USF Libraries
USF Digital Collections

Impact of area social predictors of health on Black-White disparities in stroke mortality

MISSING IMAGE

Material Information

Title:
Impact of area social predictors of health on Black-White disparities in stroke mortality
Physical Description:
Book
Language:
English
Creator:
Dark, Tyra
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Wellbeing
Resources
Access
Economic
Environment
Dissertations, Academic -- Public Health -- Doctoral -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: This dissertation investigated the area social predictors of health (ASPoH) and Black-White disparities in stroke mortality relationship. Utilizing stroke mortality data obtained from the Florida Department of Health for years 1998-2002, and social and economic data obtained from the year 2000 Census of Population, this study examined the effect of resource availability at the census tract level on Black-White disparities in stroke mortality. The influence of social class on Black-White disparities in stroke mortality and effect modification by social class of the association between Black-White disparities and ASPoH variables was also investigated. Principal component analysis produced four ASPoH scores from economic and social measures. Multiple regression analysis assessed the predictive ability of these ASPoH variables on Black-White disparities.Increases in the female Black-White ratio were significantly associated with increases in the magnitude of the ASPoH-1 and ASPoH-2 variables. When regression analyses were restricted (in terms of population count minimums) to a subset of census tracts, increases in the ASPoH-1 and ASPoH-2 variables were significantly associated with increases in all Black-White disparity measures for both males and females. Assessment of the influence of social class on Black-White disparities in stroke mortality was only feasible at the state level due to a lack of data at the census tract level. With the exception of the 65+ years age-group, Black males and females experienced higher age-group specific stroke mortality rates across each of the social class groups. Inconsistent with previous research findings, Black residents who attained a high school degree had the highest stroke death rates compared to all other educational attainment groups.In the assessment of social class as a potential effect modifier, the study hypothesis stated that the ASPoH measures would have the greatest impact on those residents in the lowest social class category. This predicted effect was only supported when the Male Black-White ratio disparity score was examined. Study findings support the conjecture that unknown and unmeasured processes influence the association between area social predictors and stroke mortality for Black Floridians. Identification of modifiable societal characteristics may be the key to unlocking the foundation of disparities in health outcomes.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2007.
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 Tyra Dark.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 155 pages.
General Note:
Includes vita.

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 - 001916914
oclc - 181158818
usfldc doi - E14-SFE0002014
usfldc handle - e14.2014
System ID:
SFS0026332:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Impact of Area Social Predictors of Hea lth on Black-White Disparities in Stroke Mortality by Tyra Dark A dissertation submitted in partial fulfillment of the requirements for the degree of Doctorate of Philosophy Department of Epidemiology and Biostatistics College of Public Health University of South Florida Co-Major Professor: Thomas J. Mason, Ph.D. Co-Major Professor: Heather Stockwell, Sc.D. Elizabeth Barnett, Ph.D. Michelle Casper, Ph.D. Douglas Schocken, M.D. Date of Approval: April 6, 2007 Keywords: wellbeing, resources, access, economic, environment Copyright 2007, Tyra Dark

PAGE 2

i Table of Contents List of Tables iv List of Figures x Abstract xi Chapter One: Introduction 1 Proposed Research Questions and Hypotheses 3 Study Purpose/Rationale 4 Proposed Pathway: ASPoH and Black-White Disparities in Stroke Mortality 8 Chapter Two: Review of Literature 14 Stroke Definition 14 Ischemic Stroke 14 Hemorrhagic Stroke 15 Public Health Importance of Stroke 16 Stroke Risk Factors 17 High Blood Pressure and Stroke 17 Diabetes Mellitus 18 Cigarette Smoking 18 High Blood Cholesterol and Other Lipids 19 Physical Inactivity 20 Overweight and Obesity 20 Stroke Mortality Trends 21 Geographical Differences in Stroke Epidemiology 23 Black-White Disparities in Mortality 26 Black-White Disparities in Stroke Mortality 28 Socioeconomic Status and Health 30 Socioeconomic Status and Stroke 33 Theories of Causation 34 Why Blacks Have Higher Stroke Mortality: Influence of Social Environment 39 Conclusion 40 Chapter Three: Methods 42 Study Design 42 Study Population 42 Data Sources 43 Stroke Mortality Data 43

PAGE 3

ii Population Data 45 Area Social Predictors of Health (ASPoH) Data 45 Statistical Products 46 Analytic Methods 46 Race and Sex Specific Mortality Rates 46 Black-White Disparity Scores 47 Statistical Methodology 48 Principal Component Analysis Methodology 50 Steps in Conducting Principal Component Analysis 51 Preliminary Analyses 53 Dividing ASPoH Measures into Quartiles 53 Research Question One Analyses 54 Research Question Two Analyses 56 Research Question Three Analyses 57 Chapter Four: Results 58 Part I. Area Social Predictors of Health 58 Descriptive Statistics 58 Principal Components Analyses Results 61 Part II. Research Question One 62 Descriptive Statistics 63 Numerator Data: Stroke Death Counts 63 Denominator Data: Florida Population Counts 64 Stroke Mortality Rates 67 Race-Sex-10 year Age Gr oup Specific Mortality Rates 67 Race and Sex Specific Age-Adju sted Stroke Mortality Rates: Census Tract Level 68 Study Outcome Scores 69 Area Social Predictors of Health (ASPoH) Quartiles 69 Descriptive Statistics 69 Regression Findings: Census Tract Level Analyses 74 Restricted Subset of Census Tracts 77 Descriptive Statistics 77 Multiple Regression Models 77 Summary of Findings 81 Part III. Research Question Two 82 Descriptive Statistics 83 Social Class Group Population Counts 83 Total Census Tract Stroke Deaths by Race and Sex 86 Stroke Death Rates and Outcom e Scores: Census Tract Level 89 All Florida Census Tracts 89 Restricted Census Tracts 90 Summary of Findings 94 Part IV. Research Question Three 95 Descriptive Statistics 96

PAGE 4

iii Regression Analyses 97 Predictability of ASPoH1 across Social Class Groups 98 Male Ratio: ASPoH-1 98 Female Ratio: ASPoH-1 98 Male Difference: ASPoH-1 99 Female Difference: ASPoH-1 100 Predictability of ASPoH2 across Social Class Groups 100 Male Ratio: ASPoH-2 100 Female Ratio: ASPoH-2 101 Male Difference: ASPoH-2 101 Female Difference: ASPoH-2 102 Predictability of ASPoH3 across Social Class Groups 102 Male Ratio: ASPoH-3 102 Female Ratio: ASPoH-3 103 Male Difference: ASPoH-3 103 Female Difference: ASPoH-3 104 Predictability of ASPoH4 across Social Class Groups 104 Male Ratio: ASPoH-4 104 Female Ratio: ASPoH-4 105 Male Difference: ASPoH-4 106 Female Difference: ASPoH-4 106 Summary of Findings 107 Chapter Five: Discussion 109 Introduction 109 Major Findings 110 Research Question One 110 Predictability of ASPoH Variables 111 Research Question Two 114 Research Question Three 116 Strengths and Limitations 118 Consistency with the Literature 124 Public Health Implications 127 Future Research 129 Conclusion 130 List of References 133 Appendices Appendix A: Residential Address Geocoding Methods 145 Appendix B: Definiti on of Study Variables 146 Appendix C: Calculation of Princi pal Component Analyses Variables 152 Appendix D: Study Acronyms 154 About the Author 155

PAGE 5

iv List of Tables Table 2.2. Trends in Stroke Death Ra tes per 100,000 Population (all ages), Florida and US 26 Table 3.1 International Classificati on of Diseases (ICD) Codes for Cerebrovascular Diseases 44 Table 3.2 Year 2000 Standard population weights 46 Table 3.3 Disparity score calculation methods 47 Table 3.4 Twelve core dimensions to understa nding social determinants of health 49 Table 4.1 Summary Statistics: Area Social Predictors of Health Variables, 2000 US Census 59 Table 4.2 Pearson Correlation Coefficien ts: ASPoH Variables, 2000 US Census 60 Table 4.3 Eigenvalues of the correlation matr ix, Principal Component s Analyses 61 Table 4.4 Factor Loadings: Principal Com ponents Retained in Further Analyses 62 Table 4.5 Percentage of Stroke D eaths by Race-Sex-Age group, Florida 1998-2002 64 Table 4.6 2000 US Census Population counts by Race-Sex-Age group 65 Table 4.7 2000 US Census population percent distribution by race-sex-age group 66 Table 4.8 Race-sex 10-year age group sp ecific stroke mortality rates*: Census tract 67 Table 4.9 Race and Sex Specific Age-Adju sted (35-74 yrs) Stroke Mortality Rates*, Florida 1998-2002 69 Table 4.10 Summary statistics for Black-White stroke mortality disparity measures 69 Table 4.11 Interquartile range of calcula ted census tract values for each ASPoH variable, 2000 US Census 70

PAGE 6

v Table 4.12 Mean Race-sex specific stroke mortality by Quartile: ASPoH-1 71 Table 4.13 Mean Black-White Stroke Mort ality Disparity by Quartile: ASPoH-1 71 Table 4.14 Mean Race-sex specific stroke mortality by Quartile: ASPoH-2 72 Table 4.15 Black-White Stroke Mortalit y Disparity by Quartile: ASPoH-2 72 Table 4.16 Mean Race-sex specific stroke mortality by Quartile: ASPoH-3 72 Table 4.17 Mean Black-White Stroke Mort ality Disparity by Quartile: ASPoH-3 73 Table 4.18 Mean Race-sex specific stroke mortality by Quartile: ASPoH-4 73 Table 4.19 Mean Black-White Stroke Mort ality Disparity by Quartile: ASPoH-4 74 Table 4.20 Regression model whic h measured the association between the male Black-White stroke mortality ratio and the ‘Area Social Predictors of H ealth’ variables. 74 Table 4.21 Regression model whic h measured the association between the female Black-White stroke mortality ratio and the ‘Area Social Predictors of Health’ variables. 75 Table 4.22 Regression model whic h measured the association between the male Black-White stroke mortality difference score and the ‘Area Social Predictors of Health’ variables. 75 Table 4.23 Regression model whic h measured the association between the female Black-White stroke mortality difference score and the ‘Area Social Predictors of Health’ variables. 76 Table 4.24 Regression model whic h measured the association between the male Black-White stroke mortality percent difference score and the ‘Area Social Predicto rs of Health’ variables. 76 Table 4.25 Regression model whic h measured the association between the female Black-White stroke mortality percent difference score and the ‘Area Social Predictors of Health’ variables. 76 Table 4.26 Descriptive Statistics for Bl ack-White stroke mortality disparity measures: Restricted Subset 77 Table 4.27 Regression model whic h measured the association between the male Black-White stroke mortality ratio and the

PAGE 7

vi ‘Area Social Predictors of Health ’ variables, Restri cted subset of census tracts 78 Table 4.28 Regression model whic h measured the association between the female Black-White stroke mortality ratio and the ‘Area Social Predictors of Health ’ variables, Restri cted subset of census tracts 78 Table 4.29 Regression model whic h measured the association between the male Black-White stroke mortality difference score and the ‘Area Social Predictors of Health’ variable s, Restricted subset of census tracts 79 Table 4.30 Regression model whic h measured the association between the female Black-White stroke mortality difference score and the ‘Area Social Predictors of Health’ variable s, Restricted subset of census tracts 79 Table 4.31 Regression model whic h measured the association between the male Black-White stroke mortality percent difference score and the ‘Area Social Predicto rs of Health’ variables, Restricted subset of census tracts 80 Table 4.32 Regression model whic h measured the association between the female Black-White stroke mortality percent difference score and the ‘Area Social Predictors of Health’ variables, Restricted subset of census tracts 81 Table 4.33 2000 US Census Percent Popul ation by race, sex and age group, Florida, All educational attain ment groups 84 Table 4.34 Black male population count by social class and age-groups (35+years), 2000 US Census, Florida population multiplied by 5 years 84 Table 4.35 Black female population c ount by social clas s and age-group (35+years), 2000 US Census, Florida population multiplied by 5 years 85 Table 4.36 Non-Hispanic White male population count by so cial class and age-group (35+years), 2000 US Ce nsus, Florida population multiplied by 5 years 85 Table 4.37 Non-Hispanic White male population count by so cial class and age-group (35+years), 2000 US Ce nsus, Florida population multiplied by 5 years 86

PAGE 8

vii Table 4.38 Percent Stroke Deaths for Race-Sex Groups By Age-Group, Florida 1998-2002 87 Table 4.39 Percent Stroke Deaths by Social Class Group, Florida 1998-2002 87 Table 4.40 Percent Stroke Deaths for Race-Sex Groups by Social Class Groups, Florida 1998-2002 88 Table 4.41 Number of census tracts ( by race, sex, and age-group) for which educational attainment data were reported, Florida, 2000 US Census, Summary File 4 89 Table 4.42 Weighted average stroke death rates and disp arity scores by social class group, Florida 1998-2002, Census tract level 90 Table 4.43 Race-Sex Specific Stroke Deat h Rates (per 100,000): 35-44 Years of Age, Florida 1998-2002 92 Table 4.44 Race-Sex Specific Stroke Deat h Rates (per 100,000): 45-64 Years of Age, Florida 1998-2002 92 Table 4.45 Race-Sex Specific Stroke Deat h Rates (per 100,000): 65-up Years of Age, Florida 1998-2002 93 Table 4.46 State level age-adjusted (35+ y ears) stroke death ra tes and disparity scores by social cl ass group, Florida 1998-2002 94 Table 4.47 State level Black-White disparity scores by social class group, Florida 1998-2002 95 Table 4.48 Florida population and stroke death counts by social cl ass category 96 Table 4.49 Effect Modification: Summary Statistics for Black-White stroke Mortality disparity measures 97 Table 4.50 Individual regression models which measured effect modification by social class of the association be tween the male Black-White stroke mortality ratio and the ‘Area Social Predictors of Hea lth-1’ variable. 98 Table 4.51 Individual regression models which measured effect modification by social class of the asso ciation between the female Black-White stroke mortality ratio and the ‘Area Social Predictors of Health-1’ variable. 99

PAGE 9

viii Table 4.52 Individual regression models which measured effect modification by social class of the association be tween the male Black-White stroke mortality difference score and the ‘Ar ea Social Predictors of Health-1’ variable. 99 Table 4.53 Individual regression models which measured effect modification by social class of the asso ciation between the female Black-White stroke mortality difference score and the ‘Ar ea Social Predictors of Health-1’ variable. 100 Table 4.54 Individual regression models which measured effect modification by social class of the association be tween the male Black-White stroke mortality ratio and the ‘Area Social Predictors of Health-2’ variable. 100 Table 4.55 Individual regression models which measured effect modification by social class of the asso ciation between the female Black-White stroke mortality ratio and the ‘Area Social Predictors of Health-2’ variable. 101 Table 4.56 Individual regression models which measured effect modification by social class of the association be tween the male Black-White stroke mortality difference score and the ‘Ar ea Social Predictors of Health-2’ variable. 101 Table 4.57 Individual regression models which measured effect modification by social class of the asso ciation between the female Black-White stroke mortality difference score and the ‘Ar ea Social Predictors of Health-2’ variable. 102 Table 4.58 Individual regression models which measured effect modification by social class of the association be tween the male Black-White stroke mortality ratio and the ‘Area Social Predictors of Health-3’ variable. 103 Table 4.59 Individual regression models which measured effect modification by social class of the asso ciation between the female Black-White stroke mortality ratio and the ‘Area Social Predictors of Health-3’ variable. 103 Table 4.60 Individual regression models which measured effect modification by social class of the association be tween the male Black-White stroke mortality difference score and the ‘Ar ea Social Predictors of Health-3’ variable. 104 Table 4.61 Individual regression models which measured effect modification by social class of the asso ciation between the female Black-White stroke mortality difference score and the ‘Ar ea Social Predictors of Health-3’ variable. 104

PAGE 10

ix Table 4.62 Individual regression models which measured effect modification by social class of the association be tween the male Black-White stroke mortality ratio and the ‘Area Social Predictors of Health-4’ variable. 105 Table 4.63 Individual regression models which measured effect modification by social class of the asso ciation between the female Black-White stroke mortality ratio and the ‘Area Social Predictors of Health-4’ variable. 105 Table 4.64 Individual regression models which measured effect modification by social class of the association be tween the male Black-White stroke mortality difference score and the ‘Ar ea Social Predictors of Health-4’ variable. 106 Table 4.65 Individual regression models which measured effect modification by social class of the asso ciation between the female Black-White stroke mortality difference score and the ‘Ar ea Social Predictors of Health-4’ variable. 106 Table 4.66 Summary results for regre ssion models which measured effect modification by social class of the association between the disparity in stroke mortality measures and the ‘Area Social Predictors of Health’ variables 108

PAGE 11

x List of Figures Figure 1. Theoretical Causal Model 12 Figure 2. Examined Theoretical Model 13

PAGE 12

xi Impact of Area Social Predictors of Hea lth on Black-White Disparities in Stroke Mortality Tyra Dark ABSTRACT This dissertation investigated the area so cial predictors of health (ASPoH) and Black-White disparities in stroke mortality re lationship. Utilizing st roke mortality data obtained from the Florida Department of Health for years 1998-2002, and social and economic data obtained from the year 2000 Census of Population, this study examined the effect of resource availability at the census tract level on Black-White disparities in stroke mortality. The influence of social class on Black-White disparities in stroke mortality and effect modification by social cl ass of the associati on between Black-White disparities and ASPoH variables was also in vestigated. Principal component analysis produced four ASPoH scores from economic a nd social measures. Multiple regression analysis assessed the predictive ability of these ASPoH variables on Black-White disparities. Increases in the female Black-White ratio were significantly associated with increases in the magnitude of the ASPoH-1 and ASPoH-2 variables. When regression analyses were restricted (in terms of populat ion count minimums) to a subset of census tracts, increases in the ASPoH-1 and ASPoH2 variables were significantly associated with increases in all Black-White disparit y measures for both males and females. Assessment of the influence of social cl ass on Black-White disparities in stroke

PAGE 13

xii mortality was only feasible at the state level due to a lack of data at the census tract level. With the exception of the 65+ years age-group, Black males and females experienced higher age-group specific stroke mortality rates across each of the social class groups. Inconsistent with previous research findings Black residents who a ttained a high school degree had the highest stroke death rates comp ared to all other educational attainment groups. In the assessment of social class as a potential effect modifier, the study hypothesis stated that the A SPoH measures would have th e greatest impact on those residents in the lowest soci al class category. This pred icted effect was only supported when the Male Black-White rati o disparity score was examined. Study findings support the conjecture th at unknown and unmeasured processes influence the association between area social predictors and stroke mortality for Black Floridians. Identification of modifiable so cietal characteristics may be the key to unlocking the foundation of dispar ities in health outcomes.

PAGE 14

1 Chapter One Introduction Quantification of neighborhood characteristics has become an important, and vastly utilized, research tool in determ ining influences on ‘small geographic area’ morbidity and mortality rates. The use of this tool is rendered possi ble by the availability of census tract population and social and ec onomic data which are typically used to construct measures of neighborhood characteri stics. Utilization of census data has proven to be an important resource for eluc idating relationships between socioeconomic position and health outcomes for the U.S. as well as for the neighborhood level (census tract).1,2 Furthermore, public health research has shown differential findings by race between socioeconomic status and health outcomes;3 consequently, the role of neighborhood socioeconomic context in the contribution and exacerbation of racial inequalities in morbidity and mortality presen ts as the next logical question for public health researchers. The challenge posed by this type of res earch is the identification and accurate definition of the aspects of the neighborhoods which are influential in health outcomes and opportunities for health promoting behavior s. Social and public health researchers have suggested using a framework of universal human needs as a basis for thinking about how places may influence health, and recomm end testing of hypotheses about specific chains of causation that might link pl ace of residence with health outcomes.4 Research

PAGE 15

2 on the resources that humans need in order to live a healthy life a nd the importance of the geographic distribution of these available resources as it re lates to the distribution of health is needed.4 Stroke mortality, an adverse health outc ome with specific socioeconomic status and geographical distributions, is an ideal health outcome that can be utilized when investigating the associa tion between available neighborhood resources and the differential distribution of health outcome s. There are many critical aspects of understanding stroke outcomes. Various fact ors have been shown to be significant contributors to stroke mortality rates.5,6 Understanding the many mechanisms leading to stroke mortality, determining the contributo rs to stroke mortality and, maybe most importantly, reconciling the relationship betw een these contributors is paramount in lessening the burden of this disease in our society. Deciphering these mechanisms may also lead to a better understanding of w hy certain groups experience a heavier disease burden. Factors consistently identified as influent ial in stroke mortality include: ethnicity, age, gender, lower socioeconomic status (as defined by occupation onl y), social class (as defined by occupation and/ or education), a nd health risk behaviors (smoking, drinking, physical activity, diet), hypertension, diabet es, end-stage renal di sease, and obesity.7,8 In addition, disparities in stroke mortality, as well as other morbidities and causes of mortality, exist between specifi c ethnic and social subgroups within our society. The health disparities persist even after contro lling for the majority of the aforementioned contributors. 7,8 Therefore, the potential relevant relationships between socioeconomic factors (education and income) and lifestyle, between lifestyle and ethnicity/race, as well

PAGE 16

3 as the relationship between ethnicity/race and so cioeconomic factors, must be taken into consideration as possible mediators in the pa thway leading to stroke mortality. Reducing the prevalence of health risk behaviors, reducing the proportion of those living in poverty, and increased education would more th an likely reduce the prevalence of stroke mortality in at risk subpopulat ions, as well as various other undesirable health outcomes. However, it is the premise of this study that differences in stroke mortality are due to a wider array of factors, many of which may be specific to contextual social and economic characteristics of small geographic areas. Th is study will explore the role of social and economic characteristics in racial disparities in stroke mortality (Refer to references 7 and 8 for rate difference between Blacks and Whites. Race specific rates are also presented in the following chapter). The specific contextual characteristics to be addressed in this study are ‘area social pr edictors of health’ (ASPoH) status at the census tract level. The ASPoH index will be fully developed within the methods chapter of this document. Proposed Research Questions and Hypotheses Research Question 1: Are Black-White disparities in st roke mortality elevated in those areas of low ASPoH status? Hypothesis: Black-White disparities in stroke mortality will be greatest at lower levels of ASPoH. Research Question 2: Are higher levels of Black-White disparities in stroke mortality associated with low levels of social class? Hypothesis: Black-White disparities in stroke mortality will be greatest for those in the lowe st social class group. Research Question 3 : Is there effect modification by soci al class of the ASPoH status and

PAGE 17

4 Black-White disparities in st oke mortality relationship? Hypothesis: ASPoH status will have a greater impact on Black-White disparities in stroke mortality for the lower social class groups. However, the association between disparities and ASPoH status will persist after controlling for individual social class (educational attainment). Study Purpose/Rationale Stroke is the third leading cause of death after heart di sease and cancer in the US.9 Stroke mortality rates for Black Americans ar e substantially higher than those of White Americans.9 Analogous to findings for other adverse health outcomes, there is consistent evidence of an unequal distributi on of stroke deaths across social class. Generally, those individuals making up the higher social class group experience better health outcomes.10 When these associations are examined separately for Whites and Blacks, worse outcomes are typically observed for Blacks at any given level of social class.11 Similar results have been found fo r the association between socioeconomic status and stroke mortality;12 not a surprising finding give n that the same or related indices of social class are used as indices of socioeconomic status. Results from these studies have varied depending on the level of socioeconomic status being investigated (individual or community SES). Consistent associations are seen at the individual level;13,14 however, results are mixed when the e ffect of SES is investigated at the community level.15,16 In an attempt to further our understa nding of these disparities, current investigations now focus on as pects of the community in which we live as a contributing factor in these continued disparities. Whether these communities are created out of

PAGE 18

5 natural growth or if the development of a community is determined by government acts/laws must be taken into considerat ion. Of concern is the lawfulness and appropriateness of land use zoning practices. Whether or not zoning practices lead to segregation by race and economic circumstance is the present question. If these practices lead to homogeneous groups of people resi ding in less healthy areas due to their economic circumstances and minority status, atte ntion must be given to these issues in order to attempt to understand th eir role in geographic and ra cial health disparities. Research on zoning laws and practices refl ect that harmful land uses tend to be disproportionately concentrated in poor a nd industrial neighborhoods which tend to have larger minority populations. 17,18 Authors have highlighted th e injustice of these zoning practices and hypothesized about the underlying belief systems (overt or covert racism, putting economic profits over the health of pe ople, or benign neglect) that lead to ‘disproportionate risk.’17,18 Also of concern in proposing a causal pathway leading from area economic and social measures to racial disparities in st roke mortality is the potential influence of residential segregation. Reside ntial segregation refers to the physical separation of the races in residential c ontexts. Residential segregation can lead to the formation of radically different environments for the se gregated group and the rest of the population. The possibility exists that this study will capture elements of residential segregation at the census tract level. If this is the result, the influence of residential segregation on the study outcomes must be considered. At issu e would be the question of whether those census tracts with greatest impact on study re sults are those tracts with relatively large Black populations. These particular census tracts may have greater impact on study

PAGE 19

6 results because they contain sufficient num bers of Black residents allowing for the calculation of stroke mortality measures. Th e next consideration would be whether those census tracts with sufficient Black population ex ist as a result of residential segregation. If residential segregation pr ocesses are influential in determining racial residential groupings within the census tracts, this may aff ect the probability that people living in the same area (census tract) may not have acces s to the same amount and quality of resources. Unspoken norms and/or rules of behavior among residents within small geographic area may dictate the one carries out his/her daily activities within a restricted area within the census tract. Residents, who travel beyond the bounds of their ‘designated’ area, may experience discomfort wi thin these locations. Therefore, residents may remain within their comfort zone and not take advantage of a ll seemingly available resources within their neighborhood. Research investigating possible health eff ects of residential segregation show that even after controlling for important risk factors (such as e ducation, income and occupational status), segregation may have a statistically signifi cant effect on various health outcomes.19,20 Residential segregation is to propos ed influence racial disparities in health because of its capacity to capture some of the effects of racism. Researchers propose the community level effects of resident ial segregation as one potential reason for the persistence of racial differences in health status even after c ontrolling for individual variations in socioeconomic status.21 Residents of disadvantag ed neighborhoods have a higher incidence of heart di sease than people who do not reside in disadvantaged neighborhoods.22 This effect persists after adjustment for education, income, occupational status, and biomed ical and behavioral risk factors for coronary heart

PAGE 20

7 disease. Residential segregation has also pr oven to be a significant predictor of mortality among adult African Americans23 as well as among Black infants.24 Environmental factors, and their diffe rential concentration within certain geographic areas, are closely related to personal behavi or and lifestyle.25 The need to “explicitly acknowledge the intimate connect ions between the social and economic conditions people live under and their biob ehavioral risk factor profile” has been expressed.25 How one interacts in his/her commun ity (social interact ion, participation, cohesion, and social networks) may have influe nces on risk factor exposure probability. This may be due to the presence of community characteristics that support or influence the probability of exposure to factors lead ing to adverse health outcomes. The “connections between these adverse {envir onmental} conditions and the adaptive responses affected communities must often make to them” needs further research.25 Investigation of economic aspects of the community (such as income inequality) in an attempt to demonstrate how relative deprivation may influence or increase the prevalence of adverse health behaviors (sm oking, alcohol, sedentarin ess, unhealthy diet) have been undertaken.26,27 The current study aims to continue this research by investigating the relationship between availability of and opportunity to obtain economic resources (captured by a measure of area social predictors of health) and racial disparities in stroke mortality. The construction of the ‘Area Social Predictors of Health’ measure incorporated dimensions of economy, empl oyment, education, a nd housing conditions. Controlling for census tract level social pos ition indicators through stratification is the technique utilized to analytical ly demonstrate an effect of the area in which one lives on health; therefore, potentially demonstrating an “independent” effect of ASPoH status.

PAGE 21

8 Proposed Pathway: ASPoH and Black-Wh ite Disparities in Stroke Mortality It is proposed that through the followi ng theoretical causal pathway, resource availability, which is hypothesized to de termine risk factor exposure potential, significantly contributes to racial disparities in stroke mortality. A relationship between ASPoH status a nd Black-White disparities in stroke mortality is proposed. The pr oposed model depicts the eff ect of ASPoH status on the local environments contributing to these dispar ities (see Figure 1). In those areas of low ASPoH status, a concentration of Blacks living in poverty is expected. According to the literature, the quality of housing in these ar eas will be of low market value and more likely to be overcrowded.28 Researchers have suggested that financial institutions determine the location of their establishments based on the credit worthiness of the local residents;4 therefore, the theoretical pathway predic ts that the availabi lity of banks and other investment institutions will be in short supply in lower ASPoH areas. The commitment of monetary resources to the phys ical maintenance of these areas will be lacking, possibly due to the ‘z oning’ laws which influence the distribution of community maintenance funds. It is believed that those residents livi ng in lower ASPoH areas will have reduced access to private transportati on and increased access to publ ic transportation services.29 Those residents without private transportation may be less likel y to possess resources that permit travel beyond their immediate residence. This restriction would thereby limit their access to the full array of resources within the census tract. The model presumes reduced availability of and/or access to medical care for residents of lower ASPoH areas. Among indi viduals with moderate or low incomes,

PAGE 22

9 those without health insurance generally have less access to medical care than those with coverage.30 The lack of annual physical exams ma y contribute to poor health, due to the inability to pay out-of-pocket medical expens es. Without periodic physical exams, early detection of disease may not be possible. Al so, the availability of important information regarding preventive stra tegies for diseases in which th e uninsured may be at increased risk will be lacking.31 Additionally, African Americans are more likely to be uninsured than White Americans.32 The aforementioned evidence coupled with the evidence that Black Americans are more likely to reside in more disadvantaged areas lends strength to the prediction that Black-White disparities in stroke mortality will be more pervasive in the lower ASPoH neighborhoods. In lower ASPoH areas, the model theorizes reduced availability of emergency medical care facilities that are properly prepared to treat stroke victims. Investigations into the role of socioeconomic status on access to health services af ter stroke, found that patients in lower income categories were le ss likely to have access to hospitals with neurologists and imagery equipment necessa ry for the diagnosis and treatment of stroke.33 Findings additionally suggest that the disparity in access to those hospitals properly staffed and equipped for stroke trea tment is related to the distribution of specialized resources in mo re affluent neighborhoods.33 It is presumed that those outcome s listed above create an environment characterized by high unemployment rates and high poverty rates among residents. These conditions, and the lack of more positive circumstances, can in turn contribute to increases in the prevalence of adverse health behaviors such as smoking and drinking. This effect has been attribut ed to racial and neighborhood sp ecific targeted marketing and

PAGE 23

10 the increased availability and concentrati on of convenience stores which supply these products.34,35,36 Conditions associated with living in lo wer ASPoH neighborhoods may also lead to sedentary lifestyles. Physical inactivity ma y be due to a lack of recreational facilities and proper sidewalks in which to exercise.37,38 Concern about personal safety issues may also lessen the probability of physical act ivity within these lower ASPoH areas. The proposed model theorizes that health food stores will be scarce in lower ASPoH areas. Research shows a lower con centration of supermarkets and a higher concentration of locally owned food stores are found in less affl uent areas and in areas in with large African Americans populations.39 Additionally, the affordability and availability of recommended foods for healt hy diets may be reduced in lower income areas.40 The lack of supermarkets and health food stores with affordable and quality foods may contribute to consuming unhealthy diets. The importance of the differential distribution of these food stor es is rendered even more si gnificant for Black Americans when it coupled with the findings that the di et of Black Americans significantly improves as the number of supermarkets in their residential area (censu s tract) increases.41 The proposed model (see Figure 1) theorize s that living in an environment with fewer resources will have adverse health outcomes at varying degrees for Black and White Floridians. It is theorized that th e low ASPoH environment will have negative effects on health behaviors accompanied by a higher prevalence of smoking, drinking, unhealthy diets and sedentary lifestyles. Th e model theorizes that these lower ASPoH groups will tend to experience a higher prevalence of stroke risk factors. Furthermore, because Black Americans are more likely to live in these disadvantaged areas,42 the

PAGE 24

11 model predicts Black-White disparities in the prevalence of stroke risk factors. It is theorized that all of the f actors mentioned above contribut e to higher incidences of hypertension, obesity, diabetes and atheroscle rosis among black residents culminating in greater Black-White disparities in stroke mortality in low ASPoH level areas versus higher ASPoH level areas. The primary purpose of this study is to identi fy contextual area characteristics that may be related to health outcomes independent ly of and/or in conj unction with social class status (see Figure 2). Specifically, the relationship between ASPoH level and Black-White disparities in stroke mortality will be investigated. In addition, it will be determined whether this association varies acros s levels of social cl ass. The effect of social class is examined due to the establ ished relationship between social class and health disparities in published literature. Consequently, a thorough examination of the effect of area resource measures on Black-White disparities in stroke mortality requires that the potential influence of social class be assessed. Many questions remain regarding the basis for racial differences in stroke mort ality. This study is an attempt to answer a number of these questions, and to sugge st directions for future research.

PAGE 25

12 Figure 1: Theoretical Causal Model ASPoH Poverty Rates % with < HS education Percentage Unemployed Median Rent Home Ownership Boarded-Up Units B-W Social Class Disparities in Stroke Risk Factors Hypertension Atherosclerosis Diabetes Obesity Adverse Health Behavior Smoking Rates Drinking Rates Sedentary Lifestyle Unhealth y Diets Physical Structures (Differential Distribution by Race/Social Class) Park/Recreational Facilities Supermarkets Alcohol Outlets Medical Care Facilities Banks Individual Social Class Defined by Educational Attainment Black-White Disparities in Stroke Mortality (Census Tract Level)

PAGE 26

13 Fi g ure 2 : ExaminedTheoretical Model Area Social Predictors of Health (ASPoH) Area Social Class BlackWhite Disparities in Stroke Mortality

PAGE 27

14 Chapter Two Review of the Literature Stroke Definition Stroke occurs when there is a sudden co mplication affecting the blood vessels of the brain.43,44 There are two major categories of stroke, ischemic and hemorrhagic, and several stroke types within each of these categories. The three major stroke types include: ischemic stroke, intr acerebral hemorrhage and subarachnoid hemorrhage. Of all strokes, 88 percent are ischemic, 9 percent are intracerebral hemorrhage and 3 percent are subarachnoid hemorrhage.45 The following section desc ribes the similarities and differences of these stroke types. Ischemic Stroke Ischemic stroke, the most common type of stroke, results from closure or blockage of an artery leading to the brain. There are several causes of ischemic stroke. The most common cause is excessive narrowing of the arteries in the neck or head, usually resulting from atherosclerosis (a gr adual cholesterol deposit ion). This narrowing of the arteries can lead to the formation of blood clots with thrombotic stroke and embolic stroke as possible consequences. Thrombosis occurs when blood clots block the artery where they are formed. A thrombotic stroke is clinically referred to as cerebral thrombosis or cerebral infarction and is responsible for almost 50% of all st rokes. There are two categories of cerebral

PAGE 28

15 infarction, large-vessel and small-vessel thro mbosis, each correlati ng to the location of the blockage within the brain. Large-vessel thrombosis occurs when blockage is located within one of the brain’s larg er blood supplying arteries, wher eas small-vessel thrombosis occurs when there is blockage in one of the brain’s smaller and deeper penetrating arteries. An embolism results when the blood clot disl odges and becomes trapped within arteries closer to the brain. In instances of embolic stroke, the clot (or embolus) was formed somewhere other than in the brain itself. These embo li, which often are released from the heart, travel the bl oodstream until they become trapped. This blockage restricts the flow of blood to the brain, and results in almost immedi ate physical and neurological deficits. Hemorrhagic Stroke Hemorrhagic strokes, intracerebral (w ithin the cerebrum, or brain) and subarachnoid (area of skull surrounding the brain) hemorrhagic, occur when there is bleeding of ruptured blood vesse ls in the brain. Intracereb ral hemorrhage occurs when a weakened blood vessel within the brain bursts, allowing blood to leak inside the brain. A sudden increase in pressure within the brai n can cause damage to the brain cells and possibly lead to unconsciousness or death. In tracerebral hemorrhage usually occurs in selected parts of the brain, in cluding the basal ganglia, cerebe llum, brainstem, or cortex. The most common cause of intracerebral hemorrhage is high blood pressure (hypertension). Less common causes include trauma, infections, tumors, blood clotting deficiencies, and abnorma lities in blood vessels. Subarachnoid hemorrhage occurs when a blood vessels just outside the brain

PAGE 29

16 ruptures. The subarachnoid sp ace rapidly fills with blood, po ssibly resulting in loss of consciousness or death. Subarachnoid hemorrh age is often caused by abnormalities of the arteries at the base of the brain, called ce rebral aneurysms. These are small areas of rounded or irregular swellings in the arteries, with the most severe swelling resulting in weakening and rupturing of the arterial wall. Public Health Importance of Stroke Approximately 700,000 people experience a ne w or recurrent stroke each year, establishing stroke as one of the major publ ic health problems in the United States today.46 About 500,000 of these are first attacks, and 200,000 are recurrent attacks. Eight to 12 percent of ischemic st rokes and 37-38 percent of he morrhagic strokes result in death within 30 days. The ageadjusted stroke incidence rate s (per 100,000) for first-ever strokes are 167 for White ma les, 138 for White females, 323 for Black males and 260 for Black females.46 Stroke is the leading cause of serious, long-term disabi lity in the United States, resulting in mounting economic costs.45 In 2004, it was estima ted that Americans would pay $54 billion in direct and indi rect cost of stroke. The mean lifetime cost of ischemic stroke in the United States is estimated at $140,048. These costs included inpatient care, rehabilitation, and follow-up care.47 Lifetime costs per patient are estimated at between $59,000 and $230,000.48 In the United States, stroke is the third leading cause of death behind diseases of the heart and cancer.45 Stroke accounted for more th an 1 of every 15 deaths in the United States in 2001. Stroke kills nearly 168,000 people a year, an average of one stroke death every three minutes Women experience three of ev ery 5 deaths from stroke.

PAGE 30

17 The overall (crude) stroke death rate for 2002 was 56.2 (p er 100,000 US population). The 2002 stroke death rates per 100,000 popul ation for specific groups were 54.2 for White males, 53.4 for White females, 81.7 for Black males and 71.8 for Black females.45 Blacks have higher stroke mortality rate th an whites. Higher stroke mortality rates among Blacks can be attributed to a greater in cidence of stroke in Blacks, given that thirty-day case-fatality rates, regardless of stroke subtype, compared between the two races is similar (see Table 2.1).49 Table 2.1. Thirty-day stroke case-fatality rates by race and stroke subtypes 1999: 30-day case fatality All Black White All Stroke subtypes 14.7% 12.8% 16.9% Ischemic stroke 10.2% 9.1% 11.5% Intracerebral Hemorrhage 37.6% 36.2% 39.0% Subarachnoid Hemorrhage 31.3% 28.2% 34.7% Stroke Risk Factors Non-traditional contributors (e.g., co mmunity economic measures, social participation, social cohesion) to stroke incidence and morta lity continue to emerge as a greater amount of literature becomes available. However, investigations have identified a consistent grouping of important risk factors for stroke. These risk factors may or may not be modifiable through actions initiate d by the individual. Stroke risk factors identified by the American Heart Associa tion include: (1) being African American, (2) older than 55 years of age, (3) male (alt hough more women die from stroke than males), (4) high blood pressure (5) heart disease, (6) diabetes mellitus, (7) prior stroke, (8) heredity and (9) cigarette smoking. Secondary risk factors for stroke include: high blood cholesterol, physical inactivity, and being overweight or obese.50 High Blood Pressure and Stroke A review of published literature of th e relationship between blood pressure and

PAGE 31

18 stroke showed that the risk of stroke incr eases continuously in a ssociation with blood pressure levels greater than 115/75 mm Hg.51 Results also demonstrated that the epidemiologically expected benefits of blood pr essure lowering for st roke risk reduction are broadly consistent across a range of different population subgroups. Depending on age, Blacks have 2 to 5 times the prevalence of hypertension of White s. The great bulk of the adult health differential between Blacks an d Whites can be ascribed to this factor. Diabetes Mellitus Diabetes is an independent risk factor for stroke, and is strongly correlated with high blood pressure. While diabetes is trea table, the presence of the disease still increases the risk of stroke,46 with the relative risk ranging from 1.8 to almost 6.0.52 Diabetes is one of the most important ri sk factors for stroke in women. In the Framingham Heart Study and in several Eur opean studies, the impact of diabetes on stroke risk is greater in women than in men.52,53 Cigarette Smoking The relative risk of stroke among heavy sm okers (more than 40 cigarettes a day) is twice that of light smokers (less than 10 cigarettes per day). Stroke risk decreases significantly after two years of smoking cessation and is at the level of nonsmokers by five years after cessation of cigarette smoking.45,54 Among Americans age 18 and older, 25.2 percent of men and 20.7 percent of wome n are smokers, putting them at increased risk of heart attack and stroke.45 In 1950 Blacks smoked less than Whites, bu t as a result of migration to large urban centers, this pattern began to change.45 Although a decline has been reported for all groups, Blacks continue to smoke more than Whites, particularly Black males. There

PAGE 32

19 is an inverse relationship between social cl ass and prevalence of smoking in the US. Smoking prevalence is higher among those with 9-11 years of education (35.4 percent) compared with those with more than 16 year s of education (11.6 percent). It is highest among persons living below the poverty level ( 33.3 percent) compared with other income groups.55 The higher prevalence of smoking among lower social class and socioeconomic groups undoubtedly contributes to higher stroke morta lity rates for these groups. High Blood Cholesterol and Other Lipids The higher a person’s high density lipoprotei n level (HDL), the better the chance of that person not experiencing stroke or hear t disease. HDL carries at least one-third of blood cholesterol away from the arteries and back to the liver where it is passed from the body. Research posits that HDL removes exces s cholesterol from plaque in arteries, which slows plaque buildup, and lessens the risk of stroke or heart disease. Low HDL cholesterol (less than 40 mg/dL in adults) is a risk factor for heart disease and stroke. The mean level of HDL chol esterol for American adults age 20 and older is 50.7 mg/dL.45 The mean level of LDL cholesterol for American adults age 20 and older is 127 mg/dL. Levels of 130-159 mg/dL are cons idered borderline high. Levels of 160-189 mg/dL are classified as high, and levels of 190 mg/dL and higher are very high.45 Among non-Hispanic Whites, 20.4 pe rcent of men and 17.0 percen t of women have an LDL cholesterol level of 160 mg/dL or higher. Among non-Hispanic Bl acks, 19.3 percent of men and 18.8 percent of women have an LDL cholesterol level of 160 mg/dL or higher. Results demonstrate that there is very little racial difference in the prevalence of this particular risk factor.

PAGE 33

20 Physical Inactivity Based on data from the 1997-2003 NHIS surv eys of the CDC/NCHS, 31.3 percent of U.S. adults age 18 and older engaged in regular leisure-time activity.56 For age groups 18-24 and 25-64, women were less likely than men to engage in regular leisure-time physical activity. The age-sex-adjusted percen t of adults who engaged in regular leisuretime physical activity was 34.0 percent for non-Hispanic Whites, 26.4 percent for nonHispanic Blacks and 21.1 percent for Hispanics. Physical inactivity is more prevalent among women than men, among Blacks and Hisp anics than Whites, among older than younger adults and among the less affl uent than the more affluent.56 A recent study of over 72,000 female nurses indicates that modera te-intensity physical activity such as walking is associated with a s ubstantial reduction in risk of ischemic stroke as well as all stroke types combined.57 Overweight and Obesity The age-adjusted prevalence of overwei ght (BMI of 25.0 or higher) increased from 55.9 percent in NHANES III (1988-94) to 64.5 percent in NHANES IV (19992000).45 The prevalence of obesity (BMI of 30.0 or higher) also increased during this period from 22.9 percent to 30.5 percent. Extreme obesity (BMI of 40.0 or higher) increased from 2.9 percent to 4.7 percent (all prevalence measures were age-adjusted).58 Increases occurred for both men and women in all age groups and for non-Hispanic Whites, non-Hispanic Blacks and Mexican Amer icans. Racial and ethnic groups did not differ significantly in the prevalence of obe sity or overweight for men. Among women, obesity and overweight prevalences were highest among non-Hispanic Black women. More than half of the women age 40 and ol der were obese, and more than 80 percent

PAGE 34

21 were overweight. The prevalence of obesity (BMI 30 or higher) in 2001 increased 5.6 percent between 2000 and 2001(BRFSS, CDC/NCHS). Re search suggests that overweight men have a greater risk of developing stroke th an those with normal levels of total body fat.59 A comparison of risk factors in both the Honolulu Heart Program and Framingham Heart Study showed a BMI increase around 3 kg/m2 raised the risk of hospitalized thromboembolic stroke 10-30 percent.60 The Health Professionals Follow-up Study examined the association of body mass index and abdominal obesity (waist/hip ratio) with stroke incidence in over 26,000 males aged 40-75. Results suggest that for men, abdominal obesity is more closely relate d to stroke risk (rather than BMI).61 A prospective cohort study of middle-aged Isra eli men sought to clarify the relationship between excess weight, its distri bution, and stroke mortality. The ratio of subscapular to triceps skinfold thickness, an indicator of tr unk versus peripheral distribution of body fat, was found to be an independent predictor of long-term stroke mortality.62 For women, BMI and weight gain are indepe ndent risk factors for stroke.57 Stroke Mortality Trends Widespread declines in stroke mortality have been observed over the past several decades. The overall decline in US stroke mortality rate accelerated in the decades between 1950 and 1980, with a mark ed acceleration noted after 1973.63 This reduction in stroke death rate occurred in both males and females for both White and Black Americans. Researchers have hypothesized th at the decline in stroke death rates may have been due to either decreased inciden ce of stroke, improved survival of stroke patients, or a combination of these effects. This downward trend in stroke mortality rates

PAGE 35

22 has also been attributed to improved treatment and control of hypertension.64 Widespread control of hypertension began to ta ke place in the early 1970s. The role of increased detection and control of hypertension in the dramatic reduction in mortality in this time period received support from several studies. Klag et al (1989) utilized US vital statistics during 1950-1972 and 1973-1981 to gather evidence of the validity of the putativ e association between the accelerated decline in stroke mortality and incr eased use of antihypertensives.65 Researchers propose the 1973 establishment of the National High Bl ood Pressure Education Program as a candidate to explain the incr ease in controlled hypertensi on, thereby resulting in the accelerated decline in stroke mortality shortl y thereafter. Stroke mortality declined throughout the study period, however, after 1973, acceleration in the ra te of decline was consistently seen in all age-race-sex groups. The rates of decline increased with age. Except in the 75-84 year olds, Blacks had greater rates of de cline than Whites. Authors attributed the ageand race-rela ted differences in the rate of decline in stroke mortality to the much higher baseline stroke mortality. However, results of the study lend no support to the proposed link between antihypertensive therapy and decline in stroke mortality. Finding no significant association, authors sugge st that treatment of hypertension may not be the principal reason for the decline in stroke mort ality after 1973. Alternative proposed candidates include: (1) some widespr ead environmental agen t (2) the targeting of hypertensives and (3) decr eased lead exposure (lead e xposure has been linked to hypertension and increased stroke incidenc e) that occurred in 1973-1980 in the US population. Expanding upon studies of the proposed an tihypertensive and stroke mortality

PAGE 36

23 decline link, Capser et al (1992) considered the influen ce of hypertension prevalence and socioeconomic profile (education, income and occupation indicat ors) on the proposed association.66 Results showed that larger changes in both stroke mortality and controlled hypertension occurred during the post-1972 year s than during the pre-1972 years. When the two study periods were combined, re sults showed an association between antihypertensive use and decline in stroke mortality. However, when time period was taken into consideration, no association wa s observed between treatment and mortality decline pre-1972. Additionally, groups with la rger accelerations in stroke mortality declines did not show larger changes in controlled hypertensio n. During the post-1972 years, the groups with the la rgest increases in prevalence of controlled hypertension experienced slightly slower rate s of decline in stroke mortality. Posing a challenge to the strength of the treatment-mortality decline hypothesis, results showed a consistent association between accelerated declines in stroke mortality and improvements in socioeconomic factors. Pre-1972, groups with the largest increases in education and income profiles experienced the slowest rates of decline in stroke mortality. Post-1972 the trend was reversed. The authors suggest that other factors may operate at the population level that either add to or detr act from the effectiveness of increased antihypertensive pharmacotherapy on declines in stroke mort ality or that influence the rates of stroke mortality directly. Geographical Differences in Stroke Epidemiology Large differences in cerebrovascular di sease mortality among geographic areas of the United States have been reported.67,68,69 Death rates were higher in the southeastern states, and lower in the plains and Roc ky Mountain regions. A study of hospitalized

PAGE 37

24 patients was performed to determine whether the mortality differences were due to a higher incidence or case fatality following a stroke in areas with high stroke death rates.70 Investigators found that the inci dence of stroke was higher in the high stroke death rate areas especially for men. The distribution of the specific types of stroke was similar among the areas. No consistent pattern in fr equency of symptoms of stroke on admission to hospital was seen. Possible differences in th e percentage of all st roke cases that were hospitalized might explain the variations in incidence among the areas. If a high percentage of all stroke cases were admitted to the hospital in the high areas, the incidence based on only hos pitalized cases would be inflated relative to the other areas. Blacks have a much higher rate than Whites. Investigator s could not determine whether race, sex and geographical differences we re due to one specific stroke type. An epidemiologic study was conducted of geographic differences in stroke mortality between areas (high, intermediate a nd low stroke rate areas) within the United States.71 Population samples of 35-54 years of age were drawn for interview and medical examination. Population samples were compared with emphasis on possible risk factors for stroke: serum chol esterol and glucose tolerance test determinations, weight and height measurements, blood pressure and cigarette smoking. The study did not explain the geographic variations in stroke mortality among the high, low and intermediate areas of the United States. Bl ack females showed the expected stepwise progressive increase in severe hypertension from the low to th e high stroke areas. White males also showed this pattern, however the differences were not as great. No other consistent pattern of increasing prevalen ce risk factors for stroke was evident. Various studies have reported consider able geographic variation of stroke

PAGE 38

25 mortality rates in the United States. The qua lity of diagnoses on de ath certificates is questionable, however, certification studies have suggested that large differences in stroke mortality between high and low rate ar eas are real and reflect differences in the same direction in incidence and possibly case fatality.72 At the level of state economic areas (SEA s) changes in the geographic distribution of stroke during 1962 –1982 (period of declin e) for White men and White women, aged 35-74 are presented.71 A cluster of SEA rates in the highest decile is observed in the Southeast (Mississippi, Alabama, Georgia, No rth and South Carolina). Most SEAs in the highest deciles were in the South. Lowest ra tes occurred in the western half of the US, particularly in the Plains and Rocky Mountain states. There were SEAs for which stroke mortality rates either increased or did not change. Patterns of higher stroke mortality rates in the eastern US and lower ra tes in the western US were observed. Socioeconomic status and living conditions have improved in the United States during the period of the decline. SES is nega tively associated with the prevalence of hypertension and with stroke mortality. Th ese associations are consistent with the concentration of high stroke rates in the South, an area economically underdeveloped in relation to the rest of the nation. This region known as the “Stroke belt” became less concentrated over the 2 decades (1962-1982).71 Stroke mortality rates, from 1970-2000, for White Floridians tend to be lower than rates for Black Floridians and for the nation as a whole.72,73 Contrastingly, not only are stroke mortality rates for Blacks slightly higher than national stroke mortality rates, the stroke mortality rates for Black Floridia ns is 1.5 to 1.9 times higher than rates for White Floridians. Florida stroke mortality ra tes, by race, compared to US rates can be

PAGE 39

26 seen in Table2.2 below. Table 2.2. Trends in Stroke Death Rate s per 100,000 Population (all ages), Florida and US 1970 1980 1990 2000 Florida US FloridaUS FloridaUS Florida US All Races 138.0 147.7 87.3 96.2 53.5 65.3 48.6 60.9 White 131.4 143.5 83.0 93.4 50.2 62.9 46.1 58.8 Black 199.3 197.1 142.5 129.3 97.5 91.7 81.6 81.9 Black-White Disparities in Mortality The issue of Black-White disparities in health and mortality is an established concern within the United States. These disp arities are consistent across many different health outcomes. African Americans die disp roportionately because of higher rates of infant mortality, cancer, substance abuse, asthma, heart disease, diabetes, AIDS, and homicide.72 Experts continue to deba te the origin of these di sparities with a decisive focus on socioeconomic influences. Given that Black Americans experience an excess burden of the majority of the adverse health outcomes, it is appropriate to begin addressing the issue of disparities with a thorough examination of the health of the US Black population. A multitude of health and mortality outcomes depict the disadvant ages experienced by the African American population. Infant mortality rates are often used as gauges of the quality of life of populations. In 1998, infants born to African Am erican women have more than twice the rate of death as infants born to non-Hispanic White women.75 Black Americans are sicker and die younger than Whites.76 The status of Black health is in decline as evidenced by several indicators. Blacks experience poorer nutrition, more untreated mental illness, more environmental exposure to toxins, and lack of quality health care for

PAGE 40

27 the elderly population.72 Black women aged forty-five to sixty-four are ten times more likely than White women of the same age to die of diseases of the heart and are five times more likely to die of diabetes Black women are three to six times more likely than White women to die from complications of pregnancy. African American men in every age group up to age sixty-five and over expe rience higher mortality rates than that of White males.77 Black Americans have higher age-adjusted rates than Whites for the majority of the leading causes of death. In the instance of diseases of the h eart, Black Americans have a higher age-adjusted rate (308.4 per 100,000) than White Americans (236.7 per 100,000) (Health US, 2004).78 Similar disparities are al so present for cerebrovascular disease death rates. For Black males (85.4 per 100,000), the age-adjusted death rate for cerebrovascular diseases is about 1.5 times that of White males (54.2 per 100,000), and the death rate for Black females (73.7 per 100,000) exceeds that of White females (54.5 per 100,000) by a similar extent (1.4 times).79 Black men experience a shorter life expectancy than do any other racial or ethn ic minority subgroup (National Vital Statistics Report). At birth, there is a difference of 5.2 years in life expectancy between Black and White Americans (both sexes). There are also extensive differences between Black and White Americans across various health indicators. Black adults 20 year s of age and older are more likely to suffer from hypertension (40 percent) th an White adults (28 percent).75 Black females are much more likely to be overweight (77.7 percent) or obese (50.4 percent) than White females (57.2 percent overweight, and 30.4 percent obese). Compared to 39.1 percent of White females, only 22 percent of Black female s achieved a healthy weight. In 1998, the

PAGE 41

28 primary and secondary syphilis case rate for Black Non-Hispanics (16.9 per 100,000) was 34 times the rate for White Non-Hispanics.75 While many studies have reported significant improvements in mortality and overall health for both Black and White Americans, the Black-White disparity pe rsists with no clear sign of convergence. Black-White Disparities in Stroke Mortality As early as the 1960’s, investigators began to report geographic differences in the distribution of stroke deaths.80 Regionally, areas in the Southeastern United States were found to experience higher str oke death rates with lower ra tes occurring in the Great Plains and Rocky Mountain areas. Later obs ervations demonstrated that not only do these geographic differences exist, however there are concomitant variations in the distribution of stroke de aths among racial groups.81 In the United States, Blacks were found to experience higher death rates than Whites, a phenomenon even more pronounced in the younger age groups. Large racial disparities in health status and health care exist between majority Whites and minority racial/ethni c groups in the United States. Data representing US cerebrovascular disease death rates, ag e-adjusted using the year 2000 standard population, demonstrate that there has been a significant decline in stroke death rates since the 1960’s.82 Generally, White females experi ence the most “favorable” stroke mortality rates, with Black males experienci ng the worst rates. For example, 1990 stroke death rates for White females and Black males were 60.3 and 102.2 per 100,000 resident population, respectively. Although these rate s declined in 2000 for each sex-race category, these racial disparities in stroke mort ality persist. Strides have been made in the effort to account for some of the Black -White disparities in health outcomes.

PAGE 42

29 Research examining the association between so cial class and premature stroke mortality demonstrated excess mortality among Blacks comp ared to Whites at every level of social class.11 Black-White stroke mortality ratios ranged from 3.9 to 4.9 for social class categories demonstrating that social class (as defined by occupation) accounts for some, but not all, of Black-White disp arities in stroke mortality. African Americans are disproportionately affected by high blood pressure and related morbidity and mortality.83 In the United States, the prevalence of hypertension increases with age, and is greater for Af rican Americans (32.4 %) than non-Blacks (23.3%) and Mexican Americans (22.6%).84 The complications of uncontrolled high blood pressure, including cerebr ovascular accident, are up to four times more prevalent among African Americans than among Whites and there are increases at any given level of blood pressure.85 Approximately 20% to 30% of deaths among African Americans is directly attributable to hypertension. Blacks develop high blood pressure at an earlier age and have more severe cases of hypertension than Whites.84 In addition, Blacks have a 1.3-fold greater rate of nonfatal stroke, a 1.8-fold greater rate of fa tal stroke, a 1.5-fold greater rate of heart disease deaths, and a fivefold greater rate of end-stage renal disease,86 each for which hypertension is a serious risk factor. Comp ared to the general public, African Americans have 80% higher rate of stroke mortality, 50% higher rate of heart disease mortality and 320% greater rate of hypertensionrelated end-stage renal disease. The elimination of these disparities w ill require a composite of strategies including enhanced efforts at preventing dise ase, promoting overall health, and delivering appropriate care. When many variables, incl uding income, are held constant, a difference

PAGE 43

30 between Black and White health status still surfaces.72 The evidence that race correlates with persistent health dispar ities among different populations in the United States rightly demands the attention of the policymakers and local, state, and nati onal health and human service heads.72 Socioeconomic Status and Health Influences of individual characteris tics and neighborhood economic structure on health has been the focus of several studi es over the past two decades. Education, income or occupation (or a combination of tw o or more of these measures) is typically used as a measure of individual social class or socioeconomic status, while an area-based socioeconomic indicator (composed of various area/neighborhood level economic and social measures obtained from census data) often represents the economic structure. Findings from this type of research fre quently support the hypot hesis that living in economically deprived areas and being a member of a lower SES group are both associated with increased prevalen ce of negative health outcomes. Athersclerosis Risk in Communities study (ARIC) data and 1990 US Census data were examined for relatedness of neighborhood (census block group) characteristics to coronary heart disease prevalence and to the di stribution of three major CHD risk factors: blood cholesterol, smoking, and systolic blood pressure.87 Results supported a relationship between living in deprived neighborhoods and increased CHD prevalence and increased levels of risk factors with resu lts persisting after adju stment for individuallevel indicators of social class (i ncome, education and occupation). The relationship between neighborhood char acteristics and mort ality (all-cause, CVD, and cancer) was investigated for African American and white participants aged 45-

PAGE 44

31 64.88 The ageand gender-adj usted mortality rate was highe st among those who lived in disadvantaged neighborhoods and who were of lower SES. All cause and CVD mortality rates decreased with increas ing neighborhood SES advantage and family income in all race-gender groups. Although the pattern ge nerally persisted after adjustment for individual socioeconomic factor s, statistically significant as sociations persisted for CVD mortality in whites only. The lack of signifi cant statistical associat ion after adjustment for individual socioeconomic factors for black participants may be due to their insufficient representation w ithin higher SES neighborhoods. A prospective study of the associations of individual occupational social class and area-based socioeconomic indicators with mo rtality revealed that both all cause and cardiovascular mortality rates showed an inverse relationship with socioeconomic position (both at the indivi dual and area based level).89 Additionally, less favorable socioeconomic position, both individually assigned and area based, were associated with cardiovascular risk disease factors including shorter hei ght, worse lung function, and higher prevalence of bronchitis and coronary heart disease. Intera ction between social class and deprivation score were not stat istically significant; however, social class differences in all cause and CVD mortality were slightly attenuated but remained substantial and statistically significant after adjustment fo r area deprivation score. Additionally, all cause and CVD mortality retained sizeable and significant associations with area deprivation after adju stment for social class. After controlling for personal income, e ducation, and occupation, a prospective study found that living in a disadvantaged nei ghborhood is associated with an increased incidence of coronary heart disease.22 Hazard ratios for coronary heart disease among

PAGE 45

32 low-income persons living in the most disa dvantaged neighborhoods, as compared with high-income persons in the most advant aged neighborhoods, were 3.1 among Whites and 2.5 among Blacks. Additionally, these associations remained unchanged after adjustment for established risk factors for coronary heart disease. An inverse association for all cause mortality with both individual and area level indicators of socioeconomic status was f ound for an American Cancer Society cohort.90 When both variables were included simultane ously in the analysis, the effect of individual level SES remained, while area level effects were somewhat diminished. Neighborhood affluence, as measured by the percentage of neighborhood residents with a household annual income of $50,000 and over, was shown to be positively correlated with the self-rated health of adult residents of the metropolitan Chicago area.91 The positive health effect of neighborhood affluence continued even after controlling for individua l-level socioeconomic (incom e, education), demographic and health-related background factors. A cross-sectional study of women from Br itish electoral wards found that adverse area-level socioeconomic characteristic s, over and above individual life-course socioeconomic position (SEP), are associated with increased coronary heart disease.92 After adjustment for age and 10 indicators of individual life-course SEP, the odds of coronary heart disease was 27% greater among those living in wards with a deprivation score above the median compared with those li ving in a ward with a deprivation score. Additionally, the size of the association between neighborhood unemployment rates (as a measure of deprivation) and all cause morta lity from samples across six countries (US, Netherlands, England, Finland, It aly, Spain) demonstrated that living in more deprived

PAGE 46

33 neighborhoods is associated with increased a ll cause mortality independent of individual socioeconomic characteristics.93 Socioeconomic Status and Stroke Stroke incidence, survival and mortality and their relation to individual and area socioeconomic status has been investigate d. A prospective study in the United Kingdom found an association of higher area deprivation with stroke at younger age, severe stroke, higher baseline systolic blood pressure, and with higher rate s of stroke risk factors.94 Stroke mortality was associated with area de privation after correc tion for age, sex and stroke risk factors. A cohort study in th e Netherlands demonstrated a statistically significant association between area socioeconomic status (p ostcode areas) and stroke incidence.16 Residents of postcode areas with below average socioeconomic status experienced a significantly higher incidence of stroke than residents of postcode areas with average or above averag e socioeconomic status. Scot tish hospital patients from the most socially deprived areas were shown to be significantly more likely to be dead or dependent 6 months after admission for an acute stroke.95 No adjustments were made for individual level socioeconomic measures Similarly, follow-up studies of stroke patients found an association between survival and individual level socioeconomic status (occupation, occupational status and income)96 and between risk of fatal stroke and having less than 12 year of education.97 The less educated, th e lower level employees, the unemployed, and the lower income gr oups, experienced higher risk of death compared to their counterparts. It is well established that the socioec onomic position of individuals, groups and places is a defining characteristic of their level of health and disease.98 Scientific

PAGE 47

34 understanding of the processe s through which neighborhood SE S influences lifestyle, health promoting opportunities, morbidity and mo rtality is essential. The goal of future research in this area must be to advance our understanding of these effects and their policy implications. Theories of Causation A multitude of theories have been proposed to address the issue of how specific aspects of society, and the pe ople who live within, work toge ther to influence population health. When researchers began to realiz e the importance of th e environment in the promotion of health and illness, many took on th e task of identifying specific quantifiable aspects of society that could be scientifically related to health outcomes. Although major steps forward have been made in reducing mo rbidity and mortality, the social class and racial divide continues to widen. The conn ection between individual level behaviors with conditions at the societal level, results in complexities which have led to the lack of more substantial improvements in health.99 Social epidemiology holds that we embody or incorporate biologically the world around us. It attempts to answer the question of who and what it is th at is responsible for population patterns of health, disease and well-be ing, as manifested in present, past and changing social inequalities in health. The three main theo ries of social epidemiology (psychosocial theory, social production of di sease and ecological th eory) are described below. The psychosocial theory is based on th e host-agent-environment relationship. The psychosocial framework directs atten tion to endogenous biological responses to stress and on stressed people in need of ps ychosocial resources. Researchers following

PAGE 48

35 this framework believe that in order to expl ain disease distribution we must investigate factors influencing susceptibility.100 The belief is that the soci al environment alters host susceptibility by affecting neuroendocrine function. Relevant psychosocial factors include social disorganization, rapid social change, bereavemen t, social support (which is believed to be a buffer to all of the above). Because of the believe d ability of social support to buffer the effects of the psychosocia l factors, the most f easible and promising interventions to reduce disease will be to impr ove or strengthen the so cial supports rather than reduce the exposure to stressors. This th eory dedicates no atte ntion to: (1) who or what generates psychosocial insults and buffers to these insults, (2) how their distribution is shaped by social, political and economic policies or (3) time. The social production of disease theory, with its Marxist origin, is also known as the political economy of health.76,101 This theory is an advocac y of materialis t analysis of health. These materialist analyses address economic and political determinants of health and disease including structural barriers to people living healthy lives. At issue are priorities of capital accumulation and their en forcement by the state so that few can stay rich while the many are poor. In this theore tical framework, determinants of health are analyzed in relation to who benefits from sp ecific policies and practi ces, at whose cost. The theory posits that economic and politic al institutions and decisions that create, enforce and perpetuate economic and social privileges and inequa lity are root or fundamental causes of social in equalities in health. The theory attempts to determine the health impacts of rising income inequali ty, and the experience of economic and noneconomic forms of racial discrimination. The call for action is for healthy public policies, especially redistribu tive policies to reduce poverty and income inequality. This

PAGE 49

36 includes an attempt to change unjust soci al and economic policies and norms and to provide systematic framework for delineati ng government accountability to promote and protect health. The ecological theory pres ents an analysis of cu rrent and changing population patterns of disease, health and well being in relation to each le vel of biological, ecological and social organization as manifested at each and every scale.101,102The theory embraces a social production of disease perspective while aiming to bring in a comparably rich biological and ecological anal ysis. It elucidates population patterns of health, disease and well being as biological expressions of social relations and proposes multilevel pathways linking expressions of stre ssors, for example (racial discrimination) and their biological consequences across the lifecourse. The theory embodies biological expressions of racism and emphasizes acc ountability. It extends beyond psychosocial explanations focused on anger and hostility to the social phenomena (interpersonal and institutional discrimination) elic iting these responses, as mediated by material pathways. There is an interplay between exposure, su sceptibility and resi stance and it advances beyond social production of diseas e analyses typically focused on racial/ethnic disparities in socioeconomic position to highlight disc rimination within class strata plus ongoing biological impact of economic de privation in early life. The social determination movement, wh ich embodies the so cial production of disease and the ecosocial theory, studies the in equality of health within a nation or among nations.103 It sees steep gradients of education, income, and so cial position as adversely affecting the health of a popul ation, not only at the bottom but throughout the entire range of the social structure. This theory holds th at inequality rather th an absolute deprivation

PAGE 50

37 in developed economies undermines the capacity of people to resist disease. The psychological concomitants of steep social positioning are emphasized both at the individual (self-esteem, hopefulness) and coll ective (community efficacy, social capital) levels. The task is to integrate the insight s of these most perspective efforts and to confront health, society, and habitat as a whole, in their full complexity. Stallones (1973) indicated a need for a broader view of the disease processes when attempting to address the issue of causation.104 The burden of disease on human populations is seen as part of an environmen tal system. The disease process is depicted as an interaction of biologic, social and phys ical factors. Stall ones proposed that the interrelatedness of the components of the system cannot be understood by pursuing research whose rationale is to divide and is olate the components in even greater detail. The belief was that disease is embedded in the environment of man, and that diseases of a society characterize the environment. It was suggested that physical environmental characteristics, demographic and social charac teristics, and disease (total mortality and morbidity) need to be brought together in order to obt ain a deeper understanding of disease as a community phenomenon. Cassel (1976) proposed the ‘s ocial environment’ as envi ronmental factors capable of changing human resistance to disease and of making subsets of people more susceptible to ubiquitous agents in the environment.100 Pyschosocial processes were presented as agents capable of altering the endocrine balance in the body, increasing susceptibility to disease. Stress, defined as either a dynamic state within the individual or as a stimulus assault (any aspect of the environment), was presented as one of these psychosocial factors. It is believed that psychosocial fa ctors should be regarded as

PAGE 51

38 predisposing to disease and not causal. The st ressor could be unfamiliarity with cues and expectations of society (immigration) leading to higher susceptibility (under conditions of social disorganization). Aut hors offered social support as a potential protective factor, buffering the individual from the physiologic or psychological consequences of exposure to the stressor situat ion. Empirical evidence suggests that Black males living in high stress areas have higher blood pr essures. Results were not th e same for Whites. Authors suggest that these results ma y reflect a subservient role of Blacks. The lack of association between high stress areas of Wh ites and blood pressure levels was explained as Whites potentially having more resources in the face of social disorganization. Additional examples of psychos ocial risks include racism, low income, physical abuse, and psychological abuse. Crawford (1977) addressed the vict im blaming ideology which emphasizes individual responsibility for health.105 Crawford asserted that th is ideology serves to reorder expectations and to jus tify a retreat from the language of rights and the policies of entitlement (to medical care). The co mmon theme of the victim blaming ideology emphasizes the need to reduce expectations and utilization of ineffective and costly medical services and instead to increase the necessity for individual responsibility. This ideology instructs people to be individually re sponsible at a time wh en they are becoming less capable as individuals of controlling their health environment. Crawford believed that this blaming ideology obscures the class structure of work, removing the focus away from influence of place in soci ety on health and well being. Stallones (1980) expressed the need fo r the development of theories of causation.106 Development of epidemiologic theory would involve the arrangement of

PAGE 52

39 facts into an orderly ch ain of inference. It was believed that this epidemiologic theory, which is likely to be unique, would guide th e collection of data and the organization of information. During this time period, eco logically based epidemiology was commonly utilized to characterize communities si multaneously by both physical and social circumstances. Individual traits were meas ured and related to the overall morbidity and mortality patterns of the communities. Instea d, Stallones proposed that the community be viewed as a social organization. The di stribution of disease in communities would therefore be considered a so cial phenomenon, and as such, mi ght be expected to have social causes. It was believed that physical and social environmental factors affect the specific etiological agents, along with the likelihood of exposure and the degree of susceptibility of the exposed persons. Why Blacks Have Higher Stroke Mortalit y: Influence of Social Environment Why are Black Americans more vulnerable to adverse health than White Americans who reside in the same area? This study theorizes that in creased vulnerability to adverse health among Black American s is differentially mediated by various environmental factors and conditions. These e nvironmental factors, measured in terms of resource availability for purpos es of this study, in turn in fluence individual lifestyle choices that may be detrimental to health. However, the availability of resources may not be truly representative of the degree of acce ss to these resources. Compromised access will result in underutilization of “available” utilization. This is a circumstance witnessed more often in the Black population. This underutilization of resources by Black Americans may be viewed in terms of unhealth y “lifestyle choices” when in reality the choices may have been extremely limited. This process possibly culminates in Black-

PAGE 53

40 White health disparities within geographi c areas where Black and White residents supposedly share resources. To investigate this problem, this study u tilizes a theoretical focus on social and economic characteristics at the census tract leve l and the influence of these characteristics beyond those typically observed at the indivi dual level. The environments in which people live their lives afford them a certain amount of opportuniti es for utilization of available resources. Explanations for illness and mortality are typically limited to the individual behaviors of Blacks and few studies address the social context in which these behaviors occur. This study aims to direct focus on social influences of Black-White health disparities, stroke mortality in particular. This study is carried out in a psychosocia l context, conceding that the study does not directly measure the influence of speci fic psychosocial risk factors on Black-White disparities in stroke mortality. This st udy builds on the perception that psychosocial influences, specifically racism, directly a nd indirectly influence access to community resources. The study hypotheses contend that research on Black health (and the BlackWhite disparities that result) should conceptu alize Black health as a complex interaction of psychosocial risks which influence access th at have a profound effect on that health. Laws based on racist ideals created these si tuations in which Blacks are more likely to experience environmental influences disadvantageous to health. Conclusion The fundamental attribute differentiati ng social class categories relates to differences in the power to access material res ources. Material factors therefore seem an obvious candidate for considera tion as an explanation of social health inequalities.107

PAGE 54

41 Because of the unequal distribution of Black and White Americans within these social class groups, the influence of race on access to material resources should be assessed. While Black Americans have made health st atus improvements, a multitude of racial disparities still exist. A major reason is that, although legislation may have made health care relatively more available and affordable the fundamental and unequal structure of American society, which is primarily responsible for racial disparitie s in health, remains unchanged. The social environmental condi tions in which a large portion of Black Americans live may not be conducive to e nduring implementation of lifestyles which promote cardiovascular healt h. The close connections between the socioeconomic conditions people live under and their adaptiv e behavioral profile must be acknowledged.

PAGE 55

42 Chapter Three Methods Study Design The study type is of a mixed design. It is fundamentally an ecological design, but this study does have elements of a retros pective cohort study de sign in that the study participants’ community resource availability is utilized as a predictor of racial disparities in stroke mortality. However, given th at the study outcome time period is 1998-2002, while the community resource in dicator, or “predictor,” wa s taken from the 2000 Census of Population and Housing, establishing that the exposure preceded the outcome cannot be achieved. A 5-year study period was chosen to increase the amount of data that would be available for small area analysis and the year 2000 chosen as the midpoint of the study period in order to use the population and so cioeconomic data from the 2000 Census of Population and Housing. Study Population The geographic study area is the State of Florida. Ou r study population consisted of Non-Hispanic White and African American (both Hispanic and non-Hispanic) adults aged 35 and older who resided in the State of Florida during the years 1998-2002. As of the 2000 census, Florida had a total population of almost 16 million. White Americans made up seventy-eight percent of the populat ion whereas African Americans make up 14.6 percent. Almost 17 percent of Florida re sidents were reported being of Hispanic

PAGE 56

43 ethnicity. Study subjects incl uded 43,945 Florida residents aged 35 and older, of which 13,605 were aged 35-74 years, who died from stroke during the 1998-2002 time period. Stroke incidence and stroke mortality is less common in those younger than age 35, while stroke incidence and mortality rates are cons iderably elevated for those beyond the age of 75. To lessen the effect of these extreme rate s, very low and very high, on study results, only those decedents aged 35-74 were included in the study. Data Sources The level of analysis utilized in this study is the census tract. Census tracts are small, relatively permanent statistical subdivisi ons of a county or statistically equivalent entity.108 A primary purpose of census tracts is to provide a stable set of geographic units for the presentation of decennial census data For Census 2000, the entire United States was covered by approximately 65,000 census tracts, while the State of Florida consisted of 3154 census tracts. The specific number of census tracts utilized for study analyses within each research question was depende nt upon the specific study outcome. The number of census tracts utilize d, in addition to a discussion of the loss of census tracts, is presented along with the appr opriate analyses results. Stroke Mortality Data Stroke mortality data was obtained from the Florida Department of Health. The Florida Department of Health provided a data set containing information on all 19982002 decedents in the study population for whom the underlying cause of death was coded as stroke. Information was not obtaine d for decedents who died from causes other than stroke; therefore, the f iles only contained Florida str oke decedents. Data on age, gender, race, Hispanic ethnicity, educatio nal attainment, and cen sus tract of usual

PAGE 57

44 residence were the only data included in the files obtained. Each stroke death was point located (geocoded) within its proper census tract by the Fl orida Department of Health (see Appendix A). Stroke deaths were id entified as those with the International Classification of Diseases (ICD)-9 (f or 1998 decedents) a nd ICD-10 (for 1999-2002 decedents) codes indicating ‘cerebrovascular di seases’ as the category for cause of death. Specific coding used for the death certif icates is presented in Table 3.1 A total of 10,799 stroke decedents, ages 35-74 years, for the study area and time period were included in the study. Data for census tract of reside nce, age, gender and cause of death were available for 100% of the stroke decedents included in the study. Race data was available for 99.96% of the deced ents, while Hispanic ethnicity data was Table 3.1. International Classification of Diseases (ICD) Codes for Cerebrovascular Diseases ICD 9 (430-438) 430 Subarachnoid hemorrhage 431 Intracerebral hemorrhage 432 Other and unspecified intracranial hemorrhage 433 Occlusion and stenosis of precerebral arteries 434 Occlusion of cerebral arteries 435 Transient cerebral ischemia 436 Acute but ill-defined cerebrovascular disease 437 Other and ill-defined cerebrovascular disease 438 Late effects of cerebrovascular ICD 10 (I60I69) I60 Subarachnoid haemorrhage I61 Intracerebral haemorrhage I62 Other nontraumatic intracranial haemorrhage I63 Cerebral infarction I64 Stroke, not specified as haemorrhage or infarction I65 Occlusion and stenosis of precerebral ar teries, not resulting in cerebral infarct I66 Occlusion and stenosis of cerebral arte ries, not resulting in cerebral infarction I67 Other cerebrovascular diseases I68 Cerebrovascular disorders in diseases classified elsewhere I69 Sequelae of cereb rovascular disease

PAGE 58

45 available for 99.72% of the decedents. For e ducational attainment, 96.57% of the stroke decedents had available data. Population Data Population counts were obtained for th e year 2000 US census from the United States Census website. The 2000 census year served as the midpoint for the five-year study period. Summary Files 3 (SF3) and 4 ( SF4) data were obtaine d at the census tract level for population count purposes. These counts were multiplied by 5 as death data were available for a five-year time period, 1998-2002. This methodology is utilized since it is the most accurate method to estimate the total population since the population typically increases each year. Utilizi ng this method, possibly overestimated the population for the years of 1998 and 1999. Howeve r, the population is more than likely underestimated for the years of 2001 and 2002. This provides for the calculation of average annual age-adjusted rates. Area Social Predictors of Health (ASPoH) Data Census tract data on population and soci oeconomic characteristics were obtained from the 2000 Census of Population and Housi ng Summary File 3 (SF3 ). Summary File 3 contains information compiled from the quest ions asked of a sample of all people and housing units. Specific information in the SF3 population files includ es: population total, urban or rural denotation, households and fa mily types, marital status, educational attainment, poverty status and many other fact ors. Summary File 3 contains a total of 813 unique tables, a subset of which is repeated by race and Hispanic or Latino ethnicity. Local area indicators of the ASPoH variable we re calculated from these summary files.

PAGE 59

46 Statistical Products SAS computer statistical package, Versi on 9.1 (SAS Institute Inc, Cary, NC) was used in the analyses of study data. Analytic Methods Race and Sex Specific Mortality Rates Race and sex specific age-adjusted stroke mo rtality rates were calculated utilizing the direct method of standardization. An ageadjusted rate is a we ighted average of the age-specific rates. The sta ndard population distribution used to adjust the stroke death rates was the 2000 US Standard Population (3574 years of age for Research Question 1 and 35 years of age and up for Research Questio ns 2 and 3). The stroke death rate for each census tract was calculated for each race-sex specific age group. The number of stroke deaths in that race-sex specific age-group was divided by the population of the same race-sex specific age-group within that census tract. The race-sex age-groupspecific stroke death rate was then mu ltiplied by the proportion of the standard population (see Table 3.2) for that specific ag e-group. The weighted age-specific rates are then summed for the census tr act to calculate the age-adjust ed rate. Direct adjustment reduces the potential for confounding by age; th erefore, comparison of death rates across racial group with different age di stributions is possible. Table 3.2. Year 2000 Standard population weights Age Group 2000 Proportion 35-44 years 0.363877 45-54 years 0.297873 55-64 years 0.191597 65-74 years 0.146652

PAGE 60

47 Black-White Disparity Scores Black-White disparity scores were calculated both on an absolute scale (difference score) and on a relative scale (ratio and percent difference scores). The absolute measure of disparity is expressed simply as the arithmetic difference between the Black stroke death rate and the Non-Hi spanic White stroke death rate (reference point). The difference score provides inform ation on the excess number of stroke deaths among Black Floridians. The ratio score is in terpreted as the relative magnitude of the Black stroke death rate compared to the NonHispanic White stroke death rate (expressed as a quotient). The ratio score is an index of how serious the stroke mortality risk is for Blacks relative to Non-Hispanic Whites. The percentage difference score is expressed as the difference between rates (Black minus N on-Hispanic White) as a percentage of the Non-Hispanic White death rate. Absolute a nd relative measures of disparity calculated from the same reference point (Non-Hispan ic White rate) should lead to the same conclusion about stroke mortal ity disparities between these groups. Utilization of both absolute and relative measures allows for a check of consistency between the disparity measures.109 Methods of calculation of the dispar ity scores are presented in Table 3.3. Table 3.3. Disparity score calculation methods Black-White Disparity Measure Formula Male Ratio Score BMAAdeathrate1 NHWMAAdeathrate2 Female Ratio Score BFAAdeathrate3 NHWFAAdeathrate4 Male Difference Score BMAAdeathrate NHWMAAdeathrate Female Difference Score BFAAdeathrate NHWFAAdeathrate Male Percent Difference Score (BMAAdeathrate NHWMAAdeathrate) NHWMAAdeathrate Female Percent Difference Score (BMAAdeathrate NHWMAAdeathrate) NHWMAAdeathrate 1 Black Male Age-Adjusted stroke death rate, 2 NHWhite Male Age-Adjusted stroke death rate 3 Black Female Age-Adjust ed stroke death rate, 4 NH-White Female Age-Adjusted stroke death rate

PAGE 61

48 Statistical Methodology This study attempted to create a vari able that would pr operly reflect the contextual characteristics of an area possibly affecting Black-W hite disparities in stroke mortality. Area Social Predictor of Health (ASPoH) describes features of social organization, structure, and stratification of the environm ent, such as socioeconomic deprivation, economic inequali ty, resource availability, a nd opportunity structure. Specifically, this study attempted to compile a set of indicators that would closely reflect both the study residents’ economic resource av ailability and their pr obability of obtaining these resources. In order to construct this area socioeconomic status measure, several indicators were statistically transformed into a smaller number of variables known as principal components. The decision to include specific meas ures was based on a core set of 12 dimensions of social determinants of health. This core set grew out of a University of Michigan School of Public H ealth project funded by the Cent ers for Disease Control and Prevention.110 Investigators representi ng a wide range of disciplines participated in a workshop to review dimensions important in und erstanding social determinants of health. Participants were able to arrive at a consen sus on a core set of 12 dimensions (4 of which are assessed in this study). Th e directory contains an extensive list of available types of data sets. Workshop participants generate d suggestions for possible data sources and specific variables that might be used to measure the components of each dimension. Researchers may choose to utilize certain elements from this list in order to evaluate how the social environment impacts the health of populations. The data sets are organized according to the 12 dimensions specified in Table 3.4.

PAGE 62

49 Table 3.4. Twelve core dimensions to understanding social determinants of health. 1 Economy 7 Medical 2 Employment 8 Governmental 3 Education 9 Public Health 4 Political 10Psychosocial 5 Environmental 11Behavioral 6 Housing 12Economy For this current project, data were available for measures representi ng four of the twelve core dimensions. The census tract level meas ures used in the construction of the ASPoH variables for this study are as follows: Economy Dimension 1. Poverty Rate 2. Median Income Employment Dimension 3. Percent Unemployed 4. Percent of workers aged 16 years or ol der using private transportation to work 5. Full vs. part-time employment Education Dimension 6. High School Graduation rates for those 25 years of age and older Housing Dimension 7. Median Rent 8. Median value of owner occupied housing units 9. Vacancy rates 10. Home Ownership

PAGE 63

50 11. Overcrowded Housing The methods of calculation for these 11 va riables is reported in Appendix C. Principal Component Analysis Methodology Principal Component Analysis (PCA) i nvolves a mathematical procedure that transforms a number of po ssibly correlated variables into a smaller number of uncorrelated variables called principal co mponents. This is accomplished by first identifying patterns in the data, followed by expressing the data in such a way as to highlight their similarities and differences. The first principal com ponent produced by the mathematical procedure accounts for as much of the variability in the data as possible, and each succeeding component accounts for as much of the remaining variability as possible.111 Direct uses of Principal Component Analysis include: (1) identification of new “meaningful” underlying variables and (2 ) reduction of numb er of variables.112 To obtain reliable results, the minimal number of subjects providing us able data for the analysis should be the larger of 100 census tracts or five times the number of variables being analyzed. In these analyses, the minimu m sample size requirement was met with 3154 census tracts contributing usable data. Generated components are thought to be representative of the underlying processes that have created th e correlations among variables.112 Variables that are correlated with one another which are also largely independent of other subsets of variables are combined into components. Comp onents may either be associated with 2 or more of the original variables (common factors) or associated with an individual variable (unique factors).

PAGE 64

51 The number of components extracted in a principal component analysis cannot exceed the number of observed variables being analyzed. The principal component is a linear combination of optimally weighted observed variables. The first component extracted accounts for a maximal amount of tota l variance in the obser ved variables. The second component extracted accounts for a ma ximal amount of variance in the dataset that was not accounted for by the first compone nt, and it will be uncorrelated (r=0) with the first component. Each remaining component accounts for a maximal amount of variance in the observed variables that was not accounted for by the preceding components and is also uncorrelated with al l the preceding components. The resulting components (all extracted components) will display varying degrees of correlation with the observed variables, but are complete ly uncorrelated w ith one another. Loadings relate the specific association between factors and or iginal variables. Therefore, it is necessary to find the loadi ngs, then solve for the factors, which will approximate the relationship be tween the original variables and underlying factors. The loadings are derived from the magnitude of eigenvalues associ ated to individual variables.112 Steps in Conducting Principal Component Analysis (1) Initial Extraction of the Components The number of components extracted is e qual to the number of variables being analyzed. An eigenvalue table is presente d. The eigenvalue repr esents the amount of variance that is accounted for by a given com ponent. The first components extracted will account for relatively large amounts of varian ce, while the later components account for relatively smaller amounts.

PAGE 65

52 (2) Determining the number of “M eaningful” Components to Retain There are four techniques that may be used to determine the number of principal components that may be retained for further analyses. One criteri on involves retaining any component with an eigenvalue greater th an 1.00. The rationale for utilizing this technique evolves from the fact that each vari able contributes one unit of variance to the total variance in the dataset. Any component that displays an ei genvalue greater than 1.00 is accounting for a greater amount of va riance than had been contributed by one variable. Such a component is therefore acc ounting for a meaningful amount of variance, and is worthy of being retained. A second criterion involves the use of the Scree test. This te st involves plotting the eigenvalues associated with each component and looking for a break between the components with relatively la rge eigenvalues and those with small eigenvalues. The components that appear before the break are assumed to be meaningf ul and are retained for rotation. A third criterion is the in terpretability criteria. This techniques involves interpreting the substantive meaning of the re tained components and verifying that this interpretation makes sense in terms of what is known about the constructs under investigation. There are four rules to follo w in doing this: (1) Are there at least three variables with significant loadings on each re tained component? (2) Do the variables that load on a given component share the same con ceptual meaning? (3) Do the variables that load on different components seem to be m easuring different constructs? (4) Does the rotated factor pattern dem onstrate “simple structure?” The final criterion, takes into account the proportion of variance accounted for by

PAGE 66

53 a specific set of components. This criterion requires that components are retained if the cumulative percent of variance accounted for is equal to some minimal value (70 to 80%). The decision was made to retain four principal components for this study. The retention of these four components satisfied each of the four criterion suggested for determining the number of principal compone nts to retain for inclusion in further analyses. All components with an eigenvalu e greater than one were included in this study (components 1, 2 and 3). The Scree test resulted in a break between principal components 3 and 4. A substantive interpretation for each of the four retained components was accomplished. Finally, the fi rst four principal components accounted for approximately 76% of the variance in the data. Preliminary Analyses Dividing ASPoH Measures into Quartiles There is no gold standard for assessing th e predictability of the ASPoH index; therefore, the association betw een the ASPoH index and Black-White disparities in stroke mortality was examined in two ways. Firs t, ASPoH categories were created. Census tracts were categorized based on group dist ribution of the ASPoH index. Therefore, categorization was as follows: (1) below the 25th percentile, (2) between the 25th and 50th percentile, (3) between the 50th and 75th percentile or (4) above the 75th percentile. The values were assigned to groups in ascending order, with the sm allest value assigned to the first quartile and so on. These methods resu lted in 25% of the census tracts being contained within each category. Therefore, each ASPoH category contains either 788 or 789 (3154 census tracts divide 4 groups) census tract values each. Black-White disparity

PAGE 67

54 scores for stroke mortality were calcul ated for each of the ASPoH categories and compared. Because of the limited range in value of each ASPoH variable, quartiles, instead of quintiles, were used. Each census tract was assigned the best -fitting category of ASPoH based on the homogeneity of the ASPoH indicators. This process allowed for the calculation of ageadjusted stroke death rates (separately fo r Blacks and Whites) for each of the ASPoH categories. SF3 data from the 2000 US Census was utilized for denominator purposes in order to obtain population counts at the cen sus tract level for race*sex*age. The ageadjusted stroke death rates were used to calculate Rela tive Risks, with the “most favorable” ASPoH category as the referent group. Research Question One Analyses Research Question 1 Do lower levels of ASPoH status result in greater black-white di sparities in stroke mortality? The multiple linear regression model was used to test the predictability of BlackWhite disparity in stroke mortality, as we ll as, race-sex specific age adjusted stroke mortality rates (ages 35-74), by the ASPoH m easures (4 principal components) at the census tract levle. ASPoH scores and Black-W hite disparity measures were calculated for each of the individual census tracts. Duri ng these analyses, the census tracts were not categorized into ASPoH quartiles as in th e previous analyses. The strength of predictability was determined at the individual census tract level. The models tested in this phase of the analyses include:

PAGE 68

55 (1) Black Female Age-Adjusted Stroke Mortality Rate = 0 + 1ASPoH1 + 2ASPoH2 + 3ASPoH3 + 4ASPoH4 + Black Male Age-Adjusted Stroke Mortality Rate = 0 + 1ASPoH1 + 2ASPoH2 + 3ASPoH3 + 4ASPoH4 + Non-Hispanic White Female Age-Adjusted Stroke Mortality Rate = 0 + 1ASPoH1 + 2ASPoH2 + 3ASPoH3 + 4ASPoH4 + Non-Hispanic White Male Age-Adjusted Stroke Mortality Rate = 0 + 1ASPoH1 + 2ASPoH2 + 3ASPoH3 + 4ASPoH4 + (5) Male Black-White Disparity Ratio = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + (6) Male Black-White Disparity Difference = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + (7) Male Percent Difference = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + (8) Female Black-White Disparity Ratio = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + (9) Female Black-White Disparity Difference = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + (10) Female Percent Difference = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 +

PAGE 69

56 Research Question Two Analyses Research Question 2 Question: Does low social class result in greater Black-White disparities in stroke mortality? Social class categorization was determined by educational attainment information extracted from the death certifi cates. Five categories of soci al class were defined in the following manner: (1) Social Class 1 (High): College graduates (with degree) and beyond (2) Social Class 2: Some college education and/or Associates Degree (3) Social Class 3: High School graduates/12 years completed (4) Social Class 4: 9-11 years of school completed (5) Social Class 5 (Low): 0-8 years of school completed Population counts were obtained from Summary File 4 data (from the 2000 US Census). Population counts stratified by race, social class, sex, and 10-year age groups (ages 35 and up) were used to calculate stroke death rates for the years 1998-2002 by social class. To be retain ed in the study, the individua l census tract must have a population count of at least one within each race-gender-10yr age group category. A total of 2156 out of the orig inal 3154 census tracts met this criterion. Each stroke decedent was assigned to the appropriate soci al class category and race specific stroke mortality rates calculated for each social class category. Black-White disparity scores were calculated for each social class group at the census tract level. Limitations of the data resulted in the calculation of the disparity scores at the state level only.

PAGE 70

57 Research Question Three Analyses Research Question 3 Question: Is there effect modification by so cial class of the ASPoH and Black-White disparities in stroke mortality relationship? For each social class category, linear re gression analyses were performed to access the association between Black-White disparity in stroke mortality and ASPoH variables (ages 35 and up). Whether the relationship (as measured by parameter estimates) between ASPoH and Black-White disp arity is stroke mortality varied across social class categories was investigated. Separate analyses were conducted for each social class group. For each social class category, the predictability of Black-W hite disparities in stroke mortality by the ASPoH variable was determined. These separate analyses were examined to determine whether the magnitude of the parameter estim ates (accessing the association between the particular disparity score and the ASPoH vari able) varied across different categories of social class.

PAGE 71

58 Chapter Four Results Part I. Area Social Predictors of Health Descriptive Statistics Principal component analysis methodology was utilized to c onstruct the ‘Area Social Predictors of Health’ variables. Su mmary statistics for each of these census tract level variables subjected to principal component analysis are presented in Table 4.1. The median employment rate at the census trac t level is 94.19%. This results in an unemployment rate close to 4.5%, which is ve ry close the national unemployment rate of 5%. Fifty-one percent of those employed residents are full-time employees. The average percent above poverty rate for Florida census tracts was 90% for the 2000 census year. The variability of the data is not unreasonable given that we are dealing with data at a small geographical unit (census tract). On average, 12% of the homes in each census tract were vacant and less than 1 percent of these owner-occupied homes were overcrowded. Seventy percent of these occu pied housing units were owner occupied. Median homes values averaged around $105,000, while renters paid an average of $677 per month. On average, nearly 80% of residents 25 years and older had received their high school diploma. The study average poverty ra te of 10.04% was sligh tly lower than that of the United States in 2001 (12.1%) and slig htly lower than the poverty level for the

PAGE 72

59 state of Florida which is 11.5% (19992000 average). The study median household income of $43322.50 is almost identical to th at of the United States in 2003 which was $43, 381.113 Table 4.1. Summary Statistics: Area Social Predictors of Health Variables, 2000 US Census, 3154 Census Tracts ASPoH Variables Mean Median Std Dev Minimum Maximum Skew Percent Employed 94.1995.465.533.33 100.00-6.45 Percent Above Poverty Rate 89.9692.969.2526.22 100.00-2.08 Percent of Occupied Homes 88.1991.0110.070.00 100.00-2.33 Percent of Noncrowded Homes 99.0799.881.8579.17 100.00-3.54 Percent Using Private Transport 90.6392.747.9317.63 100.00-3.70 Percent 25yr+ with High School Diploma 79.2482.0613.2920.46 100.00-1.01 Percent of population Employed Full-time 51.3353.0011.744.38 100.00-0.73 Median Income 47697.0043322.50203340.00 200001.001.82 Percent Home Ownership 69.7575.4221.040 100.00-1.06 Median Rent 677.78636.00248.750 2001.001.52 Median Home Value 105417.7087050.0072984.000 1000001.004.66 In the correlation matrix (Table 4.2), the major ity of the correlations were in the expected direction given that the variables were calcu lated in such a manner that the higher the score the more positive the area economic situat ion. The strongest stat istically significant correlations were between the ‘median home value’ and ‘median income’ variables and between the ‘percentage of the population 25 years and older who earned a high school diploma’ and the ‘percentage of census tract households above the poverty rate’

PAGE 73

60 Table 4.2. Pearson Correlation Coeffici ents: ASPoH Variables, 2000 US Census % Employed % Above Poverty Rate Occupied Home Rate NonCrowded Rate % Using Private Transport % High School Diploma % Employed Full-time Median Income % Home Ownership Median Rent Median Home Value % Employed 1.0 % Above Poverty Rate 0.547 1.0 Occupied Home Rate -0.036 -.006 1.0 Non-Crowded Rate 0.347 0.521 -0.074 1.0 % Using Private Transport 0.412 0.415 0.265 0.267 1.0 % High School Diploma 0.461 0.737 -.029 0.627 0.225 1.0 % Employed Full-time 0.178 0.118 0.498 -0.020 0.265 0.127 1.0 Median Income 0.415 0.622 -.021 0.339 0.077 0.674 0.119 1.0 % Home Ownership 0.391 0.617 -0.047 0.434 0.445 0.385 -0.147 0.459 1.0 Median Rent 0.296 0.467 0.045 0.169 0.113 0.499 0.151 0.634 0.299 1.0 Median Home Value 0.259 0.359 -.134 0.140 -.174 0.441 0.019 0.823 0.216 0.486 1.0 *= significant at .05 level

PAGE 74

61 variables, at 0.823 and 0.737 respectively. Va riables in which there appeared to be almost no correlation, -0.006, include the ‘c ensus tract occupied home rate’ and the ‘percentage of census tract households above the poverty rate’ variables. Principal Component Analyses Results Three of the principal components had eige nvalues above the value of one (Table 4.3). This tells us that th ese three components account fo r more than one point of variance within the data. Although it is comm on practice to only re tain those variables with an eigenvalue greater than one, othe rs have also chosen to keep as many components as you need to have a cumulative amount of variance in the data accounted for. In these analyses, four principal com ponents are retained. These components account for a total of 76.29% (range of 70-80 typically used) of the variance in the data. Table 4.3. Eigenvalues of the correlation ma trix, Principal Components Analyses Principal Component Eigenvalue Difference Proportion Cumulative 1 4.4566 2.7578 0.4051 0.4051 2 1.6987 0.2600 0.1544 0.5596 3 1.4387 0.6404 0.1308 0.6904 4 0.7982 0.0410 0.0726 0.7629 5 0.7571 0.2519 0.0688 0.8318 6 0.5052 0.0735 0.0459 0.8777 7 0.4317 0.1092 0.0392 0.9169 8 0.3224 0.0267 0.0293 0.9463 9 0.2957 0.0909 0.0269 0.9731 10 0.2047 0.1141 0.0186 0.9918 11 0.0906 0.0082 1.0000 The two variables with the largest factor loadings for each of the components are presented as the description for the principal components (The magnitude of all contributing variables, both positive and negative, can be seen in Table 4.4).

PAGE 75

62 The two variables representing princi pal component 1 are (1) the median household income and (2) the percent of house holds within the census tract that were above the poverty rate. The two variables representing principal component 2 are (1) Percent of occupied homes and (2) the percen t of residents employe d fulltime. The two variables representing principal component 3 are (1) Median home value and percent of home ownership. The two variables represen ting principal component 4 are (1) Percent of census tract residents who are employed a nd (2) the percent of census tract residents 25 years and older who are high school graduates. Table 4.4. Factor Loadings for Principal Co mponents Retained in Further Analyses Census Tract Level Variables Principal Component 1 Principal Component 2 Principal Component 3 Principal Component 4 Pct Employed 0.283 -0.114 -0.105 0.493 Pct Above Poverty Rate 0.402 -0.054 -0.166 -0.036 Occupied Home Rate 0.007 -0.582 0.243 -0.138 Non-Crowded Rate 0.274 -0.004 -0.393 -0.616 Pct Private Transport Use -0.194 0.503 0.357 -0.236 Pct High School Diploma -0.393 -0.092 -0.061 0.409 Pct Employed Full-time -0.093 0.537 -0.386 0.158 Median Income 0.424 0.132 0.223 0.015 Pct Home Ownership -0.308 -0.034 0.384 -0.249 Median Rent 0.319 0.047 0.328 0.199 Median Home Value 0.324 0.272 0.403 0.084 Throughout this dissertation, further definition of the ASPoH-1, ASPoH-2, ASPoH-3, and ASPoH-4 variables can be found in A ppendix B: Definition of Study Variables. Part II. Research Question One Question : Are Black-White disparities in stroke mortality elevated in those areas of low socioeconomic status? Hypothesis : Black-White disparities in stroke mortality will be greatest at lower

PAGE 76

63 levels of ASPoH. Descriptive Statistics Numerator Data: Stroke Death Counts Within the 1998-2002 time period, there was a total of 43,945 stroke deaths for Florida residents aged 35 years and older. These deaths were distributed across 3064 census tracts. There were no stroke deaths reported for the study time period for the remaining ninety census tracts. The distribution of the stroke deaths by race and gender are as follows: Black males 5%, Black females 7%, NH-White males 36%, NH-White Females 52%. The median age of the Black decedents was 62 years, a slightly younger age than that of White decedents, 68 years. Stroke data included in the subsequent anal yses were restricted to those decedents between the age of 35 and 74 years. Thes e remaining 10,799 deaths were distributed across these 3064 census tracts. Twenty-four percent of these decedents are Black Americans (Hispanic and non-Hispanic) and 76 % are White Americans (non-Hispanic). Males constituted the majority of the deced ents with 52.5 %. The effect of excluding deaths in the oldest age groups (75+ years) was to increase the percentage of deaths represented by Black males. This occurred because Blacks, on average, die earlier; therefore, compared to non-Hispanic White decedents, fewer Black decedents were excluded when age restrictions (35-74 years) were utilized. The percentage of female decedents decreased because females in th e oldest age group constituted a large percentage of the original subject pool. When the age restriction is introduced, the percent contribution of females to the stroke death count decreases. As expected, decedents in the oldest ag e group, 65-74 years, made up the greatest

PAGE 77

64 proportion of stroke deaths (Table 4.5). The proportion of stroke de aths decreases with younger age groups. In each of the younger ag e groups (35-44, 45-54 and 55-64), Blacks consistently contributed a higher proportion of stroke deaths than did White decedents. A particularly striking finding was that the proportion of Black males in the 45-54 year age group was almost twice that of White males, at 22.2% and 11.8%, respectively. In the 3544 year age group, the proportion of Black female stroke decedents was more than twice that of White female decedents in th e same age group, 11% and 4.2%, respectively. Table 4.5. Percentage of Stroke Deaths by Race-Sex-Age group, Florida 1998-2002 35-44 years 45-54 years 55-64 years 65-74 years Total Black Males (N= 1326) 7.7 22.2 30.0 40.1 100 (12.3%)* Black Females (N= 1265) 11.0 17.9 26.9 44.2 100 (11.7%)* NH-White Males (N= 4345) 4.8 11.8 22.0 61.4 100 (40.2%)* NH-White Females (N= 3863) 4.2 10.3 20.9 64.5 100 (35.8%)* Total (N= 10,799) 5.7 13.2 23.2 57.9 100 (100%)* = % of total study population Denominator Data: Florida Population Counts The median age for Florida residents was 38.7 years. Median age for White residents was slightly higher than that of Black resident s, 42.0 and 29.0 respectively. A large proportion of these retirees are White, re sulting in a higher median age value for this group. There is an equa l distribution of males and females within the State of Florida. Black residents were less likely than White resident s to have ever been married, 16.8% and 6.9 % respectively. The widow/di vorce rate was very si milar for both race

PAGE 78

65 groups. Black residents were le ss likely than White residents to have an education above the high school graduate level and were more likely to have not completed high school. To be retained in the stu dy for further analysis, the i ndividual census tract must have a population count of at least one within each race-g ender-10yr age group category. This resulted in the following population dist ribution: Black males 7%, Black females 9%, NH-White Males 40%, NH-White females 44% (Table 4.6). These proportions are Table 4.6. 2000 US Census Population counts by Race-Sex-Age group 2156 Census Tracts 35-44 years 45-54 years 55-64 years 65-74 years Total Black Males 155,437 109,076 61,739 38,261 364,513 Black Females 173,975 124,368 72,719 50,280 421,342 NH-White Males 597,962 515,601 377,336351,721 1,842,620 NH-White Females 584,177 526,671 414,975405,504 1,931,327 Total 1,511,5511,275,716926,769845,766 4,559,802 similar to those observed when all of the orig inal census tracts are in cluded in the study. When exclusion / inclusion criteria are applie d in the selection of census tracts for this study, only 2156 of the original 3154 census tracts remain in the study. A total population count of 4,559,802 was distributed acro ss these 2156 census tracts (Table 4.6). These population counts were multiplied by 5 (years) before being utilized as denominators in the calculati on of all study rates. The 35-44 year age group contributed the gr eatest percentage of person years to the study at 33.1% (Table 4.7). Compared to the Non-Hispanic White population, a higher proportion of the Black population made up the younger age groups. These data

PAGE 79

66 reflect the age distribution of Blacks within the State of Florida. In the State of Florida, Black residents are younger than white re sidents with median ages of 29.0 and 42.0 years, respectively. This trend was also seen in the study data. Fo r example, the 35-44 year age group, constituted 42.6% of the Bl ack male study population. In comparison to White males in the same age group at 32.4%, Black males have a younger age distribution than do White males. The same tr end was also seen for females aged 35-44 years. The percentages for Black fema les and White females are 41.3% and 30.2%, respectively. The age distributions are simila r between the race-sex groups within the 4554 and 55-64 year age groups. The percentage of the White residents in the oldest age group, 65-74 years, is almost double the percenta ge of Black residents falling into this age category. Within the State of Florida, White residents tend to live longer than do Black residents. Overall, Black resident s comprised 17.2% of th e total study population, with White residents making up 82.8% of the study population. Table 4.7. 2000 US Census population percen t distribution by race-sex-age group 35-44 years 45-54 years 55-64 years 65-74 years Total Black Males 42.6 29.9 16.9 10.5 100 (8.0%)* Black Females 41.3 29.5 17.3 11.9 100 (9.2%)* NH-White Males 32.4 28.0 20.5 19.1 100 (40.4%)* NH-White Females 30.2 27.3 21.5 21.0 100 (42.4%)* Total 33.1 28.0 20.3 18.5 100 (100%)* = percent of total study populatio n (summed across 2156 census tracts)

PAGE 80

67 Stroke Mortality Rates Race-Sex-10 year Age Group Speci fic Stroke Mortality Rates When the age-group specific stroke death rates by race and gender are examined, the trends are consistent with what is exp ected (Table 4.8). For each of the race-sex groups, the 35-44 year age-group ha s the lowest stroke death rate. The rates increase across successive 10-yr age-groups The highest stroke death rates are observed for those in the 65-74 year age group. Across racial groups, males typically ha ve the higher stroke death rates. Table 4.8. Race-sex 10-year age group specific stroke mortality rates*: Census tract N=2156 35-44 yrs 45-54 yrs 55-64 yrs 65-74 yrs Black Males 8.68 47.26 116.76 273.42 Black Females 12.85 35.47 73.21 211.56 NH-White Males 6.66 20.43 46.52 137.79 NH-White Females 5.43 15.91 33.52 114.71 *Rates per 100,000 In each of the age groups Blacks had higher stroke mortality rates than did Whites. In the 35-44 year age group, Black fe males had the highest stroke death rate at 12.85 per 100,000. This slightly higher rate for Black females, compared to Black males, at the younger age group is in accord with published data. In each of the succeeding 10 year age groups, Black males consistently had the highest rates, while Black females consistently have the second highest rates. White males and females experienced half the stroke death rate of their Bl ack counterparts in each respec tive 10-year age categories. For males, the largest stroke mortality rate difference between Blacks and Whites is seen in the 55-64 year age group. In this age gr oup, the Black male stroke death rate is 2.51 times higher than that of White males. For females, the largest stroke mortality rate

PAGE 81

68 difference between Blacks and Whites is se en in the youngest age group, 35-44. In this age-group the Black female stroke death rate is 2.36 times higher than that of White females. For males, the smallest stroke mortality rate difference between Blacks and Whites is seen in the 35-44 year age gr oup. In the 35-44 year age group Black males have a stroke death rate that is 1.30 times higher than that of White males. For females, the smallest stroke mortality rate differen ce between Blacks and Whites is seen in the 6574 year age group. In the 65-74 year age group, Black females have a stroke death rate that is 1.84 times that of White females. Race and Sex Specific Age-Adjusted Stroke Mortality Rates: Census Tract Level Census tract level average annual age-adju sted stroke mortality rates for those aged 35-74 years were calculated for each of the four race-sex groups (Table 4.9). On average, Black stroke mortality rates were twi ce that of White residents. Black males and females experienced the highest average stroke death rate at 79.70 per 100,000 and 60.29 per 100,000, respectively. Non-Hispanic White males and females experienced lower rates of 37.63 and 29.97 per 100,000, respectively. The variability in these census tract level death rates is strikingly la rge as seen in the standard de viation values (Table 4.9). These calculated stroke death rates are consid erably lower than expected given published US and state level stroke deat h rates. This finding is likely due to the comparatively smaller population size of census tracts (c ompared to US and state populations) and consequently lower number of stroke deaths (by race and sex groups) within each census tract. In instances of inade quate population and stroke death counts, calculations of racesex specific stroke death rates would pr oduce unstable results. Another possible contributor to these finding is the age-restrictions impos ed by these study analyses.

PAGE 82

69 These restrictions lower the number of study subjects included in the analyses. Table 4.9. Race and Sex Specific Age-Adjusted (35-74 yrs) Stroke Mortality Rates*, Florida 1998-2002 N= 2156 Census Tracts Mean Median Std Dev Min Max Black Male 79.70 0.00 253.91 0.00 3831.94 Black Female 60.29 0.00 222.53 0.00 5957.46 Non Hispanic White Male 37.63 22.19 75.87 0.00 2157.04 Non Hispanic White Female 29.97 17.44 60.66 0.00 1277.31 rates per 100,000 Study Outcome Scores Sex specific racial disparity measures, ra te ratio and rate difference measures, were calculated at the census tract level (see Table 4.10). On average, Black stroke death rates were twice that of White residents, with average ratios of 2.28 and 2.02 for males and females respectively. The absolute raci al difference scores were 42.07 for males and 30.33 for females. As was seen with the race-ge nder specific age-adjusted rates, there is tremendous variability with in the census tract leve l disparity scores. Table 4.10. Summary statistics for Black-Wh ite stroke mortality disparity measures N=2156 Census Tracts Mean Median Std Dev Min Max Male Black-White Ratio 2.28 0.00 10.30 0.00 229.50 Female Black-White Ratio 2.04 0.00 10.33 0.00 245.77 Male Black-White Difference 42.07 -8.86 264.40 -1900.753796.87 Female Black-White Difference 30.33 -4.45 229.88 -1277.315906.17 Male Percent Difference 173.37 -100.00 1194.00 -100.00 22850.00 Female Percent Difference 147.15 -100.00 1223.00 -100.00 24476.98 Area Social Predictors of Health (ASPoH) Quartiles Descriptive Statistics Before progressing to census tract leve l analyses of the ASPoH and racial disparities in stroke mortality relationship, this potential association was investigated

PAGE 83

70 utilizing ASPoH categories as indicators of disparity magn itude. The range of ASPoH values within each quartile are presented in Table 4.11. For example, if a census tract had a calculated value of -10 for the ASPoH1 variable, that census tract would be included in the First Quartile. If that cen sus tract had a calculated value of 1.1 for the ASPoH1 variable, that census tract would be included in the Third Quartile, and so on. Each quartile contains either 788 or 789 census tracts (3154 total census tracts). Table 4.11. Interquartile range of calculat ed census tract values for each ASPoH variable, 2000 US Census Range of Values for each ASPoH Variable N=3154 First Quartile Second Quartile Third Quartile Fourth Quartile ASPoH1 -11.08 to -1.07-1.07 to 0.190.19 to 1.35 1.35 to 6.38 ASPoH2 -7.26 to -0.82-0.82 to -0.28-0.28 to 0.49 0.49 to 9.46 ASPoH3 -8.37 to -0.78-0.78 to 0.0170.017 to 0.73 0.73 to 4.68 ASPoH4 -13.53 to -0.37-0.37 to -0.003-0.003 to 0.35 0.35 to 8.34 All race-sex specific stroke mortality ra tes are highest in the lowest quartile of the ASPoH-1 variable (see Table 4.12). The most favorable (lowest) stroke mortality rates occurred in the most affluent area as re presented by the Fourth Quartile. With the exception of the Percent Difference scores, th e remaining disparity measures followed a similar pattern of intensity. The data show ed that the male Black-White disparity in stroke mortality was more pronounced fo r the lowest ASPoH-1 quartile which represented the most deprived area (Table 4.13). Results were inconsistent for the female disparity scores.

PAGE 84

71 Table 4.12. Mean Race-sex specific stro ke mortality by Quartile: ASPoH-1 ASPoH-1 Quartile Black Male Age Adjusted Death rate Black Female Age Adjusted Death rate NH-White Male Age Adjusted Death rate NH-White Female Age Adjusted Death rate 1 (low) 89.629 66.663 42.264 32.804 2 69.120 56.668 31.675 25.234 3 55.444 44.450 26.761 20.673 4 (high) 38.715 37.588 18.890 15.329 Table 4.13. Mean Black-White Stroke Mort ality Disparity by Quartile: ASPoH-1 ASPoH-1 Quartile Male BlackWhite Ratio Female BlackWhite Ratio Male BlackWhite Difference Female BlackWhite Difference Male Percent Difference Female Percent Difference 1 (low) 2.121 2.032 47.365 33.859 112.070 103.218 2 2.182 2.246 37.446 31.435 118.219 124.574 3 2.072 2.150 28.682 23.777 107.179 115.015 4 (high) 2.049 2.452 19.825 22.258 104.954 145.200 Similar results were obtained for Non Hispanic Whites when the effect of the ASPoH-2 variable was examined, as seen in Table 4.14. For Blacks, however, the influence of the ASPoH-2 variable was in contrast to ASPoH-1 effects. The most favorable stroke mortality rates for Blacks were observed for the lowest quartile of the ASPoH-2 variable, which represen ts affluent areas, and the le ast favorable rates occurred in the highest quartile areas, which represen ts deprived areas. Accordingly, Table 4.15 demonstrates that each of the disparity scores is highest in those areas of affluence.

PAGE 85

72 Table 4.14. Mean Race-sex specific stro ke mortality by Quartile: ASPoH-2 ASPoH-2 Quartile Black Male Age Adjusted Death rate Black Female Age Adjusted Death rate NH-White Male Age Adjusted Death rate NH-White Female Age Adjusted Death rate 1 (high) 60.121 47.680 33.042 25.966 2 73.755 53.520 31.012 24.634 3 79.590 62.809 28.629 23.443 4 (low) 92.879 73.520 21.636 16.529 Table 4.15. Mean Black-White Stroke Mort ality Disparity by Quartile: ASPoH-2 ASPoH-2 Quartile Male BlackWhite Ratio Female BlackWhite Ratio Male BlackWhite Difference Female BlackWhite Difference Male Percent Difference Female Percent Difference 1 (high) 1.819 1.836 27.079 21.714 81.953 83.623 2 2.378 2.173 42.743 28.886 137.828 117.263 3 2.780 2.679 50.961 39.366 178.006 167.920 4 (low) 4.293 4.448 71.243 56.990 329.273 344.783 The race-sex specific stroke mortality ra tes scores vary slightly in magnitude across the quartiles for the ASPoH-3 measure (Table 4.16). Consequently, the magnitude of this measure has limited influence on Bl ack-White disparities in stroke mortality (Table 4.17). Table 4.16. Mean race-sex specific stroke mortality by Quartile: ASPoH-3 ASPoH-3 Quartile Black Male Age Adjusted Death rate Black Female Age Adjusted Death rate NH-White Male Age Adjusted Death rate NH-White Female Age Adjusted Death rate 1 69.895 60.231 27.120 21.957 2 86.859 59.958 28.990 23.262 3 79.929 56.064 28.760 21.772 4 67.704 60.565 25.226 19.763

PAGE 86

73 Table 4.17. Mean Black-White Stroke Mort ality Disparity by Quartile: ASPoH-3 ASPoH-3 Quartile Male BlackWhite Ratio Female BlackWhite Ratio Male BlackWhite Difference Female BlackWhite Difference Male Percent Difference Female Percent Difference 1 2.577 2.743 42.775 38.275 157.729 174.319 2 2.996 2.577 57.869 36.696 199.617 157.750 3 2.779 2.575 51.169 34.292 177.921 157.504 4 2.684 3.064 42.479 40.802 168.394 206.458 Table 4.18. Mean race-sex specific stroke mortality by Quartile: ASPoH-4 ASPoH-4 Quartile Black Male Age Adjusted Death rate Black Female Age Adjusted Death rate NH-White Male Age Adjusted Death rate NH-White Female Age Adjusted Death rate 1 88.846 63.895 31.325 25.238 2 81.718 63.131 28.899 23.096 3 68.667 53.039 25.290 20.026 4 68.743 56.243 26.176 20.143 The ASPoH-4 measure has similar influences as the ASPoH-1 measure. For Black and Non Hispanic White residents, the st roke mortality rates are lowest in the most affluent areas and highest in the most depriv ed areas. Table 4.18 demonstrates that the impact of the ASPoH-4 measure is slightly st ronger for Black residents as compared to Non Hispanic White residents. In Table 4.19 we see that the magnitude of the disparities scores shows only slight variation across the ASPoH-4 categories.

PAGE 87

74 Table 4.19. Mean Black-White Stroke Mort ality Disparity by Quartile: ASPoH-4 ASPoH-4 Quartile Male BlackWhite Ratio Female BlackWhite Ratio Male BlackWhite Difference Female BlackWhite Difference Male Percent Difference Female Percent Difference 1 2.836 2.532 57.521 38.657 183.627 153.169 2 2.828 2.733 52.819 40.035 182.772 173.344 3 2.715 2.648 43.377 33.013 171.516 164.854 4 2.626 2.792 42.568 36.100 162.622 179.221 Regression Findings: Census Tract Level Analyses The multiple regression model which was util ized to test the predictive capability of the ASPoH variables (4 principal com ponents) at the census tract level is the following: Racial_Disparity_Score = 0 + 1ASPoH1 + 2ASPoH2 + 3ASPoH3 + 4ASPoH4 + This regression model was used in si x instances, once for each of the six disparity outcome scores. The ASPoH variables were shown to be si gnificant predictors of the Female Ratio (Table 4.21) but were not significan t predictors of the Male Ratio outcome (Tables 4.20). Table 4.20. Regression model which measured the association between the male Black-White stroke mortalit y ratio and the “Area Social Predictors of Health’ variables Male Ratio F Value: 2.25 ** Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 0.168 0.135 1.24 0.2143 ASPoH-2 0.350 0.267 1.31 0.1898 ASPoH-3 -0.453 0.218 -2.07 0.0381 ASPoH-4 0.468 0.344 1.36 0.1744 ** : not statistically significant, p value > 0.05 N=1909 Census Tracts The F-Values were 2.25 and 2.38 for the Male ratio and Female ratio, respectively. The model accounted for 0.5% of the variance in the Male Ratio score and

PAGE 88

75 0.5% of the variance in the Female Ratio score. ASPoH-2 was the only significant independent predictor of the Female ratio, with the Female ratio increasing 0.557 points with every one unit increase in the ASPoH-2 variable. Table 4.21. Regression model which measured the association between the female Black-White stroke mortalit y ratio and the “Area Social Predictors of Health’ variables Female Ratio F Value: 2.38* Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 0.199 0.136 1.46 0.1442 ASPoH-2 0.557 0.364 2.10 0.0355 ASPoH-3 -0.440 0.219 -1.84 0.0665 ASPoH-4 -0.023 0.345 -0.07 0.9463 : statistically significant, p value 0.05 N=1894 Census Tracts The ASPoH variables were not shown to be significant predictors of the Male or Female Difference scores (Tables 4.22 a nd 4.23). The F-Values were 1.90 and 1.10, respectively. The amount of variance accounted for by the models was minimal. The model accounted for 0.35% of the variance in the Male Difference score and 0.20% of the variance in the Female Difference score. Table 4.22. Regression model which measured the association between the male Black-White stroke mortality difference scor e and the “Area Social Predictors of Health’ variables Male Diff F Value: 1.90 ** Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 -2.860 3.049 -0.94 0.3484 ASPoH-2 4.535 6.426 0.71 0.4805 ASPoH-3 -10.037 5.250 -1.91 0.0560 ASPoH-4 -12.695 7.969 -1.59 0.1113 ** : not statistically significant, p value > 0.05 N= 2156 Census Tracts

PAGE 89

76 Table 4.23. Regression model which measured the association between the female Black-White stroke mortality difference scor e and the “Area Social Predictors of Health’ variables Female Diff F Value: 1.10 ** Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 3.183 2.653 1.20 0.2305 ASPoH-2 2.965 5.591 0.53 0.5960 ASPoH-3 -4.479 4.568 -0.98 0.3269 ASPoH-4 -7.281 6.934 -1.05 0.2938 ** : not statistically significant, p value > 0.05 N= 2156 Census Tracts The ASPoH variables were not shown to be sign ificant predictors of the Male or Female Percent Difference scores (Tables 4.24 a nd 4.25). The F-Values were 1.72 and 2.02, respectively. The amount of variance accounted for by the models was minimal. The model accounted for 0.49% of the variance in the Male Percent Difference (MPD) score and 0.60% of the variance in the Fema le Percent Difference (FPD) score. Table 4.24. Regression model which measured the association between the male Black-White stroke mortality percent difference score and the “Area Social Predictors of Health’ variables MPD F Value: 1.72 Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 26.205 19.213 1.36 0.1728 ASPoH-2 40.643 36.506 1.11 0.2658 ASPoH-3 -37.522 30.619 -1.23 0.2206 ASPoH-4 77.139 50.574 1.53 0.1274 ** : not statistically significant, p value > 0.05 N=1415 Census Tracts Table 4.25. Regression model which measured the association between the female Black-White stroke mortality percent difference score and the “Area Social Predictors of Health’ variables FPD F Value: 2.02 Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 35.124 20.428 1.72 0.0858 ASPoH-2 81.423 38.943 2.09 0.0367 ASPoH-3 -33.890 31.935 -1.06 0.2888 ASPoH-4 -0.045 53.129 -0.00 0.9993 ** : not statistically significant, p value > 0.05 N= 1346 Census Tracts

PAGE 90

77 Restricted Subset of Census Tracts Descriptive Statistics Gender specific racial disparity measures, rate ratio and rate difference measures, were calculated at the census tract level (Tab le 4.10). On average, Black stroke death rates were twice that of White residents, with average ratios of 2.28 and 2.02 for males and females respectively. The absolute raci al difference scores were 42 for males and 30 for females. As was seen with the race-ge nder specific age-adjusted rates, there is tremendous variability within th e census tract level disparity sc ores. This is likely due to vast number of census tract with zero rates for black males and females. Dividing or subtracting by zeros (Black ageadjusted rates) leads to an attenuation of the calculated racial disparity scores. When these census tr acts with zero rates for either Blacks or Whites are excluded, the calculated disparity scores are much larger (Table 4.26). Table 4.26. Descriptive Stat istics for Black-White st roke mortality disparity measures: Restricted Subset Mean Median Std Dev Min Max Male Black-White Ratio 10.656 3.778 21.727 0.040 229.500 Female Black-White Ratio 10.299 3.314 23.315 0.075 245.770 Male Black-White Difference 249.572 117.880 495.838 -1900.755 3796.875 Female Black-White Difference 189.466 84.669 484.873 -662.725 5906.174 Male Percent Difference 965.629 277.778 2173.000 -95.952 22850.000 Female Percent Difference 929.915 231.434 2332.000 -92.470 24476.980 Multiple Regression Models Additional analyses, utilizing the previ ous regression model (Disparity Score = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + ), were performed only on those census tracts with non-zero age-adjusted rates for each of th e race groups. This

PAGE 91

78 restriction results in the u tilization of only 363 and 323 censu s tracts in the regression analyses for male and female Black-Wh ite disparity scores, respectively. The model with male ratio as the outcom e was statistically significant with an FValue of 17.78 and 17% of the variance accoun ted for by the model (Table 4.27). Both ASPoH-1 and 2 were statistically significant predictors of the Black-White male ratio score. With a one point increase in the ASPo H-1 score, the male ratio increases by 5.59. With a one point increase in the ASPo H-2 score, the male ratio increases 7.21. Table 4.27. Regression model which measured the association between the male Black-White stroke mortalit y ratio and the ‘Area Social Predictors of Health’ variables, Restricted subset of census tracts Male Ratio F Value: 17.78* Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 5.588 0.705 7.92 <.0001 ASPoH-2 7.217 1.482 4.87 <.0001 ASPoH-3 -0.819 1.027 -0.80 0.4259 ASPoH-4 3.262 1.428 2.28 0.0230 : statistically significant, p value 0.05 N=363 Census Tracts The model predicting female ratio was statistically significant with an F-Value of 16.98 and 18% of the variance accounted for by the model (Table 4.28). Both ASPoH-1 and 2 were statistically significant predictors of the Black-White female ratio score. With a one point increase in the ASPoH-1 score, the female ratio incr eases by 6.18. With a one point increase in the ASPoH-2 scor e, the female ra tio increases 10.12. Table 4.28. Regression model which measured the association between the female Black-White stroke mortalit y ratio and the ‘Area Social Predictors of Health’ variables, Restricted subset of census tracts Female Ratio F Value: 16.98* Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 6.177 0.807 7.65 <.0001 ASPoH-2 10.125 1.711 5.92 <.0001 ASPoH-3 0.039 1.189 0.03 0.9737 ASPoH-4 2.512 1.744 1.44 0.1506 : statistically significant, p value 0.05 N=323 Census Tracts

PAGE 92

79 The ASPoH variables were significant predictors of the Male difference score (Table 4.29). The model was statistically significant with an F-Value of 12.42 and 12% of the variance accounted for by the model. ASPoH-1 is a statistically significant predictor of the Black-White male difference score. With a one point increase in the ASPoH-1 score, the male difference score increases by 107.53. Table 4.29. Regression model which measured the association between the male Black-White stroke mortality difference scor e and the ‘Area Social Predictors of Health’ variables, Restricted subset of census tracts Male Diff F Value: 12.42* Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 107.531 16.511 6.51 <.0001 ASPoH-2 57.404 34.692 1.65 0.0989 ASPoH-3 -31.478 24.058 -1.31 0.1916 ASPoH-4 41.628 33.444 1.24 0.2141 : statistically significant, p value 0.05 N=363 Census Tracts The ASPoH variables were significant pred ictors of the Female difference score (Table 4.30). The model was statistically significant with an F-Value of 10.74 and 12% of the variance accounted for by the model. ASPoH-1 is a statistically significant predictor of the Black-White female differen ce score. With a one point increase in the ASPoH-1 score, the female diffe rence score increases by 103.74. Table 4.30. Regression model which measured the association between the female Black-White stroke mortality difference scor e and the ‘Area Social Predictors of Health’ variables, Restricted subset of census tracts Female Diff F Value: 10.74* Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 103.742 17.360 5.98 <.0001 ASPoH-2 50.118 36.799 1.36 0.1742 ASPoH-3 -1.510 25.569 -0.06 0.9529 ASPoH-4 -4.686 37.498 -0.12 0.9006 : statistically significant, p value 0.05 N=323 Census Tracts The ASPoH variables were significant predictors of the Male percent difference

PAGE 93

80 score (Table 4.31). The model was statistical ly significant with an F-Value of 17.78 and 16.6% of the variance accounted for by the model. ASPoH-1 is a statistically significant predictor of the Black-White male percent diffe rence score. With a one point increase in the ASPoH-1score, the male percent differe nce score increases by 558. ASPoH-2 is a statistically significant predicto r of the Black-White male pe rcent difference score. With a one point increase in the ASPoH-2 score, the male percent difference score increases by 721. Table 4.31. Regression model which measured the association between the male Black-White stroke mortality percent difference score and the ‘Area Social Predictors of Health’ variables, Re stricted subset of census tracts F Value: 17.78* Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 558.846 70.521 7.92 <.0001 ASPoH-2 721.666 148.173 4.87 <.0001 ASPoH-3 -81.908 102.751 -0.80 0.4259 ASPoH-4 326.172 142.842 2.28 0.0230 : statistically significant, p value 0.05 N=363 Census Tracts The ASPoH variables were significant pred ictors of the Female percent difference score (Table 4.32). The model was statistical ly significant with an F-Value of 16.98 and 17.6 % of the variance accounted for by the model. ASPoH-1 is a statistically significant predictor of the Black-White female percent difference score. With a one point increase in the ASPoH-1 score, the female percent di fference score increases by 617. ASPoH-2 is a statistically significant predictor of the Bl ack-White female percent difference score. With a one point increase in the ASPoH-2 score, the female percent difference score increases by 1012.

PAGE 94

81 Table 4.32. Regression model which measured the association between the female Black-White stroke mortality percent difference score and the ‘Area Social Predictors of Health’ variables, Re stricted subset of census tracts F Value: 16.98* Parameter Estimate Standard Error t-value Pr> |t| ASPoH-1 617.713 80.735 7.65 <.0001 ASPoH-2 1012.470 171.132 5.92 <.0001 ASPoH-3 3.928 118.908 0.03 0.9737 ASPoH-4 251.251 174.385 1.44 0.1506 : statistically significant, p value 0.05 N=323 Census Tracts Summary of Findings ASPoH categories (quartiles) were created to assess the relationship between level of economic advantage/disa dvantage and magnitude of Black-White disparities in stroke mortalit y. All race-sex-specific ageadjusted rates and disparity scores were lowest in the ASPoH-1 quartile (quartile 4) representi ng the highest values for economic advantage. In the assessment of the ASPoH-2 variable, Black males and females in the most economically advantaged census tracts experienced the lowest stroke mortality rates. This resulted in disparity sc ores being the greatest in these economically disadvantaged census tracts. Race-sex specific stroke mortality rates and disparity scores were very similar across quartiles for th e ASPoH-3 variable and for the ASPoH-4 variable. No inferences can be made rega rding the impact of th e ASPoH-3 and ASPoH-4 variables on the magnitude of Black-Wh ite disparities in stroke mortality. Multiple regression analysis was utilized to assess the predictive ability of the ASPoH variables on Black-White disparities in stroke mortality. Study results showed elevated age-adjusted stroke mortality rates for Black Floridians compared to NonHispanic White Floridians. For females, the Black-White ratio score was associated with significant changes in levels of the ASPoH variables. In creases in the magnitude of

PAGE 95

82 ASPoH-1 and ASPoH-2, which accounted for the highest percentage of variance in the census tract level social and economic meas ures, were associated with higher BlackWhite stroke mortality ratios. Contrastingl y, increases in the magnitude of the ASPoH-3 and ASPoH-4 variables, were associated with decreases in the Black-White stroke mortality ratios. These decreases in the magnitude of Black-White stroke mortality ratios in those areas of economic advantage s upport the study hypothesis which states that Black-White disparities in stroke mortality wi ll be greatest at lower levels (magnitudes) of the ASPoH variables. None of the rema ining multiple regression models testing the predictive ability of the ASPoH variables on Black-White stroke mortality (as measured by the disparity scores) were statistically si gnificant. When regression analyses were restricted to a subset of these same census tracts, all of the regression models were found to be statistically significan t. Increases in the ASPoH1 and ASPoH-2 variables were associated with increases in the Black-White ratio score, difference score and percent difference score for both males and females. Inconsistent results were obtained for the ASPoH-3 and ASPoH-4 variables. Additiona lly, the hypothesis was only supported when the restricted analyses were performed acces sing the predictability of the ASPoH-2 variable. In this instance, the Black-White di sparity scores decreased with elevations in the ASPoH scores. The hypothesis was not su pported when accessing the predictability of either of the remaining ASPoH variables. Part III. Research Question Two Question: Are higher levels of Black-White disparities in stroke mortality associated with low levels of social class? Hypothesis: Black-White disparities in stroke mortality will be greatest for those

PAGE 96

83 in the lowest social class group. Descriptive Statistics Social Class Groups Population Counts Educational attainment data (used as a proxy measure for social class) were reported for a total of 3138 out of the 3154 Fl orida census tracts. The census reported educational attainment data for the followi ng race, gender and age groups: Black males aged 35-44, Black males aged 45-64, and Bl ack males 65 and up. This educational attainment information was presented for the same age-groups for Black females, Non Hispanic White males and Non Hispanic White females, resulting in educational attainment data for a total of twelve race-sex-age groups. Population counts for each of the social class race-sex-age-groups are multiplied by 5(years) to estimate the population total for the 1998-2002 5-year study period. As a result, a total population of 31,884,280 was contributed to the study by all Florida residents 35 years of age and older (Table 4.33). Across all race-sex-groups, the 45-64 year age-group contributes the highest popul ation percentage at 39.51%. The 35-44 year age-group and the 65 years and up age group contribute 25.42 and 35.07 percent of the total person years, respectively. NH-White fe males and males contributed the highest percentage of person years to the stu dy, with approximately 47.47% and 41.76%, respectively. Black males contributed 4.85% of the population count and Black females contributed 5.91%% of the total population to the study. Within each race-sex group, the population distribution is slightly differe nt for NH-White females. For NH-White females, the 65+ year age-group contributes the largest percentage to the population count (39.32%). For each of the remaini ng race-sex groups, the 45-64 year age-groups

PAGE 97

84 contribute the largest percentages to the population count s. For Black males and Black females, the smallest percentage of the population is contributed by the 65 years and older age group. For the NH-White females and males, the smallest percentage of the population is contributed by the 35-44 year age group. Table 4.33. 2000 US Census Percent Population by race, sex and age group, Florida, All educational attainment groups Age-Group 35-44 45-64 65_up Total Black Males N=1,547,955 39.67 43.84 16.49 4.85 (100) Black Females N=1,885,735 37.10 41.80 21.10 5.91 (100) NH-White Males N=13,316,065 25.66 39.95 34.39 41.76 (100) NH-White Females N=15,134,525 22.30 38.38 39.32 47.47 (100) Total 25.42 39.51 35.07 100% Almost thirty-four percent of the to tal population contribut ed by Black males belonged to Social Class 3 category (Table 4.34). Th e second highest population percentage is contributed by Black males in Social Class 4 category at 25.01% (next to the last social class categor y). Only 7.94% of Black male population belonged to the Social Class 1 category (the hi ghest social class group). Black males in the social class 1 category contributed the smallest percenta ge to the Black male population count. Table 4.34. Black male population count by social class and age-groups (35+years), 2000 US Census, Florida population multiplied by 5 years Age-Group 35-44 45-64 65+ Total Social Class 1 47,610 59,595 15,670 122,875 (7.94%) Social Class 2 132,415 122,635 21,330 276,380 (17.85%) Social Class 3 250,030 221,145 49,130 520,305 (33.61%) Social Class 4 150,150 174,175 62,885 387,210 (25.01%) Social Class 5 33,835 101,090 106,260 241,185 (15.58%) Total 614,040 678,640 255,275 1,547,955 (100%)

PAGE 98

85 Thirty-two percent of th e total population contributed by Black females belonged to Social Class 3 category (Table 4.35). Th e second highest percen tage of the population count is contributed by Black females in So cial Class 4 category (23.88%). Only 9.79% of Black female population belonged to the So cial Class 1 category (the highest social class group). The social class 1 category contributed the least to the population count. Table 4.35. Black female population count by social class and age-group (35+years), 2000 US Census, Florida population multiplied by 5 years Age-Group 35-44 45-64 65+ Total Social Class 1 72,980 84,355 27,370 184,705 (9.79%) Social Class 2 199,785 160,285 30,630 390,700 (20.72%) Social Class 3 252,615 262,505 86,355 601,475 (31.90%) Social Class 4 144,395 193,710 112,155 450,260 (23.88%) Social Class 5 29,920 87,345 141,330 258,595 (13.71%) Total 699,695 788,200 397,840 1,885,735 (100%) Almost thirty-one percent of the tota l population contributed by NH-White males belonged to Social Class 3 category (Table 4.36). The s econd highest percentage of person years is contributed by NH-White male s in Social Class 2 category at 28.03%. This is in contrast to Black males, with Soci al Class 4 as the second highest percent of the population contributed. 27.48% of NH-White male population belonged to the Social Table 4.36. Non-Hispanic White male populat ion count by social class and agegroup (35+years), 2000 US Census, Flor ida population multiplied by 5 years Age-Group 35-44 45-64 65+ Total Social Class 1 867,335 1,635,445 1,156,910 3,659,690 (27.48%) Social Class 2 1,030,275 1,613,055 1,089,480 3,732,810 (28.03%) Social Class 3 1,149,720 1,531,070 1,425,735 4,106,525 (30.84%) Social Class 4 316,010 395,015 593,085 1,304,110 (9.79%) Social Class 5 53,255 145,495 314,180 512,930 (3.85%) Total 3,416,595 5,320,080 4,579,390 13,316,065 (100%) Class 1 category. This is almost 3.5 times the percentage of that for Black males. The social class category with the least amount of population contributed was the social class

PAGE 99

86 5 category (the lowest social class group). Thirty-eight percent of the population count was contributed by NH-White females belonged to Social Class 3 category (Table 4.37). The second highest percentage of the population is contributed by NH-White females in Social Class 2 category at 29.09%. This is in contrast to Black females, with Social Class 4 as the second highest percent of person years cont ributed. 19.37% of NH-White female person years belonged to the Social Class 1 category. This is 2.0 tim es higher than the per centage of that for Black females. The social class category with the least amount of the population contributed was the social class 5 cate gory (the lowest social class group). Table 4.37. Non-Hispanic White female popula tion count by social class and agegroup (35+years), 2000 US Census, Flor ida population multiplied by 5 years Age-Group 35-44 45-64 65+ Total Social Class 1 840,695 1,296,335 793,860 2,930,890 (19.37%) Social Class 2 1,190,510 1,859,595 1,353,105 4,403,210 (29.09%) Social Class 3 1,082,905 2,094,790 2,570,290 5,747,985 (37.98%) Social Class 4 224,380 453,500 849,705 1,527,585 (10.09%) Social Class 5 36,105 105,220 383,530 524,855 (3.47%) Total 3,374,595 5,809,440 5,950,490 15,134,525 (100%) Overall, Social Class 3 residents contribu ted the majority of the population to this study (34.42%). Social Class 5 residents contributed 4.82% of the total study population. Black residents were more likely then NHWhites to have less than a high school education. Consequently, NH-Whites were more likely then Black residents to continue their education beyond high school. Total Census Tract Stroke Deaths by Race and Sex A total of 42,810 stroke deat hs were documented for all Florida residents 35 years of age and older within the 1998-2002 study time period. Death records which did not

PAGE 100

87 include educational attainment information were not included in this count. NH-White females and males accounted for the highest pe rcentage of stroke deaths in the study, with approximately 51.81% and 36.29% resp ectively (Table 4.38). Black males and females accounted for 4.91% and 6.98% of the st roke deaths respectively. The 65 years and up age group contributed 89.73% of the total number of stroke deaths in the study. As expected, the youngest age group contributed th e least percentage of the stroke deaths. The 45-64 year age group accounted for the rema ining 8.87% of the stroke deaths. Within each of the race-sex groups the occurrence of stroke deaths increased with age. Table 4.38. Percent Stroke Deaths for Ra ce-Sex Groups By Age-Group, Florida 1998-2002 35-44 45-64 65_up Total Black Males 4.61 30.94 64.45 4.91 (100) Black Females 4.55 18.03 77.42 6.98 (100) NH-White Males 1.31 9.11 89.58 36.29 (100) NH-White Females 0.73 5.37 93.90 51.80 (100) Total 1.40 8.87 89.73 100 The highest percentage of st roke deaths was among reside nts of the Social Class 3 category at 44.22% (Table 4.39). Social Class 4 deaths made up the smallest percentage of total stroke deaths (10.53%). The remain ing three social class groups each contributed around 15% of the total stroke deaths. Table 4.39. Percent Stroke Deaths by Social Class Group, Florida 1998-2002 SC1 SC2 SC3 SC4 SC5 Total Total 15.45% 14.87% 44.22 % 10.53% 14.93% 100% Black males and females contributed the le ast percentage of deaths to the study, 4.91 and 6.98% respectively (Table 4.40). Over half of the st udy deaths were contributed by NH White females while NH White males ma de up 36.29 percent of the study deaths. NH Whites consistently contributed the highest percentage of deaths by social class. NH

PAGE 101

88 White females generally contri buted the highest percentage of deaths with the exception of Social Class 1, where NH White males made up 52.66% of the SC1 deaths. Black males consistently contributed the least percentage of deaths for each of the social class groups. Table 4.40. Percent Stroke Deaths for RaceSex Groups by Social Class Groups, Florida 1998-2002. SC1 SC2 SC3 SC4 SC5 Total Black Males N=2104 1.82 2.29 3.81 8.63 11.37 4.91 Black Females N=2990 3.93 4.24 4.75 11.76 16.13 6.98 NH-White Males N=15,536 52.66 39.4833.2631.74 28.35 36.29 NH-White Females N=22,180 41.58 53.9858.1847.87 44.14 51.81 Total 100 100 100 100 100 100 Seen in Table 4.41, is the number of census tracts for which educational attainment data is available for specific r ace-sex-age groups. In no instances did the US Census report social class information for the 12 race-sex-age groups for all 3138 census tracts. In particular, educational attainment data for Black Floridians is reported for only a small number of the census tracts. For Black males and females, the smallest number of census tracts with educational attainment information is for the 65 years and older age group. The exception occurs for educationa l attainment group 5, where 35-44 year old Black males and females have data reported for the least number of census tracts. For Non-Hispanic White males, the smallest number of census tracts with educational attainment information is for the 35-44 years and older age group (with the exception of social class 2). For Non-Hisp anic White females, across all age-groups, population counts for those in social class groups 4 and 5 were the least reported. Overall, less than one-third of the census tracts have complete data for reporting

PAGE 102

89 educational attainment information for Black Floridians. The opposite is true for NHWhites. In most social class categories, a larger number of the census tracts have reported data for NH-White males and females. The exception is the information Table 4.41. Number of census tracts (by ra ce, sex, age-group) for which educational attainment data were reported, Florida, 2000 US Census, Summary File 4 available for NH-White males and females for So cial Class 5 (lowest so cial class group). An average of 26% of the census tracts ha ve data for NH-Whites 35-44 years of age, nearly 53% of the census trac ts have data for NH-Whites 45-64 years of age and around 75% of the census tracts have data for NH-Whites 65 years of age and above. Stroke Death Rates and Outcom e Scores: Census Tract Level All Florida Census Tracts Stroke death rates and outcomes scores at the census tract level were calculated College Degree and Beyond Social Class Group 1 Some College / Associates Degree Social Class Group 2 High School Graduate Social Class Group 3 9-11 years of education Social Class Group 4 Less than 9 years of education Social Class Group 5 35-44 4657511055707 393 45-64 5657381029770 706 Black Males 65+ 247323604560 696 35-44 5627931057735 380 45-64 5817501052776 668 Black Females 65+ 308401721653 746 35-44 1909211326281600 763 45-64 1998212926881718 1363 NonHispanic White Males 65+ 1941207626431758 1753 35-44 1917210426251508 601 45-64 1984213027081745 1218 NonHispanic White Females 65+ 1895209327031809 1851

PAGE 103

90 for all Florida census tracts (Table 4.42). E ach of the average rates and disparity scores are weighted by the total census tract popula tion. Upon examina tion of the race-sex specific age adjusted rates, the lowest rates fo r all race-sex groups are observed for Social Class 2. For Black residents, the highest rates occur within Social Class 3, however, the highest rates for Non Hispanic White reside nts occur in the lowest social class group (Social Class 5). Black residents have highe r stroke death rates th an NH-White residents for each of the social class categories. Fo r both males and females, the largest racial difference in rates occurs in Social Class 3. Racial disparity scores (ratios, difference and percent difference scores) for males are greate st for the Social Class 3 residents. For females, the highest ratio a nd percent difference score occurs for the Social Class 2 category; however, the highest difference score occurs for the Social Class 3 category for females. Table 4.42. Weighted average stroke death ra tes and disparity scores by social class group, Florida 1998-2002, (Number of Census tracts) SC 1 (2111) SC 2 (2230) SC 3 (2892) SC 4 (2000) SC 5 (2224) Black Male Age Adjusted Rate 711.40524.191465.96 888.071292.87 Black Female Age Adjusted Rate 966.32809.891396.70 874.141329.47 NH White Male Age Adjusted Rate 208.61174.82 369.07 405.841030.17 NH White Female Age Adjusted Rate 284.15190.23 297.41 448.741151.97 Male Black-White Ratio 6.759.9014.83 11.3014.40 Female Black-White Ratio 8.5418.4315.21 13.9110.38 Male Black-White Difference 502.79349.361096.91 482.23262.69 Female Black-White Difference 682.17619.651099.28 425.41177.50 Male Black-White Percent Difference 575.35890.371382.77 1030.171340.24 Female Black-White Percent Difference 753.991742.621421.50 1291.57938.33 Restricted Census Tracts Only those census tracts that had nonzer o values (i.e., population counts not equal to zero) for each of the 12 race-sex-age-groups were retained in subsequent analyses.

PAGE 104

91 Utilization of this inclusion/exclusion criter ion resulted in the retention of only a portion of the census tracts. Please refer back to Ta ble 4.41 for the exact number of census tracts for which educational attainment (by r ace-sex-age-group) data was available. Because of the small number of census trac ts with available da ta for each of the race-sex social class groups, calculation of an accurate population (denominator data) count for the social class groups at the census tract level wa s not possible. Without the denominator counts, the calcula tion of age-adjusted stroke death rates for the 20 race-sex social class groups was not po ssible. This limitation of th e data prevented any further examination of the research question regard ing the influence of social class on the magnitude of racial disparity in .strok e mortality at the census tract level. Although reliable age adjusted stroke deat h rates could not be calculated at the census tract level, the data were sufficient for the calculation of ra tes for each of the social class groups, by race and sex, for the stat e of Florida as a whole. Three age-group specific rates were calculated (Tables 4.43, 4.44, 4.45), as well as, age-adjusted rates (Table 4.46). For those in the 35-44 y ear age-group, Black males and females experienced higher deaths rate s than their non-Hispanic White counterparts in each of the social class groups. The larges t racial disparity is seen in the Social Class 1 category. Within the Social Class 1 category, Black ma les experienced 7.1 times the stroke death rate of NH White males and Black female s had a rate 5.7 times that of NH White females. The rates for Blacks and Whites are the most similar within the Social Class 4 category. Black females have rates ranging from 3.8 to 4.5 times higher than NH White females for the remainder of the social class categories. Black male s have rates from 1.6 to 2.8 times that of NH White males for the rema inder of the social cl ass categories.

PAGE 105

92 Table 4.43. Race-Sex Specific Stroke De ath Rates (per 100,000): 35-44 Years of Age, Florida 1998-2002 Black Males Black Fe males NH White Males NH White Females Social Class 1 14.70 15.07 2.07 2.62 Social Class 2 12.08 15.52 4.27 3.44 Social Class 3 20.40 24.54 8.87 6.46 Social Class 4 11.32 17.31 10.44 12.48 Social Class 5 17.73 23.39 11.26 5.54 For those in the 45-64 year age-group, the la rgest racial disparity is seen in the Social Class 1 category (Table 4.44). Within the Social Class 1 category, Black males experienced 4.4 times the stroke death rate of NH White males and Black females had a rate 5.9 times that of NH White females. The rates for Blacks and Whites are the most similar within the Social Class 5 category, with NH white females experiencing a slightly higher rate than Black females. Black ma les and females have rates between 2 and 4 times that of NH White males and females for the remainder of the social class categories. Table 4.44. Race-Sex Specific Stroke De ath Rates (per 100,000): 45-64 Years of Age, Florida 1998-2002 Black Males Black Fe males NH White Males NH White Females Social Class 1 67.12 69.94 15.22 11.88 Social Class 2 53.00 49.29 18.35 13.77 Social Class 3 132.49 84.95 40.95 27.21 Social Class 4 85.55 58.85 38.23 28.44 Social Class 5 102.88 73.27 63.92 77.93 For those in the 65 years and up age-group, th e largest racial disparities are seen in social class group 3 for males and group 2 for females (Table 4.45). In social class group 3, Black males have rates 2.0 times higher than NH White males. In social class group 2, Black females have rates 2.2 times th at of NH White females. The rates for Blacks and Whites are the most similar within the Social Class 5 category, with NH white

PAGE 106

93 females experiencing a slightly higher rate than Black females. Black females have rates ranging from 1.5 to 2.1 times higher than NH White females for the remainder of the social class categories. Bl ack males have rates ranging from 1.5 to 1.7 times that of NH White males for the remainder of the social class categories. Table 4.45. Race-Sex Specific Stroke Death Rates (per 100,000): 65-up Years of Age, Florida 1998-2002 Black Males Black Fe males NH White Males NH White Females Social Class 1 472.24 694.19 277.98 324.24 Social Class 2 304.73 522.36 199.45 231.98 Social Class 3 767.35 711.02 390.53 403.65 Social Class 4 354.62 348.62 210.26 235.49 Social Class 5 580.65 679.26 545.23 713.63 When the age-adjusted rates are examin ed, Black males and females experienced higher deaths rates in social class groups 1 thru 4 when co mpared to the rates for NonHispanic White residents (Table 4.46). Fo r social class group 5, Non-White Hispanic females experienced the highest stroke mort ality rate at 210.53 per 100,000. For males, the largest racial disparity in stroke mortality exists within the Social Class 3 category. Within the Social Class 3 category, Black ma les experienced 2.17 times the stroke death rate of NH White males. For females, the la rgest racial disparity in stroke mortality exists within the Social Class 1 and 2 cat egories. Black females had a rate 2.35 and 2.34 times that of NH White females for social class categories 1 and 2, respectively. The rates for Black and Non-Hispanic Whites females are the most similar within the Social Class 5 category, with NH white females experi encing a slightly high er rate than Black females. The rates for Black and Non-Hispanic Whites males are also the most similar within the Social Class 5 category, with Black males experiencing a slightly higher rate than Non-Hispanic White males (1.17 times higher).

PAGE 107

94 Table 4.46. State level age-adjusted (35+ ye ars) stroke death rates and disparity scores by social class group, Florida 1998-2002 SC 1 SC 2 SC 3 SC 4 SC 5 Black Male Age Adjusted Rate 149.60101.57252.10 127.64 192.64 Black Female Age Adjusted Rate 201.00149.94211.93 111.53 199.05 NH White Male Age Adjusted Rate 75.4358.23116.36 71.47 165.01 NH White Female Age Adjusted Rate 85.4963.96112.85 74.08 210.53 Male Black-White Ratio 1.981.742.16 1.78 1.16 Female Black-White Ratio 2.352.341.87 1.50 0.94 Male Black-White Difference 74.1743.34135.74 56.17 27.63 Female Black-White Difference 115.5185.9899.08 37.45 -11.48 Male Black-White Percent Difference 98.3274.42116.65 78.59 16.74 Female Black-White Percent Difference 135.11134.4287.79 50.55 -5.45 Summary of Findings The investigation into the pot ential influence of social class on the magnitude of Black-White disparities in stroke mort ality was precluded by lack of data. Reliable age adjusted stroke d eath rates could not be calculate d at the census tract level. However, the calculation of rates for each of the social class groups, by race and sex, for the State of Florida as a whole was possibl e. As expected, stroke mortality rates increased with age for each of the race-s ex groups. In each of the three age-group categories, Black males and females consistently experienced higher stroke mortality rates across each of the social class groups. The exceptions were instances in which 4564 year old and 65+ year old NH-White female s in Social Class 5 experienced slightly higher stroke mortality rates than Black female s. Most decedents in this social class group experienced the least favorable stroke d eath rates. Of particular note is the observation that Black and Non Hispanic Wh ite residents experience similar rates only when examining the social class 5 category. The study hypothesis stated that Black-White disparities in stroke mortality would

PAGE 108

95 be greatest for those in the lowest social class group (social class group 5). The results did not support the study hypothesis; instead, st roke mortality rates were lowest for those in social class group 5 (Table 4.47). The lowe st disparity scores occurred for those in social class group 5 for both males and females for each of the three disparity scores. For each of the three disparity score outcomes, male disparities are highest for high school graduates (social class three) and female disparities are high est for the college educated. A test for trend in the disparity scores acr oss social class groups was completed. There were no statistically significant trends in any of the disparity scores across social class groups as measured by the Mantel-Haenszel Chi Square test for trend (Table 4.47). Table 4.47. State level Black-White disparity scores by social class group, Florida 1998-2002 SC 1 SC 2 SC 3 SC 4 SC 5 Trend Probability Male Ratio 1.981.742.161.781.16 0.1798 Male Difference 74.1743.34135.7456.1727.63 0.5442 Male Percent Difference 98.3274.42116.6578.5916.74 0.1816 Female Ratio 2.352.341.871.500.94 0.0528 Female Difference 115.5185.9899.0837.45-11.48 0.0651 Female Percent Difference 135.11134.4287.7950.55-5.45 0.0530 Part IV: Research Question 3 Question: Is there effect modification by soci al class of the ASPoH and BlackWhite disparities in stroke mortality relationship? Hypothesis: ASPoH will have a greater impact of Black-White disparities in stroke mortality for the lower social class groups.

PAGE 109

96 Descriptive Statistics Table 4.48. Florida population and stroke death counts by social class category Social Class Category Ages 35+ years SC1(high)SC2 SC3 SC4 SC5(low) Black Males Deaths 121 146 721 389 727 Population 123,940 277,137 521,111 387,912 242,467 Black Females Deaths 260 270 899 530 1031 Population 185,597 391,327 602,144 450,834 259,875 Non-Hispanic White Males Deaths 3483 2513 6297 1431 1812 Population 3,659,9023,732,9234,106,7661,304,615 515,279 Non-Hispanic White Females Deaths 2750 3436 11,015 2158 2821 Population 2,931,1564,403,3155,748,1461,528,106 527,428 Table 4.48 shows that the majority of the stroke deaths and population counts, for all race-sex groups, are concentr ated within social class groups 3 and 5. The Black male and female populations were lowe st in the Social Class 1 (hig hest) category. In contrast, for NH-White males and females, the populations were lowest in the Social Class 5 category. Summary statistics for the study outcome variables are presented in Table 4.49. These statistics demonstrate that there are differences between Black and White stroke mortality rates. The median statistic fo r each of the outcome variables reflects the evenness of the race specific rates occurring in at least 50% of the census tracts retained in the study analyses. The minimum scores re present those instances in which the Black age-adjusted stroke death rate was at or near zero and the White age-adjusted stroke death

PAGE 110

97 Table 4.49. Effect Modification: Summary Statistics for Black-White stroke mortality disparity measures Mean Med Std Dev Min Max Male Black-White Ratio1 9.57065.27 01809.48 Female Black-White Ratio2 9.19052.55 0860.00 Male Black-White Difference3 701.25-51.097473.58 -75476.4067202.81 Female Black-White Difference3 506.77-68.127157.29 -67815.9043472.14 Male Percent Difference1 857.55-100.006527.10 -100.00180847.70 Female Percent Difference2 819.16-100.005255.60 -100.0085900.00 1: N=2411 census tracts, 2: N=2898 census tracts, 3: N=4133 census tracts rate is either similar to or much larger th an the Black stroke death rate. The maximum scores are representative of those instances in which the Black stroke death rates are much larger than the stroke death rate for NH-Whites. Regression Analyses Simple linear regression was used to test the model: Disparity Score ASPoH. Separate regression analyses were run for each of the 5 social class groups. For example, for Social Class Group 1 only, a regression anal yses was run to test how well the ASPoH variables could predict the Male Black-White Ratio score. Next, the same analysis was performed for Social Class Group 2, only. These analyses were then completed separately for each of the three remaining Social Class groups. This technique was continued for each of the remaining Black-White disparity scores. As a consequence of using this methodology, statisti cally significant differences (of the disparity and ASPoH relationship) between social clas s groups cannot be determined. When testing the predictability of each of the disparity scores by the ASPoH variables, for each social class group, the follo wing models were found to be statistically significant (p<.05).

PAGE 111

98 1. Social Class 3: Female Bl ack-White Disparity Difference = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + 2. Social Class 4: Male Bl ack-White Disparity Difference = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + 3. Social Class 4: Female Bl ack-White Disparity Difference = 0 + 1Prin1 + 2Prin2 + 3Prin3 + 4Prin4 + Predictability of ASPoH-1 across Social Class Groups Male Ratio: ASPoH-1 There is no obvious trend in parameter es timates from the highest social class group to the lowest social cl ass group (Table 4.50). With every one unit increase in ASPoH-1, the male Black-White ratio d ecreases 0.8432 for social class group 1 and decreases 0.0093 for social class group 2. The ratio decreases 1.3773 and 0.3984 for Social Class 3 and 4, respectively. For social class 5 residents, th e ratio increases 1.8890 with every one unit increase in ASPoH-1. In the current and remaining regression models, results obtained when utilizing the Male Percent Difference Score were numerically identical to thos e obtained with the use of the Male Ratio Score. Table 4.50. Individual regression models whic h measured effect modification by social class of the association between th e male Black-White stroke mortality ratio and the ‘Area Social Predictors of Health-1’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr> t Social Class 1 (2111) -0.84321.1873-0.71 0.4780 Social Class 2 (2230) -0.00931.3287-0.01 0.9944 Social Class 3 (2892) -1.37732.3829-0.58 0.5634 Social Class 4 (2000) -0.39843.0992-0.13 0.8978 Social Class 5 (2224) 1.88901.37401.37 0.1699 Female Ratio: ASPoH-1 There is no obvious trend in parameter es timates from the highest social class

PAGE 112

99 group to the lowest social cla ss group (Table 4.51). With every unit increase in ASPoH1, the female Black-White ratio decreases 1.1784 and 1.7813 for social class groups 1 and 5 respectively; however, the ratio decrease s only slightly for social class groups 2 and 4 (0.2557 and 0.2911 respectively). The ratio increases 2.3202 for Social Class group 3. In the current and remaining regres sion models, results obt ained when utilizing the Female Percent Difference Score were num erically identical to those obtained with the use of the Female Ratio Score. Table 4.51. Individual regression models whic h measured effect modification by social class of the association between the female Black-White st roke mortality ratio and the ‘Area Social Predicto rs of Health-1’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) -1.17841.2754-0.92 0.3560 Social Class 2 (2230) -0.25572.1217-0.12 0.9041 Social Class 3 (2892) 2.32021.18211.96 0.0500 Social Class 4 (2000) -0.29112.0011-0.15 0.8844 Social Class 5 (2224) -1.78131.2689-1.40 0.1609 Male Difference: ASPoH-1 There is no trend in parameter estimates fr om the highest social class group to the lowest social class group (Table 4.52). With every one unit increase in ASPoH-1, the male Black-White difference score decreases 218.1073 and 411.8448 (statistically significant, p<.05) for social class groups 2 and 4. The difference score increases 336.9195, 204.4327 and 40.6721 for Social Class 1, 3 and 5, respectively. Table 4.52. Individual regression models wh ich measured effect modification by social class of the association between the male Black-White stroke mortality difference score and the ‘Area Social Predictors of Health-1’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) 336.9195194.02801.73 0.0832 Social Class 2 (2230) -218.1073139.1213-1.57 0.1174 Social Class 3 (2892) 204.4327176.08361.16 0.2459 Social Class 4 (2000) -411.8448179.8761-2.29 0.0223 Social Class 5 (2224) 40.6721200.81820.20 0.8395

PAGE 113

100 Female Difference: ASPoH-1 There is no trend in parameter estimates fr om the highest social class group to the lowest social class group (Table 4.53). With every one unit increase in ASPoH-1, the female Black-White difference score decrea ses for social class groups 1 and 5. The difference score increases for so cial class groups 2, 3 and 4. The increase in the Female Difference score of 427.4808 per unit increase in the ASPoH-1 variable for social class 3 is statistically significant (p<.05). Table 4.53. Individual regression models whic h measured effect modification by social class of the association between the female Black-White stroke mortality difference score and the ‘Area Social Predictors of Health-1’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) -95.3442189.8774-0.50 0.6157 Social Class 2 (2230) 27.0368141.97220.19 0.8490 Social Class 3 (2892) 427.4808126.99013.37 0.0378 Social Class 4 (2000) 58.0998182.25800.32 0.7500 Social Class 5 (2224) -193.1107220.3813-0.88 0.3811 Predictability of ASPoH-2 across Social Class Groups Male Ratio: ASPoH-2 With every one unit increase in ASPoH-2, the male Black-White ratio increases 1.0612, 0.0173 and 4.3451 for social class groups 1, 3 and 5 (Table 4.54). The ratio decreases 0.7950 and 5.2862 for Social Class 2 and 4, respectively. Table 4.54. Individual regression models whic h measured effect modification by social class of the association between th e male Black-White stroke mortality ratio and the ‘Area Social Predicto rs of Health-2’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) 1.06122.41320.44 0.6603 Social Class 2 (2230) -0.79502.7043-0.29 0.7686 Social Class 3 (2892) 0.01734.96020.00 0.9972 Social Class 4 (2000) -5.28626.2953-0.84 0.4016 Social Class 5 (2224) 4.34513.17651.37 0.1721

PAGE 114

101 Female Ratio: ASPoH-2 With every one unit increase in ASPoH-2 the female Black-White ratio increases 2.4913 for social class group 3 (Table 4.55). The ratio decreases with each unit increase of the ASPoH-2 variable for the remaining so cial class groups. The largest decrease in the female Black-White ratio (2.1904) occurs for social class group 1. Table 4.55. Individual regression models whic h measured effect modification by social class of the association between the female Black-White st roke mortality ratio and the ‘Area Social Predicto rs of Health-2’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) -2.19042.5398-0.86 0.3889 Social Class 2 (2230) -1.98464.2438-0.47 0.6402 Social Class 3 (2892) 2.49132.44281.02 0.3081 Social Class 4 (2000) -0.17064.1399-0.04 0.9671 Social Class 5 (2224) -0.71372.7448-0.26 0.7949 Male Difference: ASPoH-2 There is no trend in parameter estimates fr om the highest social class group to the lowest social class group (Table 4.56). The di fference score increases with an increase in the ASPoH-2 variable for social class groups 1, 3 and 5. The largest increase in the male Black-White difference score occurs for so cial class group 3, with an increase of 729.5325 points. However, for social class groups 2 and 4, the male Black-White difference score decreases, 192.7746 and 908.5961 (statistically sign ificant, p<.05) respectively, with every one unit increase in ASPoH-2. Table 4.56. Individual regression models whic h measured effect modification by social class of the association between the male Black-White stroke mortality difference score and the ‘Area Social Predictors of Health-2’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) 384.3310402.10290.96 0.3395 Social Class 2 (2230) -192.7746291.1515-0.66 0.5081 Social Class 3 (2892) 729.5325379.62071.92 0.0549 Social Class 4 (2000) -908.5961373.4455-2.43 0.0152 Social Class 5 (2224) 284.0109435.10220.65 0.5141

PAGE 115

102 Female Difference: ASPoH-2 There is no trend in parameter estimates fr om the highest social class group to the lowest social class group (Table 4.57). The female Black-White difference score decreases with an increase in ASPoH-2 for social class groups 2, 4 and 5. The largest decrease in the female difference score occurs for social class group 4. For social class group 4, the female difference score decrease s 769.2337 (statistically significant, p<.05) with every one unit increase in the ASPoH-2 variable. For social class groups 1 and 3, Table 4.57. Individual regression models whic h measured effect modification by social class of the association between the female Black-White stroke mortality difference score and the ‘Area Social Predictors of Health-2’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) 135.8651393.50130.35 0.7300 Social Class 2 (2230) -102.1741297.1179-0.34 0.7310 Social Class 3 (2892) 569.2816273.77942.08 0.0378 Social Class 4 (2000) -769.2337378.3905-2.03 0.0424 Social Class 5 (2224) -245.4847477.4884-0.51 0.6073 the female Black-White difference sc ore increases 135.8651 and 569.2816, respectively, with every one unit increase in ASPoH-2. Predictability of ASPoH-3 across Social Class Groups Male Ratio: ASPoH-3 With every one unit increase in ASPoH-3, the male Black-White ratio increases 1.7117 and 1.5273 for social class groups 1 and 2, respectively (Table 4.58). The ratio decreases for social class groups 3, 4 and 5, with the largest de creases occurring for groups 3 and 4. The male Black-White ratio score decreases 7.3710 and 6.1254 points for social class groups 3 and 4. Although the t-score for social class 3 is significant, p<.05, the overall model was not significant and further interpretation of this outcome is not permitted.

PAGE 116

103 Table 4.58. Individual regression models whic h measured effect modification by social class of the association between th e male Black-White stroke mortality ratio and the ‘Area Social Predicto rs of Health-3’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) 1.71171.75080.98 0.3288 Social Class 2 (2230) 1.52731.97910.77 0.4407 Social Class 3 (2892) -7.37103.4348-2.15 0.0322 Social Class 4 (2000) -6.12544.5224-1.35 0.1765 Social Class 5 (2224) -0.98492.0684-0.48 0.6342 Female Ratio: ASPoH-3 The female Black-White ratio decreases with every one unit increase in the ASPoH-3 score for social class groups 1, 2 a nd 5 (Table 4.59). The largest decrease, 4.6915, occurs for social class groups 2. The female Black-White ratio increases with every one unit increase in the ASPoH-3 score for social class groups 3 and 4. The largest increase in the female ratio score (3 .6796) occurs for social class groups 2. Table 4.59. Individual regression models whic h measured effect modification by social class of the association between the female Black-White st roke mortality ratio and the ‘Area Social Predicto rs of Health-3’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) -0.87691.8034-0.49 0.6271 Social Class 2 (2230) -4.69152.9803-1.57 0.1161 Social Class 3 (2892) 3.67961.67472.20 0.0283 Social Class 4 (2000) 0.95123.03370.31 0.7540 Social Class 5 (2224) -1.16141.8615-0.62 0.5329 Male Difference: ASPoH-3 The difference score decreases with an in crease in ASPoH-3 for all social class groups, with the exceptio n of social class 3 (Table 4.60). The largest decreases in the male Black-White difference score occur for soci al class groups 4 and 5 with decreases of 742.6002 (statistically significant, p,.05) and 622.2808, respectiv ely. For social class group 3, the male Black-White difference sc ore increases 132.1856 with every one unit

PAGE 117

104 increase in ASPoH-3. Table 4.60. Individual regression models whic h measured effect modification by social class of the association between the male Black-White stroke mortality difference score and the ‘Area Social Predictors of Health-3’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) -200.4952280.8351-0.71 0.4755 Social Class 2 (2230) -97.9421203.2443-0.48 0.6300 Social Class 3 (2892) 132.1856252.16650.52 0.6003 Social Class 4 (2000) -742.6002253.1480-2.93 0.0035 Social Class 5 (2224) -622.2808290.8366-2.14 0.0327 Female Difference: ASPoH-3 The female Black-White difference score increases with an increase in ASPoH-3 for social class groups 1 and 2(Table 4.61). For social clas s groups 1 and 3, the female Black-White difference score increases 195.8663 and 431.6962 with every one unit increase in the ASPoH-3 score. The differe nce score decreases with an increase in ASPoH-3 for social class gr oups 2, 4 and 5. The difference score decreases 603.1654 for the lowest social class group. The increas e in the Female Difference score of 431.6962 per unit increase in the ASPoH-3 variable for social class 3 is statistically significant (p<.05). Table 4.61. Individual regression models whic h measured effect modification by social class of the association between the female Black-White stroke mortality difference score and the ‘Area Social Predictors of Health-3’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) 195.8663274.82750.71 0.4763 Social Class 2 (2230) -334.6438207.4093-1.61 0.1071 Social Class 3 (2892) 431.6962181.86042.37 0.0178 Social Class 4 (2000) -429.0295256.5000-1.67 0.0948 Social Class 5 (2224) -603.1654319.1690-1.89 0.0591 Predictability of ASPoH-4 across Social Class Groups Male Ratio: ASPoH-4 With the exception of social class 4 resu lts, the male Black-White ratio decreases

PAGE 118

105 with every one unit increase in the ASPoH-4 score (Table 4.62). Th e largest decrease is seen for social class 2 with a 3.8573 decrease in the ratio score with one unit increase in the ASPoH-4 variable. For those in social class 4, the male ratio increases 13.3944 points with every one unit increase in ASPoH-4. Table 4.62. Individual regression models whic h measured effect modification by social class of the association between th e male Black-White stroke mortality ratio and the ‘Area Social Predicto rs of Health-4’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) -0.42802.6894-0.16 0.8736 Social Class 2 (2230) -3.85732.8996-1.33 0.1841 Social Class 3 (2892) -0.33784.8869-0.07 0.9449 Social Class 4 (2000) 13.39447.88451.70 0.0902 Social Class 5 (2224) -0.55102.9039-0.19 0.8496 Female Ratio: ASPoH-4 There is no obvious trend in parameter es timates from the highest social class group to the lowest social cl ass group (Table 4.63). For social class groups 1, 4 and 5, there is a decrease in the female Black-White ratio score with every one unit increase in ASPoH-4. The largest decreases are seen for those residents without a high school diploma. The female Black-White ratio incr eases for social cla ss groups 2 and 3, with the largest increase of 3.0487 observed for thos e with some college education (social class group 2). Table 4.63. Individual regression models whic h measured effect modification by social class of the association between the female Black-White st roke mortality ratio and the ‘Area Social Predicto rs of Health-4’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) -1.83922.7920-0.66 0.5104 Social Class 2 (2230) 3.04874.28810.71 0.4774 Social Class 3 (2892) 1.12072.03740.55 0.5824 Social Class 4 (2000) -3.61214.4296-0.82 0.4152 Social Class 5 (2224) -3.26242.9685-1.10 0.2722

PAGE 119

106 Male Difference: ASPoH-4 The male Black-White difference score increases 43.5065 and 193.6020 points with an increase in the ASPoH-4 score for social class groups 3 and 4 (Table 4.64). However, for social class groups 1, 2 and 5, the male Black-White difference score decreases from 133.0851 (for social class 1) to 691.3068 (for so cial class 2) points with an increase in the ASPoH-4 score. Table 4.64. Individual regression models whic h measured effect modification by social class of the association between the male Black-White stroke mortality difference score and the ‘Area Social Predictors of Health-4’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) -133.0851389.2332-0.34 0.7325 Social Class 2 (2230) -691.3068293.0601-2.36 0.0186 Social Class 3 (2892) 43.5065299.76400.15 0.8846 Social Class 4 (2000) 193.6020364.33030.53 0.5953 Social Class 5 (2224) -141.5969383.0479-0.37 0.7117 Female Difference: ASPoH-4 The female Black-White difference score increases 96.5101 (for social class 5) to 630.5164 (for social class 1) points with increas es in the ASPoH-4 score for social class groups 1, 3 and 5 (Table 4.65). Alternatel y, there is a 205.8951 point decrease in the difference score with an increase in the ASPoH-4 score for soci al class group 2 and a 368.0623 point decrease for so cial class group 4. Table 4.65. Individual regression models whic h measured effect modification by social class of the association between the female Black-White stroke mortality difference score and the ‘Area Social Predictors of Health-4’ variable. (# of census tracts) Parameter Estimate Std Error t-value Pr > t Social Class 1 (2111) 630.5164380.90691.66 0.0983 Social Class 2 (2230) -205.8951299.0664-0.69 0.4914 Social Class 3 (2892) 382.3863216.18741.77 0.0772 Social Class 4 (2000) -368.0623369.1547-1.00 0.3191 Social Class 5 (2224) 96.5101420.36320.23 0.8185

PAGE 120

107 Summary of Findings The potential for effect modification by so cial class of the association between Black-White disparities in stroke mortality and ASPoH variables was investigated. Study findings were dependent upon th e particular Black-White di sparity score and the ASPoH variable under investig ation. Patterns of association across social class groups, or the lack thereof, were not cons istent across Black-White di sparity measures. The study hypothesis stated that the A SPoH measures would have th e greatest impact on those residents in the lowest social class category, i. e. SC5 (less than 9 years of education). An increase in the disparity scor e would suggest greater differen ces in stroke mortality rates between Blacks and non-Hispanic Whites, therefore supporting the hypothesis. This hypothesized effect was identified in two in stances (Table 4.66). When utilizing the ASPoH-1 and ASPoH-2 variable s to estimate effects on the Male Black-White Ratio disparity score, the disparity score increased the greatest amount for those residents with less than 9 years of educati on. None of the remaining regression models supported the study hypothesis.

PAGE 121

108 Table 4.66. Summary results for regressi on models which measured effect modification by social class of the asso ciation between the disparity in stroke mortality measures and the ‘Area Soci al Predictors of Health’ variables Predictor Variable Black-White Stroke Mortality Difference Score Social Class Group with the greatest increase in the Disparity Measure Male Ratio SC5 Female Ratio SC3 Male Difference SC1 ASPoH-1 Female Difference SC3 Male Ratio SC5 Female Ratio SC3 Male Difference SC1 ASPoH-2 Female Difference SC3 Male Ratio SC1 Female Ratio SC3 Male Difference SC3 ASPoH-3 Female Difference SC3 Male Ratio SC4 Female Ratio SC2 Male Difference SC4 ASPoH-4 Female Difference SC1

PAGE 122

109 Chapter Five Discussion Introduction Many questions remain regarding the determin ants of racial disparities in stroke mortality. The influence of individua l characteristics a nd neighborhood economic structure on health has been the focus of se veral studies over the past two decades. Education, income or occupation (or a combina tion of two or more of these measures) are typically used as measures of individual so cial class or socioeconomic status, while an area-based socioeconomic indicator (com posed of various area/neighborhood level economic and social measures obtained from census data) often represents the economic structure. Findings from this type of re search frequently supp ort the hypothesis that living in economically deprived areas and be ing a member of a lower SES group are both associated with an increased prevalence of negative health outcomes.114 Given these research findings, there exis ts an opportunity to examine whether these area characteristics may affect race groups differentially, possibly leading to disparities in health outcomes, specifically st roke mortality. This study was an attempt to further our understanding of Black-White disparities in stroke mortality by looking beyond racial differences in i ndividual level factors common ly associated with these disparities. This study departs from this extensively investigat ed path, and instead focuses on social and economic aspects of the community as contributing factors in these

PAGE 123

110 disparities. To attempt to understand the ba sis of race-based stroke mortality patterns, this study examined variations in Black-White disparities in stroke mortality at the census tract level as a function of ar ea characteristics. Racial in equalities in stroke mortality represent a major challenge for which eff ective action must focus on the social and economic environment. 115 ‘Are there contextual social and economi c area characteristic s related to BlackWhite disparities in stroke mortality inde pendently of and/or in conjunction with individual-level variables?’ was the primary re search question investig ated in this study. In the effort to address th is issue, a progression thr ough the following questions was required: (1) Are Black-White disparities in st roke mortality elevated in those areas with lower amounts of social and economic re sources (represented by the Area Social Predictors of Health variable s)? (2) Are higher levels of Black-White disparities in stroke mortality associated with low levels of social class? (3) Is there effect modification by social class of the ASPoH meas ure and Black-White disparities in stroke mortality relationship? In response to these research questions, the study hypotheses were: (1) Black-White disparities in stroke mortality will be greatest at lower levels of ASPoH, (2) Black-White disparities in stroke mortality will be great est for those in the lowest social class group, and (3) ASPoH will have a greater impact on Black-White disparities in stroke mortality for the lower social class groups. Major Findings Research Question One Specifically, this study examined the effect of area social pred ictors of health (ASPoH) on Black-White disparities in stro ke mortality rates for Florida residents

PAGE 124

111 between 35 and 74 years of age within th e 1998-2002 five-year time period. Four measures of area social pr edictors were developed th rough the use of principal components analyses (accounting for a total of 76.29% of the variance in the data). Principal component one (ASPoH-1) was represen tative of the general economic status of the census tract (median household income and percent of households within the census tract that were above the poverty rate). Principal component two (ASPoH-2) was representative of the percen t of occupied homes and th e percent of the population employed fulltime. Principal component 3 (ASPoH-3) was representative of area affluence (Median home value and percent of home ownership). Principal component 4 (ASPoH-4) was representative of opportuni ties afforded by educational resources (Percent of census tract residents who are empl oyed and percent of census tract residents 25 years and older who are high school graduates). Predictability of ASPoH Variables Multiple linear regression models were used to test the predictability of the gender specific racial disparity scores by the ASPoH measures (4 principal components) at the census tract level. The regression model pred icting the Female Ratio score was the only statistically significant model in these analys es. Because there were census tracts with populations too small to have a ny expected stroke deaths, an alyses were performed only on those census tracts with non-ze ro age-adjusted rate s for each of the race-sex groups. This restriction results in the inclusion of only 363 census tracts in the analyses for males and 323 census tracts for females. Multiple linear regression models were then used to test the predictability of age adjusted stroke death rate s separately for Black males and females by the ASPoH

PAGE 125

112 measures for these select censu s tracts. ASPoH-1 was a statistically significant predictor of the Black male and female age adjusted stroke mortality rates. With a one point increase in the ASPoH-1 score, the ra tes increased 83.87 per 100,000 person years for males and 77.42 per 100,000 person years for female s. These results are in conflict with results obtained when examining the more incl usive group of Florida census tracts (2199 census tracts), given that only a subset of th e original census tracts are included in these analyses. It is also possible that this subset of census tr acts are more homogeneous than the originating data set which could possible result in the attenuation of any association that may be seen the area measure and the Black male and female stroke mortality rates. These higher Black male and female stroke mortality rates in more affluent areas may indicate that the ASPoH measures ar e not actually capturing levels of “area affluence.” The probability must be consider ed that the Black males and females residing in these census tracts are not as likely to have incomes in the hi gher brackets as White residents. If these assertions are true, hi gher rates of adverse health outcomes may be expected.84,85,86 These finding also may be indicative of Black males and females residing in more affluent neighborhoods yet they are not able to take advantage of the resources and services available within the community area. Perhaps those Black decedents, who resided in more affluent ne ighborhoods, nevertheless had rela tively lower incomes when compared to the White residents in these areas These findings could be reflective of the literature that describes more adverse health outcomes for societies in which there is great income inequality.116 Research of metropolitan areas suggests that in addition to the absolute amount of income, re lative disparity of income di stribution within a population is also important for health. Findings show that areas with high income inequality had

PAGE 126

113 significantly greater age-adjusted total mo rtality than those with low inequality.117 The household economic measure (ASPoH-1) was also a statistically significant predictor of the Non Hispanic-White male and fe male age adjusted stroke mortality rates. Areas with a larger proportion of residents in higher income levels and a small proportion below the poverty level experi ence significantly lower stroke death rates for both males and females. With a one point increase in the ASPoH-1 score, the age adjusted stroke death rates decreased 23.65 and 26.32 per 100,000 for males and females respectively. In contrast, higher NH-White female stroke mo rtality rates were associated with areas having more expensive homes and a larger pe rcentage of home owne rship (ASPoH-3). The Non Hispanic-White female age adju sted rate increased 13.50 per 100,000 with a one unit increase in the ASPoH-3 score. Both household economic measures (A SPoH-1) and occupied homes and employment measures (ASPoH-2) were statis tically significant pred ictors of the BlackWhite male ratio score and the Black-White fe male ratio score. More favorable measures of the ASPoH-2 variable were associated w ith lower disparity scores. ASPoH-1 was a statistically significant predicto r of the Black-White male difference score and the female difference score. Increased household economic measures were pred ictive of increased Black-White difference scores. ASPoH-1 and ASPoH-2 were statistically significant predictors of the BlackWhite female percent difference score and the male percent difference score. Increases in these measures were associated with increa ses in the percent difference scores. When this disparity score is utilized, more disa dvantaged areas, as measured by the ASPoH-2 variable, are more likely to expe rience racial disparities of a greater magnitude than more

PAGE 127

114 affluent areas. More affluent areas, as measured by the ASPoH-1 variable, are more likely to experience racial disparities of a great er magnitude than less affluent areas. This finding is in opposition to the study hypothesis. Once again, these finding may be due to the heterogeneity of various characteristics wi thin census tracts. For example, it may be possible that those census tr acts which fall within the “more affluent” category are comprised of families with vast differences in income. Possibly, these census tracts have household incomes from both extremes, with the majority of incomes in the higher brackets. Given the evidence presented rega rding health disparities in areas of high income inequality, we would expect these resu lts if the majority of the census tracts included in the analyses were of such economic circumstances. Inequalities in area resource are accomp anied by differences in life conditions which may adversely influence health.98 These health inequalities result from the differential accumulation of exposures and expe riences among those residing in different neighborhood environments. The effect of ineq uality on health reflects a combination of negative exposures and lack of resources a ccessible by individuals. Consequently, this lack of individual resources in fluences services and investme nts made available for these individuals. More equitable di stribution of public and private resources is likely to have the greatest impact on reducing Bl ack-White health disparities. Research Question Two The investigation into the pot ential influence of social class on the magnitude of Black-White disparities in stroke mort ality was precluded many times by lack of available data. Therefore, a cautionary approa ch must be taken in the interpretation of these results. Due to privacy issues, the rel ease of educational attainment data (at the

PAGE 128

115 census tract level) for particular race-sex -age groups was limited. This resulted in multiple census tracts with no educational attainment data reported for the majority of the study groups. Therefore, the accurate calculation of population (denominator) counts for many of the race-sex specific social class groups was not possible. Overall, less than one-third of the census tracts have complete data for reporting educational attainment information for Black Floridians. The opposite is true for Non-Hispanic Whites. In most social class categories, a larger majority of the census tracts have reported data for NonHispanic White males and females. This lim itation prevented any further examination of the census tract level influence of social cl ass on the magnitude of racial disparity in stroke mortality. Reliable age adjusted stroke death rates could not be cal culated at the census tract level. However, the calculation of rates for each of the social class groups, by race and sex, for the State of Florida as a whole was po ssible. As expected, stroke mortality rates increased with age for each of the race-s ex groups. In each of the three age-group categories, Black males and females consistently experienced higher stroke mortality rates across each of the social class groups. The exceptions were instances in which 45+ year old Non Hispanic white females in Social Class 5 experienced sl ightly higher stroke mortality rates than Black females. Most d ecedents in this social class group experienced the highest stroke death rates. Social and economic disadvantage is a ssociated with poor health and with increased exposure to risk fact ors for adverse health outcomes.118 It is well known that a number of factors affect a person’s health status, in cluding income, occupation, education, environment, and access to services. It has been further established that an

PAGE 129

116 additional factor, race, also has an impact. The Black-White health disparity may be a function of the overrepresenta tion of Black Americans in lower socioeconomic groups. This fact makes it difficult to ascertain whether health differentials between Black and White Americans will remain when income is held constant. Research Question 3 Given the above mentioned limitations, a cautious investig ation of effect modification by social class category was comp leted. Separate simple linear regression analyses (testing the association between area predictors and racial disparities in stroke mortality) were run for each of the 5 soci al class groups. The existence of effect modification by social class of the associa tion between Area predictors and Black-White disparities in stroke mortality was dependent upon the particular di sparity score and the Area predictor under investigation. Using multiple linear regression to measur e effect modification by social class of the association between the disparity in st roke mortality measures and ‘Area Social Predictors of Health’ variable s, three of the sixteen regres sion models were found to be statistically significant (p<.05). For soci al class group 3, the ASPoH variables were found to be significant predictors of the female Black-White difference score. For social class group 4, the ASPoH variables were found to be significant predictors of the female and male Black-White difference score. When examining the Black-White Male Ratio disparity out come, ASPoH-1 was not shown to be a statistically significant predictor for any of the Social Class groups. However, the parameter estimates increased th e most for the lowest social class group, indicating greater Black-White differences in stroke mortality rates for with less than nine

PAGE 130

117 years of education. The ratio scores were si milarly impacted for those in social class groups 2 and 4. There is an increase in the Bl ack-White male ratio score from the highest social class group to the lowest social class group, indicati ng differential effects based on social class groups. A similar effect on the ma le Black-White ratio scored occurred when measuring the effect of the ASPoH-2 variable. The greatest difference in stroke mortality rates between Blacks and Whites occurred for th ose in social class 5. For the remaining ASPoH predictors, there is a difference in effect on the outcome across social class groups, indicating that the effect of the ASPoH predictors on th e disparity score is social class dependent. Within social class group three, the four ASPoH variables were found to be significant predictors of the female Black-W hite difference score. ASPoH-1, ASPoH-2 and ASPoH-3 were each individually signifi cant predictors of the female difference score. Increases in each of the ASPoH scor es resulted in significant increases in the difference scores. If measures of the ASPoH-1 and ASPoH-3 variables accurately capture economic advantage and disadvantage, the results suggest that social class group 3 residents residing in more economically a dvantaged areas have greater female BlackWhite differences in stroke mortality rates. Within social class group four, the f our ASPoH variables were found to be significant predictors of the male and female Black-White difference score. ASPoH-1, ASPoH-2 and ASPoH-3 were each individually significant predictors of the female difference score, while only ASPoH-2 was an individually significant predictor of the male difference score. Increases in each of the ASPoH scores resulted in significant decreases in the difference scores. Results suggest that social class group 4 residents

PAGE 131

118 residing in more economically advantaged areas, as measured by ASPoH-1 and ASPoH3, have lesser male and female Black-White differences in stroke mortality rates. This study proposed a relationship betw een an ‘area social and economic measure’ and Black-White disparities in stroke mortality after adjustment for social class. Findings from this study are ambiguous, at best. There are no patterns in the results from which one may infer the slightest associati ons. Each area measure had its strongest impact on differing social class group. Agai n, these results are most likely due to the aforementioned issue of the lack of availability of educati onal attainment data. Another possibility may be that the proposed associ ation (between the area measures and BlackWhite disparity in stroke mort ality) may be, instead, mediat ed and/or attenuated through a third unmeasured aspect of the environment related to both race and social class. Disparities in stroke mortality may be refl ective of inequities in the distribution of community resources. Possible area level risk factors for stroke mortality include reduced access to specialized medical care f acilities and physicians in areas of lower socioeconomic status.13 Additionally, stroke patients w ho reside in less affluent areas may not receive emergency treatment in a similarly efficient manner as those who reside in more affluent areas. This increased time to care may be due to the quality of the roads, the accessibility of the stroke patient’s resi dential address by emergency care workers, as well as the number of emergency medical transport providers in the area. Strengths and Limitations A major strength is that I have been able to examine socioeconomic status at a smaller geographic unit than is typically i nvestigated. A composite of a multitude of census tracts level variables was used in or der to calculate the area score instead of

PAGE 132

119 simply using median income and poverty levels. Study results suggest that the new compos ite measure was meaningful. The study was able to show significant associations be tween the composite score and a selection of the Black-White disparity scor es. Possible improvements that could be made in the development of the composite score would incl ude the addition of other census tract level datasets representative of unmeasured dimensions of the social and economic environment. These additions would hopefully produce a measure with a more complete representation of social and economic resources available to census tract residents. The use of multiple disparity measures is a strength of the study design. Both absolute (difference score) and relative (ra tio and percent difference scores) Black-White disparity scores were utilized in this study. An absolute meas ure of disparity is a simple arithmetic difference between a group rate a nd a specified reference point. A relative measure of disparity expresses the differen ce between rates in terms of the chosen reference point. The percentage differen ce expresses the simple difference from the reference point as a percenta ge of the reference point. While their formulae are unique, absolute and relative measures of disparity calculated from the same reference point should lead to the same conclusion (i.e., ha ve the same direction) about disparities between groups. The use of both types of disp arity measures in this study allows for a check of the consistency in the implications of the disparity measures. A particular problem, and limitation, is that results from performing a sequence of analytical comparisons on these disparity scores is that the more comparisons conducted, the more type I errors we will make when the null hypothe sis is true. The type I familywise error rate considers the possibility that one or more type I errors are made in the group of

PAGE 133

120 comparisons. There are possible limitations to study validity with the use of census-based characteristics of residentia l areas in the study of health differentials. The 2000 census data were used to calculate Area predictor scor es for those residents who died from stroke from 1998-2002. Given that the census is take n once per decade, th e appropriateness of appending census data that ar e at least one decade old to records to proxy current socioeconomic characteristics may be in question. 119 Therefore, in some instances, the ASPoH measure was calculated from data only relevant after the resi dent had died. The potential effect on study validity is limited fo r this particular study given that the year 2000 census data is used in order to appr oximate ‘socioeconomic status’ for those residents who died from stroke within th e 1998-2002 time period. Since the census data were collected within the bounda ries of the study time frame, confidence is high in the comparability of the data to the actual resi dential social and economic situation of the Florida stroke decedents. Additionally, the e ffect of this potential bias may be limited due to the findings that socioeconomic ch aracteristics of neighbor hoods generally do not change significantly over such short time periods.119 Determination of the appropriate level of aggregation of the census data in relation to study outcome particulars was a cha llenge. This study used data aggregated at the census tract level, which typically contains 5000 residents. The census tract level was the smallest level of information available for the stroke decedents in cluded in this data. Data aggregated at a smaller more homogene ous geographic level, the census block for example, would have been preferable a nd possibly more informative, but was not available for this study.

PAGE 134

121 As each person will only die once, if there are mutually exclusive causes of death, the causes of death compete with each othe r in the same subject. Competing causes of death may influence any research on either s ubject, resulting in the competing death bias. This study does not attempt to compare different disease mortality rates to one another. Additionally, it is not possible to estima te a difference between Black and White residents in the rates of competing causes of death without death ce rtificate information on the ‘contributing causes.’ There is al so no reason to doubt the accuracy of the recording of stroke as the cau se of death. Furthermore, there is deficient reason to suppose that the underlying cause of death fo r Black and White residents would have been incorrectly categorized at differing rates. Selection bias due to missing data may ha ve occurred in this study. When there are a large number of variables, the regressi on procedure excludes an entire observation if it is missing a value for any of the variables (listwise deletion). This may result in exclusion of a considerable percentage of obs ervations and induce selection bias. In this particular study, missing data may be distri buted differentially between Black and White residents and may generate spur ious associations. In this particular study, it would be more likely that population counts for Blacks are affected more than population counts for Whites, particularly for Black men in th e 35-45 year age-group. The enumeration of this particular demographic group has been shown to be complicated.120 Missing educational attainme nt population counts at the census tract level posed a challenge. Missing educational a ttainment data was more prev alent for Black residents in the higher social class categories. In these instances, no analyses were possible do to the lack of available data. These instances were more likely to occur in census tracts with

PAGE 135

122 small Black populations. Also, gi ven that there are fewer Blac ks within the higher social class groups in general, this lack of repor ting was likely to occur within most census tracts and may not be specific to small population census tracts. Another possible limitation is the ques tion of appropriateness in using the boundaries of a census tract as a proxy to the boundaries within which resources are available. Those residents with more res ources will be able to avail themselves of additional resources outside of these boundaries. The soci al structure may be more extensive for the more affluent. The abil ity to travel and work outside of one’s immediate residential space may not be capture d by this resource availability measure. Aggregate level analyses are often critic ized for being subjec t to the ecological fallacy. Consideration should be given to the possibility that analys es at the individual level may be inappropriate when seeking to determine aggregate level social and economic correlates of health and illness.118 This study was correlational, and has the expected challenges of nonra ndomized studies. These limitati ons include selection biases and confounding by uncontrolled variables. In this instance, indivi duals within census tracts could not be assigned into socio economic groups, and, therefore, randomization was not possible. In addition, the calculated ar ea resource availabili ty measure is only a proxy for level of area economic and social we llbeing. However, the association between the calculated measure and racial stroke mo rtality rates is similar to findings from a multitude of studies using SES measures such as employment, income and education. An area SES score derived from census data is currently the only available data recorded and stored on a regular basis. Utilization of this type of data relies on the assumption that area of residence may provide additional information on social position

PAGE 136

123 connoting an aspect of status that is not captured by indi vidual SES measures. Possible correlation between individual study variab les composing the ASPoH score must be taken into consideration. Also, the associa tion between area SES and disparities in stroke mortality may be affected if census tract aggr egates differ greatly in their socioeconomic heterogeneity. Areas are not internally homogeneous, and census tracts containing a mixture of deprived and less deprived househol ds will have a middle ranking. The scores from census tracts with small populations (or rural areas) are more susceptible to small variations. There is potential problem with the analys es of the restricted subset of census tracts for research question one. Restricting the analyses to those census tracts where neither the Black stroke mortality rate nor the NH-White stroke mortality rate was equal to zero possibly resulted in the exclusion of those cen sus tracts with either very large NHWhite populations or very large Black populatio ns. Excluding those census tracts with large NH-White populations possibly resulted in the exclusion of the most affluent census tracts, whereas the exclusion of those census tracts with predominately Black populations likely resulted in the exclusion of the poorest census tracts. Consequently, the range of economic levels of the census tracts included in the restricted analyses was limited. A potential problem with u tilizing educational attain ment data obtained from death certificates is the possi bility that family members may report a higher level of educational attainment on the death certifi cate than actually achieved. Also, economic conditions make it extremely difficult, and, ther efore, less likely for poor people to live in affluent areas, resulting in a small number of poor people residing in these areas. There is also the expectation that very few rich pe ople reside in disadva ntaged areas. These

PAGE 137

124 situations potentially limit the study’s abil ity to detect a potential interaction between individual SES and ASPoH on racial disparities in stroke mortality. This study proposed that area socioeconomic structure contributes to and/or limits life choices ultimately leading to poor health outcomes. Because this study utilized ecologic data, we must take into considerati on the possibility that poor health led people to move to more-deprived areas. Economic c onditions influence reside nce in affluent and poor areas. Those who reside in poorer neighborhoods tend to have poorer health, an effect that is exacerbated in Blacks. Consistency with the Literature More affluent areas (as measured by the ASPoH-2 variable) were associated with smaller Black-White disparity scores. Consiste nt with the literature, within each racial group, residents in low SES areas experien ced increased stroke mortality rates.12 Results demonstrated higher stroke mortality rates for disadvantaged areas94 and higher rates for Black residents compared to Non-Hispanic Wh ite residents, a findings also consistent with the reported literature. As seen in previous research findings, Black males experienced the highest stroke mortality ra tes, followed by Black females, White males and females, respectively.9 Additionally, Black stroke decedents tended to be younger that White stroke decedents. Consistent with the literature, Black decedents also tended to have less education and were less likely to continue their education beyond high school and were also less likely to have ever been married.97 Inconsistencies of my study findings with the literature include findings that those Black residents who attained a high school de gree have the highest stroke death rates compared to all other educational attain ment groups. These study results were in

PAGE 138

125 opposition to the proposed study hypothesis a nd with previous research findings.11 The opposite occurred for NH-White residents where stroke mortality rates were highest for those in the lowest social class group (less than 9 years of education). For a restricted subset of census tr acts, more complete population count and stroke death count data were available. When only those census tracts in which there exists a non-zero age adjusted stroke death rate for both Black and White residents are examined, study results are less variable. The reasoning for the different pattern of results that is observed between Blacks and Whites coul d be due to the conjecture that Blacks living in the same census tract as Whites ma y not have access to the same resources as White residents. Additiona lly, results support the conjectu re that Blacks and Whites may not actually share immediate environments within the same census tract. How may the environments for Blacks diffe r from the environments of Whites? Studies have shown that hazardous material dumpsites are more likely to be located in Black neighborhoods.121 Additionally, counties with a higher percentage of Black residents and high rates of inco me inequality tend to have a higher proportion of chemical intensive facilitie s located within county boundaries.122 Black Americans are disproportionately likely to be exposed to air toxins123 and to reside clos er to the nearest industrial emission facility.124 More than poverty, home ownership or land value, race was found to be a stronger predicto r of hazardous facility placements.123 Ramification of Blacks disproportionately residing nearer to hazardous and higher risk facilities include the burden of disproportionate health risks, possibly resulting in increased Black-White disparities in adverse health outcomes. Calculating Area predictor scores for censu s tracts in which those resources for

PAGE 139

126 Whites drive the magnitude of the composite sc ores may not at all be representative of the quantity of resources Black residents actually possess. If the salaries of Whites are vastly higher than that of Blacks residing in the same census tracts, the SES status of Black residents will have very little impact on the magnitude of that composite score. Black and White workers have di fferent income potentials. White males are at least two times more likely than Black males to be em ployed in management, business or finance positions. White females are 1.5 times as likel y as Black females to possess employment in these fields. Black males and females are 2 and 1.5 times more likely to be in service oriented jobs. Therefore, Black and White residents of the same area may have inequalities in income accompanied by many diffe rences in conditions of life, both at the individual and population level, whic h may adversely influence health.125 Perhaps separate composite scores should be calculated for Blacks and Whites for each of the census tract and examined to determine if race specific composites are influentially comparable to the composite that is not race specific. If the composites are not comparable, it may be inappropriate to a ssume that individuals living within the same neighborhood have access to the same re sources. For instance, study findings demonstrate less physical activity among low-income housing units.126 These finding potentially result from the likelihood that th ese areas not supportive of physical activity for the purposes of exercising. If Blacks are more likely to live in low income areas, adverse health outcomes associated with physi cal inactivity may dispr oportionately affect the Black population. Differential rates of large food store chains by neighborhood characteristics, such as proportion of Black population, may also contribute to the racial disparities in adverse health outcomes. Pr edominantly White neighborhoods tend to have

PAGE 140

127 more supermarkets per household compar ed to predominately Black neighborhoods.127 Large food store chains are more likely to o ffer healthy foods at affordable costs than small food chains. Difficulty obtaining hea lthy foods due to the lack of available supermarkets may result in unhealthy dietary pa tterns which could lead to increased risk of disease.128 It is possible that equalization of financ ial access may not ensure receipt of equal quality treatment. Policies to address unfavor able social conditions impacting health are needed. Such policies could include reduction of income inequality through tax reform, improved housing, and expanded educational and employment opportunities for the poor.129 Understanding health from a social pers pective is important if appropriate interventions and policies are to be devel oped to eliminate disparities. This study analyzed Black-White disparities in stroke mortality from a social perspective that supports the assumption that health disparity among Blacks is related to unequal access to community resources. The key to decrea sing the disparity is the development and implementation of policies that ensure equal access and equal treatment.130 Public Health Implications Study findings suggest that raci al disparity scores are el evated in deprived areas, and in some instances, even more so for lowe r social class groups. This suggests that initiatives to lessen Black-White mortality inequalities will need to address an individual’s social class situation, while taking into account the role of residential environment in exacerbating and possibly ove rshadowing the effect of personal poverty. Study results suggest a change in the scope of interventions from a bi omedical individual

PAGE 141

128 level to interventions focusing on social determ inants of health. Progress must be made to address the disjunc tion between individual f actors and social determinants that impact racial disparities in stroke mortality. Recognizing the impor tance of the distribution of resources (as a measure of deprivation and wealth) among social groups is crucial to explaining the distribution of disease in populations and planning effective health interventions. Local-level health policies must be deve loped with the hope of improving social, economic, physical and environmental conditions in the community that affect reducing Black-White health disparities. Efforts mu st be made to insure that all community members not only have access to medical servic es, but are additionally in a position to take advantage of these health services. Local government health officials must communicate with community members with the hopes of identifyi ng barriers to and facilitators of the reception of available medi cal services. Strategies must be developed to increase access to h ealthcare services. Changes in the health care system must be implemented in order to reduce disparities in adverse health outcomes. Proposed examples of beneficial change in health care include health insurance coverage for all, and racial equality in the receipt of proper medical interventions. A health care system with adequate repres entation of African American health professionals may also provi de a positive impetus for reduction of race based health disparities by providing a more culturally se nsitive, and therefore more effective, health care system.131 Area specific local health e ducation programs must be in itiated. Health officials must direct education efforts to specific communities within levels of socioeconomic

PAGE 142

129 and/or social class. Planning must prom ote the community’s understanding of policies and activities that will improve the community’s health. It is hoped that these steps will lead to a better understanding of local commun ity health issues and how social, economic and environmental conditions affect these health issues. Proposed ch anges related to the health of communities include the promoti on of violence free-neighborhoods conducive to exercise, the addition of nutritious food stores, equality in income, educational and career opportunities.131 Finally, these findings suggest the importance of repeating these analyses at the population level in additional ar eas as complements to analyses of single areas. Future Research More research is needed to gain a be tter understanding of th e mechanisms through which the economic structure of a community in fluence the patterns of health and disease within and between communities. A clea rer understanding, and definition of, the community in which residents live and experi ence life is fundamental. This can only be accomplished through contact with individuals within a defined location, and, thereby, ascertaining the location and geogr aphic extent of social and economic interactions. Data must be compiled concerning community ava ilability of healthy and affordable food stuff, access to recreational facilities, awaren ess of community influences of health and the effectiveness of the communications of heal th related information at the local level. Identification of utilized community resources as well as an understanding of why other resources are underutilized is important. Community barriers to healthy lifestyle opportunities must be acknowledged as well as the promoters of healthy lifestyle opportunities. Future research should develop methods to identify appropriate

PAGE 143

130 populations of study within the most advant ageous graphic unit. Researchers must identify the smallest geographic unit in whic h this type of research can be accomplished and reliable data can be obt ained. Finally, efforts should be made to share research findings with the community, gove rning bodies and policy makers. Conclusion Lower ASPoH scores were predictive of hi gher Black-White disparities in stroke mortality at the Florida cens us tract level. These study results add to established literature solidifying individua l socioeconomic status as a strong predictor of stroke mortality. These study results ar e also a contribution to our knowledge of the history of Black-White disparity in stroke mortality rate s. This disparity research can be extended with the addition of information that streng thens the relationship between SES and stroke mortality by adding in the eff ect of an area measure of SES, and the influence that this measure has on the differences in stroke mortality rates between Black and White residents. With this study we are able to begin exploring census trac t level influences of the actual Black-White disparity rate. The literature suggests that SES does not fu lly account for the r acial disparity in stroke mortality rates, and this study allows for the examination of group level influences of these disparities and attemp ts to find some type of policy resolution to these racial differences in rates. The interrelatedn ess of personal health behavior, social determinants, structural inequi ties, and institutionalized raci sm suggests that eliminating disparities will require large-scale, mu ltidimensional, community-participatory interventions focused explicitly on health di sparities for specific population groups, as well as on broader dimensions of so cial equality a nd economic justice.132

PAGE 144

131 This study allows us to question the need for policy changes in resource availability and allocation at the census trac t level that will make a difference. The primary purpose of this study was to examine th e effect of area resource availability on Black-White disparities in stroke mortality. Results of this study support the conjecture that resource availability-related stroke mortality involves a complex combination of factors from a variety of avenues. This study may have only touched the surface of the influences that we should take into cons ideration when we attempt to measure the community resources that are needed to pr omote and maintain community health and reduce disparities in morbidity and mortality. “…a fundamental social cause (of dis ease) involves resources like knowledge, money, power, prestige, and social connections that strongly influe nces people’s ability to avoid risks and to minimize the consequen ces of disease once it occurs. Because of the very general utility of these social and economic resources, fundamental causes are linked to multiple disease outcomes thr ough multiple risk-factor mechanisms…In a dynamic system, fundamental causes are likely to emerge. This is because the resources embodied in fundamental causes can be transp orted from one situation to another. Consequently, as health-related situations ch ange, those with the most resources are best able to avoid diseases and their conseque nces. Thus, no matter what the profile of diseases and known risks happens to be at a ny given time, those who have greater access to important social and economic resour ces will be less afflicted by disease.”133 Black-White disparities in stroke mortality present a major challenge for which effective action must focus on the social and economic environment. Analyses of individual risk may not provide useful inform ation. Therefore, it is imperative that

PAGE 145

132 researchers continue the search for modifiable aspects of the society for which changes in both policy and attitudes may be the key to unlocking the basis of the disparities in health outcomes that have existed since data such as these have been maintained.

PAGE 146

133 List of References 1. McGrath JJ, Matthews KA, Brady SS. Individual versus neighborhood socioeconomic status and race as predictors of adolescent ambulatory blood pressure and heart rate. Soc Sci Med 2006;63(6):1442-53. 2. Wen M, Hawkley LC, Cacioppo JT. Objective and perceived neighborhood environment, individual SES and psychosoc ial factors, and se lf-rated health: an analysis of older adults in Cook County, Illinois. Soc Sci Med 2006;63(10):257590. 3. O’Malley DP, Johnston L. Racial/ethnic a nd socioeconomic status differences in overweight and health-related behavi ors among American students: National Trends 1986-2003. J Adolescent Health 2006;39(4):536-545. 4. Macintyre S, Ellaway A, Cummins S. Place effects on health: how can we conceptualise, operationalise and measure them? Am J Public Health 2002;55:125-139. 5. Hu G, Sarti C, Jousilahti P, Peltonen M, Oiao Q, Antikainen R, Tuomilehto J. The impact of history of hypertension a nd type 2 diabetes at baseline on the incidence of stroke and stroke mortality. Stroke 2005;36(12):2538-43. 6. Elkind MS, Sacco RL. Stroke risk factors and stroke prevention. Seminars in Neurology 1998;18(4):429-440. 7. Marmot MG, Syme SL, Kagan A, Kato H, Cohen JB, Belsky J. Epidemiologic studies of coronary heart disease and stroke in Japa nese men living in Japan, Hawaii and California: prevalence of cor onary and hypertensive heart disease and associated risk factors. Am J Epidemiol 1975;102(6):514-25. 8. Howard G, Anderson R, Sorlie P, et al. Ethnic differences in stroke mortality between non-Hispanic whites, Hispanic whites, and blacks: the National Longitudinal Mortality Study. Stroke 1994;25:2120-2125. 9. American Heart Association. Heart and Stroke Statistics — 2006 Update Dallas, Texas: American Heart Association; 2006.2006, American Heart Association.

PAGE 147

134 10. Franks P, Muennig P, Lubetkin E, Jia H. The burden of disease with being African American in the United States and the contribution of socio-economic status. Soc Sci Med 2006;62:2469-2478. 11. Casper ML, Barnett EB., Armstrong DL, G iles WH, Blanton CJ. Social class and race disparities in premature stroke mo rtality among men in North Carolina. Ann Epidemiol. 1997;7:146-153. 12. Cox AM, McKevitt C, Rudd AG, Wolfe CDA. Socioeconomic status and stroke. Lancet Neurol 2006;5:181-88. 13. Kapral MK, Wang H, Mandani M, Tu J. Effect of socioeconomic status on treatment and mortality after stroke. Stroke 2002;33:268-75. 14. Engstrom G, Jerntorp I, Pessah-Rasmusse n H, et al. Geographic distribution of stroke incidence within an urban population. Stroke 2001;32:1098-1103. 15. Winkelby MA, Cubbin C. Influen ce of individual and neighborhood socioeconomic status on mortality among black, Mexican-American, and white women and men in the United States. J Epidemiol Community Health 2003;57:444-452. 16. Smits J, Westert GP, van den Bos GAM. Socioeconomic status of very small areas and stroke incidence in the Netherlands. J Epidemiol Community Health 2002;56:637-640. 17. Maantay J. Zoning, equity, and public health research. Am J Public Health 2001;91:1033-1041. 18. Faber DR, Krieg EJ. Unequal exposure to ecological hazards; environmental injustices in the commonwealth of Massachusetts. Environ Health Perspect 2002;110(suppl 2):277-288. 19. Williams DR, Collins C. Racial residentia l segregation: a fundamental cause of racial disparities in Health. Public Health Reports 2001;116:404-416. 20. Collins CA, Williams DR. Segregation and mortality: the deadly effects of racism? Sociological Forum 1999;14:495-523. 21. Polednak AP. Poverty, residential segrega tion, and black/white mortality rates in urban areas. J Health Care Poor Underserved 1993;4:363-73. 22. Diez-Roux Av, Merkin SS, Arnett D, Chambl ess L, Massing M, Nieto FJ, et al. Neighborhood of residence and incide nce of coronary heart disease. N Engl J Med 2001;345:99-106.

PAGE 148

135 23. Fang J, Madhavan S, Bosworth W, Alde rman MH. Residen tial segregation and mortality in New York City. Soc Sci Med 1998;47:469-76. 24. LaVeist TA. Linking residential segrega tion and infant mortality race disparity in U.S. cities. Sociol Soc Res 1989;73:90-4. 25. James SA. Primordial prevention of cardiovascular disease among AfricanAmericans: a social epidemiological perspective. Preventive Medicine 1999;29:S84-S89. 26. Fuher R, Shipley MJ, Chastang JF, Schm aus A, Niedhammer I, Stansfeld SA, Goldberg M, Marmot MG. Socioec onomic position, health, and possible explanations: a tale of two cohorts. Am J Public Health 2002;92(8):1290-4. 27. Bennett GG, Wolin KY, Puleo E, Emmons KM. Pedometer-determined physical activity among multiethnic low-income housing residents. Med Sci Sports Exerc 2006;38(4):768-73. 28. Satterthwaite D. The impact on health of urban environments. Environ Urban 1993;5:87-111. 29. Bostock L. Pathways of disadvantage? Walking as a mode of transport among low-income mothers. Health Soc Care Community 2001;9(1):11-8. 30. Brown E, Wyn R, Teleki. Disparities in health insurance and access to care for residents across U.S. cities. The Commonwealth Fund and UCLA Center for Health Policy Research 2000:website www.cmwf.org accessed August 11, 2006. 31. Will JC, Farris RP, Sanders CG, Stockm yer CK, Finkelstein EA. Health promotion interventions for disadva ntaged women: overview of the WISEWOMAN projects. J Womens Health 2004;13:484-502. 32. Doty MM, Holmgren AL. Health care disconnect: gaps in coverage and care for minority adults. Findings from th e Commonwealth Fund Biennial Health Insurance Survey (2005). IssueBrief (Commonw Fund) 2006;21:1-12. 33. Kleindorfer DO, Lindsell CJ, Broderick JP, Flaherty ML, Woo D, et al. Community socioeconomic status and pr ehospital times in acute stroke and transient ischemic attack: do poorer patient s have longer delays from 911 call to the emergency department? Stroke 2006;37(6):1508-13.

PAGE 149

136 34. Chuang Y-C, Cubbin C, Ahn D, Wi nkleby MA. Effects of neighborhood socioeconomic status and convenience st ore concentration on individual level smoking. J Epidemiol Community Health 2005;59:568-573. 35. Cummings KM, Giovano G, Mendicino AJ. Cigarette advertising and racial differences in cigarette brand preference. Public Health Report 1987;102:698701. 36. Pollack CE, Cubbin C, Ahn D, Winkle by MA. Neighborhood deprivation and alcohol consumption: does the availa bility of alcohol play a role? Int J Epidemiol 2005;34:772-780. 37. Humpel N, Owen N, Leslie E. Environm ental factors associated with adults’ participation in physical activity: a review. Amer J Prev Med 2002;22:188-199. 38. Owen N, Humpel N, Leslie E, Ba uman A, Sallis JF. Understanding environmental influences on walking; Review and research agenda. Amer J Prev Med 2004;27:67-76. 39. Morland K, Wing S, Diez Roux A, P oole C. Neighborhood characteristics associated with the location of f ood stores and food service places. Am J Prev Med 2002;22:23-29. 40. Mooney C. Cost and availability of healthy food choices in London health district. J Hum Nutr Diet 1990;3:111-120. 41. Morland K, Wing S, Diez Roux A. The contextual effect of the local food environment on residents’ diets: the at herosclerosis risk in communities study. Am J Public Health 2002;92:1761-1767. 42. LaVeist TA. Disentangling race and socioeconomic status: a key to understanding health inequalities. J Urban Health 2005;982(2):iii26-iii34. 43. American Stroke Association. What are th e types of stroke? Retrieved July 30, 2005 from http://www.strokeassociation.org/pr esenter.jhtml? identifier=1014 44. Toole JF. Cerebrovascular Disorders 5th ed. 1999, Philadelphia, PA: Lippincott Williams & Wilkins. 45. American Heart Association. Heart Disease and Stroke Statistics -2004 Update 2004, American Heart Association: Dallas, Texas. 46. American Heart Association. Heart Disease and Stroke Statistics -2005 Update 2005, American Heart Association: Dallas, Texas.

PAGE 150

137 47. Tayor TN, et al. Lifetime cost of stroke in the United States Stroke 1996; 27:1459-1466. 48. Foulkes MA, et al. The stroke data bank: design, methods, and baseline characteristics. Stroke 1988;19:547-554. 49. Kleindorfer D, Broderick J, Khoury J, et al. The unchanging incidence and casefatality of stroke in the 1990s. A population-based study. Stroke 2006;37:24732478. 50. Stroke risk factors. (n.d.), Retrieved June 23, 2005 from http://www.americanheart.org/pr esenter.jhtml? identifier=9217 51. Lawes CM, et al. Blood pressure and str oke: an overview of published reviews Stroke 2004;35:776-785. 52. Goldstein LB, et al. Primary prevention of ischemic stroke: a statement for healthcare professional from the Str oke Council of the American Heart Association Stroke 2001;32:280-299. 53. Lukovits TG, Mazzone T, Gorelick PB. Diabetes melllitus and cerebrovascular disease Neuroepidemiology 1999;18:1-14. 54. Wolf PA, et al. Cigarette smoking as a ri sk factor of stroke. The Framingham study JAMA 1988;259:1025-1029. 55. MMWR, Vol. 48, No. 43, 1999, CDC/NCHS 56. Physical Activity and Health, U.S. Surgeon General's Report, 1996 57. Hu FB, et al. Physical activity and risk of stroke in women JAMA. 2000;283: 2961-2967. 58. Flegal KM, et al. Prevalence and trends in obesity among US adults JAMA 2002; 288:1723-1727. 59. Kurth T, et al. Body mass index and the risk of stroke in men Archives of Internal Medicine 2002;162:2557-2562. 60. Rodriguez BL, et al. Risk of hospitalized stroke in men enrolled in the Honolulu Heart Program and the Framingham Study A comparison of incidence and risk factor effects Stroke. 2002;33:230-237. 61. Walker SP, et al. Body size and fat distri bution as predictors of stroke among US males. Am J Epidemiol 1996;144:1143-1150.

PAGE 151

138 62. Tanne D, Medalie JH, Goldbourt U. Body fa t distribution and long-term risk of stroke mortality Stroke 2005;36:1021-1025. 63. Baum H.M. Stroke prevalen ce: an analysis of data from the 1977 National Health Interview Survey Public Health Reports 1982;97:24-30. 64. Tuomilheto J, et al. Effectiveness of tr eatment with antihypertensive drugs and trends in mortality from stroke in the community BMJ 1985;291:857-861. 65. Klag MJ, Whelton PK, Seidler AJ. Decline in US stroke mortality: demographic trends and antihypertensive treatment Stroke 1989;20:14-21. 66. Casper M, et al. Antihypertensive treatm ent and US trends in stroke mortality Am J Public Health 1992;82:1600-1606. 67. Mason TJ, et al. An Atlas of Mortality from Selected Diseases. 1981, Washington DC, US: US Government Printing Office. 68. Lanska DJ. Geographic distribution of st roke mortality in the United States: 1939-1941 to 1979-1981 Neurology 1993;43:1839-1851. 69. Gaines K. Regional and ethnic differences in stroke in the southeastern United States population. Ethnicity and Disease. 1997;7:150-164. 70. Kuller L, et al. Nationwide cerebro vascular disease morbidity study. Stroke. 1970;1:86-99. 71. Wing S, et al. Stroke mortality maps. United States whites aged 35--74 years, 1962--1982. Stroke 1988;19:1507-1513. 72. Health issues in the black community ed. R.L. Braithwaite and S.E. Taylor. 2001, San Francisco: Jossey-Bass. 73. Florida CHARTS. Florida Mortality Atlas. Retrieved January 29, 2007 from http://www.floridacharts.com 74. National Center for Hea lth Statistics. Health, United States, 2005 With Chartbook on Trends in the Health of Americans. 2005: Hyattsville, Maryland 75. Keppel KG, Pearcy JN, Wagener DK. Trends in racial and ethnic-specific rates for the health status indicators: Un ited States, 1990--98. Healthy people statistical notes, no 23. National Center for Health Statistics. January 2002: Hyattsville, Maryland.

PAGE 152

139 76. Krieger N, Basset M. The health of black folk: disease, class and ideology in science. Monthly Review 1986; 38:74-85. 77. Courtenay WH. Construction of mascu linity and their influences on men's wellbeing: a theory of gender and health. Soc Sci Med 2000;50:1385-1401. 78. National Center for Hea lth Statistics. Health, United States, 2004 With Chartbook on Trends in the Health of Americans. 2004: Hyattsville, Maryland. 79. Freid VM, et al. Chartbook on Trends in the Health of Americans, Health, United States, 2003. 2003, National Center for Health St atistics: Hyattsville, Maryland. 80. Bohani N. Changes and geographic distribution of mortality from cerebrovascular diseases. Am J Public Health 1965;55:673-681. 81. Kuller L, Reisler DM. An explanation for va riations in distribution of stroke and arteriosclerotic heart disease among populations and racial groups. Am J Epidemiol 1971;93:1-9. 82. National Center for Health Statistics. Monitoring the nation's health. Fastats A to Z (2005). Retrieved June 23, 2005, from htttp://www.cdc.gov/nchs/fastats/stroke.htm 83. Flack JM, et al. Ethnicity and renal disease: lessons from the Multiple Risk Factor Intervention Trial and the Tr eatment of Mild Hypertension study. Am J Kidney Disease 1993;21:31-40. 84. Burt VL, et al. Prevalence of hypertensi on in the U.S. adult population: results from the Third National Health and Nutrition Examination Survey, 1988-1991. Hypertensio., 1995;25:305-313. 85. Wolf PA, et al. Preventing ischemic str oke in patients with prior stroke and transient ischemic attack: a statement for healthcare professionals from the Stroke Council of the Amer ican Heart Association. Stroke. 1999;30:1991-1994. 86. Coresh J, Jaar B. Further trends in th e etiology of end-stag e renal disease in African-Americans. Curr Opin Nephrol Hypertens 1997;6:243-249. 87. Diez-Roux A., et al. Neighborhood environmen ts and coronary heart disease: a multilevel analysis. Am J Epidemiol 1997;146:48-63. 88. Borrell LN, et al. Neighbourhood char acteristics and mortality in the Atherosclerosis Risk in Communities study. Int J Epidemiol 2004;33(2):398407.

PAGE 153

140 89. Smith GD, et al. Individual social cla ss, area-based deprivation, cardiovascular disease risk factors, and mortal ity: the Renfrew and Paisley study. J Epidemiol Community Health. 1998;52:399-405. 90. Steenland K, et al. Individua land area-level socioeconomic status variables as predictors of mortality in a cohort of 179,383 persons. Am J Epidemiol. 2004; 159: 1047-1056. 91. Wen M, Browning CR, Cagney KA. Povery, affluence, and income inequality: neighborhood economic structure and its implications for health. Soc Sci Med 2003;57:843-860. 92. Lawlor DA, et al. Life-course soci oeconomic position, area deprivation, and coronary heart disease: findings from the British Women's Heart and Health Study. Am J Public Health 2005; 95(1):91-97. 93. van Lenthe FJ, et al. Neighbourhood unemp loyment and all cause mortality: a comparison of six countries. J Epidemiol Community Health. 2005;59: 231-237. 94. Aslanyan S, et al. Effect of area based deprivation on the severity, subtype, and outcome of ischemic stroke. Stroke 2003;34:2623-2629. 95. Weir NU, et al. Study of the relationship between social deprivation and outcome after stroke. Stroke 2005;36:815-819. 96. Arrich J, Lalouschek W, Mullner M. Influence of socioeconomic status on mortality after stroke. Retrospective cohort study. Stroke 2005;36:310-314. 97. Qureshi AI, et al. Educa tional attainment and risk of stroke and myocardial infarction. Medical Science Monitor. 2003;9(11):CR466-473. 98. Lynch J, Kaplan G. Socioeconomic Position in Social Epidemiology L. Berkman and I. Kawachi, Editors. 2000, Oxford University Press: Oxford. 99. Prager handbook of Black American health : policies and issues behind disparities in health, Ivor Lensworth Livingsto n, Editors. 2004, Prager Publishers, Westport,CT. 100.Cassel J. The contribution of the social environment to host resistance. Am J Epidemiol. 1976;10:107-123. 101.Krieger N. Epidemiology and the web of causation: has anyone seen the spider? Soc Sci Med 1994;39:887-903.

PAGE 154

141 102.Krieger N. Discrimination and health in Social epidemiology L. Berkman and I. Kawachi, Editors. 2000, Oxford University Press: Oxford. 103.Levins R, Lopez C. Toward an ecosocial view of health. Int J Health Services. 1999;29(2):261-293. 104.Stallones RA. Epidemiologists as environmentalists. Int J Health Services. 1973;3:29-33. 105.Crawford R. You are a danger to your heal th: the ideology and politics of victim blaming. Int J Health Services. 1977;7:663-680. 106.Stallones RA. To advance epidemiology. Ann Rev Public Health 1980;1:69-82. 107.Macleod J, Smith GD. Psychosocial factor s and public health: a suitable case for treatment? J Epidemiol Community Health 2003;57:565-570. 108.Census Tracts and Block Numbering Areas. Retrieved January 4, 2006 from http://www.census.gov/geo/www/centract.html 109.Harper S, Lynch J. Methods for Measuring Cancer Disparities: Using Data Relevant to Healthy People 2010 Cancer-Related Objectives. NCI Cancer Surveillance Monograph Series, Number 6. Bethesda, MD: National Cancer Institute, 2005. NIH Publication No. 05-5777. 110.Hillemeier MM, Lynch J, Harper S, and Casper. Measuring contextual characteristics for community health. Health Services Research 2003;38(6 part 2):1645-1717. 111.Principal Components Analysis Retrieved July 22, 2005 from http://www.fon.hum.uva.nl/praat/manual/P rincipal_Component_Analysis.html 112.Principal Components and Factor Analysis. Retrieved July 22, 2005 from http://www.pfc.forestry.ca/profiles/wulder/mvstats/pcafae.html 113.DeNavas-Walt C, Proctor BD, Mills RJ. U.S Census Bureau, Current Population Reports, P60-226, Income, Poverty, and Health Insuranc e Coverage in the United States: 2003, U.S. Government Pr inting Office, Washington, DC, 2004. 114.Stafford M, Marmot M. Neighborhood depr ivation and health: does it affect us all equally. Int J Epidemiol. 2003;32:357-366. 115.Kindig, DA, Seplaki CL, Libby DL. Death rate variation in US subpopulations. Bull World Health Organ, 2002, vol 80(1):9-15.

PAGE 155

142 116.Rodgers, GB. Income and inequality as determinants of mortality: an international cross-section analysis. Popul Stud 1979;33:343-351. 117.Lynch JW, et.al. Income inequality a nd mortality in metropolitan areas of the United States. Am J Public Health. 1998; 88(7)1074-1080. 118.Marmot MG. Improvement of social environment to improve health. The Lancet .1998;351:57-60. 119.Geronimus AT, Bound J. Use of census-ba sed aggregate variab les to proxy for socioeconomic group: evidence from national samples. Am J Epidemiol. 1998;148:475-486. 120. What is the role of demographic analy sis in the 2000 Unite d States census? Last accessed February 26, 2007 from http://www.census.gov/population/www/documentation/1996/symposium96.html 121.Davidson 2000: Davidson P, Anderton DL. Demographics of dumping. II: A national environmental equity survey a nd the distribution of hazardous materials handlers. Demography. 2000 Nov; 37(4):461-6. 122.Elliot MR, Wang Y, Lowe RA, Kleindor fer PR. Environmental justice: frequency and severity of US chemical industry accidents and the socioeconomic status of surrounding communities. J Epidemiol Community Health. 2004;58:2430. 123.Brown P. Race, class, and environmenta l health: a review and systematization of the literature. Environmental Research 1995;69:15-30. 124.Perlin S, Wang D, Sexton K. Residentia l proximity to industrial sources of air pollution: interrelationships among race, poverty, and age. J Air Waste Manag Assoc. 2001;51:406-21. 125.Lynch JW, Smith GD, Kaplan GA, House JS Income inequality and mortality: importance to health of individual income, psychosocial environment, or material conditions. BMJ 2000;320:1200-1204. 126.Bennett GG, Wolin KY, Puleo E, Emmons KM. Pedometer-determined physical activity among multiethnic low-income housing residents. Medicine and Science in Sports and Exercise 2006;38(4):768-773. 127.Shaffer A. The persistence of L.S.’s gr ocery gap: The need for a new food policy and approach to market development. Center for Food and Justice, Urban and Environmental Policy Institute (UEPI), Occidental College.

PAGE 156

143 128.Powell LM, Slater S, Mirtcheva D, Bao Y, Chaloupka FJ. Food store availability and neighborhood characteristics in the United States. Preventive Medicine 2006, doi:10.1016/j.ypmed.2006.08.008. 129.Lasser KE, Himmelstein DU, Woolhandler S. Access to care, health status and health disparities in the United States and Canada: results of a cross-national population-based survey. Am J Public Health 2006;96(7):1300-1307. 130.Plowden KO, Thompson LS. Sociological perspective of black American health disparity: implications for social policy. Policy Politics Nurs Prac 2002;3(4):325-332. 131.Satcher D, Fryer GE, McCann, et al. Wh at if we were equa l? A comparison of the black-white mortality gap in 1960 and 2000. Health Affairs 2005;24(2):459464. 132.Williams DR. Racial/et hnic variations in women’s health: the social embeddedness of health. Am J Public Health 2002;92(4):588-97. 133.Link BG, Phelan JC. Understanding so ciodemographic differences in health— the role of fundamental social causes. Am J Public Health 1996;86:471-473.

PAGE 157

144 Appendices

PAGE 158

145 Appendix A Residential Address Geocoding Methods Personal Communications with Bill Alfr ed, Florida Department of Health Tallahassee, FL Date: August 8, 2005 Regarding: Geo-coding of death certificate residential addresses Florida Department of Health began geocoding death certifi cates using deaths reported in 1995. Death certificates are moved into Access from Sequale Server Address correction software (Accumail) is then used to correct addresses to the postal address This process provides the Zip code + 4 digits, if possible Not all addresses can be corrected The Geo_result variable is an indication of how well the Accumail sort performed Addresses may be passed through Accumail again Accuracy for Accumail is 90-95% Addresses are then sent through Map Make r Plus, which provides latitude and longitude information From this information census tract info rmation and geo_result information can be obtained o S5 as a georesult: Most accurate o Z5 as a georesult: coded to the zip+4; exact CT may or may not be good Data results from the Accumail sort is then run through the Map Marker Plus software in 3 to 4 batches. The diffe rence between batched is the level of strictness utilized and the criteria is loosened for each successive batch. Usually take the results that get S5 as a geo_result This geo-coding is perfor med on a statewide basis Total death certificate records in which geo-coding was attempted o 1998: 157,172 o 1999: 162,152 o 2000: 162,840 o 2001: 161,974 o 2002: 163,024 Accuracy in the geo-coding of death cer tificate residential addresses for 19982002 o 1998: 93.7% o 1999: 93.0% o 2000: 93.3% o 2001: 87.0% o 2002: 94.1%

PAGE 159

146 Appendix B Definition of Study Variables Exposures: Black or African American. A person having origins in any of the Black racial groups of Africa. It includes peopl e who indicate their race as ‘‘B lack, African Am., or Negro,’’ or provide written entries such as African American, Afro-American, Kenyan, Nigerian, or Haitian. White. A person having origins in an y of the original peoples of Europe, the Middle East, or North Africa. It includes people who indi cate their race as ‘‘Wh ite’’ or report entries such as Irish, German, Italian, Lebane se, Near Easterner, Arab, or Polish. Area. For the purposes of this study, ar ea is defined as a census tract ASPoH. Socioeconomic conditions define the cont ext within which the distributions of physiological and behavioral risk factors are determined. ASPoH describes features of social organization, structure, stratificati on, or environment, such as socioeconomic deprivation, economic inequalit y, resource availability, or op portunity structure. This ASPoH variable is a linear combination of the original census tract level variables subjected to principal component analysis. ASPoH-1 is principal component number 1 ( accounts for the highest percentage of variance within the cens us tract level variables) and thus is a linear combination of the original census tract level variables derived from principal component analysis. ASPoH-2 is principal component number 2 (accounts for the second highest percentage of variance within the census tract level variables) and thus is a linear

PAGE 160

147 combination of the original census tract level variables derived from principal component analysis. ASPoH-3 is principal component number 3 (accounts for the third highest percentage of variance within the census tract level variables) and thus is a linear combination of the original census tract level variables derived from principal component analysis. ASPoH-4 is principal component number 4 (accounts for the fourth highest percentage of variance within the census tract level variables) and thus is a linear combination of the original census tract level variables derived from principal component analysis. Since ASPoH-1, ASPoH-2, ASPoH-3 and A SPoH-4 are derived from principal component analysis, they are, by de finition, new independent variables. Census Tract. Census tracts are small statistical subdivisions of a county designed to be relatively permanent. The goal is for census tr acts, when originally designated, to have between 2,500 and 8,000 people and to be hom ogeneous with respect to population characteristics, economic status, and living conditions. Census tracts never cross county boundaries. Census tract size va ries depending on the density of the population. They are designed to be fixed to allow comparisons over time but are occasionally split or combined to reflect significant changes in geography (such as the construction of an interstate) or population (rapid growth). Dimensions of Social Determinants of Health Economy Dimension 1. Poverty Rate. To determine a person’s poverty status, one compares the

PAGE 161

148 person’s total family income with the poverty threshold appropriate for that person’s family size and composition. If the total income of that person’s family is less than the threshold appropriate for that family, then the person is considered poor, together with every member of his or her family. If a person is not living with anyone related by bi rth, marriage, or adoption, then the person’s own income is compared with his or her poverty threshold. Poverty rate will be determined as a percenta ge of the total census tract population living in poverty. 2. Median Family Income. The median divides the income distribution into two equal parts: one-half of the cases falling below the median income and one-half above the median. For households and families, the median income is based on the distribution of the total number of households and families including those with no income. The medi an income for individuals is based on individuals 15 years old and over with income. Median income for households, families, and individuals is computed on the basis of a standard distribution. Median income is rounded to the nearest whole dollar. Median income figures are calculated using lin ear interpolation if the width of the interval containing the estim ate is $2,500 or less. If th e width of the interval containing the estimate is greater th an $2,500, Pareto interpolation is used. Employment Dimension 3. Percent Unemployed. All civilians 16 years old a nd over were classified as unemployed if they were neither ‘‘at wo rk’’ nor ‘‘with a job but not at work’’ during the reference week, were looking for work during the last 4 weeks, and

PAGE 162

149 were available to start a job. Also in cluded as unemployed were civilians 16 years old and over who: di d not work at all during the reference week, were on temporary layoff from a job, had been in formed that they would be recalled to work within the next 6 months or ha d been given a date to return to work, and were available to return to work during the reference week, except for temporary illness. 4. Transportation system: This measur e represents the percent of workers aged 16 years or older using various m eans of transportation (public versus private) to travel to work. 5. Full vs. part-time employment: This measure represents the percent of workers who work part-time compared to those workers who have full time employment. Education Dimension. 6. Graduation rates: This measure in cludes the percent of population over 25 years of age without a high school degree Housing Dimension 7. Median Rent. Median gross rent divides the gross rent distribution into two equal parts: one-half of the cases falling below th e median gross rent and one-half above the median. Median gross rent is computed on the basis of a standard distribution 8. Median housing value (often utiliz ed as a measure of wealth). (Median value of owner occupied housing units) 9. Vacancy rates: Percent of housing units vacant

PAGE 163

150 10. Home Ownership: Percent of occ upied housing units that are owner occupied 11. Overcrowded Housing: This value will be determined based on the mean number of persons per room Social Class. For purposes of this study, social class will be based on educational attainment. Educational attain ment was chosen as the measur e of social class due to its availability on the death certificates and the be lief that education is more reliable than the recorded occupation of the decedents. D ecedent occupation may be considered not reliable because spouses sometimes overstate the occupation of their loved ones. Also, the categories may be over inclusive or not specific enough. For example, both a chemical engineer and an assembly-line engin eer would be categorized as engineer, even though there are obvious differences in income and relative position within their respective occupations. Outcomes : Stroke. For year 1998, stroke (cereb rovascular disease) is de fined as code numbers 430 to 438 of the International Cl assification of Diseases (ICD) Ninth Revision. For years 1999-2002, codes I60 to I69 of the ICD Tenth Re vision are used to denote death from stroke. Age-Adjusted Stroke Mortality Rate. Age-adjusted rates are computed by the direct method by applying age-specific rates in a population of interest to a standardized age distributi on (year 2000), in order to eliminate differences in observed rates that result from ag e differences in population composition (National Center for Health Statistics

PAGE 164

151 http://www.cdc.gov/nchs/datawh/nchsdefs /ageadjustment.htm ). Age-adjusted rates are calculated by the direct method as follows: Disparities. Health disparities are differences in the incidence, prevalence, mortality, and burden of diseases and othe r adverse health conditions that exist among specific population groups in the United States (NIH Definition). Black-White Ratio Score : expressed as a quotient and interpreted as the relative magnitude of the Black stroke death rate compared to the NonHispanic White stroke death rate. Black-White Difference Score : the absolute measure of disparity expressed simply as the arithmetic difference between the Black stroke death rate and the Non-Hispanic White stroke death rate (reference point). Black-White Percent Difference Score : the difference between mortality rates (Black minus Non-Hispanic White ) expressed as a percentage of the Non-Hispanic White death rate.

PAGE 165

152 Appendix C: Calculation Strategy for Principa l Component Analyses Variables 1. Poverty rate: number of families within th e census tract below the poverty level Total census tract families 2. Non-poverty rate: 100 minus poverty rate 3. Median Income: median family incomes for census tracts 4. Percent unemployed: (Males and Female s in the labor force and the civilian unemployed) (Males and Females in the labor force (minus those in the armed forces)*100 5. Percent employment: 100 minus percent unemployed 6. Percent full-time employed: (Males a nd Females employed fulltime) (Total population 16 years and older)*100 7. Percent utilizing private transportation to work: (employed persons using private transportation to work) (employed pe rsons using either private or public transportation to work) 8. Percent 25 years and older with High Sc hool education: (Male and Female high school graduates) (Total population 25 years and older) 9. Median rent: census tract me dian rent paid by renters 10. Median home value: Median valu e for owner-occupied housing units 11. Vacancy rate: number of vacant housing units total housing units in the census tract 12. Non-vacancy rate: 100 minus vacancy rate 13. Home ownership rate: number of ow ner occupied housing units number of

PAGE 166

153 occupied housing units w ithin the census tract 14. Overcrowded housing rates: (number of ow ner/renter occupied housing units with 2.01 or more occupants per room Total occupied housing units) *100 15. Non-Overcrowded housing rates: 10 0 minus overcrowded housing rates

PAGE 167

154 Appendix D Study Acronyms 1. ASPoH: Area Social Predictors of Health 2. BFAAdeathrate: Black female ag e-adjusted stroke death rate 3. BMAAdeathrate: Black male age-adjusted stroke death rate 4. CVD: cardiovascular disease 5. FPD: Female Percent Difference 6. MPD: Male Percent Difference 7. NH: Non -Hispanic 8. NHWFAAdeathrate: Non-Hisp anic White female age-adjusted stroke death rate 9. NHWMAAdeathrate: Non-Hispanic White male age-adjusted stroke death rate 10. PCA: Principal Component Analysis 11. SES: Socioeconomic Status 12. SF3: Summary File 3 13. SF4: Summary File 4

PAGE 168

155 About the Author Tyra Dark obtained a Bachelors degree in Psychology from the University of Alabama, located in Tuscaloosa, Alabama. She received a Masters degree in Psychology from the University of South Florida. Her Masters work, within the Cognitive and Neural Sciences program, invol ved the study of drug effects a nd drug interactions. Prior to beginning the doctoral program in the Ep idemiology and Biostatistics department, she conducted research of central nervous system responses to injury and worked as a Biological Scientist.


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 001916914
003 fts
005 20071114113803.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 071114s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002014
035
(OCoLC)181158818
040
FHM
c FHM
049
FHMM
090
RA425 (ONLINE)
1 100
Dark, Tyra.
0 245
Impact of area social predictors of health on Black-White disparities in stroke mortality
h [electronic resource] /
by Tyra Dark.
260
[Tampa, Fla.] :
b University of South Florida,
2007.
520
ABSTRACT: This dissertation investigated the area social predictors of health (ASPoH) and Black-White disparities in stroke mortality relationship. Utilizing stroke mortality data obtained from the Florida Department of Health for years 1998-2002, and social and economic data obtained from the year 2000 Census of Population, this study examined the effect of resource availability at the census tract level on Black-White disparities in stroke mortality. The influence of social class on Black-White disparities in stroke mortality and effect modification by social class of the association between Black-White disparities and ASPoH variables was also investigated. Principal component analysis produced four ASPoH scores from economic and social measures. Multiple regression analysis assessed the predictive ability of these ASPoH variables on Black-White disparities.Increases in the female Black-White ratio were significantly associated with increases in the magnitude of the ASPoH-1 and ASPoH-2 variables. When regression analyses were restricted (in terms of population count minimums) to a subset of census tracts, increases in the ASPoH-1 and ASPoH-2 variables were significantly associated with increases in all Black-White disparity measures for both males and females. Assessment of the influence of social class on Black-White disparities in stroke mortality was only feasible at the state level due to a lack of data at the census tract level. With the exception of the 65+ years age-group, Black males and females experienced higher age-group specific stroke mortality rates across each of the social class groups. Inconsistent with previous research findings, Black residents who attained a high school degree had the highest stroke death rates compared to all other educational attainment groups.In the assessment of social class as a potential effect modifier, the study hypothesis stated that the ASPoH measures would have the greatest impact on those residents in the lowest social class category. This predicted effect was only supported when the Male Black-White ratio disparity score was examined. Study findings support the conjecture that unknown and unmeasured processes influence the association between area social predictors and stroke mortality for Black Floridians. Identification of modifiable societal characteristics may be the key to unlocking the foundation of disparities in health outcomes.
502
Dissertation (Ph.D.)--University of South Florida, 2007.
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 155 pages.
Includes vita.
590
Advisor: Thomas J. Mason, Ph.D.
653
Wellbeing.
Resources.
Access.
Economic.
Environment.
690
Dissertations, Academic
z USF
x Public Health
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.2014