2003). Overweight and obesity in women: health risks and consequences. Journal of Womens Health (Larchmt), 12 (2), 163-172. Huff, N. C., & Gray, D. (2001). Coronar y heart disease inequalities: deaths and the socio-economic environm ent in Nottingham, England. Health & Place, 7 (1), 57-61. Humpel, N., Owen, N., & Leslie, E. (2002) Environmental factors associated with adults' participation in physical activity: A review. American Journal of Preventive Medicine, 22 (3), 188-199. Jackson, C., Jatulis, D. E., & Fortmann, S. P. (1992). The Behavioral Risk Factor Survey and the Stanford Five-City Project Survey: A comparison of cardiovascular risk behavior estimates. American Journal of Public Health, 82 (3), 412-416. Jia, H., Muennig, P., Lubetkin, E. I., & Gold, M. R. (2004). Predicting geographical variations in behavioural ri sk factors: An anal ysis of physical

PAGE 263

249 and mental healthy days. Journal of Epidemiology and Community Health, 58 150-155. Johnson, J. G., Cohen, P., Pine, D. S., Kl ein, D. F., Kasen, S., & Brook, J. S. (2000). Association between cigarette smoking and anxiety disorders during adolescence and early adulthood. Journal of the American Medical Association, 284 (18), 2348-2351. Jones, K., & Duncan, C. (1995). Indivi duals and their ecologies: analysing the geography of chronic disease within a multilevel modeling framework. Health & Place, 1 27-40. Jun, H., Subramanian, S. V., Gortmaker, S., & Kawachi, I. (2004). A multilevel analysis of women's status and self-ra ted health in the United States. Journal of the American Medi cal Women's Association, 59 172-180. Kahn, H. S., Tatham, L. M., Pamuk, E. R., & Heath, C. W., Jr. (1998). Are geographic regions with high income inequality associated with risk of abdominal weight gain? Social Science & Medicine, 47 (1), 1-6. Kahn, R. S., Wise, P. H., Kennedy, B. P., & Kawachi, I. (2000). State income inequality, household income, and mate rnal mental and physical health: Cross-sectional national survey. British Medical Journal, 321 (7272), 13111315. Kannel, W. B., Belanger, A., D'Agostino, R ., & Israel, I. (1986). Physical activity and physical demand on the job and risk of cardiovascular disease and death: The Framingham Study. American Heart Journal, 112 (4), 820-825.

PAGE 264

250 Kannel, W. B., & Sorlie, P. (1979). Some health benefits of physical activity: The Framingham Study. Archives of Internal Medicine, 139 (8), 857-861. Kannel, W. B., Wilson, P. W., Nam, B. H., & D'A gostino, R. B. (2002). Risk stratification of obesity as a coronary risk factor. American Journal of Cardiology, 90 (7), 697-701. Kaplan, G. A., & Lynch, J. W. (1999) Socioeconomic considerations in the primordial prevention of cardiovascular disease. Preventive Medicine, 29 (6 Pt 2), S30-35. Kaplan, G. A., Pamuk, E. R., Lynch, J. W. Cohen, R. D., & Balfour, J. L. (1996). Inequality in income and mortality in the United States: Analysis of mortality and potential pathways. British Medical Journal, 312 (7037), 999-1003. Kaplan, G. A., Strawbridge, W. J., Cohen, R. D., & Hungerford, L. R. (1996). Natural history of leisure-time phy sical activity and its correlates: Associations with mortality from all c auses and cardiovascular disease over 28 years. American Journal of Epidemiology, 144 (8), 793-797. Kaufman, J. S., Kaufman, S., & Pool e, C. (2003). Causal inference from randomized trials in social epidemiology. Social Science & Medicine, 57 2397-2409. Kawachi, I., Colditz, G. A. Stampfer, M. J., Willett, W. C., Manson, J. E., Rosner, B., Speizer, F. E., & Hennekens, C. H. (1994). Smoking cessation and time course of decreased risks of coronary heart disease in middle-aged women. Archives of Internal Medicine, 154 (2), 169-175.

PAGE 265

251 Kawachi, I., Kennedy, B. P., Lochner, K., & Prothrow-Stith, D. (1997). Social capital, income inequality, and mortality. American Journal of Public Health, 87 (9), 1491-1498. Kawachi, I., & Kennedy, B. P. (1997a). Health and social cohesion: Why care about income inequality? British Medical Journal, 314 (7086), 1037-1040. Kawachi, I., & Kennedy, B. P. (1997b). T he relationship of income inequality to mortality: Does the choice of indicator matter? Social Science & Medicine, 45 (7), 1121-1127. Kawachi, I. (1999). Social capita l and community effects on population and individual health. Annals of the New York Academy of Sciences, 896 120130. Kawachi, I., & Kennedy, B. P. (1999) Income inequality and health: Pathways and mechanisms. Health Services Research, 34 (1 Pt 2), 215-227. Kawachi, I., Kennedy, B. P., & Glass, R. (1999). Social capital and self-rated health: A contextual analysis. American Journal of Public Health, 89 (8), 1187-1193. Kawachi, I., Kennedy, B. P., & Wilkin son, R. G. (1999). Crime: Social disorganization and relative deprivation. Social Science & Medicine, 48 (6), 719-731. Kawachi, I. (1999). Social capita l and community effects on population and individual health. Annals of the New York Academy of Sciences, 896 120130.

PAGE 266

252 Kawachi, I., & Berkman, L. F. (2000). Soci al cohesion, social capital, and health. In L. F. Berkman & I. Kawachi (Eds.), Social Epidemiology (pp. 174-190). New York: Oxford University Press. Kawachi, I., & Berkman, L. F. (20 01). Social ties and mental health. Journal of Urban Health, 78 (3), 458-467. Kelleher, C., Timoney, A., Friel, S., & McKeown, D. (2002). Indicators of deprivation, voting patterns, and health st atus at area level in the Republic of Ireland. Journal of Epidemiology & Community Health, 56 (1), 36-44. Kennedy, B. P., Kawachi, I., Glass, R. & Prothrow-Stith, D. (1998). Income distribution, socioeconomic status, and self rated health in the United States: multilevel analysis. British Medical Journal, 317 (7163), 917-921. Kennedy, B. P., Kawachi, I., & Prothrow-S tith, D. (1996). Inco me distribution and mortality: Cross-sectional ecological st udy of the Robin Hood index in the United States. British Medical Journal, 312 (7037), 1004-1007. Kennedy, B. P., Kawachi, I., Prothrow-Sti th, D., Lochner, K., & Gupta, V. (1998). Social capital, income inequalit y, and firearm violent crime. Social Science & Medicine, 47 (1), 7-17. Kleinschmidt, I., Hills, M., & Elliott, P. (1995). Smoking behaviour can be predicted by neighbourhood deprivation measures. Journal of Epidemiology & Community Health, 49 (Suppl 2), S72-77. Kobau, R., Safran, M. A., Zack, M. M., Moriarty, D. G., & Chapman, D. (2004). Sad, blue, or depressed days, health behaviors and health-related quality of

PAGE 267

253 life, Behavioral Risk Factor Surveillance System, 1995-2000. Health & Quality of Life Outcomes, 2 (1), 40. Kreft, I. & de Leeuw, J. (1998). Introducing Multilevel Modeling London: Sage. Krieger, N. (1994). Epidemio logy and the web of causation: Has anyone seen the spider? Social Science & Medicine, 39 (7), 887-903. Krieger, N. (2001). Theories for social ep idemiology in the 21st century: an ecosocial perspective. International Journal of Epidemiology, 30 (4), 668677. Kubzansky, L. D., & Kawachi, I. (2000). Affective states and health. In L. F. Berkman & I. Kawachi (Eds.), Social Epidemiology (pp. 213-241). New York: Oxford University Press. Labarthe, D. R. (1998). Epidemiology and Prevention of Cardiovascular Diseases Gaithersburg, MD: As pen Publishers, Inc. LaCroix, A. Z., Leveille, S. G., Hecht, J. A., Grothaus, L. C., & Wagner, E. H. (1996). Does walking decrease the risk of cardiovascular disease hospitalizations and death in older adults? Journal of the American Geriatric Society, 44 (2), 113-120. LaVeist, T.A. (1993). Segregation, poverty, and empowerment: Health consequences for African Americans. Milbank Quarterly 71, 41-64. Lantz, P. M., House, J. S., Lepkowski, J. M., Williams, D. R., Mero, R. P., & Chen, J. (1998). Socioeconomic factor s, health behaviors, and mortality:

PAGE 268

254 Results from a nationally representat ive prospective study of US adults. Journal of the American Medical Association, 279 (21), 1703-1708. Lasser, K., Boyd, J. W., Woolhandler, S. Himmelstein, D. U., McCormick, D., & Bor, D. H. (2000). Smoking and m ental illness: A population-based prevalence study. Journal of the American Medical Association, 284 (20), 2606-2610. LeClere, F. B., Rogers, R. G., & Pete rs, K. (1998). Neighborhood social context and racial differences in wom en's heart disease mortality. Journal of Health & Social Behavior, 39 (2), 91-107. Lee, R. E., & Cubbin, C. (2002). Neighborhood contex t and youth cardiovascular health behaviors. American Journal of Public Health, 92 (3), 428-436. Leon, A. S., & Connett, J. (1991). Physical ac tivity and 10.5 year mortality in the Multiple Risk Factor In tervention Trial (MRFIT). International Journal of Epidemiology, 20 (3), 690-697. Leon, A. S., Myers, M. J., & Connett, J. (1997). Leisure time physical activity and the 16-year risks of morta lity from coronary heart di sease and all-causes in the Multiple Risk Factor Intervention Trial (MRFIT). International Journal of Sports Medicine, 18 (Suppl 3), S208-215. Leventhal, T., & Brooks-Gunn, J. (2003). Mo ving to opportunity: an experimental study of neighborhood effect s on mental health. American Journal of Public Health, 93 (9), 1576-1582.

PAGE 269

255 Leyden, K. M. (2003). Social capital and t he built environment: the importance of walkable neighborhoods. American Journal of Public Health, 93 (9), 15461551. Lindstrom, M., Hanson, B. S., & Oste rgren, P. O. (2001). Socioeconomic differences in leisure-time physical acti vity: the role of social participation and social capital in shaping health related behaviour. Social Science & Medicine, 52 (3), 441-451. Link, B. G., & Phelan, J. (1995). Social conditions as fundamental causes of disease. Journal of Health & Social Behavior, Extra Issue 80-94. Link, B. G., & Phelan, J. (1996). Understandi ng sociodemographic differences in health – the role of fundamental causes. American Journal of Public Health, 86 (4), 471-473. Lobmayer, P., & Wilkinson, R. G. (2002). Inequality, re sidential segregation by income, and mortality in US cities. Journal of Epidemiology & Community Health, 56 (3), 183-187. Lochner, K., Pamuk, E., Makuc, D., Kennedy B. P., & Kawachi, I. (2001). Statelevel income inequality and individual mo rtality risk: a prospective, multilevel study. American Journal of Public Health, 91 (3), 385-391. Lochner, K. A., Kawachi, I., Brennan, R. T., & Buka, S. L. (2003). Social capital and neighborhood mortality rates in Chicago. Social Science & Medicine, 56 (8), 1797-1805.

PAGE 270

256 Lopez, R. (2004). Income inequality and self-rated health in US metropolitan areas: A multi-level analysis. Social Science & Medicine, 59 (12), 24092419. Lynch, J., & Kaplan, G. (2000) Socioeconomic position. In L. F. Berkman & I. Kawachi (Eds.), Social Epidemiology (pp. 13-35). New York: Oxford University Press. Lynch, J. W., Kaplan, G. A., Pamuk, E. R., Cohen, R. D., Heck, K. E., Balfour, J. L., & Yen, I. H. (1998). Income inequality and mortality in metropolitan areas of the United States. American Journal of Public Health, 88 (7), 1074-1080. Lynch, J., Smith, G. D., Harper, S., Hillemeier, M., Ro ss, N., Kaplan, G. A., & Wolfson, M. (2004). Is income ine quality a determinant of population health? Part 1. A systematic review. The Milbank Quarterly, 82 (1), 5-99. Lynch, J. W., Smith, G. D., Kaplan, G. A., & Hous e, J. S. (2000). Income inequality and mortality: Importance to health of individual income, psychosocial environment, or material conditions. British Medical Journal, 320 (7243), 1200-1204. Macintyre, S. (2000). The social patterni ng of exercise behaviours: The role of personal and local resources. British Journal of Sports Medicine, 34 (1), 6. MacIntyre, S., & Ellaway, A. (2000). Ecological appr oaches: Rediscovering the role of the physical and so cial environment. In L. F. Berkman & I. Kawachi (Eds.), Social epidemiology (pp. 332-348). New York: Oxford University Press.

PAGE 271

257 Macintyre, S., & Ellaway, A. (2003). Neighb orhoods and health: An overview. In I. Kawachi & L. F. Berkman (Eds.), Neighborhoods and health (pp. 20-42). New York: Oxford University Press. Macintyre, S., Maciver, S., & Sooman, A. (1993). Ar ea, class and health: Should we be focusing on places or people? Journal of Social Policy, 22 (2), 213234. Malmstrom, M., Sundquist, J., & Johansson, S. E. (1999). Neighborhood environment and self-reported health status: A multilevel analysis. American Journal of Public Health, 89 (8), 1181-1186. Manderbacka, K., Lundberg, O., & Martikainen, P. (1999). Do risk factors and health behaviours contribute to self-ratings of health? Social Science & Medicine, 48 (12), 1713-1720. Mann, S. J., James, G. D., Wang, R. S., & Pickering, T. G. (1991). Elevation of ambulatory systolic blood pressure in hypertensive smokers: A case-control study. Journal of the American Journal Association, 265 (17), 2226-2228. Mansson, N. O., & Merl o, J. (2001). The relati on between self-rated health, socioeconomic status, body mass index and disability pension among middle-aged men. European Journal of Epidemiology, 17 (1), 65-69. Marmot, M., & Wilkinson, R. G. (2001). Psychosocial and material pathways in the relation between income and health: A response to Lynch et al. British Medical Journal, 322 (7296), 1233-1236.

PAGE 272

258 Marmot, M. G., Bobak, M., & Smith, G. D. (1995). Explanat ions for social inequalities in health. In B. C. Amick III, S. Levine, A. R. Tarlov, & D. Chapman Walsh (Eds.), Society and health (pp. 172-210). New York: Oxford University Press. Massey, D. S., & Denton, N. A. (1 988). The dimensions of residential segregation. Social Forces, 67 281-315. Massey, D., White, M., & Phua, V. ( 1996). The dimensions of segregation revisited. Sociological Methods and Research, 25 172-206. McCall, R. B., & Applebaum, M. I. (1991) Some issues of conducting secondary analysis. Developmental Psychology, 27 911-917. McCulloch, A. (2001). Social environment s and health: Cross-sectional national survey. British Medical Journal, 323 (7306), 208-209. McCulloch, A. (2003). An exam ination of social capital and social disorganisation in neighbourhoods in the Br itish household panel study. Social Science & Medicine, 56 (7), 1425-1438. McEwen, B. S. (1998). Prot ective and damaging effects of stress mediators. New England Journal of Medicine, 338 (3), 171-179. McKechnie, R., & Mosca, L. (2003). Physi cal activity and coronary heart disease: Prevention and effect on risk factors. Cardiology Review, 11 (1), 21-25. McKenzie, K., Whitley, R., & Weich, S. (2002). Social capital and mental health. British Journal of Psychiatry, 181 280-283.

PAGE 273

259 McLaughlin, D. K., & Stokes, C. S. (2002). Income inequality and mortality in US counties: Does minority racial concentration matter? American Journal of Public Health, 92 (1), 99-104. Mellor, J. M., & Milyo, J. (2002). Income inequality and heal th status in the United States: Evidence form the Current Population Survey. Journal of Human Resources, 37 (3), 510-539. Mellor, J. M., & Milyo, J. (2003). Is exposure to inco me inequality a public health concern? Lagged effects of income i nequality on individual and population health. Health Services Research, 38 (1 Pt 1), 137-151. Merlo, J., Yang, M., Chaix, B., Lynch, J., & Ra stam, L. (2005a). Journal of Epidemiology and Community Health, 59 (9), 729-736. Merlo, J., Yang, M., Chaix, B., Lynch, J., & Ra stam, L. (2005b). Journal of Epidemiology and Community Health, 59 (12), 1022-1028. Meyer, H. E., Sogaard, A. J., Tverdal, A., & Selmer R. M. (2002). Body mass index and mortality: the influence of physical activity and smoking. Medicine & Science in Sports & Exercise, 34 (7), 1065-1070. Millen, B. E., Quatromoni, P. A., Nam, B. H., O'Horo, C. E., Polak, J. F., & D'Agostino, R. B. (2002). Dietary patterns and the odds of carotid atherosclerosis in women: the Framingham Nutrition Studies. Preventive Medicine, 35 (6), 540-547. Mitchell, C., & LaGory, M. (2002). Soci al capital and mental distress in an impoverished community. City & Community, 1 (2), 199-222.

PAGE 274

260 Morland, K., Wing, S., & Diez Roux, A. (2002). The contextu al effect of the local food environment on residents' diet s: The atherosclerosis risk in communities study. American Journal of Public Health, 92 (11), 1761-1767. Muntaner, C., Li, Y., Xue, X ., O'Campo, P., Chung, H. J ., & Eaton, W. W. (2004). Work organization, area labor-market characteristics, and depression among U.S. nursing home work ers: A cross-classified multilevel analysis. International Journal of Occupat ional & Environmental Health, 10 (4), 392400. Muramatsu, N. (2003). County-level income inequality and depression among older Americans. Health Services Research, 38 (6 Pt 2), 1863-1883. National Heart, Lung, and Blo od Institute (NHLBI). ( 2002). NHLBI Strategic Plan FY2002-2006 http://www.nhlbi.nih.gov/res ources/docs/plan/index.htm (retrieved July 2002). National Institute of Diabetes and Di gestive and Kidney Diseases (NIDDKD). (1998). Clinical Guidelines on the Indentific ation, Evaluation, and Treatment of Overweight and Obesity in Adults NIH. Nelson, D. E., Holtzman, D., Bolen, J. Stanwyck, C. A., & Mack, K. A. (2001). Reliability and validity of measures from the B ehavioral Risk Factor Surveillance System (BRFSS). Soz Praventivmed, 46 (Suppl 1), S3-42. Nordstrom, C. K., Dwyer, K. M., Merz, C. N., Shircore, A., & Dwyer, J. H. (2003). Leisure time physical activity and ear ly atherosclerosis: The Los Angeles Atherosclerosis Study. American Journal of Medicine, 115 (1), 19-25.

PAGE 275

261 Norman, J. E., Bild, D., Lewis C. E., Liu, K., & West, D. S. (2003). The impact of weight change on cardiovascular diseas e risk factors in young black and white adults: The CARDIA study. International Journal of Obesity Related Metabolic Disorders, 27 (3), 369-376. Oakes, J. M. (2004). The (mis)estima tion of neighborhood effects: Causal inference for a practical social epidemiology. Social Science & Medicine, 58 (10), 1929-52. O'Campo, P., Xue, X., Wang, M. C ., & Caughy, M. (1997). Neighborhood risk factors for low birthweight in Ba ltimore: A multilevel analysis. American Journal of Public Health, 87 (7), 1113-1118. Oguma, Y., Sesso, H. D., Pa ffenbarger, R. S., Jr., & Lee, I. M. (2002). Physical activity and all cause mortality in women: A review of the evidence. British Journal of Sports Medicine, 36 (3), 162-172. Onwuegbuzie, A. J. (2003). Expanding the framework of internal and external validity in quantitative research. Research in the Schools, 10 (1), 71-89. Osler, M., Christensen, U., Due, P., Lund, R., Andersen, I., Diderichsen, F., & Prescott, E. (2003). Income inequality and ischaemic heart disease in Danish men and women. International Journal of Epidemiology, 32 (3), 375380. Owens, J. F., Matthews, K. A., Raikkonen, K., & Kuller, L. H. (2003). It is never too late: Change in physical activity fosters change in cardiovascular risk factors in middle-aged women. Preventive Cardiology, 6 (1), 22-28.

PAGE 276

262 Paffenbarger, R. S., Jr., Wing, A. L., & Hyde, R. T. (1978). Physical activity as an index of heart attack risk in college alumni. American Journal of Epidemiology, 108 (3), 161-175. Palinkas, L. A., Wingard, D. L., & Barrett-Connor, E. (1996). Depressive symptoms in overweight and obese older ad ults: A test of the "jolly fat" hypothesis. Journal of Psychosomatic Research, 40 (1), 59-66. Parks, S. E., Housemann, R. A., & Browns on, R. C. (2003). Differential correlates of physical activity in urban and rura l adults of various socioeconomic backgrounds in the United States. Journal of Epidemiology & Community Health, 57 (1), 29-35. Patt, M. R., Yanek, L. R., Moy, T. F., & Becker, D. M. (2003). Assessment of global coronary heart disease risk in overweight and obese AfricanAmerican women. Obesity Research, 11 (5), 660-667. Pearlin, L. I. (1989). The so ciological study of stress. Journal of Health & Social Behavior, 30 (3), 241-256. Picciotto, M. R., Brunzell, D. H., & Caldarone, B. J. (20 02). Effect of nicotine and nicotinic receptors on anxiety and depression. Neuroreport, 13 (9), 10971106. Pickett, K. E., & Pearl, M. (2001) Multilevel analyse s of neighbourhood socioeconomic context and health outcomes: A critical review. Journal of Epidemiology & Community Health, 55 (2), 111-122.

PAGE 277

263 Pittilo, M. (2000). Cigarette smoking, endothelial injury and cardiovascular disease. International Journal of Experimental Pathology, 81 (4), 219-230. Poland, B. D. (2000). The 'considerate' smoker in public space: The micropolitics and political economy of 'doing the right thing'. Health & Place, 6 (1), 1-14. Portes, A. (1998). Social capi tal: its origins and applicatio ns in modern sociology. Annual Review of Sociology, 24 1-24. Putnam, R. D. (2000). Bowling Alone New York: Simon & Schuster. Raphael, D., Renwick, R., Br own, I., Steinmetz, B., Sehdev, H., & Phillips, S. (2001). Making the links between community structure and individual wellbeing: Community quality of life in Riverdale, Toronto, Canada. Health & Place, 7 (3), 179-196. Rashid, M. N., Fuentes, F., Touchon, R. C., & Wehner, P. S. (2003). Obesity and the risk for cardiovascular disease. Preventive Cardiology, 6 (1), 42-47. Raudenbush, S. W. (2003). The quantitative assessm ent of neighborhood social environments. In I. Kawachi & L. F. Berkman (Eds.), Neighborhoods and health (pp. 112-131). New York: Oxford University Press. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage Publications Inc.

PAGE 278

264 Raudenbush, S. W., & Samps on, R.J. (1999). Ecomet rics: Toward a science of assessing ecological settings, with applic ation to the systematic social observation of neighborhoods. Sociological Methodology, 29 1-41. Reidpath, D. D. (2003). Love thy neighbour"-it's good fo r your health: A study of racial homogeneity, mortality and soci al cohesion in the United States. Social Science & Medicine, 57 (2), 253-261. Reidpath, D. D., Burns, C., Garrard, J. Mahoney, M., & Townsend, M. (2002). An ecological study of the relationshi p between social and environmental determinants of obesity. Health & Place, 8 (2), 141-145. Reijneveld, S. A. (1998). The impact of individual and area characteristics on urban socioeconomic differences in health and smoking. International Journal of Epidemiology, 27 (1), 33-40. Robert, S. A. (1998). Community-level socioeconomic status effects on adult health. Journal of Health & Social Behavior, 39 (1), 18-37. Robert, S. A. (1999). So cioeconomic position and health: The independent contribution of community socioeconomic context. Annual Review of Sociology, 25 489-516. Roberts, R. E., Deleger, S., Strawbri dge, W. J., & Kapl an, G. A. (2003). Prospective association between obes ity and depression: Evidence from the Alameda County Study. International Journal of Obesity Related Metabolic Disorders, 27 (4), 514-521.

PAGE 279

265 Roberts, R. E., Kaplan, G. A., Shema, S. J., & Strawbridge, W. J. (2000). Are the obese at greater risk for depression? American Journal of Epidemiology, 152 (2), 163-170. Roberts, R. E., Strawbridge, W. J., Del eger, S., & Kaplan, G. A. (2002). Are the fat more jolly? Annals of Behavioral Medicine, 24 (3), 169-180. Rodgers, G. (1979). Income and inequality as determinants of mortality: An international cross-section analysis. Population Studies (Camb), 33 (2), 343351. Rogers, R. G., Hummer, R. A., & Krueger, P. M. (2003) The effect of obesity on overall, circulatory diseaseand diabetes-specific mortality. Journal of Biosocial Science, 35 (1), 107-129. Rohrer, J. E., Arif, A. A ., Pierce, J. R., Jr., & Bla ckburn, C. (2004). Unsafe neighborhoods, social group activity, and self-rated health. Journal of Public Health Management & Practice, 10 (2), 124-129. Roper Center for Public Opinion Research. (2001). Social Capital Community Benchmark Survey Cambridge, MA: Roper C enter for Public Opinion Research. Rose, G. (1992). The strategy of preventive medicine New York: Oxford University Press. Ross, C. E. (1994). Over weight and depression. Journal of Health & Social Behavior, 35 (1), 63-79.

PAGE 280

266 Ross, C. E. (2000a). Neighborhoo d disadvantage and adult depression. Journal of Health and Social Behavior, 41 (June), 177-187. Ross, C. E. (2000b). Walking, exer cising, and smoking: Does neighborhood matter? Social Science & Medicine, 51 (2), 265-274. Ross, C. E., & Mirowsky, J. (2001). Neighborhood disadvantage, disorder, and health. Journal of Health & Social Behavior, 42 (3), 258-276. Sallis, J. F., Kraft, K., & Linton, L. S. (2002). How the en vironment shapes physical activity: A transdi sciplinary research agenda. American Journal of Preventive Medicine, 22 (3), 208. Salmon, J., Owen, N., Craw ford, D., Bauman, A., & Sa llis, J. F. (2003). Physical activity and sedentary behavior: A popul ation-based study of barriers, enjoyment, and preference. Health Psychology, 22 (2), 178-188. Salmon, P. (2001). Effects of physica l exercise on anxiety, depression, and sensitivity to stress: A unifying theory. Clinical Psychology Review, 21 (1), 33-61. Sampson, R. J. (2003). The nei ghborhood context of well-being. Perspectives in Biology & Medicine, 46 (3 Suppl), S53-64. Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel st udy of collective efficacy. Science, 277 (5328), 918-924. Sanmartin, C., Ross, N. A., Tremblay, S., Wolfson, M., Dunn, J. R., & Lynch, J. (2003). Labour market income inequalit y and mortality in North American

PAGE 281

267 metropolitan areas. Journal of Epidemiology & Community Health, 57 (10), 792-797. Sartorius, N. (2003). Social capital and mental health. Current Opinion in Psychiatry, 16 (Suppl 2), S101-S105. Schmid, T. L., Pratt, M., & Howze, E. (1995). Policy as intervention: Environmental and policy approaches to the prevention of cardiovascular disease. American Journal of Public Health, 85 (9), 1207-1211. Schulz, A., Israel, B., Williams, D., Park er, E., Becker, A., & James, S. (2000). Social inequalities, stressors and self reported health status among African American and white women in t he Detroit metropolitan area. Social Science & Medicine, 51 (11), 1639-1653. Schulz, A., Williams, D., Israel, B., Be cker, A., Parker, E., James, S. A., & Jackson, J. (2000). Unfair treatmen t, neighborhood effects, and mental health in the Detroit metropolitan area. Journal of Health & Social Behavior, 41 (3), 314-332. Scully, D., Kremer, J., Meade, M. M., Graham, R., & Dudgeon, K. (1998). Physical exercise and psychological well being: A critical review. British Journal of Sports Medicine, 32 (2), 111-120. Sehili, S., Elbasha, E. H., Moriarty, D. G., & Zack, M. M. (2005). Inequalities in self-reported physical health in the United States, 1993-1999. Health Economics, 14 (4), 377-89.

PAGE 282

268 Shi, L., & Starfield, B. (2000). Primar y care, income ine quality, and self-rated health in the United States : A mixed-level analysis. International Journal of Health Services, 30 (3), 541-555. Shi, L., Starfield, B., Politzer, R., & Regan, J. (2002). Primary care, self-rated health, and reductions in social disparities in health. Health Services Research, 37 (3), 529-550. Singh, G. K., & Siahpush, M. (2002). In creasing inequalities in all-cause and cardiovascular mortality among US adults aged 25-64 years by area socioeconomic status, 1969-1998. International Journal of Epidemiology, 31 (3), 600-613. Skrabski, A., Kopp, M., & Kawa chi, I. (2003). Social capi tal in a changing society: cross-sectional associations with mi ddle aged female and male mortality rates. Journal of Epidemiology & Community Health, 57 (2), 114-119. Soobader, M. J., & LeClere, F. B. (1999). Aggregation and the measurement of income inequality: Ef fects on morbidity. Social Science & Medicine, 48 (6), 733-744. Sorensen, G., Emmons, K., Hunt, M. K., Barbeau, E., Goldman, R., Peterson, K., Kuntz, K., Stoddard, A., & Berkman, L. (2003). Model for incorporating social context in health behavior inte rventions: Applications for cancer prevention for working-class, multiethnic populations. Preventive Medicine, 37 (3), 188-197.

PAGE 283

269 Sowers, J. R. (2003). Obesity as a cardiovascular risk factor. American Journal of Medicine, 115 (Suppl 8A), 37S-41S. Stein, A. D., Lederman, R. I., & Shea, S. (1993). The Behavioral Risk Factor Surveillance System questionnaire: It s reliability in a statewide sample. American Journal of Public Health, 83 (12), 1768-1772. Steptoe, A., & Feldman, P. J. (2001) Neighborhood problems as sources of chronic stress: Devel opment of a measure of neighborhood problems, and associations with socioeconomic status and health. Annals of Behavioral Medicine, 23 (3), 177-185. Stewart, D. W., & Kamins, M. A. (1993). Secondary research: Information sources and methods. Newbury Park, CA: Sage Publications Inc. Strine, T. W., Balluz, L., Chapman, D. P., Moriarty, D. G., Owens, M., & Mokdad, A. H. (2004). Risk behaviors and healthcare coverage among adults by frequent mental distress status, 2001. American Journal of Preventive Medicine, 26 (3), 213-216. Strine, T. W., Okoro, C. A., Chapman, D. P., Balluz, L. S., Ford, E. S., Ajani, U. A., & Mokdad, A. H. (2 005). Health-related quality of life and health risk behaviors among smokers. American Journal of Preventive Medicine, 28 (2), 182-187. Sturm, R., & Gresenz, C. R. (2002). Relations of in come inequality and family income to chronic medical conditions and mental health disorders: National survey. British Medical Journal, 324 (7328), 20-23.

PAGE 284

270 Sturm, R., & Wells, K. B. (2001). Does obesity contribute as much to morbidity as poverty or smoking? Public Health, 115 (3), 229-235. Subramanian, S. V., Kawach i, I., & Kennedy, B. P. (2 001). Does the state you live in make a difference? Multilevel analysis of self-rated health in the US. Social Science & Medicine, 53 9-19. Subramanian, S. V., Kim, D. J., & Kawachi, I. (2002). Social trust and self-rated health in US communities: A multilevel analysis. Journal of Urban Health, 79 (4 Suppl 1), S21-34. Subramanian, S. V., & Kawa chi, I. (2003). The association between state income inequality and worse health is not confounded by race. International Journal of Epidemiology, 32 (6), 1022-1028. Subramanian, S. V., Jones, K., & Dunc an, C. (2003). Multilevel methods for public health research. In I. Kawa chi & L. F. Berkman (Eds.), Neighborhoods and health (pp. 65-111). New York: Oxford University Press. Subramanian, S. V., & Kawachi, I. (2004) Income inequality and health: What have we learned so far? Epidemiologic Reviews, 26 78-91. Sundquist, J., Malmstrom, M., & Johanss on, S. E. (1999). Cardiovascular risk factors and the neighbourhood envir onment: A multilevel analysis. International Journal of Epidemiology, 28 (5), 841-845. Sundquist, K., Lindstrom, M., Malmstrom, M., Johansson, S. E., & Sundquist, J. (2004). Social participation and coronary heart disease: A follow-up study of

PAGE 285

271 6900 women and men in Sweden. Social Science & Medicine, 58 (3), 615622. Taylor, S. E., Repetti, R. L., & Seeman, T. (1997). Health psychol ogy: What is an unhealthy environment and how does it get under the skin? Annual Review of Psychology, 48 411-447. Terris, M. (1999). The development and pr evention of cardiovascular disease risk factors: Socioenvironmental influences. Preventive Medicine, 29 (6 Pt 2), S11-17. Thomas, J. C., & Thomas, K. K. (1999). Thi ngs ain't what they ought to be: Social forces underlying racial disparities in rates of sexually transmitted diseases in a rural North Carolina county. Social Science & Medicine, 49 (8), 10751084. Treloar, C., Porteous, J., Hassan, F., Ka sniyah, N., Lakshmanudu, M., Sama, M., Sja'bani, M., & Heller, R. F. (1999). The cross cultural context of obesity: An INCLEN multicentre collaborative study. Health & Place, 5 (4), 279-286. Turner, R., Wheaton, B., & Ll oyd, D. (1995). The epidemio logy of social stress. American Sociological Review, 60 (February), 104-125. Turshen, M. (1989). The politics of public health New Brunswick, New Jersey: Rutgers University Press. US Census Bureau. (2000). U.S. Census. http://www.census.gov/main/www/cen2000.html (retrieved May 2003).

PAGE 286

272 US Department of Health and Hu man Services (USDHHS). (1983). The health consequences of smoking cardiovascular disease: A report of the Surgeon General. CDC NCCDPHP. US Department of Health and Hu man Services (USDHHS). (1996). Physical activity and health: A report of the Surgeon General. CDC NCCDPHP. US Department of Health and Hu man Services (USDHHS). (2001). Women & smoking: A report of the Surgeon General. CDC NCCDPHP. Vandegrift, D., & Yoked, T. (2004). Obesity rates, in come, and suburban sprawl: An analysis of U.S. states. Health & Place, 10 (3), 221-9. Veenstra, G. (2002). Social capital and health (plus wealth, income inequality and regional health governance). Social Science & Medicine, 54 (6), 849868. Waitzman, N. J., & Smith, K. R. (1998a). Phantom of the area: Poverty-area residence and mortality in the United States. American Journal of Public Health, 88 (6), 973-976. Waitzman, N. J., & Smith, K. R. (1998b). Separate but lethal: The effects of economic segregation on mortalit y in metropolitan America. Milbank Quarterly, 76 (3), 341-373, 304. Wakefield, S. E., Elliott, S. J., Cole, D. C., & Eyles, J. D. (2001). Environmental risk and (re)action: Air quality, health and civic involvement in an urban industrial neighbourhood. Health & Place, 7 (3), 163-177.

PAGE 287

273 Walsh, D. C., Sorensen, G., & Leonard, L. (1995). Gender, health, and cigarette smoking. In B. C. Amick III, S. Levi ne, A. R. Tarlov, & D. Chapman Walsh (Eds.), Society and health (pp. 131-171). Oxford: Oxfo rd University Press. Weich, S., Lewis, G., & Jenkins, S. P. (2001). Income inequality and the prevalence of common mental disorders in Britain. British Journal of Psychiatry, 178 222-227. Weich, S., Lewis, G., & Jenkins, S. P. (2002). Income inequality and self rated health in Britain. Journal of Epidemiology & Community Health, 56 (6), 436441. Weich, S., Twigg, L., Ho lt, G., Lewis, G., & Jones, K. (2003). Contextual risk factors for the common mental disorders in Britain: A multilevel investigation of the effects of place. Journal of Epidemiology & Community Health, 57 (8), 616-621. Weitzman, E. R., & Kawach i, I. (2000). Giving means re ceiving: The protective effect of social capital on binge drinking on college campuses. American Journal of Public Health, 90 (12), 1936-1939. Whelton, S. P., Chin, A., Xi n, X., & He, J. (2002). Effe ct of aerobic exercise on blood pressure: A meta-analysis of randomized, controlled trials. Annals of Internal Medicine, 136 (7), 493-503. Wilkinson, R. G. (1992). Income distribution and life expectancy. British Medical Journal, 304 (6820), 165-8. Wilkinson, R.G. (1996). Unhealthy societies London: Routledge.

PAGE 288

274 Williams, D. R. ( 1997). Race and health: Basic questions, emerging directions. Annals of Epidemiology, 7 (5), 322-333. Williams, S. M., Chapman, D., & Lando, J. (2005). The role of public health in mental health promotion. Morbidity and Mortality Weekly Report, 54 (34), 841-842. Williams, J. M., & Ziedonis, D. (2004) Addressing tobacco among individuals with a mental illne ss or an addiction. Addictive Behaviors, 29 (6), 1067-1083. Williams, S. M., Chapman, D., & Lando, J. (2005). The role of public health in mental health promotion. Morbidity and Mortality Weekly Report, 54 (34), 841-842. Winkleby, M. A., & Cubbin, C. (2003). Influence of individual and neighbourhood socioeconomic status on mortality among black, Mexican-American, and white women and men in the United States. Journal of Epidemiology & Community Health, 57 (6), 444-452. Wister, A. V. (1996). T he effects of socioeconomic status on exercise and smoking: age-related differences. Journal of Aging & Health, 8 (4), 467-488. Wolfson, M., Kaplan, G., Lynch, J., Ro ss, N., & Backlund, E. (1999). Relation between income inequality and morta lity: Empirical demonstration. British Medical Journal, 319 (7215), 953-955. Wolk, R., Berger, P., Lennon, R. J., Brilaki s, E. S., & Somers, V. K. (2003). Body mass index: a risk factor for unstable angina and myocardial infarction in

PAGE 289

275 patients with angiographica lly confirmed coronary artery disease. Circulation, 108 (18), 2206-2211. Yen, I. H., & Kaplan, G. A. (1999a). Nei ghborhood social environment and risk of death: multilevel evidence from the Alameda County Study. American Journal of Epidemiology, 149 (10), 898-907. Yen, I. H., & Kaplan, G. A. (1999b). Po verty area residence and changes in depression and perceived health status : Evidence from the Alameda County Study. International Journal of Epidemiology, 28 (1), 90-94. Zack, M. M., Moriarty D. G., Stroup, D. F., Ford, E. S., & Mokdad, A. H. (2004). Worsening trends in adult health-related quality of life and self-rated healthUnited States, 1993-2001. Public Health Reports, 119 (5), 493-505. Zahran, H. S., Kobau, R., Mo riarty, D. G., Zack, M. M. Holt, J., & Donehoo, R. (2005). Health-related quality of life surveillance – United States, 19932002. Morbidity & Mortality Weekly R eport Surveillance Summaries, 54 (4), 1-35. Ziersch, A. M., Baum, F. E., Macdougall, C., & Putland, C. (2005). Neighbourhood life and social capital: The implications for health. Social Science & Medicine, 60 (1), 71-86

PAGE 290

276 Appendix A1: Sample Distribution Comparison SPONSOR: C.F. of Greater Birmingham (AL) STATE/Counties: Alabama/ Jefferson, Shelby BRFSSSCCBS Census Population 496 500 805340 Sex Male 36.7%46.9% 47.4% Female 63.3%53.1% 52.6% Race White 65.4%65.4% 64.4% Black 31.7%30.2% 34.0% Hispanic 1.0% 3.1% 1.7% Other 2.8% 4.4% 2.5% Education 0 – 12 (no diploma) 10.9%9.2% 18.0% 12 30.3%28.3% 27.0% 13-15 29.2%39.6% 28.2% > 16 29.5%22.9% 26.8% Age 20 – 34 26.4%28.1% 29.1% 35 – 44 20.0%22.3% 22.2% 45 – 64 33.6%31.0% 31.1% 65 + 20.1%18.6% 17.6% Income <$20,000 26.4%15.0% 24.5% $20,000 – $50,000 41.5%44.4% 35.7% $50,000 $75,000 15.2%21.0% 18.2% >$75,000 16.8%19.6% 21.4% Marital Status Married 50.9%57.8% 52.9% Separated/Widowed/Divorc ed 31.0%22.3% 20.8% Never Married 18.1% 19.9% 26.2%

PAGE 291

277 Appendix A2: Sample Distribution Comparison SPONSOR: Arizona Community Foundation STATE/County: Arizona/Maricopa BRFSSSCCBS Census Population 856 501 3072149 Sex Male 42.5%49.5% 50.0% Female 57.5%50.5% 50.0% Race White 82.4%77.3% 79.8% Black 4.1% 4.4% 4.3% Hispanic 13.8%20.2% 24.8% Other 13.6%18.4% 19.0% Education 0 – 12 (no diploma) 8.9% 19.2% 17.5% 12 25.5%20.2% 23.1% 13-15 33.3%31.3% 33.6% > 16 32.3%29.3% 25.9% Age 20 – 34 26.8%30.8% 33.1% 35 – 44 21.7%23.5% 22.1% 45 – 64 32.1%29.2% 28.2% 65 + 19.5%16.5% 16.7% Income <$20,000 13.2%17.1% 24.5% $20,000 – $50,000 45.8%44.1% 35.7% $50,000 $75,000 17.5%17.3% 18.4% >$75,000 23.5%21.5% 21.4% Marital Status Married 54.4%60.7% 54.9% Separated/Widowed/Divorc ed 27.7%15.3% 26.8% Never Married 17.9% 24.0% 18.4%

PAGE 292

278 Appendix A3: Sample Distribution Comparison SPONSOR: California C.F. STATE/County: Californi a/Los Angeles BRFSSSCCBS Census Population 1002 515 9519338 Sex Male 43.1%48.4% 49.4% Female 56.9%51.6% 50.6% Race White 75.8%50.4% 52.8% Black 11.4%10.2% 10.5% Hispanic 34.4%39.7% 44.6% Other 12.8%39.3% 42.0% Education 0 – 12 (no diploma) 17.0%27.5% 30.0% 12 21.4%14.2% 18.8% 13-15 26.5%27.9% 26.2% > 16 35.2%30.4% 24.9% Age 20 – 34 31.7%32.6% 34.7% 35 – 44 22.4%25.0% 23.1% 45 – 64 32.0%28.0% 28.1% 65 + 13.9%14.4% 14.1% Income <$20,000 27.7%23.2% 23.3% $20,000 – $50,000 31.5%35.5% 33.6% $50,000 $75,000 16.5%15.8% 17.8% >$75,000 24.4%25.6% 25.3% Marital Status Married 46.0%49.1% 48.8% Separated/Widowed/Divorc ed 28.2%20.2% 17.1% Never Married 25.8% 30.7% 34.1%

PAGE 293

279 Appendix A4: Sample Distribution Comparison SPONSOR: The San Diego Foundation STATE/County: Californi a/San Diego BRFSSSCCBS Census Population 346 504 2813833 Sex Male 41.3%50.1% 50.3% Female 57.7%49.9% 49.7% Race White 87.8%65.9% 70.3% Black 4.1% 6.3% 6.6% Hispanic 22.3%18.8% 26.7% Other 8.1% 27.8% 28.1% Education 0 – 12 (no diploma) 9.3% 18.5% 17.4% 12 24.4%17.7% 19.9% 13-15 28.1%30.5% 33.2% > 16 38.3%33.4% 29.6% Age 20 – 34 22.7%29.8% 33.7% 35 – 44 22.6%26.1% 22.8% 45 – 64 30.1%29.2% 27.8% 65 + 19.6%14.9% 15.7% Income <$20,000 21.3%15.2% 18.2% $20,000 – $50,000 33.0%43.1% 34.3% $50,000 $75,000 20.6%15.2% 20.2% >$75,000 25.1%26.5% 27.3% Marital Status Married 56.9%53.2% 52.0% Separated/Widowed/Divorc ed 24.7%21.9% 17.8% Never Married 18.4% 24.9% 30.2%

PAGE 294

280 Appendix A5: Sample Distribution Comparison SPONSOR: Walter & Elise Haas Fund STATE/County: California/ San Francisco BRFSSSCCBSCensus Population 95 500 776733 Sex Male 48.4%49.8%50.8% Female 51.6%50.2%49.2% Race White 74.7%52.2%53.0% Black 7.4% 4.7% 8.6% Hispanic 14.7%20.1%14.1% Other 17.9%43.1%43.1% Education 0 – 12 (no diploma) 6.3% 8.3% 18.8% 12 9.5% 14.2%13.9% 13-15 24.2%37.5%22.4% > 16 60.0%40.0%45.0% Age 20 – 34 34.0%44.9%36.4% 35 – 44 18.1%23.1%20.6% 45 – 64 33.0%24.0%26.6% 65 + 14.9%8.0% 16.3% Income <$20,000 19.8%12.5%19.0% $20,000 – $50,000 25.3%31.1%26.6% $50,000 $75,000 16.5%15.4%17.7% >$75,000 38.5%41.0%36.7% Marital Status Married 33.8%37.7%38.7% Separated/Widowed/Divo rced 30.0%13.1%16.6% Never Married 36.3% 49.3%44.8%

PAGE 295

281 Appendix A6: Sample Distribution Comparison SPONSOR: C.F. Serving Boulder County STATE/County: Colorado/Boulder BRFSSSCCBSCensus Population 124 500 291288 Sex Male 44.4%46.8%50.6% Female 55.6%53.3%49.4% Race White 90.2%89.9%90.5% Black 0.8% 0.8% 1.2% Hispanic 7.3% 8.3% 10.5% Other 9.0% 9.3% 10.6% Education 0 – 12 (no diploma) 3.2% 6.2% 7.2% 12 12.1%15.2%15.1% 13-15 25.8%36.6%25.3% > 16 58.9%42.1%52.4% Age 20 – 34 30.8%29.3%34.9% 35 – 44 25.8%27.3%24.0% 45 – 64 29.2%31.3%30.0% 65 + 14.2%12.1%10.6% Income <$20,000 14.2%11.9%15.0% $20,000 – $50,000 40.7%32.0%29.2% $50,000 $75,000 16.8%22.7%20.5% >$75,000 28.3%33.4%35.5% Marital Status Married 57.4%54.7%52.1% Separated/Widowed/Divo rced 15.7%13.5%15.0% Never Married 27.0% 31.7%32.9%

PAGE 296

282 Appendix A7: Sample Distribution Comparison SPONSOR: Denver Foundation/Ro se C.F./Piton Foundation STATE/County: Colorado/Denver BRFSSSCCBS Census Population 228 501 554636 Sex Male 39.9%46.1% 50.5% Female 60.1%53.9% 49.5% Race White 76.1%66.9% 68.3% Black 12.0%11.6% 12.1% Hispanic 22.8%28.7% 31.7% Other 12.0%21.6% 23.5% Education 0 – 12 (no diploma) 17.1%17.0% 21.1% 12 18.0%18.7% 20.0% 13-15 21.1%28.1% 24.4% > 16 43.9%36.1% 34.5% Age 20 – 34 33.5%40.5% 38.0% 35 – 44 19.4%23.2% 20.6% 45 – 64 31.8%19.7% 26.5% 65 + 14.3%16.5% 14.9% Income <$20,000 24.2%13.0% 22.6% $20,000 – $50,000 40.6%48.3% 38.5% $50,000 $75,000 16.4%18.0% 18.3% >$75,000 18.8%20.7% 20.6% Marital Status Married 39.5%49.0% 43.2% Separated/Widowed/Divorc ed 29.4%18.4% 20.8% Never Married 31.2% 32.6% 35.9%

PAGE 297

283 Appendix A8: Sample Distribution Comparison SPONSOR: Delaware Division of Stat e Service Centers/Delaware C.F. STATE: Delaware BRFSSSCCBS Census Population 3514 1379 783600 Sex Male 38.7%48.0% 48.6% Female 61.3%52.0% 51.4% Race White 81.0%75.2% 75.9% Black 14.6%18.5% 20.1% Hispanic 2.8% 5.4% 4.8% Other 4.5% 6.3% 5.9% Education 0 – 12 (no diploma) 9.3% 17.3% 17.4% 12 34.5%28.4% 31.4% 13-15 25.7%31.1% 26.1% > 16 30.5%23.2% 25.0% Age 20 – 34 23.3%27.1% 28.4% 35 – 44 21.8%24.7% 22.6% 45 – 64 33.3%30.0% 31.0% 65 + 21.6%18.2% 18.0% Income <$20,000 20.1%13.9% 17.7% $20,000 – $50,000 40.4%40.6% 35.0% $50,000 $75,000 17.9%20.3% 21.3% >$75,000 21.6%25.3% 26.0% Marital Status Married 53.5%59.2% 54.0% Separated/Widowed/Divorc ed 27.7%18.4% 18.7% Never Married 18.8% 22.4% 27.2%

PAGE 298

284 Appendix A9: Sample Distribution Comparison Sponsor: C.F. For Greater Atlanta State/Counties: Georgi a/ DeKalb, Fulton, Cobb, Rockdale, Henry BRFSSSCCBS Census Population 646 510 2279074 Sex Male 40.3%46.9% 49.1% Female 59.8%53.1% 50.9% Race White 59.3%60.5% 54.7% Black 34.1%31.9% 38.9% Hispanic 2.0% 4.0% 6.8% Other 6.6% 7.6% 8.2% Education 0 – 12 (no diploma) 7.0% 8.8% 14.4% 12 17.5%23.8% 21.1% 13-15 22.5%35.3% 26.7% > 16 53.1%32.1% 37.8% Age 20 – 34 28.5%29.9% 35.9% 35 – 44 27.6%25.5% 24.3% 45 – 64 31.6%31.2% 28.8% 65 + 12.2%13.4% 11.0% Income <$20,000 11.5%9.6% 16.0% $20,000 – $50,000 36.2%36.1% 32.0% $50,000 $75,000 16.1%21.5% 20.5% > $75,000 36.3%32.7% 31.5% Marital Status Married 49.5%56.4% 48.6% Separated/Widowed/Divorc ed 22.2%19.1% 17.9% Never Married 28.4% 24.6% 33.5%

PAGE 299

285 Appendix A10: Sample Distribution Comparison SPONSOR: Indiana Grantmakers Alliance STATE: Indiana BRFSSSCCBS Census Population 3993 1001 6080485 Sex Male 40.4%48.3% 49.0% Female 59.6%51.7% 51.0% Race White 90.7%88.5% 88.6% Black 6.2% 6.3% 8.8% Hispanic 2.9% 5.3% 3.5% Other 3.0% 5.3% 3.9% Education 0 – 12 (no diploma) 10.4%18.2% 17.9% 12 38.7%28.2% 37.2% 13-15 24.1%28.5% 25.5% > 16 26.8%25.1% 19.4% Age 20 – 34 26.4%25.5% 29.1% 35 – 44 21.4%25.4% 22.3% 45 – 64 33.1%32.7% 31.2% 65 + 19.1%16.4% 17.4% Income <$20,000 18.7%16.3% 20.8% $20,000 – $50,000 47.0%43.2% 38.5% $50,000 $75,000 18.7%22.2% 21.4% >$75,000 15.6%18.3% 19.5% Marital Status Married 56.7%60.1% 56.3% Separated/Widowed/Divorc ed 28.1%17.4% 18.8% Never Married 15.3% 22.5% 24.8%

PAGE 300

286 Appendix A11: Sample Distribution Comparison SPONSOR: Forum 35/Baton Rouge Area Foundation STATE/County: Louisiana/East Baton Rouge Parish BRFSSSCCBS Census Population 461 500 412852 Sex Male 39.9%46.1% 47.9% Female 60.1%53.9% 52.1% Race White 63.8%59.1% 56.9% Black 32.1%36.3% 40.5% Hispanic 1.7% 2.7% 1.8% Other 4.0% 4.6% 3.7% Education 0 – 12 (no diploma) 7.6% 8.3% 16.1% 12 24.6%21.8% 26.4% 13-15 28.1%41.4% 26.7% > 16 39.7%28.5% 30.8% Age 20 – 34 36.3%30.0% 34.4% 35 – 44 16.8%23.8% 21.3% 45 – 64 32.7%30.7% 30.0% 65 + 14.3%15.5% 14.3% Income <$20,000 20.1%19.8% 27.7% $20,000 – $50,000 41.7%37.1% 34.6% $50,000 $75,000 16.6%19.4% 17.1% >$75,000 21.6%23.7% 20.7% Marital Status Married 48.1%52.2% 47.4% Separated/Widowed/Divorc ed 25.2%18.2% 18.3% Never Married 26.7% 29.6% 34.2%

PAGE 301

287 Appendix A12: Sample Distribution Comparison SPONSOR: Kalamazoo C.F. STATE/County: Michigan/Kalamazoo BRFSSSCCBS Census Population 89 500 238603 Sex Male 44.9%50.0% 48.4% Female 55.1%50.0% 51.6% Race White 85.4%86.0% 86.5% Black 9.0% 9.4% 10.8% Hispanic 2.3% 2.2% 2.6% Other 5.6% 4.6% 5.3% Education 0 – 12 (no diploma) 12.4%5.6% 11.3% 12 23.6%23.3% 25.9% 13-15 22.5%43.4% 31.6% > 16 41.6%27.7% 31.1% Age 20 – 34 36.1%30.3% 33.8% 35 – 44 15.1%22.6% 20.6% 45 – 64 30.2%30.6% 29.6% 65 + 18.6%16.6% 16.0% Income <$20,000 27.6%12.4% 22.1% $20,000 – $50,000 38.2%41.2% 36.5% $50,000 $75,000 6.6% 21.3% 20.0% >$75,000 27.6%25.1% 21.5% Marital Status Married 54.7%54.9% 51.6% Separated/Widowed/Divorc ed 23.3%16.6% 16.5% Never Married 22.1% 28.5% 31.9%

PAGE 302

288 Appendix A13: Sample Distribution Comparison SPONSOR: C.F. for S outheastern Michigan STATE/Counties: MICHIGAN/Way ne, Oakland, Macomb, St.Clair, Washtenaw, Monroe, Livingston BRFSSSCCBS Census Population 1554 501 4833493 Sex Male 38.2%50.0% 48.7% Female 61.8%50.0% 51.3% Race White 72.8%69.5% 73.8% Black 20.5%21.7% 22.5% Hispanic 3.2% 3.1% 2.8% Other 6.7% 8.8% 5.9% Education 0 – 12 (no diploma) 8.5% 9.9% 17.2% 12 27.7%28.4% 28.3% 13-15 30.5%41.2% 29.8% > 16 33.4%20.5% 24.7% Age 20 – 34 24.7%27.8% 29.2% 35 – 44 23.9%22.9% 23.1% 45 – 64 34.1%31.4% 31.2% 65 + 17.4%17.8% 16.5% Income <$20,000 14.7%9.9% 18.5% $20,000 – $50,000 39.6%37.7% 31.7% $50,000 $75,000 18.8%23.7% 20.4% >$75,000 26.9%28.7% 29.6% Marital Status Married 50.7%54.8% 51.3% Separated/Widowed/Divorc ed 27.8%19.5% 18.9% Never Married 21.6% 25.7% 29.8%

PAGE 303

289 Appendix A14: Sample Distribution Comparison Sponsor: The St. Paul Foundation State/Counties: Minnesota/Dakota, Ramsey, Washington BRFSSSCCBS Census Population 844 503 1068069 Sex Male 41.2%46.9% 48.9% Female 58.8%53.1% 51.1% Race White 91.7%88.6% 86.9% Black 4.4% 1.9% 5.6% Hispanic 1.9% 4.1% 3.9% Other 3.9% 9.5% 9.9% Education 0 – 12 (no diploma) 5.1% 7.9% 9.3% 12 21.1%17.7% 24.8% 13-15 32.0%25.8% 31.4% > 16 41.8%48.6% 34.4% Age 20 – 34 26.2%35.8% 30.6% 35 – 44 27.0%26.4% 25.2% 45 – 64 31.0%25.7% 30.6% 65 + 15.8%12.1% 13.5% Income <$20,000 8.4% 6.1% 13.2% $20,000 – $50,000 36.7%37.9% 33.4% $50,000 $75,000 20.7%25.2% 23.2% >$75,000 34.2%30.8% 32.2% Marital Status Married 55.5%62.0% 55.6% Separated/Widowed/Divorc ed 25.1%12.0% 15.1% Never Married 19.5% 26.0% 29.2%

PAGE 304

290 Appendix A15: Sample Distribution Comparison SPONSOR: Montana C.F. STATE: Montana BRFSSSCCBS Census Population 3338 502 902195 Sex Male 42.6%49.1% 49.8% Female 57.4%50.9% 50.2% Race White 86.4%89.4% 92.2% Black 0.2% 0.6% 0.5% Hispanic 2.5% 3.9% 2.0% Other 13.5%10.0% 9.2% Education 0 – 12 (no diploma) 10.9%13.9% 12.9% 12 34.1%29.2% 31.3% 13-15 29.1%29.6% 31.5% > 16 25.9%27.3% 24.4% Age 20 – 34 19.8%27.6% 25.1% 35 – 44 20.1%24.8% 22.0% 45 – 64 37.6%31.7% 34.2% 65 + 22.5%15.9% 18.8% Income <$20,000 25.5%22.8% 28.8% $20,000 – $50,000 52.7%46.9% 42.2% $50,000 $75,000 12.5%18.8% 17.2% >$75,000 9.3% 11.5% 11.9% Marital Status Married 56.3%63.9% 57.3% Separated/Widowed/Divorc ed 30.4%17.8% 18.7% Never Married 13.3% 18.3% 24.0%

PAGE 305

291 Appendix A16: Sample Distribution Comparison SPONSOR: New Hampshire Charitable Foundation STATE: New Hampshire BRFSSSCCBS Census Population 4068 711 1235786 Sex Male 42.5%49.6% 49.2% Female 57.5%50.4% 50.8% Race White 96.0%94.9% 97.0% Black 0.4% 0.9% 1.0% Hispanic 1.6% 1.6% 1.7% Other 3.6% 4.2% 3.2% Education 0 – 12 (no diploma) 7.4% 7.0% 12.6% 12 29.8%31.4% 30.1% 13-15 26.5%37.4% 28.7% > 16 36.4%24.3% 28.7% Age 20 – 34 22.0%25.2% 25.7% 35 – 44 25.5%24.2% 24.8% 45 – 64 35.3%32.4% 32.9% 65 + 17.2%18.2% 16.6% Income <$20,000 12.3%13.4% 16.0% $20,000 – $50,000 40.4%39.7% 34.6% $50,000 $75,000 21.6%24.5% 23.1% >$75,000 25.7%22.4% 26.5% Marital Status Married 59.4%57.5% 57.3% Separated/Widowed/Divorc ed 25.2%19.6% 17.8% Never Married 15.4% 22.9% 24.9%

PAGE 306

292 Appendix A17: Sample Distribution Comparison SPONSOR: Central New York C.F. STATE/County: New York/Onondaga BRFSSSCCBS Census Population 106 541 458336 Sex Male 37.7%45.7% 47.8% Female 62.3%54.3% 52.2% Race White 93.3%82.4% 86.4% Black 4.8% 9.2% 10.3% Hispanic 2.8% 2.7% 2.4% Other 1.9% 8.4% 5.3% Education 0 – 12 (no diploma) 7.7% 5.0% 14.3% 12 26.0%29.5% 29.1% 13-15 27.9%40.4% 28.1% > 16 38.5%25.1% 28.4% Age 20 – 34 20.0%27.8% 27.0% 35 – 44 23.0%22.3% 22.6% 45 – 64 34.0%30.3% 31.0% 65 + 23.0%19.7% 19.4% Income <$20,000 19.8%16.7% 23.9% $20,000 – $50,000 34.1%39.9% 35.4% $50,000 $75,000 22.0%20.7% 19.4% >$75,000 24.2%22.8% 21.3% Marital Status Married 51.5%53.1% 50.9% Separated/Widowed/Divorc ed 30.1%19.1% 18.6% Never Married 18.5% 27.8% 30.4%

PAGE 307

293 Appendix A18: Sample Distribution Comparison SPONSOR: Rochester Area C.F. STATE/Counties: NEW YORK/Monroe, Wayne, Ontario, Livingston, Genesee, Orleans BRFSS SCCBS Census Population 164 988 1098201 Sex Male 34.2% 45.7% 48.6% Female 65.9% 54.3% 51.4% Race White 84.7% 84.4% 85.3% Black 9.2% 9.4% 11.1% Hispanic 4.3% 2.7% 4.3% Other 6.1% 6.3% 5.4% Education 0 – 12 (no diploma) 5.5% 7.2% 15.7% 12 26.2% 30.7% 29.1% 13-15 28.1% 38.9% 28.1% > 16 40.2% 23.3% 27.1% Age 20 – 34 26.0% 27.8% 26.9% 35 – 44 24.1% 22.4% 23.0% 45 – 64 32.9% 31.3% 32.0% 65 + 17.1% 18.5% 18.1% Income <$20,000 16.3% 13.2% 20.6% $20,000 – $50,000 38.8% 45.1% 35.6% $50,000 $75,000 15.0% 22.3% 20.9% >$75,000 29.9% 19.4% 22.9% Marital Status Married 50.9% 54.7% 52.6% Separated/Widowed/Divorced 25.8% 21.6% 18.6% Never Married 23.3% 23.7% 28.8%

PAGE 308

294 Appendix A19: Sample Distribution Comparison SPONSOR: Winston-Salem Foundation STATE/County: North Caro lina/Forsyth BRFSSSCCBS Census Population 454 750 306067 Sex Male 38.8%48.0% 47.8% Female 61.2%52.0% 52.2% Race White 70.7%71.8% 69.5% Black 26.9%23.8% 26.2% Hispanic 1.8% 5.3% 6.4% Other 2.5% 4.4% 5.7% Education 0 – 12 (no diploma) 13.5%11.7% 18.0% 12 27.0%28.1% 27.0% 13-15 24.6%35.4% 26.4% > 16 35.0%24.9% 28.7% Age 20 – 34 22.3%27.4% 29.6% 35 – 44 19.8%22.8% 22.1% 45 – 64 33.5%31.6% 31.1% 65 + 24.5%18.1% 17.2% Income <$20,000 20.8%16.0% 21.4% $20,000 – $50,000 46.8%41.0% 36.6% $50,000 $75,000 17.8%21.7% 20.4% >$75,000 14.6%21.3% 21.5% Marital Status Married 49.9%58.5% 54.9% Separated/Widowed/Divorc ed 31.2%21.6% 19.4% Never Married 18.9% 19.9% 25.7%

PAGE 309

295 Appendix A20: Sample Distribution Comparison SPONSOR: C.F. of Greater Greensboro STATE/County: North Carolina/Guilford BRFSSSCCBS Census Population 413 750 421048 Sex Male 35.6%48.0% 47.9% Female 64.4%52.0% 52.1% Race White 68.7%68.5% 65.5% Black 26.7%25.6% 29.9% Hispanic 2.4% 3.1% 3.8% Other 4.7% 5.9% 6.1% Education 0 – 12 (no diploma) 11.4%8.3% 17.1% 12 25.6%27.8% 25.1% 13-15 26.3%38.1% 27.6% > 16 36.7%25.8% 30.3% Age 20 – 34 24.6%27.4% 31.7% 35 – 44 22.6%23.1% 21.8% 45 – 64 31.8%32.1% 30.3% 65 + 20.9%17.5% 16.1% Income <$20,000 20.8%14.6% 20.1% $20,000 – $50,000 45.2%41.5% 37.5% $50,000 $75,000 16.0%21.7% 19.9% >$75,000 18.1%22.2% 22.4% Marital Status Married 46.4%57.6% 52.9% Separated/Widowed/Divorc ed 31.2%17.9% 18.6% Never Married 22.4% 24.5% 28.5%

PAGE 310

296 Appendix A21: Sample Distribution Comparison SPONSOR: Cleveland Foundation STATE/County: Ohio/Cuyahoga BRFSSSCCBS Census Population 459 1100 1393978 Sex Male 37.7%46.6% 47.2% Female 62.3%53.4% 52.8% Race White 73.7%68.5% 68.7% Black 22.6%25.6% 28.2% Hispanic 3.5% 0.7% 3.4% Other 3.8% 5.9% 5.0% Education 0 – 12 (no diploma) 8.3% 7.7% 18.4% 12 30.1%33.2% 30.0% 13-15 28.2%37.6% 26.5% > 16 33.4%21.6% 25.2% Age 20 – 34 21.7%24.2% 26.3% 35 – 44 23.9%21.8% 21.7% 45 – 64 33.6%31.7% 30.6% 65 + 20.8%22.4% 21.4% Income <$20,000 16.3%15.8% 24.9% $20,000 – $50,000 49.0%45.4% 36.5% $50,000 $75,000 14.5%21.1% 18.4% >$75,000 20.2%17.7% 20.4% Marital Status Married 46.8%51.1% 49.4% Separated/Widowed/Divorc ed 29.2%21.4% 21.9% Never Married 24.1% 27.6% 30.6%

PAGE 311

297 Appendix A22: Sample Distribution Comparison SPONSOR: Greater Cincinnati Foundation STATE/Counties: OHIO/Butler, Clermont, Hamilton, Warren STATE/Counties: Kentucky/B oone, Campbell, Kenton STATE/Counties: Dearborn BRFSSSCCBS Census Population 1038 1001 1886650 Sex Male 40.9%46.4% 48.5% Female 59.1%53.6% 51.5% Race White 87.5%84.5% 85.6% Black 10.5%11.8% 12.7% Hispanic 1.8% 2.2% 1.1% Other 2.1% 3.7% 2.9% Education 0 – 12 (no diploma) 10.6%10.8% 17.0% 12 32.6%34.5% 31.1% 13-15 26.3%34.1% 26.2% > 16 30.5%20.7% 25.8% Age 20 – 34 28.6%26.7% 29.3% 35 – 44 21.2%23.2% 23.4% 45 – 64 30.8%31.6% 30.7% 65 + 19.4%18.5% 16.6% Income <$20,000 17.3%13.3% 19.5% $20,000 – $50,000 42.6%40.5% 34.9% $50,000 $75,000 17.8%23.4% 21.0% >$75,000 22.3%22.8% 24.6% Marital Status Married 52.3%59.9% 54.5% Separated/Widowed/Divorc ed 27.0%20.2% 18.4% Never Married 20.7% 19.9% 27.2%

PAGE 312

298 Appendix A23: Sample Distribution Comparison SPONSOR: Northwest Area Foundation STATE/County: OREGON/Crook, Deschutes, Jefferson BRFSSSCCBS Census Population 99 500 153558 Sex Male 47.5%48.9% 49.8% Female 52.5%51.2% 50.2% Race White 96.9%90.5% 93.3% Black 0.0% 1.7% 0.4% Hispanic 5.1% 5.0% 5.7% Other 3.1% 7.8% 8.5% Education 0 – 12 (no diploma) 3.0% 10.1% 14.1% 12 38.4%34.8% 29.2% 13-15 34.3%37.3% 34.7% > 16 24.2%17.8% 22.0% Age 20 – 34 29.8%22.1% 24.9% 35 – 44 17.0%22.3% 21.5% 45 – 64 34.0%33.4% 35.3% 65 + 19.2%22.2% 18.4% Income <$20,000 12.4%15.1% 20.2% $20,000 – $50,000 55.1%47.1% 41.8% $50,000 $75,000 16.9%20.0% 20.3% >$75,000 15.7%17.9% 17.8% Marital Status Married 62.0%67.9% 61.6% Separated/Widowed/Divorc ed 22.8%15.5% 18.2% Never Married 15.2% 16.7% 20.1%

PAGE 313

299 Appendix A24: Sample Distribution Comparison SPONSOR: York Foundation STATE/County: Pennsylv ania/York BRFSSSCCBS Census Population 127 500 381751 Sex Male 40.9%46.6% 49.2% Female 59.1%53.5% 50.8% Race White 92.9%93.5% 93.7% Black 4.0% 1.9% 4.2% Hispanic 2.4% 1.8% 3.0% Other 3.2% 4.6% 3.2% Education 0 – 12 (no diploma) 8.7% 11.0% 19.3% 12 40.9%42.7% 41.6% 13-15 23.6%32.5% 20.7% > 16 26.8%13.8% 18.4% Age 20 – 34 30.9%24.8% 25.0% 35 – 44 15.5%22.9% 23.6% 45 – 64 40.7%32.3% 33.0% 65 + 13.0%19.9% 18.5% Income <$20,000 16.4%13.7% 16.7% $20,000 – $50,000 50.9%40.9% 38.9% $50,000 $75,000 23.3%24.2% 24.8% >$75,000 9.5% 21.2% 19.5% Marital Status Married 61.8%67.0% 60.3% Separated/Widowed/Divorc ed 23.6%16.7% 17.4% Never Married 14.6% 16.4% 22.3%

PAGE 314

300 Appendix A25: Sample Distribution Comparison SPONSOR: Greater Houston C.F. STATE/County: Texas/Harris BRFSSSCCBS Census Population 802 500 3400578 Sex Male 41.0%47.8% 49.8% Female 59.0%52.2% 50.2% Race White 68.7%63.7% 61.2% Black 16.0%19.6% 19.0% Hispanic 24.9%26.9% 32.9% Other 15.3%16.8% 22.9% Education 0 – 12 (no diploma) 15.7%18.5% 25.4% 12 22.0%22.3% 21.6% 13-15 23.7%31.5% 26.0% > 16 38.7%27.7% 26.9% Age 20 – 34 31.5%30.8% 35.7% 35 – 44 25.8%26.1% 24.3% 45 – 64 30.7%30.4% 29.1% 65 + 12.0%12.7% 10.9% Income <$20,000 19.7%20.4% 20.9% $20,000 – $50,000 38.5%39.9% 35.9% $50,000 $75,000 17.2%15.7% 18.4% >$75,000 24.6%24.0% 24.6% Marital Status Married 51.7%54.7% 53.8% Separated/Widowed/Divorc ed 26.7%21.2% 17.4% Never Married 21.6% 24.2% 28.7%

PAGE 315

301 Appendix A26: Sample Distribution Comparison SPONSOR: Northwest Area Foundation STATE/County: Washington/Yakima BRFSSSCCBS Census Population 119 500 222581 Sex Male 39.5%46.9% 49.9% Female 60.5%53.1% 50.1% Race White 93.2%77.4% 68.6% Black 0.9% 2.4% 1.4% Hispanic 14.3%30.1% 35.9% Other 5.9% 20.2% 33.7% Education 0 – 12 (no diploma) 17.7%23.5% 31.4% 12 35.3%30.3% 27.4% 13-15 26.9%32.1% 26.0% > 16 20.2%14.1% 15.3% Age 20 – 34 21.7%26.2% 30.8% 35 – 44 18.3%21.7% 21.7% 45 – 64 27.8%31.6% 30.3% 65 + 32.2%20.5% 17.2% Income <$20,000 25.2%23.0% 26.8% $20,000 – $50,000 48.5%49.0% 40.6% $50,000 $75,000 17.5%17.7% 18.4% >$75,000 8.7% 10.4% 14.5% Marital Status Married 58.0%62.2% 56.9% Separated/Widowed/Divorc ed 26.9%20.6% 17.5% Never Married 15.1% 17.3% 25.7%

PAGE 316

302 Appendix A27: Sample Distribution Comparison SPONSOR: Greater Ka nawha Valley Foundation STATE/Counties: WEST VIR GINIA/Kanawha, Putnam, Boone BRFSSSCCBS Census Population 497 500 277197 Sex Male 40.4%46.1% 48.0% Female 59.6%53.9% 52.0% Race White 93.5%90.3% 93.6% Black 3.8% 5.0% 5.7% Hispanic 1.0% 2.3% 0.6% Other 2.6% 4.8% 1.8% Education 0 – 12 (no diploma) 14.5%11.2% 21.2% 12 36.9%35.7% 37.1% 13-15 26.4%37.6% 22.9% > 16 22.1%15.5% 19.2% Age 20 – 34 22.5%22.6% 24.8% 35 – 44 21.1%21.8% 20.9% 45 – 64 36.4%34.3% 34.0% 65 + 20.0%21.3% 20.2% Income <$20,000 22.9%19.7% 29.1% $20,000 – $50,000 47.0%44.6% 38.0% $50,000 $75,000 16.8%22.4% 17.4% >$75,000 13.4%13.3% 15.7% Marital Status Married 57.1%64.0% 56.8% Separated/Widowed/Divorc ed 27.7%18.1% 21.9% Never Married 15.3% 17.9% 21.4%

PAGE 317

303 About the Author Caroline Mae McKay received her B.A. in E nglish from University of Miami. She attended Florida State University School of Social Work, where she graduated with honors with a master’s degree in clinical social work. After practicing in Miami with the homeless mentally-ill veteran population, she came to the doctoral program in the Department of Community and Family Healt h, University of South Florida College of Public Health. Focus areas of her work include social structural and contextual influences on health and health behavior, th eoretical development in Public Health, and multilevel modeling.