USF Libraries
USF Digital Collections

Analysis of the role of residential segregation on perinatal outcomes in Florida, Georgia, and Louisiana

MISSING IMAGE

Material Information

Title:
Analysis of the role of residential segregation on perinatal outcomes in Florida, Georgia, and Louisiana
Physical Description:
Book
Language:
English
Creator:
Akintobi, Tabia Henry
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Births
Ethnic disparities
Infant health
Social determinants of health
Multilevel modeling
Dissertations, Academic -- Public Health -- Doctoral -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: The purpose of this study was to investigate the relationship between residential segregation (the physical separation of Blacks and Whites in residential contexts) and adverse perinatal outcomes (low birth weight, preterm delivery and small for gestational age births) in Florida, Georgia and Louisiana. The study determined the independent effect of the level of residential segregation on the likelihood of adverse perinatal outcomes after controlling for contextual and individual factors. The study also assessed whether the relationship between residential segregation and adverse perinatal outcomes were moderated by ethnicity and median income.The studied employed an observational, cross-sectional study design that utilized secondary data. Live birth certificates between 1999 and 2001 provided information on individual covariates and perinatal outcomes. Structural indicators of residential segregation and contextual covariates were obtained from the U.S. Census Bureau. Th e nested data structure for each birth outcome model was composed of individual, contextual, and structural data. Three-level, hierarchical generalized linear models were used to test research hypotheses.The study population consisted of non-Hispanic White and Black primaparous women between 15 and 49 years of age experiencing singleton live births delivered at less than or equal to 45 weeks gestation. The final sample consisted of 255,548 women nested within 4,360 census tracts and 63 Metropolitan or Micropolitan Statistical Areas. Residential segregation did not have a direct relationship with low birth weight, preterm delivery or small for gestational age, after controlling for other variables in multilevel models. Models testing the moderating effects of ethnicity indicated that increased Isolation decreased the risk of LBW among Black women. Several contextual --level variables and the majority of individual-level variables were significantly associated with perinatal outcome risk .Findings indicate that effects of residential segregation may be birth outcome and ethnic group specific. Relationships between individual factors, contextual factors and adverse perinatal outcomes signal the importance of proximal factors to perinatal outcomes. There is a need for specification of a broader constellation of biological, social and spatial factors and a thorough assessment of residential preferences and experiences in order to better understand the associations between neighborhoods and perinatal outcomes.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2006.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Tabia Henry Akintobi.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 192 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 - 001790090
oclc - 144377776
usfldc doi - E14-SFE0001484
usfldc handle - e14.1484
System ID:
SFS0025803:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001790090
003 fts
005 20070613105852.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 070613s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001484
040
FHM
c FHM
035
(OCoLC)144377776
049
FHMM
090
RA425 (ONLINE)
1 100
Akintobi, Tabia Henry.
0 245
Analysis of the role of residential segregation on perinatal outcomes in Florida, Georgia, and Louisiana
h [electronic resource] /
by Tabia Henry Akintobi.
260
[Tampa, Fla] :
b University of South Florida,
2006.
3 520
ABSTRACT: The purpose of this study was to investigate the relationship between residential segregation (the physical separation of Blacks and Whites in residential contexts) and adverse perinatal outcomes (low birth weight, preterm delivery and small for gestational age births) in Florida, Georgia and Louisiana. The study determined the independent effect of the level of residential segregation on the likelihood of adverse perinatal outcomes after controlling for contextual and individual factors. The study also assessed whether the relationship between residential segregation and adverse perinatal outcomes were moderated by ethnicity and median income.The studied employed an observational, cross-sectional study design that utilized secondary data. Live birth certificates between 1999 and 2001 provided information on individual covariates and perinatal outcomes. Structural indicators of residential segregation and contextual covariates were obtained from the U.S. Census Bureau. Th e nested data structure for each birth outcome model was composed of individual, contextual, and structural data. Three-level, hierarchical generalized linear models were used to test research hypotheses.The study population consisted of non-Hispanic White and Black primaparous women between 15 and 49 years of age experiencing singleton live births delivered at less than or equal to 45 weeks gestation. The final sample consisted of 255,548 women nested within 4,360 census tracts and 63 Metropolitan or Micropolitan Statistical Areas. Residential segregation did not have a direct relationship with low birth weight, preterm delivery or small for gestational age, after controlling for other variables in multilevel models. Models testing the moderating effects of ethnicity indicated that increased Isolation decreased the risk of LBW among Black women. Several contextual --level variables and the majority of individual-level variables were significantly associated with perinatal outcome risk .Findings indicate that effects of residential segregation may be birth outcome and ethnic group specific. Relationships between individual factors, contextual factors and adverse perinatal outcomes signal the importance of proximal factors to perinatal outcomes. There is a need for specification of a broader constellation of biological, social and spatial factors and a thorough assessment of residential preferences and experiences in order to better understand the associations between neighborhoods and perinatal outcomes.
502
Dissertation (Ph.D.)--University of South Florida, 2006.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 192 pages.
Includes vita.
590
Adviser: Melinda S. Forthofer, Ph.D.
653
Births.
Ethnic disparities.
Infant health.
Social determinants of health.
Multilevel modeling.
690
Dissertations, Academic
z USF
x Public Health
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1484



PAGE 1

Analysis of the Role of Residential Segreg ation on Perinatal Outcomes in Florida, Georgia and Louisiana by Tabia Henry Akintobi A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Community and Family Health College of Public Health University of South Florida Co-Major Professor: Meli nda S. Forthofer, Ph.D. Co-Major Professor: Carol Bryant, Ph.D. Elleen Daley, Ph.D. Wendy N. Nembhard, Ph.D. Kathleen O’Rourke, Ph.D. Date of Approval: April 3, 2006 Keywords: births, ethnic disp arities, infant health, social determinants of health, multilevel modeling Copyright 2006, Tabia Henry Akintobi

PAGE 2

DEDICATION I dedicate this work, first, to my Lord Jesus Christ. To my husband, Adebayo A. Akintobi, M.D.; you have been my biggest ch eerleader, a professional colleague, and the man of my dreams. To my daughter, Ifelol a; you were a good baby and are now a big girl. I love who you are and eag erly anticipate all that you wi ll become. To my parents, Leila and Zadoc Henry; your steadfast examples of faith and excelle nce in education will be legacies I will pass on. Thanks for always believing in me, not matter what I thought. To my first brother, David Henry; when I think of your life, I know that absolutely nothing is impossible and that giving up is ne ver an option. To my baby brother, Joshua Henry; your tireless energy, enormous talent a nd tunnel vision have always motivated me to be more.

PAGE 3

ACKNOWLEDGMENTS I publicly acknowledge and thank members of my dissertation committee (Melinda S. Forthofer, Ph.D., Carol Bryant, Ph.D., Wendy N. Nembhard, Ph.D., Ellen Daley Ph.D., and Kathleen O’Rourke, Ph.D). Thank you for your professional examples and expertise. Special appreciation goes to Drs. Bryant Forthofer, Quinn, and Yancey for your investments in me and allowing me to prof essionally grow and flourish. Each of you epitomizes what it means to be a mentor. Special appreciation goes to Mr. Roderick Hale, Dr. Morehouse and the Florida Educati on Fund for providing critical professional and fiscal support throughout my tenure as a doctoral student.

PAGE 4

i TABLE OF CONTENTS List of Tables iv List of Figures vii Abstract viii Chapter One: Introduction 1 Purpose of the Study and Study Significance 2 Challenges Associated With Use of the Term Race 4 Study Rationale 6 Structural Explanations for Ethnic Health Disparities 7 Contextual Factors Associated with Peri natal Outcomes 8 Gaps in the Scientific Knowledge 9 Research Questions and Hypotheses 9 Overview of Study Design 12 Data Sources 13 Implications for Public Health 14 Delimitations 16 Limitations 17 Definitions 18 Chapter Two: Literature Review 19 Theoretical Framework 19 Conceptual Framework 20 Significance of Perinatal Outcomes 21 National Trends in Adverse Perinatal Outcomes 22 Significance of Small for Gestational Age Bi rths 23 Disparities in Perinatal Outcomes 24 Individual Factors and Perinatal Outcomes 27 Birth Certificate Data Validity 28 Gestational Age Estimation Specificity 31 Contextual Factors and Perinatal Outcomes 34 Significance of Health Disparities 35 Structural Explanations for Health Disparitie s 37 Residential Segregation 37 Significance 37 Historical Context 38 Conceptualization and Operationalization Dimensi ons 38 Index of Dissimilarity 39

PAGE 5

ii Isolation Index 40 Trends in Residential Segregation 41 Consequences of Residential Segregation 43 Employment Opportunities 43 Family Structure 44 Housing Quality and Ownership 44 Inappropriate Land Uses and Maladaptive Behavior 45 Socioeconomic Status 45 Pathways 46 Residential Segregation and Health 47 Adult Mortality 47 Crime 48 Infant Mortality 49 Research Questions and Hypotheses 53 Chapter Three: Methodology 56 Study Design 56 Study Population 59 Sample Size 59 Nested Data Structure 61 Power Implications for Multilevel Models 61 Data Sources 64 Variable Measures 65 Residential Segregation 69 Control Variables 70 Outcome Variables 71 Moderating Variables 73 Level of Aggregation 74 Analysis Procedures 74 Hierarchical Generalized Linear Models 78 Statistical Models 79 Strengths and Limitations 81 Chapter Four: Results 84 Univariate Analysis 87 Individual Covariates 87 Contextual Covariates and Moderating Vari ables 90 Independent Variables 94 Outcome Variables 101 Bivariate Analysis 103 Multilevel Modeling Analysis 109 Hypothesis One 109 Low Birth Weight 119

PAGE 6

iii Preterm Delivery 113 Small for Gestational Age Birth 115 Hypotheses Two 118 Low Birth Weight 119 Preterm Delivery 121 Small for Gestational Age Birth 123 Hypothesis Three 126 Low Birth Weight 126 Preterm Delivery 128 Small for Gestational Age Birth 131 Summary of Findings 134 Chapter Five: Discussion 135 Major Research Findings 135 Possible Explanations for Research Findings 138 Study Strengths 142 Study Limitations 143 Directions for Future Research 145 Implications for Public Health 150 Conclusion 151 References 152 Appendix A: Conceptual Model 191 About the Author End Page

PAGE 7

iv LIST OF TABLES Table 1. Multilevel Influences on Health Outcomes 3 Table 2. Perinatal Outcomes for the United States and Selected 57 Geographical Divisions Table 3. Variable Descripti ons and Scales of Measurement 67 Table 4. Analysis of the Distri butions of Individual Covariates Amongst Mothers Included and Excluded from Study Sample 85 Table 5. Analysis of the Distributi ons of Adverse Perinatal Outcomes Amongst Mothers Included and Excluded fr om Study Sample 86 Table 6. Distributions, P-Values and Unadjusted Odds Ratios for Individual Covariates, Overall Sample and by Ethnicity, for Sample of Flor ida, Georgia and Louisiana Mothers 1999-2001 (N=255,548) 89 Table 7. Distributions of Contextual Covariates in Census Tracts of Residence for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=4360) 91 Table 8. Distributions, P-Values and Chi Square Analyses for Contextual Covariates, by Proportion of Black Residents, in Census Tracts of Residence for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=4,360) 93 Table 9. Distribution of Dissimilar ity Indices, by Statistical Area Name and Increasing Dissimilarity, for Metropolitan and Micropolitan Statistical Areas Occu pied by Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=63) 96

PAGE 8

v Table 10. Distribution of Isolati on Indices, by Statistical Area Name and Increasing Isolation, for Metropolitan and Micropolitan Statistical Areas Occupied by Samp le of Florida, Georgia and Louisiana Mothers 1999-2001 (N=63) 99 Table 11. Distributions and Unadju sted Odds Ratio for Adverse Perinatal Outcomes Overall and by Ethnicity for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=255,548) 102 Table 12. Bivariate Analysis of Low Birth Weight Live by Individual Covariates for Sample of Flor ida, Georgia and Louisiana Mothers 1999-2001 (N=20,258) 104 Table 13. Bivariate Analysis of Pr eterm Live Births by Individual Covariates for Sample of Flor ida, Georgia and Louisiana Mothers 1999-2001 (N=20,258) 106 Table 14. Bivariate Analysis of Sm all for Gestational Age Births by Individual Covariates for Sample of Florida Georgia and Louisiana Mothers 1999-2001 (N=28,998) 108 Table 15. Parameter Estimates for Total Sample: Multilevel Analysis of the Effects of Residential Segregation on Low Birth Weight for Sample of Florida, Geor gia and Louisiana Mothers 1999-2001 111 Table 16. Parameter Estimates for Total Sample: Multilevel Analysis of the Effects of Residential Segregation on Preterm Delivery for Sample of Florida, Geor gia and Louisiana Mothers 1999-2001 114 Table 17. Parameter Estimates for Total Sample: Multilevel Analysis of the Effects of Residential Segregation on Small for Gestational Age Births for Samp le of Florida, Georgia and Louisiana Mothers 1999-2001 116 Table 18. Parameter Estimates for Random Effects of Ethnicity: Multilevel Analysis of M oderating Effects on the Relationship between Residential Segregation and Low Birth Weight for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 119

PAGE 9

vi Table 19. Parameter Estimates for Random Effects of Ethnicity: Multilevel Analysis of Modera ting Effects on the Relationship between Residential Segregati on and Preterm Delivery for Sample of Florida, Georgia and L ouisiana Mothers 1999-2001 122 Table 20. Parameter Estimates for Random Effects of Ethnicity: Multilevel Analysis of Moderati ng Effects on the Relationship between Residential Segregati on and Small for Gestational Age Births for Sample of Flor ida, Georgia and Louisiana Mothers 1999-2001 124 Table 21. Parameter Estimates for Random Effects of Median Income: Multilevel Analysis of the Modera ting Effect on the Relationship between Residential Segregati on and Low Birth Weight for Sample of Florida, Georgia and L ouisiana Mothers 1999-2001 127 Table 22. Parameter Estimates for Random Effects of Median Income: Multilevel Analysis of the Modera ting Effect on the Relationship between Residential Segregati on and Preterm Delivery for Sample of Florida, Georgia and Loui siana Mothers 1999-2001 129 Table 23. Parameter Estimates for Random Effects of Median Income: Multilevel Analysis of the Moderating Effect on the Relationship between Residentia l Segregation and Small for Gestational Age for Sample of Fl orida, Georgia and Louisiana Mothers 1999-2001 132

PAGE 10

vii LIST OF FIGURES Figure A. Population Characteristics of Whites for Selected States 2002-2003 58 Figure B. Population Characteristics of Blacks for Selected States 2002-2003 58 Figure C. Sample Size and Relative Risk Calculations for Selected Perinatal Outcomes 60 FigureD. Distribution of Dissimila rity Indices for Metropolitan and Micropolitan Statistical Areas occupied by Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=63) 95 Figure E. Distribution of Isolat ion Indices for Metropolitan and Micropolitan Statistical Areas occupied by Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=63) 98 Figure F. Distribution of the Proportion of Bl acks for Metropolitan and Micropolitan Statistical Areas occupied by Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=63) 101

PAGE 11

viii Analysis of the Role of Residential Segreg ation on Perinatal Outcomes in Florida, Georgia and Lousiana Tabia Henry Akintobi ABSTRACT The purpose of this study was to investig ate the relationship between residential segregation (the physical sepa ration of Blacks and Whites in residential contexts) and adverse perinatal outcomes (low birth weight preterm delivery and small for gestational age births) in Florida, Georgia and Louisi ana. The study determined the independent effect of the level of residential segreg ation on the likelihood of adverse perinatal outcomes after controlling for contextual and individual factors. Th e study also assessed whether the relationship between residential segregation and adverse perinatal outcomes were moderated by ethnicity and median income. The studied employed an observational, cros s-sectional study de sign that utilized secondary data. Live birth certificates between 1999 and 2001 provided information on individual covariates and perinatal outcom es. Structural indicators of residential segregation and contextual c ovariates were obtained from the U.S. Census Bureau. The nested data structure for each birth outc ome model was composed of individual, contextual, and structural data Three-level, hierarchical ge neralized linear models were used to test research hypotheses. The study population consiste d of non-Hispanic White and Black primaparous women between 15 and 49 years of age experi encing singleton live births delivered at

PAGE 12

ix less than or equal to 45 weeks gestation. The final sample consisted of 255,548 women nested within 4,360 census tracts and 63 Metropolit an or Micropolitan Statistical Areas. Residential segregation did not have a di rect relationship with low birth weight, preterm delivery or small for gestational age, after controlling fo r other variables in multilevel models. Models testing the modera ting effects of ethnicity indicated that increased Isolation decreased the risk of LBW among Black women. Several contextual – level variables and the majority of individual-level variables were significantly associated with perinatal outcome risk. Findings indicate that effects of residen tial segregation may be birth outcome and ethnic group specific. Relationships between individual factors, c ontextual factors and adverse perinatal outcomes signal the impor tance of proximal factors to perinatal outcomes. There is a need for specification of a broader constellation of biological, social and spatial factors and a thorough assessment of residential preferen ces and experiences in order to better understand the associat ions between neighborhoods and perinatal outcomes.

PAGE 13

1 CHAPTER ONE: INTRODUCTION The health status of mothers and infants are critical indicators of the nation’s health. Public health professionals view in fant mortality as “a measure of community health, economic efficiency, collective mora l well being, and future military strength” (Brosco, 1999). Healthy People 2010 objectives to reduce low birth weight to 5.0 % and preterm deliveries to 7.6% are critical because they represent perinata l health indicators that are the leading causes of neonatal death and compromised quality of life to infants and families (United States Department of Health and Human Services, 2000). Regional findings indicate that low birth weight, neonata l mortality rates (death of live born from birth to < 28 days of life), and infant mortality rates (dea th of live born before first birthday) are highest in the Southern United States1 (U.S.) (National Center for Health Statistics, 2004). Adverse perinatal outcomes that contribute to increased infant mortality include low birth weight (LBW) (birth weight < 2,500 grams), preterm delivery (delivery <37th complete weeks of gestation), and small fo r gestational age (SGA) birth (birthweight below the 10th percentile for gestational age and sex). LBW infants are 40 times more likely to die during their firs t month of life (Health Beat, 1998). Preterm infants are 70 times more likely to die than infants born at term (37-41 weeks) (Matthews, Menacker & MacDorman, 2002). SGA infants experience a 6-fold risk for neonatal mortality when compared to infants born appropriate for gestational age (Doctor, et al., 2001). 1 Delaware, District of Columbia, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, West Virginia, Alabama, Kentucky, Mississippi, Tennessee, Arkansas, Louisiana, Oklahoma, Texas

PAGE 14

2 National trends indicate that the preval ences of adverse perinatal outcomes has increased and ethnic disparitie s have persisted over time. LBW rates have increased 15% (from 6.8%) since the middle of the 1980s (National Center fo r Health Statistics, 2004). In 2002, LBW increased to 7.8%, the highest repo rted rate in three decades (Martin, et al., 2003). Infants born preterm have incr eased 29% since 1981 and 14% since 1990. Preterm births reached a high of 12.1% of all births in 2002 (2003). The largest disparities in birth outcomes are by maternal ethnicity. Despite a 6% decrease in preterm deliveries among Black women and an increase of 29% among White women since 1990, rates were hi gher among Black women (17.7%) when compared to White women (11%) in 2002 ( 2003). Black-White trends in SGA births indicate that while rates have decreased for both Black s and Whites since 1990, Blacks maintained a rate of term SGA (16%) that was 1.5 to 2.0 times higher than Whites (9%) in 2000 (Ananth, Balsubramanian, Demissie & Kinzler, 2004). Black women are almost twice as likely to have a LBW infant (Guyer, Martin, MacDorman, Anderson & Strobino, 1997). While LBW rates have increased for White women (from 5.6%) and rates among Black women have remained relatively unchanged (from 13.6%) since 1990, the LBW rate for Black and White infants in 2002 we re 13.4% and 6.9%, respectively (Martin et al., 2003). Purpose of the Study and Study Significance The purpose of this study was to inves tigate the direct relationship between residential segregation (the physical separation of Black s and Whites in residential contexts) and adverse perinatal outcomes (L BW, preterm delivery a nd SGA births) in the Southern states Florida, Ge orgia and Louisiana. This st udy specifically determined the

PAGE 15

3 independent effect of residence in an area characterized by high re sidential segregation on the likelihood of adverse perinatal outcomes after controlling for contextual factors (e.g., poverty level, unemployment ) and individual factors (e .g., prenatal care, substance use). The study also determined whether the di rect relationship betw een residence in an area with high residential segregation and a dverse perinatal outcomes were moderated by ethnicity and median income, respectively. The variables of interest in this study were captured by concepts that represent the multilevel nature of factors that influence health. Structural factors operate at the macro level (in this study, dimensions of residential segregation) and are indicative of broader social, political or economic conditions (e .g. discrimination, racism). Contextual or mezzo level factors (e.g. poverty level, proportion of female-headed households) represent characterist ics of the social and physical e nvironment that are a product of structural effects. Individua l factors represent genetic, behavioral, demographic and economic characteristics at the micro level (Table 1.0). Table 1 Multilevel Influences on Health Outcomes LEVEL OF INFLUENCE CONCEPTUAL TERM OPERATIONAL DEFINITION STUDY VARIABLE Macro Structural Indicators of broad social, economic or political conditions Residential Segregation Mezzo Contextual Featur es of the immediate residential environments Social, Physical and Economic Environments Micro Individual Individual behaviors, risk factors of health outcomes Maternal Characteristics and Perinatal Outcomes

PAGE 16

4 The majority of etiologic research on in fant mortality and morbidity has focused on individual factors. The iden tification of individual determin ants of perinatal outcomes, including maternal age, pari ty, weight gain, smoking and pr enatal care, have been the foundation upon which numerous national and community-based ini tiatives have been developed (Goldenberg, et al., 1996; Schiono, Rauh, Park, Lederman, & Zuskar, 1997). Ethnic disparities in perinatal outcomes conti nue to exist, despite these efforts (Ananth, Balsubramanian, Demissie & Kinzler, 2004; Guyer, Martin, MacDorman, Anderson & Strobino, 1997; Marti n, et al., 2003). Challenges Associated With Use of the Term Race The term ethnicity is used instead of r ace in this study due to numerous problems with the concept of racial differences. While genetically determined diseases that are more prevalent among certain ethnic groups (e.g., Crohn's disease among Whites; Sickle Cell among Blacks) do exist, studies indicate th at individuals are not identical within a group, and the perceived physical traits used for such purposes may not be biological in origin (Cooper, Kaufman & Ward, 2003; He ssol, Fuentes-Afflic k, & Bacchetti, 1998; Jones, 2001; Rowley, 1994). Limitations associated with the term race are further complicated by changing definitions and terms. The U.S. Census Bur eau's use of 26 different terms since 1990 to identify populations may contribute to limited understanding and/or delineation of terms (Hayes-Bautista & Chapa, 1987). Changes in terminology over time have depended upon an unreliable mixture of fact ors, including national origi n, language, surname, minority status and physical characteristics. Terms ha ve been ambiguously used with no clarity or consistency by investigators in research studies or by th e lay public when completing

PAGE 17

5 surveys (Gerber & de la Puenta, 1996; Kissi m, Herrara, & Nakamoto, 1993; Rodriguez & Cordero-Guzman, 1992). One problem with use of the term race rather than ethnicity is related to the limited biological connotations with which it is associated, despite its social meanings. Race is associated with the power to determine a group’s identity through a unidimensional label and the way such labeli ng translates to a group’s experience with others. Health is not affected by ethnic or racial identity, but by the respective consequences of these identities on social economic and political positions (American Anthropological Association, 1997). Most public health investigators agre e that consideration of social and environmental factors are crucia l to investigating the determinants of health outcomes in general, and ethnic health disparities in par ticular. Efforts to eliminate ethnic health disparities can begin with avoidance of the te rm race and use of more appropriate labels for social and cultura l population subgroups, In contrast to race, traditionally understood to represent perceived biological traits, ethnicity refers to social groups characterized by distinctive cultural traditions that are ma intained across generations, and a common history or origin (Last, 1995). For this reason, ethnicity is used instead of race in this dissertation. The current state of the perinatal literature sets the foundation for further expansion of the field's understanding of ethni c health disparities. Contextual factors associated with preterm delivery and LBW have been identified with evidence that risk factors vary across space a nd ethnic groups (Buka, Bre nnan, Rick-Edwards, Raudenbush, & Earls, 2002; Diez-Roux, et al., 1997; Gorman, 1999; O’Campo, Xue & Wang, 1997;

PAGE 18

6 Pearl, Braveman & Abrams, 2001; Rauh, A ndrews & Garfinkel, 2001; Roberts, 1997; Sloggett & Joshi, 1994). Far less is known about factors placing groups at risk for SGA births. Further, macro level structures associ ated with increased lik elihood of perinatal morbidity and mortality, rather than the out come of infant mortality, have not been previously explored in the perinatal literature. Study Rationale Lower socioeconomic status (SES) (e.g., median income) and ethnicity have been associated with a variety of health outcomes. SES has been established in the literature as one of the primary determinants of health outcomes (Lynch & Kapl an, 2000). Disparities in life expectancy and health status ar e widest between Blacks and Whites, with disproportionate mortality rela ted to cardiovascular and cere brovascular diseases, cancer, homicide, infant death, diab etes, and AIDS among Blacks. Disparities in perinatal outcomes mirror these trends. Most health disparities pers ist at all levels of SES despite the disproportionate numbers of minorities represented in lowe r SES groups in the U.S. Researchers challenging the notion that health differences are solely explained by poverty have found that disparities exist at every level of the social hierarchy, reflec ting a social gradient, rather than a threshold, effect (Macintyr e, 1994; Marmot, Bobak & Davey Smith, 1995; Wilkinson, 1992). If Black-White differences in health are not simply attributable to group differences in SES, research is needed to understand the factor s that influence the relationship between ethnicity and health.

PAGE 19

7 The identification of predisease pathways are an emerging public health priority with research emphasis placed on the linka ges between behavioral, psychological and social influences that precede disease/health specific morbidity and mortality (Singer & Ryff, 2001). Among predisease influences are prenatal and early risk factors. The Ecosocial framework provided a macro-level unde rstanding of the processes that generate ethnic health disparities (K rieger, 1994, 1999, 2001). This conceptualization was selected for this study due to its emphasis on social, pol itical and economic processes that account for population patterns of healt h, disease, and well-being. Structural Explanations fo r Ethnic Health Disparities Residential segregation is the physical separation of ethnic groups in residential contexts or the differentiation of two or more groups among dimensions of a given social space (Acevedo-Garcia & Lochner, 2003). Resi dential segregation has been primarily conceptualized and studied in the context of the separation of ethnic groups. Residential segregation in this study is specified to reflect the phys ical separation between Blacks and Whites. Segregation historically reflect ed individual and grouplevel discriminatory practices based on racism, and was driven by majority perceptions of minority group inferiority and practices focused on maintain ing social distance be tween defined groups (Pettigrew & Meertens, 1995). Discriminatory policies led by the re al estate industry, federal housing authorities and banking institut ions sought to ensure the restriction of Blacks from housing choices, which relegated them to substandard residential areas (Williams & Collins, 2001).

PAGE 20

8 Residential segregation has received incr eased attention as scholars have sought to better understand ethnic he alth disparities. Residential segregation effects have been examined among primarily Black minority gr oups. These studies show that greater residential segregation is associated with increased stress and poorer health (Collins & Williams, 1999; Frazier, 1957; Jiobu, 1972; Kr ivo, 1999; LaVeist, 1989; Poldenak, 1993; 1996; Shihadeh & Flynn, 1996; Wilson, 1980, 1987; Yankauer, 1950). Studies that have explored the effects of resi dential segregation on non-Blac k ethnic groups have yielded inconsistent findings (Collins & Willi ams, 1999; Fang, Madhaven, Bosworth, & Alderman, 1998). Contextual Factors Associated with Perinatal Outcomes A relatively small number of studies have examined the association between local area contexts and perinatal outcomes. St udies have focused on the city (Buka et al., 2002; O’Campo, Xue & Wang, 1997; Pearl, Br aveman & Abrams, 2001; Rauh et al., 2001; Roberts, 1997) national (Gorman, 1999) and international levels (Sloggett & Joshi, 1994). Evidence from these studies has not been conclusive. Some st udies have employed measures of residential compos ition taken from the U.S. Census Bureau, such as ethnic composition, age and immigrant composition (G orman, 1999; Roberts, 1997). None have advanced a coherent theoretical framework that considers the macro level structures that shape residential contexts and may be associ ated with adverse perinatal outcomes. Much of the evidence linking residential disadvantag e to adverse perinatal outcomes has been based on conventional regression analysis meth ods that are not appropriate for data at multiple levels (i.e., individuals nested in residential areas).

PAGE 21

9 Gaps in the Scientific Knowledge The majority of the compelling evidence supporting the contribution of residential segregation to population health has been limited by cross-se ctional and ecological study designs analyzed at individual or aggregate levels. Further, c ontextual factors that may be statistically controlled in the relationship be tween structural factor s and health outcomes often have been excluded in single level studi es. Evidence indicates that the relationships between residential contexts/structures a nd health outcomes are not homogeneous and require more specificity in investigation. Previously studied health outcomes br oadly include self-rated health, life expectancy, all-cause mortality, and infant mortality. Identification of the ways that residential segregation may be associated with different health outcomes requires targeted examination and analysis (Lynch & Davey Sm ith, 2003). Virtually unexplored is how the effects of residential segregation may vary by dimension and perinatal outcome investigated. Examination of the moderating effects of ethnicity and median income on the relationship between residential segregation and cause-specific health outcomes will contribute to better understanding of whom and under what conditions macro level factors may be protective or adverse. Research Questions and Hypotheses The purpose of this study was to investig ate the relationship between residential segregation and adverse perina tal outcomes (LBW, preterm delivery and SGA births) in Florida, Georgia, and Louisiana. Specifical ly this study will determine the independent effect of the level of residential segreg ation on the likelihood of adverse perinatal outcomes after controlling for contextual and individual factor s. The study also

PAGE 22

10 determined whether the relati onship between level of resi dential segregation and the likelihood of adverse perinatal outcomes was moderated by et hnicity and median income, respectively. The following research questions and hypotheses were investigated: Research Question 1: Is there a relationship between residential se gregation and adverse perinatal outcomes among mothers in Florida, Georgia and Louisiana? Hypothesis 1: The level of re sidential segregation is posi tively associated with the likelihood of adverse birth outcomes after c ontrolling for contextual and individual factors. Hypothesis 1A: The level of residential segregation is positively associated with the likelihood of preterm delivery after c ontrolling for contextual and individual factors. Hypothesis 1B: The level of residential segregation is positively associated with the likelihood of low birth weight after controlling for contextual and individual factors. Hypothesis 1C: The level of residential segregation is positively associated with the likelihood of small for gestational age births after controlling for contextual and individual factors.

PAGE 23

11 Research Question 2: Is the relationship be tween residential segr egation and perinatal outcomes moderated by ethnicity? Hypothesis 2: The relationship between re sidential segregation and the increased likelihood of adverse pe rinatal outcomes is moderated by ethnicity. Hypothesis 2A: The relationship between residential segr egation and the increased likelihood of preterm de livery is moderated by ethnicity. Hypothesis 2B: The relationship between residential segr egation and the increased likelihood of low birth we ight is moderated by ethnicity. Hypothesis 2C: The relationship betw een residential se gregation and the increased likelihood of small for gesta tional birth is moderated by ethnicity. Research Question 3: Is the relationship be tween residential segr egation and perinatal outcomes moderated by median income? Hypothesis 3: The relationship between re sidential segregation and the increased likelihood of adverse perinatal outcom es is moderated by median income.

PAGE 24

12 Hypothesis 3A: The relationship between residential segr egation and the increased likelihood of preterm delivery is moderated by median income. Hypothesis 3B: The relationship between residential segr egation and the increased likelihood of low birth weight is moderated by median income. Hypothesis 3C: The relationship between residential segr egation and the increased likelihood of small for gestati onal age birth is moderated by median income. Overview of Study Design The research employed an observational, cr oss-sectional study de sign that utilized secondary data. The study was observationa l in that groups are formed by sample population status for each respective adverse perinatal outcome rather than randomization into treatment and control groups. The study was cross-sectional due to the proposed association between prevalence of advers e perinatal outcomes and prevalence of exposure to residential segreg ation and groups are not identif ied on the basis of exposure or outcome. The use of preexisti ng birth certificate and U.S. Census Bureau data explains the secondary nature of the study. A multilevel analysis plan was used because the majority of the compelling evidence supporting the contributions of resi dential segregation to population health trends has been limited by data analyzed at aggregate levels. One of the criticisms of these studies is that they fall prey to the ecological fallacy, which involves making individual inferences based on population or group-level measures (Judge, 1999). The atomistic fallacy is a limitation of res earch studies which may make inferences

PAGE 25

13 regarding variability ac ross units defined at a higher le vel of analyses based on data collected at a lower level of analysis or solely examine i ndividual characteristics to the exclusion of factors associated the context or structures within wh ich individual risk is generated (Alker, 1969; Curtis & Jone s, 1998; Diez-Roux, 1998; Diez-Roux, 2003; Scheuch, 1969). Through utilizing multile vel modeling, individual risk may be simultaneously distinguished from populati on, contextual risk, minimizing concerns related to the atomistic and ecological fallacies. Finally, multilevel analyses can potentially contribute greatly to public health practice throug h the identification of risk profiles that account for both individual risk and contextual factors. Data Sources Birth certificate records were used to co llect information on individual covariates and adverse perinatal outcomes. Selection of southern states was initially based on regional findings which show higher rates of LBW, neonatal and infant mortality in the South (National Center for Health Statistics 2004). Florida, Georgi a and Louisiana were the states analyzed in this st udy due to their inclusion residential census tract information on live birth certificates. The census tract was th e selected unit of analysis because it is a small, relatively permanent statistical subdivision. Designed to be relatively homogeneous with respect to population char acteristics, economi c status and living conditions, census tracts capture within ci ty variation and the immediacy of social context that is lost in cross-metro anal yses (Guest, Almgren, & Hussey, 1998). Census tract information on birth certificates was used to match residential segregation indices and contextual covariates calculate d by the 2000 U. S. Census Bureau.

PAGE 26

14 Residential segregation indices and contex tual covariates were derived from the U.S. Census Bureau data. The Household a nd Economics Statistics Division (HESD) of the U.S. Census Bureau provided data on resi dential segregation indices (Census Bureau of the United States, 2000). The HESD cal culated residential segregation through examination of the distribution of populati ons across census tracts within Metropolitan Areas (MAs) and Primary Metropolitan Statistical Areas (PMSAs) (Census Bureau of the United States, 2000). The 2000 Neighborhood Change Database (1970-2000) (NCD) (GeoLytics, 2004) provided data on contextual characteristics of residential areas. The NCD included Long Form U.S. Census data from 1970, 1980, 1990 and 2000 at the census tract level. A unique feature of the dataset is that geogra phical identifiers allow census tracts to be summarized into larger geographical levels including counties and MSAs similar to processes used to calculated resident ial segregation through the HESD (2004). Implications for Public Health This research has several potential impli cations for perinatal research. First, no identified studies have examined the relativ e contributions of va rious dimensions of residential segregation on traditionally studied (LBW and preterm delivery) and more newly investigated perinatal outcomes (S GA). The proportion of variance in each outcome that is explained by each residential segregation dimension is important to better understand the relative roles of structural dime nsions of residential segregation factors on perinatal outcomes.

PAGE 27

15 Second, with regard to ethnicity and median income, the majority of studies that have investigated determinants of health, beyond the micro level, have focused on the contextual level rather than investigations of the determinants of residential contexts. Research on the effects of residential areas on child and adolescent development outcomes generally find less powerful resident ial effects among Blacks when compared to Whites (Brooks-Gunn, et al., 1997). There is some evidence that living in residential areas of high SES enhances educational atta inment of Black males if neighbors are also Black (Duncan et al., 1997). Williams & Co llins (1995) found that although individual SES greatly mediates the association between ethnicity and health, some ethnic effects on health remain. Others studies found that ethnic differences in health and mortality that persisted after controlling for individual SES, were eliminated upon examination of community socioeconomic factors (Haan, Ka plan, & Camacho, 1997; LeClere, Rogers, & Peters, 1997; Roberts, 1998). While these studi es shed light on the role of residential contexts in explaining relative ethnic di sadvantage (Jargowsky, 1997; Massey & Denton, 1993; Wilson, 1987) the degree to which ethnicity moderates the effect of residential segregation on perinatal outcomes has been unexplored. Results have implications for public hea lth advocacy and practic e. If residential segregation contributes someth ing unique to the perinatal health of individuals in identified geographical areas, residential and individual-focused interventions are critical. The multilevel analysis strategies employed in this study will help to identify multilevel risk profiles to initiate research investigati on of how structural fact ors (e.g. dimensions of residential segregation) contribute to perina tal health outcomes.

PAGE 28

16 This study was developed with carefu l consideration of delimitations and limitations. Delimitations included are those criterion that are used to determine the geographical level, unit of analysis and characteristics of the population that was studied. Limitations noted are factors th at affect the generalizability of findings and include issues related to use of preexisting data. Operational terms used in the study follow listed limitations. Delimitations 1. The study is limited to investigation of births to women of White and Black, non-Hispanic ethnicity. 2. The study included women delivering in th e southern states of Florida, Georgia and Louisiana. 3. The study was limited to states that include census trac t of residence in live birth data certificate records. 4. The study included census tr acts that are repr esented in the 2000 U.S. Census Bureau.

PAGE 29

17 Limitations 1. The presence of an association between residential segregation dimensions and adverse perinatal outcomes were examined in the study; causality could not be determined. 2. The study examined data available thr ough live birth certificat es and the U.S. Census Bureau; no psychosocial, attitudina l or stress measures were assessed. 3. The study was limited to live births Florid a, Georgia, and Louisianna; results cannot be generalized to popul ations beyond those examined. 4. U.S. Census Bureau data provided sample estimates and were subject to both sampling and non-sampling errors. 5. Live birth certificate data may have been subject to non-sampling error occurring during initial data entry (at the hospital le vel) and in processi ng birth certificates at each health department. 6. Live birth data did not contain SES information for individual women (e.g., income level, employment and occupational status of mother). 7. The use of 2000 U.S. Census Bureau data may not have reflected characteristics of women delivering infants in 1999 and 2001. 8. The residential segregation window of expos ure associated with increased risk for perinatal outcomes and risk factors prior to time of pregnancy was unknown. 9. Misclassification of low birth wei ght and gestational age that may have influenced the analysis and repor ting of birth outcome trends.

PAGE 30

18 Definitions Adverse Perinatal Outcomes pregnancy outcomes that are associated with increased infant mortality and morbidity; preterm deliver y; low birth weight; small for gestational age births Ecosocial frame of reference (Krieger, 1994, 1999, 2001) which emphasizes social, political and economic processes that account for population patterns of health, disease, and well-being Low Birth Weight birth weight <2500 grams Macro Level structural factors that shape contextu al risk factors and are indicators of broader social, economic or political conditions Mezzo Level contextual factors characteristics of the social and physical environment that are a by-product of structural effect s; factors operating at the more immediate residential level that affect indivi dual risks and health outcomes Micro Level -genetic, behavioral, he alth-related, sociodemographic and socioeconomic factors attributable to individuals Neighborhood -census tract; small, relatively permanent statistical subdivision; geographical unit that is relatively ho mogeneous with respect to population characteristics, economic status and living conditions (G uest et al., 1998) Preterm Delivery delivery <37 complete d weeks of gestation Residential Segregation the physical separation of ethnic groups in reside ntial contexts; the differentiation of two or more groups am ong dimensions of a given social space Small for Gestational Age Birth less than the 10th percentile of birthweight for gestational age and sex

PAGE 31

19 CHAPTER TWO: LITERATURE REVIEW Theoretical Framework The Ecosocial framework structured e xplanations for relationships hypothesized in this study. This framework provided a m acro-level understanding of the processes that generate ethnic health disparities. This theoretical framewor k (Krieger, 1994, 1999, 2001) is selected due to its emphasis on social, political and economic processes that account for population patterns of health, disease a nd well-being. The linkages between primary constructs of this large-scal e perspective and the proposed re search indicate its relevance to more specific lines of public health inquiry. Embodiment, a central construct of the Ecosoc ial perspective, is conceptualized as the biological manifestation of social expe riences (Krieger, 1999). This concept is relevant to research investigations of how populations' social and environmental contexts are biologically interpreted and processed. Vari ations in population health morbidity and mortality are the consequences of embodiment Pathways of embodiment include the 1) social arrangement of power/property a nd contingent patterns of production, 2) consumption and reproduction and 3) the ba rriers and potential human biology enabled by human history, ecological cont ext and individual history of biology and development. This investigation was aligned with the c onstruct of embodiment in that it centered on how residential segregat ion, with historical and current im plications, influenced perinatal outcomes.

PAGE 32

20 The theory stresses the importance of the interplay of exposures at multiple levels (e.g. individual, neighborhood, regional). Th e multilevel nature of exposures were examined in this study through analysis of thre e levels. Live birth statistics (individual), census tract contexts (neighborhood) and censusbased data used to measure residential segregation (structural) were studied thr ough construction of hier archical generalized linear models to understand contributions of each to adverse perinatal outcomes. The Ecosocial perspective provides a fr amework that supports the relationship between residential segregation and health. Th is study did not measure all of the factors related to residential segregation that may include perceptions of institutionalized or personally experienced racism and personal pr eferences regarding integrated residence. The framework does, however, set the stage for increased understanding of the issues that must be considered when examining factors associated with ethnic health disparities. Conceptual Framework Residential segregation was conceptualized as a macro level structure that is an important determinant of adverse perinatal outcomes at the micro level guiding this research (Appendix A). Residential segreg ation is a socially, economically and politically motivated structur e when viewed through the Ecosocial frame of reference. The spatial separation of groups (macro level) fostered by historical discriminatory practices relegate, primarily, Black minorities to economic and social contexts that are less health promoting than their majority c ounterparts. Environmental contexts influence micro level risk factors. The relationshi p between residential segregation and the likelihood of adverse perinatal outcomes is differentially experienced by ethnic group and median income.

PAGE 33

21 The examination of distinct etiological processes leading to adverse perinatal outcomes is critical to the development of appropriate prevention and intervention strategies. Trends and disparities in perina tal outcomes warrant further examination to determine the differential influences of macr o and micro level factor s that increase the likelihood of adverse perinatal ou tcomes in residential areas. Significance of Perinatal Outcomes The health status of mothers and infants are critical indicators of the nation’s health. Public health professionals view in fant mortality as “a measure of community health, economic efficiency, collective mora l well being, and future military strength” (Brosco, 1999). Healthy People 2010 objectives to reduce low birth weight to 5.0 % and preterm deliveries to 7.6% are critical because they represent perinata l health indicators that are the leading causes of neonatal death and compromised quality of life to infants and families (United States Department of Health and Human Services, 2000). National infant mortality rates declined 11% between 1995 (7.6 infant deaths per 1,000 live births) and 2001 (6.8 infants deaths per 1,000 live bi rths) (Centers for Disease Control and Prevention, 2005). A decline of 36% would be necessary to achieve the Healthy People 2010 target despite declines for all ethnic groups. Regional findings confirm that LBW rates, neonatal mortality rates and infant mort ality rates are highest in the South (National Center for Health Statistics, 2004). Infants born too soon (preterm) or too small (LBW) are at increased morbidity and mortality risk. Preterm infants account for the relatively high perinatal mortality rate in the U.S. (Lumley, 2003). Preterm infants are 70 times more likely to die than infants born at term (37 to 41 weeks) (Matthew s, Menacker & MacDorman, 2002). Preterm

PAGE 34

22 delivery has been associated with neurode velopmental handicaps, chronic respiratory problems and infection (Berkowitz & Paperi nik, 1993; United States Department of Health and Human Services, 2000). Individu al risks factors for adverse perinatal outcomes include prior delivery of a LBW infant, a female infant, a multiple pregnancy, cigarette smoking, short maternal height, low maternal weight and preeclamptic hypertension (Collins & David, 1997; Control, 2000; Goldenberg et al., 1996). LBW infants are 40 times more likely to die during their first month of life (Health Beat, 1998). LBW has been associated with long-term di sabilities such as cer ebral palsy, autism, mental retardation, vision and hearing impair ments and other developmental disabilities (United States Department of H ealth and Human Services, 2000). National Trends in Adve rse Perinatal Outcomes The proportion of infants born with increased morbidity and mortality risk due to traditionally investigated adverse perinatal outcomes of LBW and preterm delivery have increased over the past thr ee decades. LBW rates increased from 7.7% in 2001 to 7.8% in 2002; a 15% increase from rates in the mi ddle of the 1980s (6.8%) (Martin et al., 2003; National Center for Health Statistics, 2002). The preterm birth rate increase from 11.9% in 2001 to 12.1% of all birt hs in 2002. The proportion of preterm births in 2002 has increased 29% since 1981 (9.4 %) and 14% since 1990 (10.6%) (2003). Trends in SGA have been more recently track ed, by ethnicity, and are incl uded in the discussion of disparities in perinatal outcomes.

PAGE 35

23 Significance of Small for Gestational Age Births The identification and tracking of SGA birt hs by perinatal researchers have been more recent than LBW and preterm delivery, but it is a birth outcome with noteworthy infant mortality and morbidity implications. SGA infants are at increased risk of infant mortality and are more likely to sustain shor t and long-term disabili ties than appropriate for gestational age infants. SGA infants expe rience a 6-fold risk for neonatal mortality and a nearly 3-fold risk for neonatal morb idity (birth asphyxia, meconium aspiration syndrome, hypoglycemia, and polycythemia) wh en compared to infants born appropriate for gestational age (Doctor, et al., 2001).The World Health Organization defines SGA as birth weight below the 10th percentile for gestational age and sex (Lee, Chernausek, Hokken-Koelega & Czernichow 2003). Although term (> 37 weeks) SGA rates have decreased among both Whites and Blacks, preterm (<37 weeks) SGA rates have increased for both groups (Ananth, Demissie, Kramer et al., 2003). The association between SGA and Intrau terine Growth Retardation (IUGR) is important for identification of infants at incr eased health risk. IUGR characterizes fetuses that do not achieve their gene tically determined potential size due to an unfavorable intrauterine environment (Regev et al., 2003) SGA infants who are pathologically small and at risk for modifiable, poorer ou tcomes may be the result of IUGR.

PAGE 36

24 The significance of SGA partially lies in the fact that LBW, the strongest predictor of infant survival is constructed of two components: preterm delivery and reduced fetal growth. Preterm and term SGA infants may be overlooked if the LBW measure is employed in isolation. The crude LBW measure examined in the majority of perinatal research does not differentiate be tween infants born due to short gestation (preterm delivery), SGA or sub-variations of these measures. Further, the measure does not allow for assessment of growth poten tial and growth achieved. The etiological significance of SGA further stre ngthens the importance of its inclusion in examinations of risk factors associated with adverse peri natal outcomes to potentially decrease infant morbidity and mortality rates. Disparities in Perinatal Outcomes The largest disparities in perinatal outc omes are by ethnicity of mother. Preterm delivery rates were higher among Black women (17.7%) when compared to White women (11%) in 2002, despite a 6% decreas e among Black women and an increase of 29% among White women since 1990 (Martin et al., 2003). The rate of term SGA (16%) among Blacks was 1.5 to 2.0 times higher than Whites (9%) in 2000 (Ananth, Balsubramanian, Demissie & Kinzler, 2004), in spite of decreases among both ethnic groups since 1990. While LBW rates have in creased for White women (from 5.6%) and rates among Black women have remained re latively unchanged (f rom 13.6%) since 1990, the LBW rate for Black and White infants in 2002 were 13.4% and 6.9%, respectively (2003).

PAGE 37

25 Explanations for disparities in perinata l outcomes are largely unknown with most research focus placed on factors during pregnancy. Far less is known about how experiences and contexts of women prior to pregnancy may impact perinatal outcomes. Lu and Halfon (2003) present a life course perspective to better understand ethnic disparities in perinatal outcomes. Th e model emphasizes early programming and cumulative pathway perspectives to model how health trajectories are formed and cumulative life stressors shape risk. Th is longitudinal con ceptualization views reproductive health as a product of developm ent, exposure and coping, with sensitive periods that heighten susceptibility to risk and protective factors. Sensitive periods occur in utero, puberty, and pregnancy and predict initial ri sk trajectory formation. Trajectories are later altered by cumulati ve allostatic lo ad through the life course (2003). Early fetal programming hypotheses exam ine causal associations between fetal development and later onset of disease. Under nutrition in utero has been associated with increased risk of insulin resistance, hype rtension and cardiovascu lar heart disease in adulthood. Barker’s Hypothesis is one of th e most frequently cited perspectives surrounding the association between LBW and adult onset of disease. The hypothesis predicts that the highest risk of heart disease and Type 2 diabetes (the insulin resistance syndrome or impaired glucose tolerance) are experienced by infants who experience compromised fetal growth. Increased cortisol levels during intrauteri ne life are thought to cause the endothelial damage related to car diovascular disease, as well as insulin resistance, which is related to other parame ters of the metabolic syndrome (Clark, 1998; Edwards, Coulter, Symonds, & Mc Millen, 2001; Phillips, 2001).

PAGE 38

26 The cumulative pathway perspective is associated with how biological consequences of stress can accumulate over time to affect health and function (McEwen,1995). Higher corticotrophin-releasing hormone levels were found in women who have preterm babies compared to wo men who delivery full term (Wadhwa, Porto, Garite, Chicz-DeMet, & Sa ndman, 1998; Warren, Patrick, & Goland, 1992). Stress may also increase proinflammatory cytoki ne production which leads to increased prostaglandin production, increased uterine contractility and pr eterm labor (Chouros, 1995; Turnball & Rivier, 1999). Because of higher stress levels among Black and Hispanic women in studies of ethnically a nd socioeconomically diverse samples, some propose that perceived ethnic and gender disc rimination and their as sociated stressors contribute to greater cumulative stress and in creased risk of advers e perinatal outcomes (Krieger, Rowley, Herman, Avery, & Phillip s, 1993; McLean, Hatfield Timajchy, Wingo, & Floyd, 1993; Rich-Edwards et al., 2001; Stancil, Hertz-Picciotto, Schramm, & Watt-Morse, 2000). The life course explanation of repr oductive disparities among Black and White women is related to initial trajectories that are lower for Black women due to intergenerational effects rela ted to smaller acceleration a nd greater deceleration during sensitive periods and greater cumulative exposur e to stress and risk factors later in life (Lu & Halfon, 2003). The strength of intergen erational effects is illustrated by a study which found that women who are raised in low socioeconomic conditions and marry into higher socioeconomic positions have higher rates of adverse birth outcomes when compared to women born into high SES (Ills ley, 1995). Higher risk of LBW and preterm delivery among high SES Black women after two generations of afflue nce indicate that

PAGE 39

27 high SES may not provide the same level of protection for Blacks (Foster, Bracken, Semenya, Thomas, & Thomas, 2000). These findings indicate that SES and associated living conditions, during both early life and pregnancy, influence perinatal outcomes. This may be particularly relevant for Bl ack women, whose higher SES position is more likely to have been attained within the past generation (Jones, 1998). While the life course perspective will not be directly tested in this study, it supports heightened understanding of proposed explanations for varying trends in perinatal outcomes among Blacks and Whites. Its emphasis on the social, economic, and biological implications of ethnicity are the founda tion upon which the proposed investigation is built. While the window of exposure for the infl uence of residential segregation on perinatal outcomes is not known, the lifecourse pe rspective underscores the importance of subsequent longitudinal investigation in to how experiences, coupled with contextual and structural fact ors influence perinatal outcomes. Individual Factors and Perinatal Outcomes The majority of etiologic research on in fant mortality and morbidity has focused on individual factors. The iden tification of individual determin ants of perinatal outcomes, including maternal age, pari ty, weight gain, smoking and pr enatal care, have been the foundation upon which numerous national and community-based ini tiatives have been developed (Goldenberg et al ., 1996; Schiono, Rauh et al., 19 97). Ethnic disparities in perinatal outcomes continue to exist, de spite these efforts (Ananth, Balsubramanian, Demissie, & Kinzler, 2004; Guyer, Ma rtin, MacDorman, Anderson, & Strobino, 1997; Martin, Hamilton, Sutton, Ventura, Menack er, & Munson, 2003). While these studies have focused primarily on LBW and preterm delivery, th ey provide a foundation and

PAGE 40

28 rationale upon which the study is built. Birth Certificate Data Validity The validity of birth certifi cate data has been examined due to its importance to researchers and decision-makers in assessment of the status of care delivered to pregnant women, individual risk factors a nd the birth outcomes of infant s. Errors have been noted in coding of birth weight (B runskill, 1990) and in undere stimation of birth defects (Watkins et al., 1996); vaginal births after cesarean section and primary cesarean section rates (Green et al., 1998) and hyaline membrane diseas e (Hamvas, Kwong, DeBaun, Schramm & Cole, 1998). Findings reveal that variability exists in the specificity, sensitivity and predictive valu e of birth certificate elem ents. The agreement between medical records and birth certificates vary widely by the data element in question. The greatest variance between birth certificates and medical record s exist in the estimation of maternal medical risk factors, complicat ions of labor and delivery, number of interventions or procedures, congenital anomalie s, and measures of prenatal care. Studies examining the relative validity of birth certific ate data elements ar e acknowledged at the onset of this study. Several noteworthy studies have examined the validity of birth certificate data. An early study of 379 North Carolina birth certif icates and maternal medical records revealed: very accurate reporting for birthwei ght, Apgar scores and method of delivery; fair to good reporting for tobacco use, pren atal care, weight gain during pregnancy, obstetrical procedure and events of labor and delivery; and poor reporting for medical history and alcohol use (Buesche r, Taylor, Davis, & Bowling, 1993). Piper et al. (1993) found low sensitivity for maternal medical ri sk factors, complications of labor and

PAGE 41

29 delivery, and abnormal conditions of the newbor n in a study that exam ined the validity of 1998 Tennessee birth certificates. Clark, Chun-Me i, & Burnett (1997) study revealed that birth certificates overestimated prenatal ca re. Data abstraction from prenatal and intrapartum clinic and hospital records were used to assess birth certificate validity in a study conducted by Dobie et al. (1998). Resu lts showed significant underestimation of complications of pregnancy, number of interven tions and procedures, and prenatal visits. Reichman and Hade’s (2001) assessment of th e validity of birth cer tificate data between 1989 and 1992 from a sample of high risk New Jersey women revealed that sensitivity was low for maternal medical risk factors, co mplications of labor a nd delivery, obstetrical procedures and transfer status. DiGuise ppe, Aron, Ranbom, Harper, & Rosenthal’s (2002) examination of the reliability of bi rth certificate data for women admitted to teaching and non-teaching hospitals in No rtheast Ohio between 1993 and 1995 revealed: perfect agreement for measures of prior obstetrical histor y, delivery type, and infant Apgar scores; substantial agreement for tobacco use, gestational ag e and prenatal care and; slight to moderate agreement for most ma ternal risk factors and co morbidities, and many complications of pregnanc y and/or labor and delivery. A recent study by Roohan et al. (2003) a ssessed the sensitivity, specificity and positive predictive value for birth certificates in four New York State counties. Findings showed that all maternal medical risk factor s and risk factors related to pregnancy and lifestyle risk factors had high levels of specificity, but varying levels of sensitivity, primarily attributable to ra re conditions that are ofte n not documented on the birth certificate. Similar to results of Piper et al (1993), Clark et al. ( 1997) and Debbie et al. (1998), the numbers of prenatal care visits were poorly reported with fewer visits

PAGE 42

30 reported on the medical record. Limitati ons of studies include questionable generalizability due to small geographical region of analysis and hospital variations that may have skewed results. Explanations for lower sensitivity among cer tain birth certificate elements are vast and include issues surro unding site variations in data ma nagement and self-report bias. Risk factors related to subs tance abuse, smoking and drinki ng during pregnancy are also likely to be underreported on both medical reco rds and birth certificate due to self-report. Although medical records are considered in most validation studies as the referent, gold standard, hospital staff have access to a broa der number of information sources when completing the birth certificate, including the mother and father. Estimates of prenatal care are speculated to be less accurate on me dical records because records are often last updated during the 36th week of pregnancy, when a potential of 4 additional visits may occur by 40 weeks, or full term. Cross-site vari ations in protocol for collecting data, nonclinical personnel recording da ta, little or no auditing to ensure data validity and reliability in hospitals are also among explana tions for lower sensitivity of specific birth certificate elements. Limited funding warrants ta rgeted standardization efforts for data elements that are found to be less reliable. Despite noted limitations, birth certificate data is a potentially rich source of information, with wide application for both research and practice and will provide the best means, to date, for examination of the proposed research questi ons. Birth certificates are among the most recognized and commonly used data sources for research and program planning. More states are implemen ting error-checking initiatives to improve quality or vital data (Alexa nder & Petersen, 1997). Birth cer tificate data cover multiple

PAGE 43

31 years and contain large amount s of standardized information about U.S. women and potential risk factors for a dverse perinatal outcomes. Gestational Age Es timation Specificity Accurate estimation of gesta tional age has important implications for research and program planning. Resources are allocated and programs are developed and evaluated based on data obtained from vital records. They are used to understand regional trends, to calculate the proportion of pret erm delivery, term and post te rm births in the population and to determine individual newborn’s risk stat us. They are also used in the computation of measures of intrauterine gr owth and adequacy of prenatal care. Certificates of live births are among the most freque ntly utilized sources of ge stational age data. Prenatal errors in the estimation of gestational age im pact clinical decisions regarding delivery timing. Classification of infants as term or preterm, based on live birth certificates may subsequently result in misclassification errors that influence the anal ysis and reporting of birth outcome trends. Prior to the 1980s gestational age (GA) wa s calculated from the date of the last menstrual period (LMP). Subsequently, clin ical estimation (CE), particularly early ultrasound examination (EUS), the measurement of the fetal biparietal diameter between 16-18 weeks of gestation, has resulted in di fferences in methods of gestational age estimation (Blondel et al., 2002). It is widely reco gnized that EUS-based gestational age estimates result in lower average estimate of GA and a higher pr evalence of preterm delivery when compared to LMP gestationa l age estimations (Alexander, Tompkins, Petersen, Hulsey ,& Mor, 1995; Goldenbe rg et al., 1989; Hogberg & Larsson, 1997; Kramer, McLean, Boyd, & Usher, 1988; Yang et al., 2002). Due to discrepancies in

PAGE 44

32 LMP and EUS estimates both among and with in countries, geographic and temporal trends in preterm and post term birth will re main difficult to interpret. (Blondel et al., 2002). Limitations of the LMP gestational measure have been reported in the scientific literature (Hall, Carr-Hill, Fraser, Campbe ll, & Samphier, 1985; Milner, & Richards, 1980; Taffel, Johnson, & Heusser, 1982;). LMP-based measures produce values inconsistent with birth we ight (Alexander, Tompkins, & Cornely, 1990; David, 1980). The majority of errors in the estimation of LMP-based gestational age are due to biological errors primarily attributable to ina ccurate reporting (identif ication or recall) of the last menstrual period (LMP), more freque nt occurrence of delayed rather than early ovulation, sporadic bleeding dur ing pregnancy and unrecogni zed abortions (Battaglia, Frazier, & Hellegers, 1966; Joseph et al., 1998; Milner, & Richards, 1974; Treloa, Behn, & Cowan, 1967). These reporting problems have been observed more frequently among women of lower educational and SES. There is also a question regard ing the accuracy of this technique, particularly for determin ing preterm delivery am ong ethnically diverse populations (Alexander, Kogan, & Nabukera, 1992; Sanders et al., 1991; Spinnato, Sibai, Shaver, & Anderson, 1984). Women who lack in formation on their LMP are more likely to have adverse pregnancy outcomes (B ueckens, Delvoye, & Robyn, 1984; Wenner & Young, 1974). These studies may be biased in that women who experience adverse birth outcomes are excluded from these studies, underestimating the proportion of adverse perinatal outcomes. When compared to EU S, LMP is more likely to overestimate gestational age with a significant number of errors occurring in either direction. LMP distribution contains a prominen t postterm-tail in live birth data, primarily attributable

PAGE 45

33 to maternal errors in re cognizing or reporting LMP. The CE measure of gestational age is widely available on state vital record databases. The National Center for Vital Stat istics has substituted the CE value for the LMP based gestational age when the LMP date is incomplete or incompatible with birth weight. The extent to which public health pla nners and researchers fo llow this strategy is unclear. The CE measure provides a gestational age distribution that is closer to conventional expectations. Fewer implausible out-of-range values exist and the measure is more highly correlated with BW. Some attr ibutes of the measure are, however, sources of concern. To the extent that the CE meas ure more closely corre sponds to BW and in some or many cases could conceivably have been estimated after delivery based upon knowledge of the BW, the variation in BW for each gestational age de creases. It has been shown that variations in BW by gestati onal age, for example SGA, are important indicators of morbidity and mortality ris k. Any reliance on BW to estimate gestational age could result in overly censo red intrauterine growth curves and in the loss of important risk information about the newborn. The CE measure is limited to what is known by the person completing the Certificate of Live Bi rth. Ultrasound and othe r obstetric measures, including fundal height and fetal heart tones and pediatric examina tions of the physical and neurological characteris tics of the newborn are among th e types of information used to establish the CE (Alexander, Tompkins, Pe tersen, Hulsey, & Mor, 1995). It is not clear which methods are given the most emphasis acro ss clinical sites and how availability and use may vary. It may be argued that errors inherent in the LMP-ba sed measure are more

PAGE 46

34 random and less influenced by these fact ors although the LMP-based measure is perceived to be less reliab le than the CE measure. In spite of the evidence on the limita tions and strengths of gestational age estimations, the majority of gestational age estimations on birth certificates is based on self-reported LMP, or the difference between the date of delivery and the last menstrual period. Only a small fraction (5%) is ba sed on the CE (Ananth, Balasubramanian, Demissie & Kinzler, 2004; Ananth & Pla tt, 2004; Taffel, Ventura, & Gay, 1989). Contextual Factors and Perinatal Outcomes A relatively small number of studies ha ve focused attention on the association between local area characteristics and birt h weight. Socioeconomic characteristics associated with areas of residence (educa tion, income, poverty leve l and housing value) have been increasingly used to examine social inequalities in health and have been found to be related to mortality and other health outcomes (Diez-Roux et al., 1997). Roberts (1997) found that neighborhood economic hard ship and housing costs were positively associated with LBW, while community SE S, crowded housing and high percentages of young and African American residents were ne gatively associated with LBW. Studies have been examined at the city (Buka et al., 2002; O’Campo et al.,, 1997; Pearl, Braveman & Abrams, 2001; Rauh et al., 2001 ; Roberts, 1997) national (Gorman, 1999) and international levels (Sloggett & Joshi, 1994). None have advanced a coherent theoretical framework that considers the macr o level structures that shape residential contexts and effect perinatal outcomes. Much of the evidence linking residential disadvantage to LBW also util ize conventional regression met hods that are inappropriate

PAGE 47

35 for multilevel data (e.g., individual s nested in residential areas). Multilevel analyses of determinants of a dverse perinatal outcom es reveal distinct risk factors associated with subgroups of preterm delivery and LBW infants. Research findings suggest that LBW and gestational age have differential indi vidual and contextual predictors (English et al., 2003). Others illust rate that individual and population effects on health may vary by ethnicity (Kaufman, Dole, Savitz, & Herring, 2003; Rauh et al., 2001). Most studies have been limited to the investigation of determinants of LBW and preterm delivery. Significance of Health Disparities SES and ethnicity have been associated with a variety of health outcomes. Disparities in life expectancy and health st atus are widest between Blacks and Whites. Blacks have disproportionate mortality from cardiovascular and cereb rovascular diseases, cancer, homicide, infant death, diabetes, and AIDS. These persistent trends help to explain the increased focus of public health research agendas on ethnic disparities in health. Most public health investigators agre e that consideration of social and environmental factors are crucia l in investigating the determinants of health outcomes in general, and ethnic health disparities in particular. Genetically determined diseases that are more prevalent among certain ethnic groups (e.g., Crohn's Disease among Whites; Sickle Cell Anemia among Blacks) do exist. These designations, how ever, are socially constructed and do not capture within-group va riability for many health morbidities and mortalities (Cooper et al., 2003; Hessol, 1998; Jones, 2001; Rowley, 1994). Cooper et al. (2003) cited wide variation in susceptibil ities to chronic diseases among people of

PAGE 48

36 African-descent in all large c ontinental populations. Palloto, Collins, and David, (2000) found that among American and Caribbean Black women, American women had higher rates of LBW infants.. A comparison of Bl ack women who were American, Haitian, West Indian, Cape Verdean, Hispanic and Af rican revealed that American women were at increased risk for adverse perinatal outco mes (Friedman et al., 1993). Further research indicates that foreign born women have in fants with better peri natal outcomes when compared to U.S. born Black mothers (Cabral, Fried, Levenson, Amaro, & Zuckerman, 1990; Fang, Madhavan & Alderman, 1997). These studies indicate the heterogeneity in perinatal outcomes within ethnic groups a nd the unique contribution of structural and contextual factors to the hea lth of U.S. born Black women. Most health disparities pers ist at all levels of SES despite the disproportionate numbers of minorities represented in lowe r SES groups in the U.S. Researchers challenging the notion that health differences are solely explained by poverty have found that the disparities exist at every level of the social hierarchy, exhibi ting a social gradient rather than a threshold effect (Wilkins on, 1992; Macintyre, 1994; Marmot, Bobak, & Davey Smith, 1995). If Black-White differences in health are not simply attributable to group differences in SES, research is needed to understand the factor s that influence the relationship between ethnicity and health.

PAGE 49

37 Structural Explanations for Health Disparities Residential Segregation Significance Residential segregation is the physical separation of gr oups in residential contexts or the differentiation of two or more groups among dimensions of a given social space (Acevedo-Garcia & Lochner, 2003). Resident ial segregation has received increased attention as scholars have sought to better understand et hnic health disparities. Residential segregation effects have been examined primarily am ong Black minorities. These studies show that greater segregation leads to increas ed stress and poorer health. Residential segregation in urban areas has a significant effect on both Black infant and adult mortality rates, after adjusting for other socioeconomic and demographic characteristics (Collins & Williams, 1999; Massey, White & Phua, 1996; Peterson & Krivo, 1999; Poldenak, 1993, 1996; Shihad eh & Flynn, 1996). Research that has explored the effects of resi dential segregation among Whites have yielded inconsistent findings (Collins & Williams, 1999; Fang, et al., 1998). Residential segregation has historical and current implications for di sparate educational attainment, employment opportunities and housing quality that may me diate the relationship between social position and health (Williams & Collins, 2001). Few studies have investigated the association between resi dential segregation and perinatal outcomes.

PAGE 50

38 Historical Context Residential segregation reflects d ecades of individual and group-level discrimination that is based on racism, w hose primary goal was to maintain social distance between defined groups (Pettigrew & Meertens, 1995). Historically, segregation was driven by majority perceptions of mi nority group inferiority with discriminatory policies led by real estate industry, federal housing policy, banking institutions and organizations that sought to ensure the restriction of Blacks from housing choices relegating them to substandard residential areas (Williams & Collins, 2001). Conceptualization and Operational Dimensions The broad definition of residential segr egation masks the different dimensions by which groups may be segregated. Massey and Denton (1988) conducted a seminal study that has provided the foundation upon which the conceptualization and measurement of residential segregation has been operationaliz ed in subsequent research. Residential segregation is conceptuali zed across five dimensions: Centralization, Concentration, Clustering, Evenness and Exposure. Centralization is the degr ee to which members of an ethnic group live near to an urban area. Concen tration measures the physical space that is occupied by an ethnic group in a geographical area. Clustering measures the degree to which ethnic minority areas cluster together in space. Evenness represents the degree of spatial separation between ethnic groups. E xposure measures a group's experience of segregation by the degree to which members co me into contact with one another or with another designated group. While Evenness a nd Exposure are correlated, they measure different markers: Exposure, rath er than Evenness, is dependen t on the relative size of the comparison groups.

PAGE 51

39 Massey and Denton (1988) utilized prin cipal components factor analysis to evaluate 20 residential segregation measures and to categorize them according to the five segregation dimensions. Results indicated that Evenness and Exposure cumulatively explained approximately 72% of the variance in residential segreg ation. Centralization, Clustering and Concentration segregati on dimensions explained only 28% (1988). Evenness and exposure indices are the reside ntial segregation dimensions that were examined in this study. Index of Dissimilarity. The Index of Dissimilarity, an evenness di mension of residential segregation, is the degree to which ethnic groups share areas in common. This meas ure represents the evenness dimension of residen tial segregation, which is the relative distribution of two groups within a given geographical unit or the degree of spatial separation between groups. The Index is derived from the Lore nz curve, through plotting the cumulative proportion of minority groups against the cumulative proportion of majority groups across areal units (Massey & Denton, 1988). St ated another way, the Index measures the proportion of a given minority group that would have to relocate (move either in or out of a given geographic area) to achieve an even distribution of ethnic groups. Index values ranged from 0-1, with 0 representing no segregation and 1 representing complete segregation (Massey, 1996; Massey & De nton, 1988). Massey and Denton's (1988) analysis of correlation and factor pattern matrices among various evenness indicators showed that little additional information is provided by other m easures of evenness compared to that which is included in the I ndex of Dissimilarity. The Index represents the most comprehensive, frequently examined and easily calculated measure of evenness

PAGE 52

40 (1988). Current residential segregation researchers believe that the Index of Dissimilarity has less relevance to studies related to h ealth outcomes (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia, Lochner, Osypuk, & S ubramanian 2003) and is less associated with individual and residential indicators when compared to other measures (Denton, 1994). These considerations are used to par tially justify th e inclusion of two exposure measures: The Isolation Inde x and the Interaction Index. Isolation Index. The Isolation Index an exposure dimensi on of residential se gregation, measures the degree to which minority group members are exposed to one another (and not the majority group). The index is operationalized as the probability that a randomly drawn minority group member will share an area of residence with another similar minority group member. The index represents the mi nority-weighted average of each unit's minority proportion with values that ranged from 0 and 1. This index is utilized in this study du e to potential race-specific effects. The Isolation Index is a theoretica lly relevant dimension of resi dential segrega tion that may be an important mechanism through which Black perinatal trends are manifested. Because residential segregation is based on effo rts to avoid social contact, the degree of isolation experienced by ethnic minority groups are distinct. Findings have proven more deleterious effects of residen tial segregation on both individua l and areas of residence for Blacks (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia et al., 2003; Collins & Williams, 1999).

PAGE 53

41 Trends in Residential Segregation The United States has experienced a stea dy increase in residential segregation with declines that have only occurred within the last three decades. Residential segregation rose during each decade between 1890 and 1970 (Cutler, Glaeser, & Vigdor, 1999). Segregation began to decline in the 1970s with sharpest av erage decreases of almost 10% across metropolitan st atistical areas (MSAs) (1999). While national trends indicate that residential segregation has decreased, investigation of trends by ethnicity, MSA size and regi on are not homogeneous. An examination of residential segregation am ong Blacks, Hispanics, Asians and Pacific Islanders, American Indians and Alaska Na tives from 1990 to 2000 indicated that only Blacks experienced declines across all di mensions described by Massey and Denton (1988). Despite this trend, segregation rates are highest am ong Blacks (Iceland, Weinberg, & Steinmetz, 2002). Size of the geographical unit influences levels of residen tial segregation. The largest MSAs (populations >1,000,000) have high er residential segregation levels when compared to those that are respectively smalle r. In turn, mid-sized MSAs (populations of 500,000 to 999,999 persons) are more segregated th an smaller MSAs (Glaeser & Vigdor, 2001; Iceland, Weinberg, & Steinmetz 2002). Regional trends in residential segrega tion over the past two decades vary. The West is more integrated followed by the Sout h, while the Northeast and Midwest are both highly segregated when measured using the Dissimilarity Index (2002). With regard to change over time, while residential segr egation has decreased both nationally and regionally, the South experienced the largest regional reducti on, with equal change across

PAGE 54

42 the Midwest, Northeast and the West (G laeser & Vigdor, 2001) between 1990 and 2000. Investigators attribute declines in resident ial segregation to a number of factors. First, an increasing number of Blacks reside in MSAs that were previously composed of predominantly White residents; the percentage of census tracts that are <1% Black has decreased from 61.8% in 1960 to 23.1% in 2000 (2001). The argument for the influence of Black movement to White residential area s (rather than White movement into Black residential areas) is supported by a relatively constant number of census tracts with a Black share of at least 80% between 1990 and 2000. Further, MSAs with both rapid declines and large influxes of Black resident s have had the most rapidly decreasing levels of residential segregat ion (7% and 6.4 % average declines, respectively). Interpretation of these findings are that Black flight from areas that are generally more highly segregated result in declines in levels of residentia l segregation, while new MSAs to which Blacks move are characterized by more integrated settlement patterns (Glaeser & Vigdor, 2001). Trends in residential segregation indicat e that the U.S. has made great strides toward greater integration across the nation. Despite thes e promising findings, highly segregated areas are character ized by Dissimilarity indices >0.6 and the nati onal Index of Dissimilarity was 0.652 in 2000 (2001). The conseque nces of this structural force point to the lingering and noteworthy effects that ar e described in the section that follows.

PAGE 55

43 Consequences of Residential Segregation Residential segregation has been associat ed with ethnic disparities in individual well-being (e.g. employment or education) a nd the contexts of residential areas (e.g. concentrated poverty or poverty exposure) (G alster & Killen, 1995; Galster & Mikelsons, 1995; Logan, 2002; Massey, Condran, & Dent on, 1987) across metropolitan areas. Ethnic disparities in the effects of residential segreg ation at the individual and contextual levels has been hypothesized to account for the poor er health outcomes of Blacks (AcevedoGarcia & Lochner, 2003; Acevedo-Garcia et al., 2003; Collins & Williams, 1999; Ellen, 2000; Williams, 1997, 2001). Although residential segregation varies across geographical areas and groups, few studies have specifically examined whether health disparities between Blacks and Whites are related to variations in reside ntial segregation. Employment opportunities. The negative effects of segregation center on limited opportunities for overall health and well-being. Segregation is associ ated with more restricted employment, particularly for ethnic minorities (Collin s & Williams, 1999; Frazier, 1957; Wilson, 1980, 1987). The determinants of reduced job opportunities include rapid urbanization, industrialization and immigration. The historic al migration of Blacks from the South to the Northeast and Midwest was followed by Whites and middle-class Blacks leaving cities to live in the suburbs. A spatial mismatch in urban areas was the result; as urban centers were populated by larger proportions of Black residents, the availability of lowskilled jobs declined and high skilled jobs increased (Kasarda, 1989). This trend has contributed to Black joblessness over time.

PAGE 56

44 Family structure. Compromised quality of life follows th e restricted employment opportunities for minorities, particularly in urban, highly se gregated areas. Research linkages between marriage rates, employment opportunities and income for males indi cate that high rates of Black male unemployment in segregated urban areas has led to increased rates of female-headed households (Testa, Astone, Krogh, & Neckerman, 1993). The association between employment and marital status for males has long been established (Bishop, 1977). Housing quality and ownership. Smaller real estate return s are another consequence of segregation, particularly for Blacks. Despite increases in housing equality in the U.S., fewer Blacks have experienced this primary source of wealth for the American Family. This is due, in part, to the increased number of minoriti es (particularly Black) in less desirable housing areas (Logan & Alba, 1993; Oliver & Shapiro, 1997). Owning a home may be an avenue of assimilation for minority groups in mainstream society. Homeownership is a proxy for wealth accumulation and is frequently a prerequisite for living in certain locations especially suburban areas where the housing stock is predominantly composed of owne r-occupied single family homes (Alba & Logan, 1992). In an examination of Black mobility, South and Crowder (1998) demonstrated that homeownership increas es the likelihood of moving to a more ethnically integrated residential area.

PAGE 57

45 Inappropriate land uses and maladaptive behaviors. Neglect and deterioration of physical environments and a disproportionate representation of undesirable la nd use also characterize highly segregated areas. Policy makers are more likely to cut important sp ending on social services in areas where residents are unlikely to mount political opposition (Shihadeh & Flynn, 1996; Wallace, 1990, 1991). Withdrawal of fire and police servic es from such areas can trigger migration of arsonists, drug dealers a nd other individuals who may in itiate high-risk, maladaptive behaviors (Greenberg & Schneider, 1994). Socioeconomic status. Trends indicate that middle class status does not alleviate Blacks from the plight of segregation. Education and income level do not substantially change the value of the segregation index for Blacks. In contrast to other minorities (Asian and Hispanic), middle class status does far less to ameliorate segregation effects among Blacks (Denton & Massey 1988, 1993). Darden (1987) calculated re sidential segregation indices (Measured by the Index of Dissimilarity) by Blacks and Wh ites within similar income strata in a Kansas City, Missouri metropolitan area in 1980. Residential segregation among the most affluent (those with median income of more than $75,000) Blacks and Whites was nearly 10 points higher than residential segregati on in the lowest income categories (1987). Denton and Massey (1988) demonstrated th at Black segregation does not vary by affluence when the Black-White Index of Dissi milarity is calculated by income level. Massey and Fischer's (1999) analysis of 1990 income segregation patterns revealed persistently higher levels of segregation for Blacks such th at compared to their minority counterparts (e.g. Asian and Hisp anics) affluent Blacks were most segregated from non-

PAGE 58

46 Hispanic Whites (1999). Pathways Research on ethnicity and SES has f ound persistent, independent effects of ethnicity on health outcomes (LaVei st, 1999; Polednak, 1996, 1997; Williams, 1996). Discrimination along ethnic lines is a widely recognized, yet newly investigated area in studies of factors related to health dispari ties. The existence of non-genetic explanations of ethnicity's contribution, inde pendent of SES, are implied by research findings in this area (Kaufman, Cooper, & Mc Gee, 1997; Williams, 2002). The role of discrimination and related st ressors are important in discussions of ethnic health disparities. Discrimination has been a well-documented determinant of the marginalization of ethnic minority and lo w SES groups (Gee, 2002; Krieger 2001; Ren, Amick & Williams, 1999). The stressors that result from individually experienced and institutionally sanctione d discriminatory practices are impor tant to discussions of health disparities. Stress is a well r ecognized correlate of physical and mental health (LaVeist, 1993; 2001; Lillie-Blanton & LaVeist, 1996; Li nk & Phelan, 1995). A socially deprived life experience may also result in psychologica l stress responses that are deleterious to health (Farmer & Ferraro, 1997; Krie ger & Sidney, 1996; Livingston, 1994). The isolation resulting from increased residential segregatio n may keep persons within these areas of residence from role models of stable employment and the social networks that may provide health promoti ng information and behavior. Williams' (1996). framework conceptualizes the influence of societal structures on ethnic discrimination These social structures and institutions differentially influence Whites and ethnic minorities through social stratification, geogra phic isolation, decrea sed opportunities for

PAGE 59

47 social ties, lack of medical care and et hnic bias (Collins & W illiams, 1999; Williams, 1997). The structural conditions in residen tially segregated environments may induce cultural responses that weaken the commitmen t to norms and values that are important for socioeconomic mobility. Other mechanisms that link residential se gregation to health have been identified. Poverty is associated with poorer nutrition and less access to medical services. Increased stress levels associated with poor economic conditions are also related to weaker social support systems in communities characterized by the concentrated poverty that follows segregation (Roberts, 1997). Individual and collective organizational participation that foster social support networks are lower in concentrated poverty areas (Shihadeh & Flynn, 1996). The section that follows descri bes the association between residential segregation and health outcomes. Residential Segregation and Health Adult Mortality The relationship between residential segreg ation and adult all-cause mortality has been illustrated in severa l studies (Collins & Williams, 1999; Hart, Kunitz, Sell, & Mukamel, 1998; Jackson, Anderson, Johns on & Sorlie, 2000; LaVeist, 1992, 1993; Polednak, 1996, 1997). Most investigations employed cross-sectional study designs. Varying geographical units were examined to explore residential segregation and mortality associations. Polednak's (1993) investigation of allcause mortality among SMSAs showed that the Black/White ratio of age-adjusted mo rtality was significantly associated with Black/White residential segregation. A pos itive association for Black men and women

PAGE 60

48 ages 15-44 was also evident. Hart et al.’s (1998) examin ation of SMSAs showed higher mortality rates for Blacks and Whites in cities characterized by higher residential segregation. Fang et al. (1998) investigat ed New York City and found that Black mortality rates significantly differed by location, despite higher mortality rates of Blacks compared to Whites. Differences were positively associated with the city's ethnic residential segregation (measured using the Index of Dissimilarity) patterns. Collins and Williams (1999) conducted an authoritative investigation of the effects of ethnic residential segregation (measured using the Inde x of Dissimilarity and the Index of Social Isolation) on a variety of mortality indicators (age-adjusted all-cause mortality, cancer mortality, cardiovascular disease mortality and homicide). Results demonstrated that residential segregation was nega tively related to the health of Whites and Blacks. Patterns of the association were weaker for the former group than they were for the latter (1999). Jackson et al. (2000) investigated census tracts and found that all-cause mortality increased as residential segregation increased in Black (ages 25-44) and non-Black (ages 45-64) populations. Crime Positive associations have been found in studies examining the relationship between residential segregation and crime. Re sults yielded mixed results similar to the studies described above. Positive associations have been demonstrated between residential segregation indi ces and: all-group homicide rates (Logan & Messner, 1987; Rosenfield, 1986); Black-White differences in homicide rates (Potter, 1991); and Black homicide rates (Peterson & Krivo, 1993). The positive association between segregation and homicide rates also has been shown for Whites and not Blacks (Sampson, 1987).

PAGE 61

49 Shihadeh and Flynn (1996) found a positive association between segregation and Black homicide when using two measures (The Isol ation Index and The Inde x of Dissimilarity). The Isolation Index was a st ronger predictor of homicide rates than the Index of Dissimilarity (1996). Investigation of the influence of community violence (measured using the number of violent crimes per 1000 re sidents in each police district) on perinatal outcomes revealed a significant increase in S GA births in residentia l areas characterized by higher violence, a characteri stic of highly segregated geographical areas (Collins & David, 1997). Infant Mortality Research on the association between residential segregation and perinatal outcomes has focused primarily on infant mort ality. The earliest empirical investigation of the association between residential segr egation and perinatal outcomes demonstrated that both Black and White infant mortality ra tes were highest in highly segregated areas with a larger proportion of Bl ack residents (Yankauer, 1950). Jiobu's (1972) path analysis also showed that a positive association existe d between residential segregation and infant mortality. An investigation of Black-White di fferences in infant mortality within large and mid-size U.S. cities revealed a positive association between segregation (measured using the Index of Dissimilarity) and Black infant mortality (LaVeist, 1989). This relationship persisted independent of povert y rates. When rates of LBW among Black infants and unwed mothers were controlled for, the association between residential segregation and infant mortality remained. White infant mortality rates were largely unaffected by residential segr egation indices (1989).

PAGE 62

50 Two related cross-sectional studies i nvestigated the association between residential segregation and infant mortality in standardized metropolitan statistical areas (SMSAs) using multiple regression analyses. Both demonstrated mixed effects. The initial study (Polednak, 1991) employed mo rtality data between 1982 and 1986. Findings demonstrated that the residential segregat ion index (measured using the unevenness of residential distribution of African-Americans) was the only significant predictor of total infant mortality rates for census tracts in each SMSA when other factors (female householder, poverty prevalence and median fami ly income) were included in regression models. Similar results were found in analysis among African Americans (1991). The second study assessed differences in infant mortality rates between Blacks and Whites over a nine-year (1982 1991) period (Pol ednak, 1996). While the 1980 residential segregation index was significan tly related to infant mortality from 1982 to 1986, the rate of unmarried Black women (and not the 1990 se gregation index) was the only significant predictor of infant mortality from 1989 to 1991. Noteworthy in both studies were high correlations between residentia l segregation indices and the prevalence of poverty and the rate of unmarried mothers, respectively (1996). These ecological investigations we re characterized by well-recognized limitations. The lack of indivi dual risk factors provided onl y partial understanding of the context within which identified trends exis ted. In addition, by utilizing census measures of SES such as the prevalence of poverty a nd median family income, a population based measure of SES was used as a proxy for th e individual level. Notwithstanding, these studies provided a new direction of inves tigation concerning the role of residential segregation in predicting risk factors for adve rse perinatal outcomes.

PAGE 63

51 Studies linking residential segregation to h ealth outcomes suffer from a number of research limitations. First, the effects of residential segregation often differ by ethnic group. Black-White socioeconomic disparities ha ve clear spatial expr essions, previously cited at the metropolitan area level along bot h central-city/suburban lines and across residential areas (Altshuler, Morrill, Wo lman, Mitchell & The Committee on Improving the Future of U.S. Cities through Improved Metropolitan Area Governances, 1999). Further, the significantly more adverse soci oeconomic and residential quality measures among Blacks when compared to Whites have b een attributed to differential effects of residential segregation (Massey, 2001; Massey et al., 1987; Massey & Denton, 1993; Williams & Collins, 2001). Little is known about the effects of resi dential segregation on perinatal outcomes that increase risk for in fant mortality and morbidity risk. Further, whether the differential trends in the adve rse effects of reside ntial segregation along ethnic lines are also applicable to disparitie s in perinatal outcomes remains to be seen. Second, while the influences of residential segr egation are manifested at both individual (e.g., education, employment opportunities) and contextual levels (e.g. concentrated poverty) the majority of the re search on the residential segreg ation and health relationship are based on single level, aggregate analys es (Acevedo-Garcia et al., 2003; AcevedoGarcia & Lochner, 2003; Ellen, 2000). Third, the majority of studies have examined mortality rates as the outcome of interest The myriad of social and environmental consequences of ethnic residential segregati on infer the need for in creased investigation of how specific characteristics of segregated areas may be associated with risk for causespecific health outcomes.

PAGE 64

52 Finally, less is known about for whom and under what conditions residential segregation may be protective. The conseque nces of residential segregation for nonBlack ethnic groups are not well understood. In Polednak’s (1993) study, the residential segregation Index of Dissimilarity was positiv ely associated with all cause mortality for Blacks, but inversely related for Whites (P olednak, 1993). Others have suggested that residential segregation will benefit Whites who reside outside of segregated areas (Massey & Denton, 1993). Evidence also shows th at segregation is costly for Whites on a broad range of dimensions (Roisman, 1995) Some have found that residential segregation was unrelated or inversely rela ted to mortality for Whites (LaVeist, 1989; Polednak, 1991, 1993, 1996). Others have found th at segregation was positively related to homicide rates for Whites but not for Bl acks (Sampson, 1985). LeClere et al. (1997) found that residence in segreg ated areas predicted higher mortality rates for Blacks as well as for Whites. Research Questions and Hypotheses The purpose of this study was to investig ate the relationship between residential segregation and adverse perina tal outcomes (LBW, preterm delivery and SGA births) in Florida, Georgia and Louisiana. Specifica lly, this study determined the independent effects of the level of resi dential segregation on the like lihood of adverse perinatal outcomes after controlling for contextual and individual factor s. The study also determined whether the relati onship between residential segr egation and the likelihood of adverse perinatal outcomes adverse perinata l outcomes is modera ted by ethnicity and median income, respectively. The following research questions and hypotheses were investigated:

PAGE 65

53 Research Question 1: Is there a relationship between residential se gregation and adverse perinatal outcomes among mothers in Florida, Georgia, and Louisiana? Hypothesis 1: The level of re sidential segregation is posi tively associated with the likelihood of adverse birth outcomes af ter controlling for contextual and individual factors. Hypothesis 1A: The level of residential segregation is positively associated with the likelihood of preterm delivery after c ontrolling for contextual and individual factors. Hypothesis 1B: The level of residential segregation is positively associated with the likelihood of low birth weight after controlling for contextual and individual factors. Hypothesis 1C: The level of residential segregation is positively associated with the likelihood of small for gestational age births after controlling for contextual and individual factors. Research Question 2: Is the relationship be tween residential segr egation and perinatal outcomes moderated by ethnicity?

PAGE 66

54 Hypothesis 2: The relationship between re sidential segregation and the increased likelihood of adverse pe rinatal outcomes is moderated by ethnicity. Hypothesis 2A: The relationship between residential segr egation and the increased likelihood of preterm de livery is moderated by ethnicity. Hypothesis 2B: The relationship between residential segr egation and the increased likelihood of low birth we ight is moderated by ethnicity. Hypothesis 2C: The relationship between residential segr egation and the increased likelihood of small for gesta tional birth is moderated by ethnicity. Research Question 3: Is the relationship be tween residential segr egation and perinatal outcomes moderated by income? Hypothesis 3: The relationship between re sidential segregation and the increased likelihood of adverse perinatal ou tcomes is moderated by income. Hypothesis 3A: The relationship between residential segr egation and the increased likelihood of preterm de livery is moderated by income. Hypothesis 3B: The relationship between residential segr egation and the increased likelihood of low birth weight is moderated by income.

PAGE 67

55 Hypothesis 3C: The relationship between residential segr egation and the increased likelihood of small for gestati onal age birth is moderated by income.

PAGE 68

56 CHAPTER THREE: METHODOLOGY Study Design The research employed an observational study design that was cross-sectional utilizing secondary data. The study was obs ervational in that groups are formed by sample population status for each respective adverse perinatal outcome rather than randomization into treatment and control gr oups. The study selected was cross-sectional due to the proposed association between prev alence of adverse perinatal outcomes and prevalence of exposure to residential segrega tion. The use of preexis ting birth certificate and U.S. Census Bureau data explains the secondary nature of the study. Study Population The inclusion criteria for the selection of birth certificate data sources were based on regional rates of adverse perinatal outcomes and the availability of data needed to analyze residential segregation effects at the census tract level. Regional findings confirmed that LBW, neonatal mortality rates (death of live born from birth to < 28 days of life) and infant mortality (death of live born before first birthday) rates were highest in the Southern United States (N ational Center for Health Statistics, 2004) (Table 2).

PAGE 69

57 Table 2 Perinatal Outcomes for the United Stat es and Selected Geographical Divisions LOW BIRTH WEIGHT RATE NEONATAL MORTALITY RATE INFANT MORTALITY RATE United States *7.69 **4.6 **6.9 South Atlantic Delaware, District of Columbia Florida, Georgia, Maryland North Carolina, South Carolina Virginia, West Virginia *8.63 **5.5 **8.0 East South Atlantic Alabama, Kentucky, Mississippi, Tennessee *9.45 **5.5 **8.7 West South Central Arkansas, Louisiana, Oklahoma Texas *8.00 **4.1 **6.7 Source: National Center for Health Statistics, 2004 *Selected Years 2000-2002 **Selected Years 1999-2001 Census tracts were conceptualized as proxies for the residential contexts that shape individual health opport unities and risks. Justification for selection of this level of aggregation is discussed in detail later in this chapter. The Southern United States meeting the criteria for inclusion were Flor ida, Georgia and Louisiana. These states represent similar demographic profiles and bi rth trends (Figure 1 and Figure 2). Birth certificate data was examined from January 1, 1999 to December 31, 2001.

PAGE 70

58 0% 10% 20% 30% 40% 50% 60% 70% Florida GeorgiaLouisianaUnited States Proportion of Population >100% of Poverty Medicaid Enrolled Adapted from The Henry J. Kaiser Family Foundation, 2004 Figure A. Population Characteristics of Whites for Selected States 2002-2003 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% FloridaGeorgiaLouisianaUnited States Proportion of Population >100% of Poverty Medicaid Enrolled Adapted from The Henry J. Kaiser Family Foundation, 2004 Figure B. Population Characteristics of Blacks for Selected States 2002-2003

PAGE 71

59 The study population consis ted of: non-Hispanic White and Black women; singleton live births; and subjects who are pr imaparous. The population was restricted to singleton live births to primaparous women in order to control for unique risk factors associated with previous pregnancies and multip le gestations. The decision to exclusively include primaparous women was also driven by the fact that the birth certificates examined provided no information on the structur al or contextual factors that characterize prior pregnancies. Exclusion criteria for study participan ts included: women who were foreign residents at the time of delivery due to mo re positive birth outcomes among women who were not born in the U.S. (Cabral et al.,1990; Friedman et al., 1993: Palloto et al., 2000), births to women less than 15 years old or gr eater than 49 years old since there are very few pregnancies in these extreme age strata; birt hs delivered at greate r than or equal to 45 weeks due to increased errors in gestational age at indicated postterm gestations (Kramer, McLean, Boyd, et al., 1988); stillbirths; a nd births with improbable birthweight and gestational age combinations. Sample Size National and state-specific (e.g., Florida, Georgia and Louisiana) power analyses were conducted for each adverse perinatal outcome to determine the required sample size for this study. Assumptions for all analyses included a type 1 e rror rate of 5% and analyses that would yield 80% power. National and state-specif ic live birth ratios (5.13:1 and 2.28:1, respectively) were used to repres ent the proportion of unexposed (White) to exposed (Black) (National Center for Hea lth Statistics, 2002, 2003). The power analyses

PAGE 72

60 for SGA births were conducted based on nationa l rates due to no identified state-specific rates in the reviewed literature. National S GA birth rates were taken from a recent study conducted by Ananth, Balasubramanian, Demi ssie, & Kinzler (2004). Power analyses were conducted using Epi Info versio n 3.3.2. Figure 3 illustrates sample size computations for each perinatal outcome based on available state and national rates. Figure C Sample Size and Relative Risk Calculations for Selected Perinatal Outcomes 0 20000 40000 60000 80000 100000 Relative Risk Sample Size National-Preterm State-Preterm State-LBW National-SGA National-LB W National-Preterm 30333138027944519936872765216217421437 State-Preterm 496652254610311845359894481349428142317 State-LBW 504602298013235866961574622361529162408 National-SGA 6083427628158701036623445499429134512844 National-LBW 8122936903212041358498147356573846223813 0.10.150.20.250.30.350.40.450.5 Based on the aforementioned power analyses assumptions and calculations, a sample size of 36,903 live births, comprised of at least 6,020 Black neonates, was necessary for this study. This sample size estimation provides the most generous estimate when comparing national and state-specific rate s calculated to detect a rela tive risk difference of 15%.

PAGE 73

61 Nested Data Structure A three-level nested data structure was used in this study. Level-1 (individual) data was taken from certificates of live birth and contains individual risk factors (control variables) and outcomes variables. Level 2 (contextual) data was composed of census tracts and contains neighborhood risk factors (control variable s). Level 3 (structural) data was composed of Metropolitan and Micropolitan St atistical Areas which are used by the U.S. Census Bureau to calculate residentia l segregation indicators. Each Metropolitan Statistical Area has at least one urbani zed area and a population of 50,000 or more persons. Each Micropolitan Statistical Ar ea has at least one urban cluster and a population of at least 10,000 but less than 50,000 persons. Census tracts are the units of analysis for Metropolitan and Mi cropolitan Statistical Areas. This three-level model can be decomposed into submodels (Byrk & Raudenbush, 2002). Power Implications for Multilevel Models Although straightforward and practical techniques for power calculations involving two-level models (Snijders & Bosker, 1993) are available, these procedures are not precisely generalizeable to three-level models. An appropriate method for sample size and power calculations for three-level mo dels would involve conducting a simulation study using applicable estimates (Muthen & Muthen, 2002; Sastry, Ghosh-Dastidar, Adams, & Pebley 2003), preferably from a similar study or pilot study. Estimates are not available to identify and run a simulation study. Instead, published simulation studies and available studies concerning sample size sensitivity in multilevel analysis have been reviewed to provide rules of thumb for these types of models and to help approximate the sample size needed to achieve statisti cal power of 0.80 (Cohen, 1988). The studies

PAGE 74

62 described below serve as instructive guides. For multilevel models Hox described the “50/20 rule” for researchers interested in evaluating random variance and covariance para meters (1998). There should be at least 50 higher level groups and 20 in each gr oup (amounting to a total sample of approximately 1000 individuals) based on this ru le. Similarly, for multilevel structural equation models, Hox (2001) performed a simulation study in which he varied four conditions: 1) balanced versus unbalanced, 2) number of groups (50, 100, and 200), 3) group size (10, 20, 50), and 4) in traclass correlation [low (0.20) versus high (0.33)]. For a balanced design with 50 groups and a 95% confidence interval, there was 92% coverage of the parameter estimates. In addition, with a high in traclass correlation and 50 groups, inadmissible solutions were found in 5.70% of the samples. Hox (2001) ultimately concluded that “a low number of groups are only partially compensated by having large groups, a high ICC, or balanced data (p. 170).” Based on these studies, there should be at least 50 clusters at level 3 (MSAs). Furthermore, an average of 10 un its per cluster is warranted. In other words, there should be a minimum of 10 census tracts per MSA or a total of 1,000 census tracts should be sampled. Applying this rule again, there should be a minimum of 10 individuals per census tract or a total of 10,000 individuals. The eligibility criteria for th is study’s sample was discusse d earlier in this chapter. After applying these criteria, the initial sa mple comprised 298,926 women nested within 5,518 census tracts and 63 metropolitan or micropo litan statistical ar eas. Individuals (Level 1) without census tract information (Level 2) and census tracts with missing metropolitan or micropolitan area information (Lev el 3) were then excluded due to HLM

PAGE 75

63 statistical software requirements that there ar e datasets for each le vel of analysis and linking variables across each level of data The final sample was 255,548 women nested within 4,360 census tracts and 63 metropol itan or micropolitan districts. Analyses of geographic variables at each level of analysis indicate the HLM sample size objectives described above have been met based on the following: An average of 58.61 indivi duals per census tract: Minimum number of indivi duals per census tract: 1.00 Maximum number of indivi duals per census tract: 628.00 Standard Deviation: 45.23 Census tracts with 10 or more individuals: 95.92% An average of 69.21 census tracts per MSA: Minimum number of indivi duals per census tract: 4 Maximum number of indivi duals per census tract: 689.00 Standard Deviation: 120.74 MSAs with 20 or more census tracts: 93.65% A total of 63 Metropolita n and Micropolitan Areas

PAGE 76

64 Data Sources Birth certificate records were used to co llect information on individual covariates and adverse perinatal outcomes Birth certificate data is a potentially rich source of information, with wide application for both re search and practice and provides the best means, to date, for examination of the rese arch questions. Birth certificate data cover multiple years and contain large amounts of standardized information about U.S. women and potential risk factors fo r adverse perinatal outcomes. The birth certificate is among the most recognized and commonly used data sources for research and program planning. More states are implementing error-checking initiatives to improve quality of vital records (Alexander & Petersen, 1997). Residential segregation indices and contextu al covariates were obtained from data sources derived from the U.S. Census Bureau. The Household and Economics Statistics Division (HESD) of the U.S. Census Bureau provided data on residential segregation indices (Census Bureau of the United States, 2000). The HESD calculated residential segregation through examination of the dist ribution of populations across census tracts within Metropolitan Areas (MSAs) and Prim ary Metropolitan Statistical Areas (PMSAs) (Census Bureau of the United States, 2000). The 2000 Neighborhood Change Database (1970-2000) (NCD) (GeoLytics, 2004) provided data on contextual characteristics of residential areas. The NCD includes Long Form U.S. Census data from 1970, 1980, 1990 and 2000 at the census tract level. A unique feature of the dataset is that geogr aphical identifiers enable the researcher summarize census tracts into larger geogra phical levels includi ng counties and MSAs (2004) similar to processes used to calculate residential segrega tion through the HESD.

PAGE 77

65 Variable Measures Independent, dependent, potential co nfounders, and moderating variables included in the analysis are described in this section. The Appendix includes the conceptual model for the proposed relations hips among variables. Table 3.1 includes a list of study vari ables and scales of measurement. A set of theoretically relevant independe nt variables at each level were included in the analyses. The measure of concentrat ed poverty at the census tract level was excluded from the analyses because data were not available for a large number of the census tracts. Two variables were excluded fr om the analysis due to unacceptably high multicollinearity. In particul ar, the total proportion of pe rsons below the poverty level was excluded because that variable was not mutually exclusive from the proportion of Whites or Blacks under the poverty level. In addition, the proportion of Whites was found to be highly correlated ( r = -.94, p < 0.001) with the proportion of Blacks residing in census tract. Furthermore, the Isolati on Index was perfectly correlated (r = -1.00, p < .001) with the Interaction I ndex at the metropolitan area level. For the remaining variables, univariate descriptiv e statistics were performed. To aid in the interpretation of the resu lts, some variables were dichotomized. More specifically, highly skewed or kurtot ic variables were dichomotized near the median. Each variable that did not meet the assumption of univariate normality were recoded into binary variables. These variables were split so that approxim ately half of the sample was coded as 0 and the other half coded as 1. For instance the variable that examined the proportion of the population within the census tract that was in correctional facilities possessed a preponde rance of zeros and very little variance. Due to the

PAGE 78

66 preponderance of zeros and highly skewed dist ribution, the values were divided at the median value of 0: all 0’s were coded 0 and all other values were coded as 1. There is some debate regarding how to handle ordered categorical variables (Knapp, 1990). Nevertheless, it is common practice to treat ordi nal variables with five or more categories as continuous variables. Median income, age of housing and prenatal care variables in the current st udy met this criterion and were an alyzed accordingly. It is important to note that the dichotomization or the creation of artificial categories was purposefully avoided whenever possible because this practice can of ten decrease effect size and statistical power (MacCallum, Zhang, Preacher, & Rucker, 2002). A list of variables and the coding ar e provided in Table 3.

PAGE 79

67 Table 3 Variable Descriptions and Scales of Measurement *Represents original coding of variable. Ordinal variable with > 5 categories treated as continuous because variable dichotomization or creation of artificial cate gories can often decrease ef fect size and statistical power (MacCallum, Zhang, Preacher, & Rucker, 2002). VARIABLE SCALE OF MEASUREMENT LEVEL OF ANALYSIS AND CODING Independent Variables Level 3 Index of Dissimilarity or Evenness Continuous Isolation Index or Exposure Continuous Proportion of Black Continuous Area Size Binary 0 = Metropolitan-population >50,000 1 = Micropolitan-population >10,000 to <50,000 Control Variables-Indi vidual Covariates Level 1 Maternal Age 15-19 years 20-24 years 25-29 years 30-34 years > 35 years Ordinal 5 Categories Maternal Education High school Ordinal 3 Categories Marital Status Binary 0 = Married 1 = Unmarried *Prenatal Care Adequacy Intensive Adequate Intermediate Inadequate No Care Ordinal 5 Categories Cigarette Use During Pregnancy Binary 0 = No 1 = Yes Outcome Variables Preterm Delivery Binary 0=> 37 weeks gestation 1=< 37 weeks gestation Low Birth Weight Binary 0=> 2,500 grams 1=<2,500 Small for Gestational Age Birth Binary 0= > 10th percentile for gestational age and sex 1= < 10th percentile for gestational age and sex

PAGE 80

68 Table 3 (Continued) *Represents original coding of variable as measured by U.S. Census Bureau. Ordinal variables with > 5 categories treated as continuous because variable di chotomization or creation of artificial categories can often decrease effect size and statistical power (MacCallum, Zhang, Preacher, & Rucker, 2002) Residential Segregation VARIABLE SCALE OF MEASUREMENT LEVEL OF ANALYSIS AND CODING Moderating Variable Maternal Ethnicity Binary 0=White 1=Black Control Variables-Contextual Covariates Level 2 Total Owner-occupied Homes Continuous Household Vacancy Binary 0 = < 10% 1 = > 10% **Age of Housing Built 1999 to March 2000 (< 1 year) Built 1995 to 1998 (2-5 years) Built 1990 to 1994 (6-10 years) Built 1980 to 1989 (11-20 years) Built 1970 to 1979 (21-30 years) Built 1960 to 1969 (31-40 years) Built 1950 to 1959 (41-50 years) Built 1940 to 1949 (51-60 years) Built 1939 or earlier (> 61 years) Ordinal 10 categories Proportion Female-Headed Households Binary 0 = < 10% 1 = > 10% Proportion in Correctional Institutions Binary 0 = = 0% 1 = > 0% Proportion of Resident Civilians Unemployed Binary 0 = < 5% 1 = > 5% Proportion of Residents with High School Diploma Continuous Proportion of Blacks Below Poverty Binary 0 = < 10% 1 = > 10% Proportion of Whites Below Poverty Binary 0 = < 10% 1 = > 10% Moderating Variable **Median Household Income 0-$24999 $25,000 to $59,999 $60,000 or higher Ordinal 3 categories

PAGE 81

69 Residential segregation was measured us ing the Index of Di ssimilarity and the Isolation Index. The selected indicators repres ented two of the five dimensions described and examined by Massey and Denton (1988): Evenness and Exposure. The Black-White Dissimilarity Index and the Black Isolati on Index was measured using residential segregation calculated by the Household and Economics Statis tics Division of the U.S. Census Bureau (Census Bureau of the Un ited States, 2000). Residential segregation estimates were calculated through examinati on of the distribution of populations across census tracts within Metropolitan and Micropo litan Statistical Areas Indices calculated for non-Hispanic Blacks use non-Hispanic Wh ites as the reference group. The U.S. Census Bureau has calculated publicly accessi ble residential segregation indices that were utilized in this study Formulas used by the U.S. Census Bureau to calculate selected indices are included below. The formula for the Index of Dissimilarity is the specified as: P TP P p tn i i i 1 21 Where, n number of areas (census tracts) in the metropolitan area, ranked smallest to largest by land area ti the total popul ation of area i pi the ratio of xi to ti (proporti on of area I's population that is minority) P the ratio of X to T (proportion of the metropolitan area's population that is minority) T the sum of all ti (the total population) The formula for the Index of Is olation is the specified as:

PAGE 82

70 Where, n number of areas (census tracts) in the metropolitan area, ranked smallest to largest by land area xi the minority population of area i X the sum of all xi (the total minority population) ti the total popul ation of area i Formulas and definitions are taken fro m Massey and Denton (1988) and Iceland, Weinberg, & Steinmetz (2002). Dissimilar ity measures >0.6 are considered hypersegregated (Glaeser & Vigdor, 2001). Control Variables Unlike traditional models that simply in clude multiple intercorrelated individuallevel variables for the purpose of contro lling for individual-le vel confounding, multilevel models allow for simultaneous examination of the association of individual-level and group-level variables with perinatal outcomes. In the full multilevel model, coefficients for individual-level variables quantify how each variable is associated with the outcomes after adjusting for the group characteristics included and moderated by group random effects. Hence, the examination of the associat ion between variables th at are "controlled" for and health outcomes are possible through multilevel models (Diez-Roux, 2004). Hypotheses testing of the ma in relationship between residential segregation and adverse perinatal outcomes controlled for individual and contextual covariates. Birth certificate data provided information on individu al covariates that in clude: ethnicity; age; education; marital status; pren atal care; and substance use. Contextual covariates were n i ii it x X x1

PAGE 83

71 obtained from the 2000 Neighborhood Change Database (1970-2000) (GeoLytics, 2004). The Graduate Index of Prenatal Care Utili zation (GINDEX) was used to quantify the adequacy of prenatal care ut ilization. This index accounts fo r month that prenatal care began and the total number of prenatal car e visits, adjusted for gestational age (Alexander, Kogan & Nabukera, 2002; Alexander & Kotelchuck, 1996). Selected covariates fell into one of three environm ental categories: the physical, social, and economic. Physical environmental factors in cluded total owner-occupied homes, housing unit vacancy, and age of housing. The social environment was measured through factors including the percentage of female-headed households and the proportion of residents in correctional facilities. Characteristics of the economic environment included: unemployment rates; residents with a high sc hool degree; total povert y rates; and racespecific poverty rates. Outcome Variables Perinatal outcomes examined in this st udy were LBW, preterm delivery and SGA. LBW was measured as a dichotomous variable of < 2,500 grams and > 2,500 grams. Studies on the reliability of birth certificate data have shown that the birth weight measure is more accurate and complete when compared with medical records (Buescher, Taylor, Davis, & Bowling, 1993). Preterm de livery was measured as a dichotomous variable designated by delivery <37 complete weeks of gestation and term delivery > 37 complete weeks of gestation. SGA births was a similarly dichotomized variable designated by birthweight that is < 10th percentile of birthweigh t for gestational age and sex non-SGA birth represented by > 10th percentile of birthweight for gestational age and sex. Birth-weight and gestati onal age combinations were ca lculated using an algorithm

PAGE 84

72 developed by Alexander Hines, Kaufman, Mor and Kogan (1996). Alexander et al. (1996) de veloped an algorithm to provide a U.S. reference for fetal growth, to calculate birth weight-ges tational age combinations and to identify combinations that are implausible. Birth weig hts that appear to be normally distributed may have inaccurate gestational age values l eading to errors in preterm classification (David, 1980; Wang, Guyer, & Paige, 1994; Alexander, Tompkins, Altekruse, & Hornung, 1985). For example, infants with the same gestational age may have birth weights that place them at varying risk for SGA. Two infants of the same birth weight may have very different morbidity and mort ality risk classifica tions, contingent upon their gestational age. In order to account for implausible birt h weight-gestational ag e data, gestational age distributions were examined for births grouped into 125-g birth weight intervals. Birth weights were used as a reference because of its greater reliabl y than gestational age on vital records. Gesta tional age values of + 2.5 standard deviations from the mean were treated as cut points for implausible bi rth weight-gestational age combinations. Percentiles of the birth wei ght distribution were calculated for each completed week of gestational age. A resistant nonlinear sm oothing technique, labeled 4325H, was used twice to dampen irregularities due to random variation in the fe tal growth percentile curves across gestational age groups (Him es & Hoaglin, 1989; Ryan, Joiner, & Ryan, 1985). This data smoothing process uses a se ries of running medians and presumes no functional shape of fetal growth curves (Velleman, 1980). Moderating Variables The designation of ethnic groups on the birth certificate is an important

PAGE 85

73 consideration in this study. The inclusion of non-Hispanic White and Black women in this research is related to heterogeneous race and ethnic group classifications. Births categorized by race may be either Hispanic or nonHispanic. These groups may not be homogeneous with regard to pregnancy-rela ted risk factors or birth outcomes. For example, while the majority of Hispanic-origi n births were to White women, the majority of Black births were to non-Hispanic women (97% and 96%, respectively) in 2003 (Sutton & Matthews, 2004). The broader structural and contextual for ces that contribute to Black-White health disparities are critical to the current study and the limitati ons of the classification of groups by the term race have been carefully considered. Using the term race perpetuates a limited conceptualization of health disparit ies that is based on skin color or genetic predisposition. Median income was used to represent SE S for census tracts within MSAs. Median income was measured in the U.S. Census th rough division of the income distribution in a geographical area into two groups; one with income above the median and the other below the median. Median income was measured as the aggregate average household income with residential areas. Level of Aggregation The geographical unit of an alysis was the census tract. Census tracts are geographically defined areas of 3000-5000 person s. This unit of analysis was selected due to the fact that it is a small, relatively permanent statistical subdi vision. Census tracts capture within city variation, which can be used to conceptualize and measure the

PAGE 86

74 immediacy of social contexts that are lost to cross-metro analyses, and are relatively homogeneous with respect to population char acteristics, economi c status and living conditions (Guest et al., 1998). A recent study evaluating area based measures used to measure social inequalities found that censu s tract and block group data consistently detected equivalent and typically stronger so cioeconomic gradients than their zip code level counterparts (Krieger et al., 2003). Limitations of other studies examining the impact of areas of residence on birth outco mes have been the size of the geographical unit of analysis. According to Massey and De nton (1988) ethnic groups are more readily available using census tracts and are the leve l of aggregation most widely used in residential segregation studies. Analysis Procedures All statistical analyses were perfor med using the SAS version 9.0 program. A series of exploratory, descriptive data anal yses was performed on a ll variables described above to assess variable distributions (mean s, variance, frequency distribution), missing values and the magnitude of outliers. Analys es were performed for the overall population of each census tract, and for Black and White women, respectively. Descriptive and explanatory analysis were also performed se parately for each dimension of residential segregation and each adverse perinatal outcom e. For continuous variables, univariate analysis (PROC UNIVARIATE) were perfor med to determine normality. Categorical data were explored throu gh 2 by 2 tables and chi-squa re analysis (PROC FREQ). Bivariate analysis were conducted to dete rmine if there are associations between the macro level variables (residential segregati on dimensions), micro le vel individual risk factors (e.g., smoking, maternal age) and each adverse perina tal outcomes overall, and for

PAGE 87

75 the total populations of Black and White women separatel y, including correlations and collinearity. T-tests (PROC TTEST) and P earson correlations (PROC CORR) were used on parametric continuous data. Nonparametric data were subjected to one-way Wilcoxon and median testing (PROC ANOVA). Odds ratios and chi-square analysis were performed for categorical data (PROC FREQ). The geographical distribution of each a dverse perinatal outcome was calculated. In order to assess the variability of outcome rates, determination of the precision of the true population rates estimation is important Ninety-five percent confidence intervals were constructed around sample rates in order to assess this variability. The sample mean is a maximum-likelihood estimator of the tr ue population. When a pplied to small areadata (in this case, the censu s tract) the maximum-likelihood estimate tends to be quite variable (Biggeri, Marchi, Lagazio, Martuzzi, & Bohning, D ., 2000). Bayesian analysis combines sample information with other av ailable information from which inferential procedures are then based. Multilevel logistic regression was pe rformed following the identification of statistical confounders. Multileve l linear regression was utilized for continuous variables. Independent variables included in analysis included statistically significant contextual and individual factors describe d above. Multilevel models were used to test whether the level of residential residential segregation was positively associated with increased likelihood of adverse perinatal outcomes after controlling fo r contextual and individual factors (Hypothesis 1). Models included individual (e .g. maternal characteristics, medical/obstetric conditions), contextual (e.g., median poverty, % incarcerated) and structural factors (e.g., reside ntial segregation). Parameters were considered for both

PAGE 88

76 fixed and random effects. This determination was made prior to performing multilevel analysis. Parameters were fixed when all indepe ndent variables are present or if the levels present are the only ones of inte rest (Littrell, 1996). In this analysis, fixed effects were typically those at the individual level that are categorical (i.e., ye s/no). Variables were treated as when the levels of an independent variable are considered to represent a sample of the levels available in the population (1996). Proportion Black, median income, or residential segregation indices are examples of random effects in this study because these factors are only a sampling of all possible levels or proportions in the population. The determination of variables that were examined as fixed or random effects occurred prior to performing multilevel analysis. The lines of inquiry for this investiga tion centered on simultaneously gauging the unique contributions of residential areas a nd individual factors on the macro, structural level and the likelihood of adve rse perinatal outcomes at the micro level and are the primary justifications for the use of a mu ltilevel analysis procedures. As previously indicated, the use of a multile vel design was also driven by the limited scope of analysis strategies employed in previous research. Research studies at aggregate levels and falling prey to ecological and atomistic fallacies are among the major criticisms of existing studies. Further, while multilevel modeling has been employed to examine mortality, self-rated health, depressive symptoms, and health behaviors (Daly, Duncan, Kaplan, & Lynch, 1998; Deaton, 2001, Fiscella, & Franks 1997; Lochner, Pamuk Makuc, Kennedy, & Kawachi, 2001), no investiga tion of influences on perinatal outcomes related to increased risk for infant mortality have b een identified. Examini ng the influences of residential segregation on adverse perinata l outcomes over and above known individual

PAGE 89

77 (micro) and contextual (mezzo) is critical to co ncurrent examination of multiple levels of influence on adverse perinatal outcomes. The statistical framework of the hierar chical linear model (HLM) is suited for research involving classification and conti nuous variables, fixed and random effects, individual difference variables, time-varian t and invariant covariates or predictors, repeated measures, and naturally occurring hi erarchies (e.g., students within teachers, teachers within schools). It is important to note that these models are mathematically equivalent to mixed models, random regressi on models, random effect models, random coefficient models, multilevel models, and growth curve models. Although the notation and conceptualization can vary, they involve the same basic model (Hedeker, Gibbons, & Flay, 1994). HLMs are broadly applicable and have several statistical advantages over traditional ANOVA and MANOVA. In particular, HLMs are generally more flexible, handle missing data more effectively, and can yield more precise parameter estimates and standard errors when data ar e nested. Consequently, this approach can provide more statistical power and improve the ability to make statistical inferences when used appropriately (Burton, Gurrin, & Sl y, 1998; Singer & Willett, 2003). Moderation can also be tested in multilevel models. A significant main effect for a moderator variable indicates the mean re sponse depends upon the moderator variable. The direction of significant moderator eff ects is assessed by probi ng simple regression lines (Aiken, 1991; Tien, Sandler, MacKi nnon, & Wolchik, 2004). In other words, moderator effects can be asse ssed by examining the slope of the outcome at different levels of the moderator variables.

PAGE 90

78 Hierarchical Generalized Linear Models The assumptions of linearity and multivariate normality required for the standard HLM are not always tenable. In these cas es, hierarchical generalized linear models (herein referred to as HGLM) have been developed to handle non-normal and nonlinear models (i.e., binary, count, ordinal, and multinomial). Incorporation of such data is done by using one of several link f unctions to transform the outco me such that the dependent variable is substituted so that it adheres to the assumptions of the linear model. Hierarchical generalized lin ear models are mathematically equivalent to the standard HLM when the identity link functi on is used. The identity link function would be used for a model in which no transformation of the dependent variable is performed. However, in the case of binary outcomes, for instance, it is common to use a logit link function to serve as the distribution of th e dependent variable. For a HGLM with a binary outcome, the interpretation is similar to that found in logistic regression where the estimates represent the predicted log-odds whic h can be converted into an odds ratio by taking the exponent of the estimate. Individual, census tract, and metropolitan area data were entered into three separate datasets in SPSS 14.0 and imported in to HLM 6.0. To run the proposed analysis a multivariate data matrix (MDM) file from the raw data was generated to supply HLM 6.0 with the appropriate information. More specifically, an MDM file was constructed for a three-level hierarchical linear model fo r cross-sectional data or persons within groups. It is also necessary to sort identifica tion variables before generating the MDM

PAGE 91

79 file to connect the different datasets. Statistical Models The following models are related to ma jor hypothesis outlined in Chapter 2: Hypothesis One: The level of residential segregation is po sitively associated with the likelihood of adverse birth outcomes after controlling for cont extual and individual factors. To test the primary hypothesi s concerning the impact of residential segregation variables on birth outcomes, multilevel models were tested for each of the dependent variables using the total sample. Hypothesis Two: The relationship between residential segregation and the increased likelihood of adverse perinatal outcom es is moderated by ethnicity. A random intercept model was developed to allow the slope fo r ethnicity (Level 1) to randomly vary across all structural variables (Level 3) to examine the impact of residential segregation and et hnicity on perinatal outcomes. In other words, Level 3 variables served as predictors of the slope of ethnicity. Hypothesis Third: The relationship between residential segrega tion and the increased likelihood of adverse perinatal outcom es is moderated by median income. A random intercept model was developed to allow the slope for median income (Level 2) to randomly vary acros s all structural variables (Lev el 3) to examine the impact of residential segregat ion and ethnicity on peri natal outcomes. In other words, Level 3

PAGE 92

80 variables served as predictors of the slope of ethnicity. For each analysis, level-1 and level-2 inte rcepts was allowed to vary randomly. The level-1 intercept will serve as the outcome for level-2 variables and level 2 intercept will serve as the outcome for level-3 variables. The basic statistical model can be expressed by the following equations: Level 1: Individual log[P/(1-P)] = P0 + P1*(AGE) + P2*(EDUC A) + P3*(MARITAL) + P4*(PRENATAL) + P5*(SMOKE) + P6*(ETHNICIT) where the Prob(Y=1|B) = P Level 2: Contextual P0 = B00 + B01*(P_BLACK) + B02*(INCOME) + B03*(OCCUPY) + B04*(VACANT) + B05*(HOUSEAGE) + B06*(FEMHOUSE) + B07*(CO RRECT) + B08*(UNEMPLOY) + B09*(DIPLOMA) + B010*(BPOVERTY) + B011*(WPOVERTY) + R0 P1 = B10 P2 = B20 P3 = B30 P4 = B40 P5 = B50 P6 = B60 Level 3: Structural B00 = G000 + G001(METMIC) + G002(DISSIM) + G003(ISOLA) + G004(P_BLACK) + U00 B01 = G010 B02 = G020 B03 = G030 B04 = G040 B05 = G050 B06 = G060 B07 = G070 B08 = G080

PAGE 93

81 B09 = G090 B010 = G0100 B011 = G0110 B10 = G100 B20 = G200 B30 = G300 B40 = G400 B50 = G500 B60 = G600 Strengths and Limitations Multilevel modeling of the independent e ffects of residentia l segregation on the likelihood of adverse perinatal outcomes of women is a majo r contribution to research on determinants of ethnic disparities, due to studies limited by an alysis at individual or aggregate levels. Limitations of previous studies also include employing traditional regression analysis techniques. The lines of inquiry for this study centered on simultaneously gauging the unique contributions of the structural le vel and the likelihood of adverse perinatal outcom es at the micro level. The study also expands knowledge on the in fluence of structural factors through investigation of cause-specifi c outcomes. Health outcomes previously investigating area effects on health outcomes broadly include se lf-rated health, life expectancy, all-cause mortality, and infant mortality. The etiological precursors of adverse perinatal outcomes are heterogenous and may be differentially influenced by contextual and structural factors. This study was developed to test models that were spec ific to each birth outcome. Several assumptions underlie the primary fo cus of residential se gregation research on minority populations. Among them include the notions that (1) the whiter the residential area the better and (2) that Whites exclusively benefit from living in areas with lower numbers of minority residents. Much of the residential segregation

PAGE 94

82 literature does not acknowledge that White residential area s may also be segregated. Without discussion of the contex t outcomes of the segregation of Whites, the analysis of its effects on health outcomes are not comple te, pointing to the need for more research examining the effects of segregation on the re lationship between ethnicity and health for all groups of interest. Through the identification of pathwa ys through which residential segregation affects populations differently, this investiga tion may contribute to the development of policies to improve social conditions that shape health outcomes and accompanying ethnic disparities. There are a number of limitations associated with use of U.S. Census Bureau data and live birth certificates. The decennial nature of U.S. Census is a recognized limitation when used with other sources that capture annual data; 2000 Census Bureau data may not have reflected characteristics of women delivering infants in 1999 and 2001. Both data sources utilized in this study provide no info rmation on psychosocial, attitudinal or stress measures that may be associated with pe rinatal outcomes. Higher stress levels among Black and Hispanic women in studies of ethnically and socioeconomically diverse samples have led some to conclude that perceived ethnic and ge nder discrimination and their associated stressors contribute to an in creased risk for preterm birth (Rich-Edwards et al., 2001; Stancil 2000). The duration and length of exposure to area of residence were unknown in this study and may have important imp lications for contextual and structural risk assessment. Measurement at one point in time may underestimate the effects of residential segregation. Limitations related to the use of birth certificate data are acknowledged and have been previously discussed in detail in Chapter Two.

PAGE 95

83 Ethnic misclassification in data collec tion are also recognized in the proposed study. The rates of infants born to parents of tw o different ethnicitie s have increased from 1.4% in the 1970s to 4.3% in 1998 (Atkins on, MacDorman, & Parker, 2001; Parker, 2001). In response to revisions to the Offi ce of Management Bureau's Directive 15, a person could designate more than one categor y to describe themselves in the 2000 U.S. Census (1977; 1997). While this change wa s an important move toward acknowledging the fluid and multilevel nature of self identity, most national health surveys only allow individuals to select only one ethnic designation and cited studies indicate that the effect of ethnic misclassification is minimal among Black and Whites. Maternal and infant ethnic trends are al so challenging to measure due to varying data collection methods. An infant's ethnicity is based on the perceived identity of the mother by an unspecified hospital observer. Ethnicity on the Census is self-reported. This presents noteworthy considerations for cal culation of birth statis tics that are based on the division of number of events (taken from vital statistics), by the population at risk (taken from the U.S Census Bureau reside nt file) (Dutch & Madams, 1997; Ventura, 2001). Imprecision and variance in data colle ction existed prior to the inclusion of multiple race designations, but may be further complicated by it (Rosenberg et al., 1999). However, due to relatively small numbers of persons selecting more than one ethnic group, researchers estimate that the biases associated with th is change is negligible for Blacks and Whites, when compared to As ian American and Alaska Natives who represent the largest pr oportion of those select ing more than one ethnic group (Parker, et al., 2004).

PAGE 96

84 CHAPTER FOUR: RESULTS The following chapter presents results of the study. Univariate and bivariate statistics are reported for each level of da ta used in multilevel models. Control and moderating variables are descri bed through narrative and illu strative presentation and are followed by discussion of independent variab les. Outcome variable distributions are described overall sample and by ethnicity. The chapter concludes with findings for hierarchical generalized linear models used to test research hypotheses. After applying eligibility criteria for this study, the initial sample comprised 298,926 women nested within 5,518 census tracts and 63 metropolitan or micropolitan statistical areas. Individuals (Level 1) without census tract information (Level 2) and census tracts with missing metropolitan or micropo litan area information (Level 3) were then excluded due to HLM statistical software requirements that there are datasets for each level of analysis and geographical linking variables across each level of data. The final sample consisted of 255,548 women nested within 4,360 census tracts and 63 metropolitan or micropolitan statistical areas. Tables 4 and 5 compare women incl uded (N=255,548) and excluded (N=43,378) from the final sample by individual covariates Given the large sample size, significant pvalues may not indicate meaningful differe nces between these groups. Frequencies and percentages infer that included and excluded women were similar with regard to both individual covariates and bi rth outcomes. This observati on is further strengthened by initial decisions to exclude women primar ily on missing geographical linking variables and not individual covariates.

PAGE 97

85 Table 4 Analysis of the Distributions of Indivi dual Covariates Amongst Mothers Included and Excluded from Study Sample Note. Ordinal variable with > 5 categories treated as continuous because variable dichotomization or creation of artificial categor ies may threaten effect size and statisti cal power (MacCallum, Zhang, Preacher, & Rucker, 2002). aPrenatal Care Adequacy, Mean (Sta ndard Deviation). Measured using The Graduate Index of Prenatal Care Utilization Index. *Reference Group INCLUDED (N = 255,548) N (%) EXCLUDED (N = 43,378) N (%) STATISTIC P -VALUE Maternal Age, mean (sd) 24.70 (6.07) 25.21 (6.54) t = 15.82 <.001 Maternal Age 15-19 years 61,561 (24.1%) 10,524 (24.3%) 2 = 649.10 <.001 20-24 years 75,871 (29.7%) 11,910 (27.5%) *25-29 years 59,013 (23.1%) 8,781 (20.2%) 30-34 years 40,457 (15.8%) 7,859 (18.1% > 35 years 18,646 (7.3%) 4,304 (9.9%) aPrenatal Care Adequacy, mean (sd) 1.34 (0.74) 1.42 (0.77) t = 22.28 <.001 Maternal Education < High School = High School > High School 50,923 (19.93%) 79,686 (31.2%) 124,939 (48.9%) 8,988 (20.72%) 14,077 (32.5%) 20,313 (46.8%) 2 = 63.17 <.001 Marital Status *Married Unmarried 138,934 (54.37%) 116,614 (45.63%) 22,851 (52.68%) 20,527 (47.32%) 2 = 42.57 <.001 Cigarette Use During Pregnancy Yes *No 23,299 (9.12%) 232,249 (90.88%) 3,054 (7.04%) 40,324 (92.96%) 2 = 198.97 <.001 Ethnicity *White Black 178,960 (70.03%) 76,588 (29.97%) 29,661 (68.38%) 13,717 (31.62%) 2 = 48.00 <.001

PAGE 98

86 Table 5 Analysis of the Distributions of Adverse Perinatal Outcomes Amongst Mothers Included and Excluded from Study Sample INCLUDED (N = 255,548) N (%) EXCLUDED (N = 43,378) N (%) STATISTIC P -VALUE SGA Yes No 28,995 (11.3%) 226,553 (88.7%) 5,166 (11.9%) 38,212 (88.1%) 2 = 11.62 <.001 LBW Yes No 20,258 (7.9%) 235,290 (92.1%) 3,341 (7.7%) 40,037 (92.3%) 2 = 2.59 > .05 Preterm Yes No 24,965 (9.8%) 230,583 (90.2%) 3,962 (9.1%) 39,416 (90.9%) 2 = 17.14 <.001

PAGE 99

87 Univariate Analysis Control Variables Individual Covariates A total of 255,548 women were included in the sample. Overall, 29.97% of the sample was Black and 70.03% were White. The mean age of the sample population was 24.70 years. The marital status of the sa mple was mixed: 54.37% were married and 45.63% were not married. Completion of hi gh school was reported by 80.07% of the sample. The prenatal care adequacy was measured using the Graduate Index of Prenatal Care Utilization Index (GINDEX), a measure that accounts for both month that prenatal care began and the total number of prenatal care visits, adjusted for gestational age (Alexander, Kogan & Nabukera, 2002; Alexander & Kotelchuck, 1996). Each unit increase in the GINDEX scale is associated with decreased pren atal care adequacy (scores range from intensive to inadequate). This variab le was treated as continuous, because variable dichotomiza tion or creation of artificial categories can often decrease effect size and statistical power (MacCa llum, Zhang, Preacher, & Rucker, 2002). Mean prenatal adequacy for women in study wa s 1.34 (SD 1.34) suggesting the average women in the sample received inte nsive prenatal care. The majo rity (90.88%) of women did not smoke during pregnancy. Table 6 illustrates the distribution of individual covariates overall and by ethnicity. Black and White women significantl y differed on the majority of individual covariates. Black women were more likely to be younger than White women and were less likely (OR 1.56, 95% CI: 1.52, 1.60) to have less than a high school education when

PAGE 100

88 compared to White women. Black women we re more likely to be single (OR 7.65, 95% CI: 7.50, 7.80) and were significantly less likely to receive adequate prenatal care (p=< .001) when compared to White women. With regard to smoking, Black mothers were more likely not to smoke when compared to White women (OR 5.37, 95% CI: 5.12, 5.63).

PAGE 101

89 Table 6 Distributions, P-Values and Unadjusted Odds Ratios for Individual Covariates, Overall Sample and by Ethnicity, for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=255,548) VARIABLE TOTAL N=255,548 WHITE (N=178,960) BLACK (N=76,588) P-VALUE OR OR (95% CI) Maternal Age, mean (SD) 24.70 (6.07) 25.77 (6.00) 22.21 (5.46) <.001 Maternal Age, n (%) 15-19 years 61,561 (24.1%) 31,620 (17.7%) 29,941 (39.1%) 4.00 (3.91-4.11) 20-24 years 75,871 (29.7%) 49,676 (27.8%) 26,195 (34.2%) 2.23 (2.18-2.29) 25-29 years 59,013 (23.1%) 47,737 (26.7%) 11,276 (14.7%) 1.00 30-34 years 40,457 (15.8%) 34,290 (19.2%) 6,167 (8.1%) 0.76 (0.74-0.85) > 35 years 18,646 (7.3%) 15,637 (8.7%) 3,009 (3.9%) 0.81 (0.78-0.85) Prenatal Care Adequacy Index, mean (sd) 1.34 (0.74) 1.29 (.71) 1.44 (.79) <.001 Maternal Education, n (%) < High School 50,923 (19.9%) 28,522 (15.9%) 22,401 (29.2%) 1.56 (1.52-1.60) = High School 79,686 (31.2%) 52,998 (29.8%) 26,688 (34.8%) 1.00 > High School 124,939 (48.9%) 97,440 (54.4%) 27,499 (35.9%) 0.56 (0.55-0.57) Marital Status, n (%) Unmarried 116,614 (45.63%) 56,823 (31.75%) 59,791 (51.27%) 7.65 (7.50, 7.80) *Married 138,934 (54.37%) 122,137 (68.25%) 16,797 (21.93%) 1.00 Cigarette Use During Pregnancy, n (%) No 232,249 (90.88%) 157,552 (88.04%) 74,697 (97.53%) 5.37 (5.12, 5.63) *Yes 23,299 (9.12%) 21,408 (11.96 %) 1,891 (2.47%) 1.00 Note. OR=odds ratio; CI=confidence interval. Means and standard deviations are presented for ordinal variables with > 5 categories that were treated as continuous because variable dichotomization or creation of artificial categories can often decrease effect size an d statistical power (MacCallum, Zhang, Preacher, & Rucker, 2002). Whites serve as referent group for odds ratio comparisons.

PAGE 102

90 Contextual Covariates and Moderating Variable The distributions of categor ical covariates representing census tracts are listed in Table 7. Over half of census tracts were composed of those that had a median income of $25,000-$59,999. Sixty-one percent of the sample liv ed in census tracts where less than or equal to 10% of homes were vacant. The me an age of housing for census tracts in this study was 21-30 years. Fiftynine percent of census tracts were composed of homes where greater than 10% of hous eholds were female-headed. Individuals in correctional institutions made up 30.32% of census tracts. The majority (69.82%) of the sample lived in census tracts where less th an or equal to 5% of the population were unemployed. Sixty percent of census tracts were characterized by more than 10% of Blacks who lived below the poverty level. In contrast, 61.58% of censu s tracts were composed of less than or equal to 10% of Whites who lived below the poverty level.

PAGE 103

91 Table 7 Distributions of Contextual Covariates in Census Tracts of Residence for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=4360) Note. SD=Standard Deviation *Continuous Variable Bivariate analyses were conducted to examine associations between the proportion of Blacks in census tracts and contextual covariates. The continuous, Proportion Black variable was di chotomized at the median. Census tracts with an average proportion of Blacks that was less than or equal to 10% (Low Group) were assigned a value of 0 and census tracts w ith average proportions of Blacks that was greater than 10% (High Group) were assigned a value of 1. The dichotomized variable VARIABLE MEAN (SD) OR N (%) Median Income n (%) 0-$24999 1,208 (27.71%) $25,000 to $59,999 2,657 (60.94%) $60,000 or higher 495 (11.35%) *Proportion of Owner-O ccupied Homes, mean (SD) .67 (0.22) *Age of Housing, mean (SD) 5.93 (1.51) *Proportion of Residents with High School Diploma, mean (SD) .29 (0.09) Proportion Female-Headed Households n (%) More than 10% 2,582 (59.22%) Household Vacancy n (%) >10% 1,689 (38.74%) Proportion of Residents in Co rrectional Institutions n (%) >0% 1,322 (30.32%) Proportion of Resident Civilians Unemployed >5% 871 (19.98%) Proportion of Blacks Below Poverty >10% 2,835 (65.02%) Proportion of Whites Below Poverty >10% 1,675 (38.42%)

PAGE 104

92 served as the primary indepe ndent variable and all other level 2 variables served as dependent variables (Table 8). Continuous variables were analyzed using independent samples t-tests. Analyses revealed significant differences between the High and Low groups for all dependent variables. All findings were statistically si gnificant (p=<.01) but due to the large study sample, p-values may not represent significan t differences. Sample means are displayed in Table 8. Census tracts representing the High Group had lower average median incomes, a lower mean number of owner-occupied homes, older housing. Categorical variables for bivariate analysis of contextual variables were subject to chi-square tests. All associations were statistically significan t (p<.01). The High Group had a higher proportion of female-headed hous eholds (85.2%) when compared to the Low Group (31.5%). Similar results were found when groups were compared by the proportion of residents in correctional institutions (Low Group, 28.1%; High Group, 32.4%). Proportions of Blacks and Whites in poverty were higher for the High Group (85.9%, 53.3%, respectively) when co mpared to the Low Group (42.8%, 22.6%, respectively). These findings point to associ ations between ethnicity and disparities in economic and physical environments that characterize neighborhoods.

PAGE 105

93 Table 8 Distributions, P-Values and Chi Square Anal yses for Contextual Covariates, by Proportion of Black Residents, in Census Tracts of Resi dence for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=4,360) VARIABLE LOW ( 10%) (n = 2,110) HIGH (>10 %) (n = 2,250) STATISTIC P -VALUE *Median Income, mean (SD ) 8.27 (2.23) 6.01 (2.40) t = 32.16 <.001 Median Income, n (%) 0-$24999 212 (10.0) 996 (44.3) 2 = 744.24 <.001 $25,000 to $59,999 1,494 (70.8) 1,163 (51.7) $60,000 or higher 404 (19.1) 91 (4.0) *Proportion of Owner-Occupied Homes, mean (sd) 0.77 (0.16) 0.58 (0.23) t = 31.98 <.001 *Age of Housing, mean (sd) 5.46 (1.40) 6.38 (1.47) t = -21.29 <.001 *Proportion of Residents with High School Diploma 0.29 (0.96) 0.30 (0.78) t = -5.47 <.001 Household Vacancy, n (%) 10% > 10% 1,342 (63.6) 768 (36.4) 1,329 (59.1) 921 (40.9) 2 = 9.44 <.001 *Proportion of Female-Headed Households n (%) 10% > 10% 1,446 (68.5) 664 (31.5) 332 (14.8) 1,918 (85.2) 2 = 1303.85 <.001 Proportion of Resident Civilians in Correctional Institutions n (%) = 0% > 0% 1,517 (71.9) 593 (28.1) 1,521 (67.6) 729 (32.4) 2 = 9.51 <.01 Proportion of Resident Civilians Unemployed n (%) 5% > 5% 2,035 (96.4) 75 (3.6) 1,454 (64.6) 796 (35.4) 2 = 689.79 <.001 *Proportion of Blacks Below Poverty, n (%) 10% > 10% 1,207 (57.2) 903 (42.8) 318 (14.1) 1,932 (85.9) 2 = 888.15 <.001 Proportion of Whites Below Poverty, n (%) 10% > 10% 1,634 (77.4) 476 (22.6) 1051 (46.7) 1,199 (53.3) 2 = 434.62 <.001 Note. SD=Standard Deviation. Means and Standard Devia tions are included for continuous variables. For all dichotomous variables, the first group serves as referent group.

PAGE 106

94 Independent Variables Sixty-three metropolitan and micropolitan areas were analyzed in this study. Sixty-five percent of areas were metropolitan statistical areas a nd 35% of areas were micropolitan statistical areas. The Index of Di ssimilarity captures the extent to which Blacks are unevenly distributed relative to an ideal degree of integration with Whites. The mean Dissimilarity Index for the study sample was 0.47 (SD .12; scores ranged from 0.20-0.69). This score indicates that 47% of Black residents of metropolitan and micropolitan statistical areas would have to move to achieve perfect representation and evenness in reference to Whites. Dissimilarity measures greater than 0.06 are considered hypersegregated (Glaeser & Vigdor, 2001). The di stribution of Dissimilarity for the study sample was negatively skewed and relatively peaked, or leptokurtic (see Figure D). Leptokurtosis is associated with probability density functions that are peaked and have “fat” tails (Hatcher & Stepanski, 1994).

PAGE 107

95 0.70 0.60 0.50 0.40 0.30 0.20 Indices of Dissimilarity 14 12 10 8 6 4 2 0 Frequency Figure D. Distribution of Dissimilarity I ndices for Metropolitan and Micropolitan Statistical Areas occupied by Sample of Fl orida, Georgia and Louisiana Mothers 19992001 (N=63) Table 9 includes metropolitan and micropo litan areas in order by increasing Dissimilarity. The most dissimilar areas were New-Orleans-Metairie-Kenner, Louisiana (0.69), followed by Chattanooga-Tennessee-Georgi a (0.69) and Cape Coral-Fort Myers, Florida (0.67. Least dissimila r areas were Hinesville-Fort Stewart, Georgia (0.20), followed by Statesboro, Georgia (0.2 4) and Americus, Georgia (0.24).

PAGE 108

96 Table 9 Distribution of Dissimilarity Indices, by Statistica l Area Name and Increasing Dissimilarity, for Metropo litan and Micropolitan Statistical Areas occupied by Sample of Florida, Georgia and L ouisiana Mothers 1999-2001 (N=63) METROPOLITAN/ MICROPOLITAN STATISTICAL AREA DISSIMILARITY Hinesville-Fort Stewart, GA 0.20 Statesboro, GA 0.24 Americus, GA 0.24 Homosassa Springs, FL 0.25 Douglas, GA 0.27 Fort Walton Beach-Crestview-Destin, FL 0.30 Warner Robins, GA 0.30 Milledgeville, GA 0.32 Dublin, GA 0.34 Waycross, GA 0.37 Hammond, LA 0.37 Minden, LA 0.37 LaGrange, GA 0.38 Thomasville, GA 0.38 Punta Gordo, FL 0.39 Opelousas-Eunice, LA 0.39 Morgan City, LA 0.39 Gainesville, FL 0.41 Palatka, FL 0.41 New Iberia, LA 0.42 Tallahassee, FL 0.43 Bogalusa, LA 0.43 Athens-Clarke County, GA 0.44 Valdosta, GA 0.44 Augusta-Richmond County, GA-SC 0.44 Houma-Bayou Cane-Thibodaux, LA 0.46 Abbeville, LA 0.47 Palm Bay-Melbourne-Titusville, FL 0.47 Dalton, GA 0.47 Panama City-Lynn Haven, FL 0.48 Sebring, FL 0.48 Lafayette, LA 0.49 Ocala, FL 0.49 Crowley, LA 0.49 Natchez, MS-LA 0.49

PAGE 109

97 Table 9 (Continued) METROPOLITAN/ MICROPOLITAN STATISTICAL AREA DISSIMILARITY Key West-Marathon, FL 0.49 Ruston, LA 0.49 Pensacola-Ferry Pass-Brent, FL 0.50 Lakeland-Winter Haven, FL 0.51 Gainesville, GA 0.52 Jacksonville, FL 0.53 Macon, GA 0.53 Orlando, FL 0.54 Albany, GA 0.54 Savannah, GA 0.56 Brunswick, GA 0.56 Rome, GA 0.56 Shreveport-Bossier City, LA 0.56 Deltona-Daytona Beach-Ormond Beach, FL 0.56 Vero Beach, FL 0.57 Columbus, GA-AL 0.57 Port St. Lucie-Fort Pierce, FL 0.60 Baton Rouge, LA 0.60 Lake Charles, LA 0.62 Alexandria, LA 0.62 Tampa-St. Petersburg-Clearwater, FL 0.63 Atlanta-Sandy Springs-Marietta, GA 0.64 Sarasota-Bradenton-Venice, FL 0.65 Monroe, LA 0.66 Naples-Marco Island, FL 0.66 Cape Coral-Fort Myers, FL 0.67 Chattanooga, TN-GA 0.69 New Orleans-Metairie-Kenner, LA 0.69 The Isolation Index captures the extent to which Black residents are primarily surrounded by Whites or other Blacks. The me an Isolation Index for the study sample was 0.45 (SD 0.15; scores ranged from 0.04-0.74). This index reveals that the average Black resident lives in a cen sus tract in which the Black share of the population exceeds the overall metropolitan and micropolitan average by approxiamately 45%. The distribution of Isolation for the study sample was negatively skewed and relatively

PAGE 110

98 peaked, or leptokurtic (see Figure E). 0.80 0.60 0.40 0.20 0.00 Indices of Isolation 25 20 15 10 5 0 Frequency Figure E Distribution of Isolati on Indices for Metropolitan an d Micropolitan Statistical Areas Occupied by Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=63) Table 10 includes metropolitan and micropolitan areas in order by increasing Isolation. The most isolated areas were Ne w Orleans-Metairie-Kenner, Louisiana (0.74), followed by Monroe, Louisiana (0.70) and Albany, Georgia (0.69). Homosassa Springs, Florida (0.04) was the least isolated area, followed by Punt a Gordo Florida, (0.09) and

PAGE 111

99 Dalton, Florida (0.13). Table 10 Distribution of Isolation Indi ces, by Statistical Area Name and Increasing Isolation, for Metropolitan and Micropolitan Statistical Areas Occupied by Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=63) METROPOLITAN/ MICROPOLITAN STATISTICAL AREA ISOLATION Homosassa Springs, FL 0.04 Punta Gordo, FL 0.09 Dalton, GA 0.13 Fort Walton Beach-Crestview-Destin, FL 0.16 Key West-Marathon, FL 0.18 Palm Bay-Melbourne-Titusville, FL 0.26 Sebring, FL 0.28 Gainesville, GA 0.29 Panama City-Lynn Haven, FL 0.31 Houma-Bayou Cane-Thibodaux, LA 0.32 Douglas, GA 0.33 Ocala, FL 0.34 Abbeville, LA 0.35 Palatka, FL 0.35 Warner Robins, GA 0.35 Statesboro, GA 0.36 Crowley, LA 0.36 Rome, GA 0.38 Pensacola-Ferry Pass-Brent, FL 0.39 Sarasota-Bradenton-Venice, FL 0.40 Lakeland-Winter Haven, FL 0.41 Gainesville, FL 0.41 Waycross, GA 0.41 Deltona-Daytona Beach-Ormond Beach, FL 0.42 Athens-Clarke County, GA 0.42 Hammond, LA 0.42 Naples-Marco Island, FL 0.42 LaGrange, GA 0.45 Orlando, FL 0.46 Vero Beach, FL 0.46 Tampa-St. Petersburg-Clearwater, FL 0.47

PAGE 112

100 Table 10 (Continued) METROPOLITAN/ MICROPOLITAN STATISTICAL AREA ISOLATION Morgan City, LA 0.48 Dublin, GA 0.48 Lafayette, LA 0.48 Brunswick, GA 0.48 Minden, LA 0.48 Bogalusa, LA 0.48 New Iberia, LA 0.49 Cape Coral-Fort Myers, FL 0.50 Thomasville, GA 0.50 Port St. Lucie-Fort Pierce, FL 0.51 Hinesville-Fort Stewart, GA 0.51 Tallahassee, FL 0.52 Valdosta, GA 0.53 Jacksonville, FL 0.53 Americus, GA 0.53 Augusta-Richmond County, GA-SC 0.54 Opelousas-Eunice, LA 0.55 Chattanooga, TN-GA 0.56 Milledgeville, GA 0.58 Lake Charles, LA 0.58 Ruston, LA 0.59 Alexandria, LA 0.60 Macon, GA 0.63 Savannah, GA 0.63 Baton Rouge, LA 0.64 Natchez, MS-LA 0.64 Shreveport-Bossier City, LA 0.64 Atlanta-Sandy Springs-Marietta, GA 0.66 Columbus, GA-AL 0.66 Albany, GA 0.69 Monroe, LA 0.70 New Orleans-Metairie-Kenner, LA 0.74 The mean proportion of Blacks in select ed metropolitan and micropolitan areas was 0.24 (SD 0.13; scores ranged from 0.03-0.50). The distribution of the proportion of Blacks was positively skewed and relativel y flat, or platykurtic (see Figure F). Platykurtosis is associated w ith probability density functi ons that are simultaneously

PAGE 113

101 peaked and have thin tails (Hatcher & Stepanski, 1994). Figure F. Distribution of the Proportion of Blacks for Metropolitan and Micropolitan Statistical Areas Occupied by Sample of Fl orida, Georgia and L ouisiana Mothers 19992001 (N=63) Outcome Variables Table 11 illustrates the di stribution of outcome variab les overall and by ethnicity. Overall, 9.12% of the sample delivered low birth weight infants, 9.77% delivered preterm infants, and 11.0% delivered small for gesta tional age infants. Black women were more likely to experience all adverse perinatal outcomes when compared to White women. Specifically, Black women had more than a tw o-fold likelihood of having LBW (95% CI: 2.17, 2.30) and SGA infants (95% CI: 2.22, 2.33) and were 1.50 times (95% CI: 1.50, 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Proportion Blac k 7 6 5 4 3 2 1 0 Frequency

PAGE 114

102 1.58) as likely to deliver a preterm in fant when compared to White women. Table 11 Distributions and Unadjusted Odds Ratio for Adverse Perinatal Outcomes Overall and by Ethnicity for Sample of Flori da, Georgia and Louisiana Mothers 1999-2001 (N=255,548) VARIABLE TOTAL (N=255,548) N (%) OR (95% CI) LBW (<2,500 grams) White 10,721 (5.99%) 1.00 Black 9,537 (12.45%) 2.23 (2.17-2.30) Preterm Delivery (< 37 weeks gestation) White 15,309 (8.55%) 1.00 Black 9,656 (12.61%) 1.54 (1.50-1.58) SGA (<10th percentile of birthweigh t for gestational age) White 15,440 (8.63%) 1.00 Black 13,555 (17.70%) 2.28 (2.22-2.33) Note. OR= Odds Ratio. CI=Confidence Interval The mean LBW infant in the sample weighed 1981.45 grams (birthweight ranged from 500-2498 grams). The mean gestational age among preterm infants in the sample was 33.87 weeks. The distributions of pr eterm delivery and LBW were negatively skewed and leptokurtic. The majority of SGA births (85.90%) were term deliveries

PAGE 115

103 Bivariate Analysis Table 12 displays the frequency and pe rcentage of LBW by each individual covariate. Black women were 2.23 times as likely (95% CI: 2.17, 2.30) to have a LBW infant when compared to White women. The mean age of women delivering LBW infants was 24.15 years. The greatest risk of LB W was found among wome n who were 15-19 years of age (OR 1.66, 95% CI: 1.59, 1.74) and those were at least 35 years old (OR 1.56, 95% CI: 1.47, 1.65). Women with less than a high school education (OR 1.28, 95% CI: 1.23, 1.33) were most likely to have a LBW infant when compared to those who had a high school diploma or higher. Similar findings were illustra ted when marital status was examined; unmarried mothers were more likely (OR 1.74, 95% CI: 1.69, 1.79) to have a LBW infant when compared to those who were married. The risk of low birth weight decreased (OR 0.83, 95% CI: 0.81, 0.84) as pren atal care adequacy decreased. This finding may be due to the increased intensity of prenatal care targeted towards those who present with behavioral risk factors such as tobacco use, which increased the likelihood of LBW delivery (OR 1.47, 95% CI: 1.41, 1.54) wh en compared to non-smokers in this study.

PAGE 116

104 Table 12 Bivariate Analysis of Low Bi rth Weight Live by Individua l Covariates for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=20,258) VARIABLE LOW BIRTH WEIGHT (N=20,258) OR (95% CI) Ethnicity n (%) White 10,721 (52.92%) 1.00 Black 9,537 (47.08%) 2.23 (2.17-2.30) Maternal Age, n (%) 15-19 years 6,167 (30.4%) 1.66 (1.59-1.74) 20-24 years 5,896 (29.1%) 1.26 (1.21-1.31) 25-29 years 3,702 (18.3%) 1.00 30-34 years 2,731 (13.5%) 1.08 (1.03-1.14) > 35 years 1,762 (8.7%) 1.56 (1.47-1.65) Maternal Education, n (%) < High School 6,349 (26.9%) 1.28 (1.23-1.33) = High School 8,004 (33.9%) 1.00 > High School 9,246 (39.2%) 0.74 (0.71-0.76) Marital Status, n (%) Unmarried 8,442 (41.67%) 1.74 (1.69-1.79) Married 11,816 (58.33%) 1.00 (1.00, 1.00) aPrenatal Care Adequacy, mean (SD) 1.24 (0.78) 0.83 (0.81-0.84) Cigarette Use During Pregnancy, n (%) No 2,530 (12.49%) 1.00 Yes 17,728 (87.51%) 1.47 (1.41-1.54) Note. OR=Odds Ratio. CI=Confidence Interval. SD=Standard Deviation. Means and standard deviations are presented for ordinal variables with > 5 categories that were treated as continuous because variable dichotomization or creation of artificial categories can often decrease effect size and statistical power (MacCallum, Zhang, Preach er, & Rucker, 2002). aPrenatal Care Adequacy, Mean (Sta ndard Deviation). Measured using The Graduate Index of Prenatal Care Utilization Index. Table 13 displays the frequency and pe rcentage of preterm delivery by each individual covariate. Blac k women were 1.54 times more likely to deliver a preterm infant than White women (OR 1.54, 95% CI: 1.50, 1.58). The mean age of women delivering preterm infants was 24.69 years. The greatest risk of preterm delivery was found among women who were 15-19 years of age (OR 1.28, 95% CI: 1.23, 1.33) and those were at least 35 year s old (OR 1.40, 95% CI: 1.33, 1.48). Mothers with less than a

PAGE 117

105 high school education (OR 1.17, 95% CI: 1.13, 1.21) and those who were unmarried (OR 1.27, 95% CI: 1.24, 1.30) were more likely to have a preterm infant when compared to those who were high school graduates or we re married, respectively. Decreased prenatal care adequacy resulted in the decreased lik elihood of preterm delivery (OR 0.68, 95% CI: 0.67, 0.69). This finding may be related to in creased prenatal care intensity targeted towards those who may be at risk for adve rse birth outcomes based on sociodemographic or behavioral factors. Mothers who smoked we re more likely to delivery preterm infants than those who did not smoke (OR 1.05, 95% CI: 1.01, 1.10).

PAGE 118

106 Table 13 Bivariate Analysis of Preterm Live Births by Individual Cova riates for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 (N=20,258) VARIABLE PRETERM DELIVERIES (N=24,965) OR (95% CI) Ethnicity White 15,309 (61.32%) 1.00 Black 9,656 (38.68%) 1.54 (1.50-1-58) Maternal Age, n (%) 15-19 years 6,721 (26.9%) 1.28 (1.23-1.33) 20-24 years 6,900 (27.6%) 1.04 (1.00-1.08) 25-29 years 5,162 (20.7%) 1.00 30-34 years 3,976 (15.9%) 1.14 (1.09-1.19) > 35 years 2,206 (8.8%) 1.40 (1.33-1.48) Maternal Education, n (%) < High School 6,736 (23.3%) 1.17 (1.13-1.21) = High School 9,227 (31.9%) 1.00 > High School 12,964 (44.8%) 0.91 (0.89-0.94) Marital Status, n (%) Unmarried 12,726 (50.98%) 1.27 (1.24-1.30) Married 12,239 (49.02%) 1.00 aPrenatal Care Adequacy, mean (SD) 1.16 (0.75) 0.68 (0.67-0.69) Cigarette Use During Pregnancy, n (%) No 2,374 (9.51%) 1.00 Yes 22,591 (90.49%) 1.05 (1.01-1.10) Note. OR=Odds Ratio. CI=Confidence Interval. SD=Standard Deviation. Means and standard deviations are presented for ordinal variables with > 5 categories that were treated as continuous because variable dichotomization or creation of artificial categories can often decrease effect size and statistical power (MacCallum, Zhang, Preach er, & Rucker, 2002). aPrenatal Care Adequacy, Mean (Sta ndard Deviation). Measured using The Graduate Index of Prenatal Care Utilization Index

PAGE 119

107 Table 14 displays the frequencies an d percentages of SGA births by each individual covariate. Blac k women were 2.27 times as li kely (95% CI: 2.22, 2.33) to deliver an SGA infant when compared to White women. The mean age of women delivering SGA infants was 23.64 years. Th e greatest risk of SGA was found among women who were 15-19 years of age (O R 1.87, 95% CI: 1.80, 1.93) followed by those were 20-24 years of age (OR 1.43, 95% CI: 1.38, 1.49). Those with less than a high school education (OR 1.31, 95% CI: 1.27, 1.35) and those who were unmarried (OR 1.91, 95% CI: 1.86, 1.96) were more likely to have an SGA infant when compared to those with at least a high school di ploma or those who were married, respectively. Only for the SGA birth outcome did decreased prenatal car e adequacy result in increased risk (1.12, 95% CI: 1.10, 1.14). Smokers were 1.80 times mo re likely (95% CI: 1.74, 1.87) to deliver an SGA infant when compared to those who did not smoke during pregnancy.

PAGE 120

108 Table 14 Bivariate Analysis of Small for Gestational Age Births by Indi vidual Covariates for Sample of Florida Georgia and Louisiana Mo thers 1999-2001 (N=28,998) VARIABLE TOTAL SGA BIRTHS (N=28,998) OR (95% CI) Ethnicity, n (%) White 15,440 (53.25%) 1.00 Black 13,555 (46.75%) 2.28 (2.22-2.33) Maternal Age, n (%) 15-19 years 9,249 (31.9%) 1.87 (1.80-1.93) 20-24 years 9,078 (31.3%) 1.43 (1.38-1.49) 25-29 years 5,108 (17.6%) 1.00 30-34 years 3,496 (12.1%) 1.00 (0.95-1.04) > 35 years 2,064 (7.1%) 1.31 (1.24-1.37) Maternal Education, n (%) < High School 9,504 (27.8%) 1.31 (1.27-1.35) = High School 11,860 (34.7%) 1.00 > High School 12,797 (37.5%) 0.67 (0.65-0.69) Marital Status, n (%) Unmarried 11,648 (40.17%) 1.91 (1.86-1.96) Married 17,347 (59.83%) 1.00 aPrenatal Care Adequacy, mean (SD) 1.39 (.778) 1.12 (1.10-1.14) Cigarette Use During Pregnancy, n (%) No 24,852 (85.71%) 1.00 Yes 4,143 (14.29%) 1.80 (1.74-1.87) Note. OR=Odds Ratio. CI=Confidence Interval. SD=Standard Deviation. Means and standard deviations are presented for ordinal variables with > 5 categories that were treated as continuous because variable dichotomization or creation of artificial categories can often decrease effect size and statistical power (MacCallum, Zhang, Preach er, & Rucker, 2002). aPrenatal Care Adequacy, Mean (Sta ndard Deviation). Measured using The Graduate Index of Prenatal Care Utilization Index. Each unit increase indicating decreased prenatal care adequacy

PAGE 121

109 Multilevel Modeling Analysis Hypothesis One: The level of residential segregation is positively associated with the likelihood of adverse birth outcomes after controlling for cont extual and individual factors. Multilevel models were used to determine the association between residential segregation on LBW, preterm birth deliver y and SGA birth. In each model, Level 1 variables served as predictors for the out come, while Level 2 and Level 3 variables served as predictors of the intercept (i.e., the intercept was allowed to vary across the Level 2 and Level 3 variables2). Parameter estimates were not appreciably different across outcomes. The intercepts for each mode l were significant (p <.01), indicating that the grand intercepts were different from zer o. This represents the average log odds of each outcome. Low Birth Weight Table 15 provides the parameter estimates for the association between residential segregation and LBW. Given the large sample size, there was adequate statistical power to detect small effects. Each variable represented its effect on the average probability of LBW, or the slope of the intercept. Ther e was no statistically significant evidence (p > .05) supporting the relationship between reside ntial segregation indi ces (Dissimilarity, p=0.49; Isolation, p=0.51), ot her structural variables (Area Size, p=0.36; Proportion Black, p=0.71) and average LBW, after accoun ting for other variables in the model. 2 In a three-level multilevel model, the Level I intercept is set by the researcher to vary randomly across the Levels II and III groups so the extent to which the dependent variable varies across the Level II groups (e.g. neighborhoods) can be assessed.

PAGE 122

110 Median income, the proportion of owne r-occupied homes, the proportion of residents with a high school diploma and the proportion of Wh ites living below the poverty level were significant c ontextual predictors of LBW (p <.01). Findings indicated that both low (0-$24,000) and high (> $60,000 or more) median income tertiles were associated with increased risk of LB W (OR 1.12, 95% CI: 1.05, 1.21 and OR 1.14, 95% CI: 1.04, 1.26, respectively). Women who lived in census tracts where greater than 10% of homes were owner-occupied were 1.14 times more likely to have a LBW infant (95% CI: 1.04, 1.26) when compared to those in ne ighborhoods where <10% of residents were Black. The odds of LBW increased 1.34 times (95% CI: 1.10, 1.62) for each unit increase in the proportion of high school diplomas a ttained in a census tract. Women living in census tracts where greater than 10% of White residents were in poverty were more likely to have LBW infants (OR 1.06, 95% CI: 1.03, 1.10). All individual-level variables were si gnificantly associated with LBW (p<0.01). Women 35 years old or older were most likely to have a LBW infant (OR 1.65, 95% CI: 1.55, 1.76) when compared to women 25-29 years of age. Mothers with less than a high school education were most likely to ha ve a LBW infant (OR 1.18, 95% CI: 1.12, 1.24) when compared to women with at least a high school education. Unmarried mothers were 1.24 times as likely (95% CI: 1.20, 1.28) as married mothers to deliver a LBW infant. Increased prenatal care adequacy resulted in 1.34 times ( 95% CI: 1.27, 1.41) the odds of LBW, after accounting for other variab les in the model. Smokers were 1.71 times more likely to have a LBW infant when compared to non-smokers (95% CI: 1.63, 1.80). With regard to ethnicity, Black women we re 2.20 times as likely (95% CI: 2.11, 2.30) as White women to have a LBW infant in the sample.

PAGE 123

111 Table 15 Parameter Estimates for Total Sample: Multilevel Analysis of the Effects of Residential Segregation on Low Birth Weight for Sampl e of Florida, Georgia and Louisiana Mothers 1999-2001 LOW BIRTH WEIGHT OR 95% CI LEVEL 3 Intercept 0.01 0.01, 0.02 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.04 .95, 1.15 *Dissimilarity 0.81 0.42, 1.56 *Isolation 1.27 0.54, 3.00 *Proportion Black 1.14 0.53, 2.48 LEVEL 2 *Proportion Black 0.99 0.89, 1.09 Median Income 0-$24,999 1.12 1.05, 1.21 $25,000-$59,999 1.00 1.00, 1.00 > $60,000 1.12 1.07, 1.18 *Proportion of Owner-Occupied Homes 1.14 1.04, 1.26 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 1.01 0.98, 1.05 *Age of Housing 1.00 0.99, 1.01 Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 1.02 0.98, 1.06 Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.00 0.96, 1.04 Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.04 0.99, 1.09 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 124

112 Table 15 (Continued) LOW BIRTH WEIGHT OR 95% CI LEVEL 1 *Proportion of Residents with High School Diploma 1.34 1.10, 1.62 Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.03 0.99, 1.06 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.06 1.03, 1.10 Maternal Age 15-19 years 0.90 0.84, 0.96 20-24 years 0.89 0.84, 0.95 25-29 years 1.00 1.00, 1.00 30-34 years 1.19 1.13, 1.25 > 35 years 1.65 1.55, 1.76 Maternal Education < High School 1.18 1.12, 1.24 = High School 1.00 1.00, 1.00 > High School 0.82 0.80, 0.84 Marital Status Unmarried 1.24 1.20, 1.28 Married 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.34 1.27, 1.41 Cigarette Use During Pregnancy No 1.00 1.00, 1.00 Yes 1.71 1.63, 1.80 Ethnicity Black 2.20 2.11, 2.30 White 1.00 1.00, 1.00 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 125

113 Preterm Delivery Similar to findings for the LBW model, residential segr egation and other structural variables were not associated with the increased likelihood of preterm delivery after controlling for other f actors. These findings were il lustrated by no statistically significant relationship (p > .05) between each residential segregation variable (Dissimilarity, p=0.19; Isolation, p=0.38), othe r structural variable s (Area size, p=0.44; Proportion Black, p=0.84) and pret erm delivery (Table 16). The likelihood of LBW was significantly a ssociated with several contextual and individual variables in this model. The propor tion of residents with a high school diploma were significantly (p<.05) associated with preterm delivery. The odds of preterm delivery increased 1.54 times (95% CI: 1.21, 1.96) for each unit increase in the proportion of residents with a high school diploma. All i ndividual-level variables were significantly associated with preterm delivery (p<0.01). Wo men 35 years or older were most likely to experience preterm delivery (OR 1.45, 95% CI : OR 1.39, 1.51) when compared women 25-29 years of age. Women with less than a high school education were more likely to have a preterm delivery (OR 1.18, 95% CI: 1.11, 1.25) when compared to women with a high school diploma. Unmarried mothers were 1.17 times more likely (95% CI: 1.12, 1.22) to experience preterm delivery as compared to married mothers. Each unit increase in prenatal adequacy was associated with a 1.59 times the likelihood of preterm delivery (95% CI: 1.48, 1.71). Women who smoked ciga rettes during pregnancy were 1.14 times as likely (95% CI: 1.06, 1.22) to have a pret erm infant as compared to non-smokers. Blacks were 1.53 times (95% CI: 1.47, 1.59) more likely to have a preterm infant as compared to Whites in the sample, after cont rolling for other variables in the model.

PAGE 126

114 Table 16 Parameter Estimates for Total Sample: Multilevel Analysis of the Effects of Residential Segregation on Preterm Delivery for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 PRETERM DELIVERY OR 95% CI LEVEL 3 Intercept 0.01 0.01, 0.02 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 0.95 0.84, 1.07 *Dissimilarity 0.50 0.22, 1.17 *Isolation 1.71 0.56, 5.18 *Proportion Black 1.14 0.38, 3.44 LEVEL 2 *Proportion Black 0.97 0.84, 1.12 Median Income 0-$24,999 1.03 0.95, 1.13 $25,000-$59,999 1.00 1.00, 1.00 > $60,000 1.04 1.00, 1.08 *Proportion of Owner-Occupied Homes 1.08 1.00, 1.17 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 1.02 1.00, 1.05 *Age of Housing 0.99 0.98, 1.00 Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 1.04 1.00, 1.08 Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.01 0.96, 1.06 Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.03 0.99, 1.08 *Proportion of Residents with High School Diploma 1.54 1.21, 1.96 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 127

115 Table 16 (Continued) Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.01 0.98, 1.04 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.01 0.98, 1.05 LEVEL 1 Maternal Age 15-19 years 0.93 0.89, 0.99 20-24 years 0.89 0.86, 0.92 25-29 years 1.00 1.00, 1.00 30-34 years 1.20 1.15, 1.24 > 35 years 1.45 1.39, 1.51 Maternal Education < High School 1.18 1.11, 1.25 = High School 1.00 1.00, 1.00 > High School 0.90 0.87, 0.93 Marital Status No 1.17 1.12, 1.22 Yes 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.59 1.48, 1.71 Cigarette Use During Pregnancy Unmarried 1.00 1.00, 1.00 Married 1.14 1.06, 1.22 Ethnicity Black 1.53 1.47, 1.59 White 1.00 1.00, 1.00 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable Small for Gestational Age Birth No residential segregation (Dissimilarit y, p=0.76; Isolation, p= 0.69) or structural variables (Area Size, p=0.57; Proportion Blac k, p=0.22) were signifi cantly associated with SGA births (Table 17). Several contextual variables in this model were significantly associated with SGA. Census tracts with an average income of $0-$24,999 were associated with increased risk of SGA bi rths (OR 1.13, 1.06, 1.21). Each unit increase in the proportion of individuals w ith at least a high school dipl oma was associated with an increase an SGA births (OR 1.38, 95% CI: 1.14, 1.67). All individual level factors were

PAGE 128

116 significantly associated with SGA with the exception of pren atal care adequacy (OR 1.00, 95% CI: 0.98, 1.02). Table 17 Parameter Estimates for Total Sample: Multilevel Analysis of the Effects of Residential Segregation on Small for Gestational Age Births for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 SMALL FOR GESTATIONAL AGE BIRTH OR 95% CI LEVEL 3 Intercept 0.06 0.05, 0.07 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.02 0.95, 1.10 *Dissimilarity 1.08 0.64, 1.82 *Isolation 0.89 0.49, 1.63 *Proportion Black 1.45 0.79, 2.69 LEVEL 2 *Proportion Black 0.97 0.91, 1.02 Median Income 0-$24,999 1.13 1.06, 1.21 $25,000-$59,999 1.00 1.00, 1.00 > $60,000 1.07 1.03, 1.12 *Proportion of Owner-Occupied Homes 1.09 1.03, 1.16 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 0.99 0.96, 1.02 *Age of Housing 1.01 1.00, 1.03 Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 0.98 0.94, 1.02 Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.00 0.97, 1.03 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 129

117 Table 17 (Continued) SMALL FOR GESTATIONAL AGE BIRTH OR 95% CI Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.02 0.98, 1.07 *Proportion of Residents with High School Diploma 1.38 1.14, 1.67 Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.01 0.97, 1.04 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.04 1.01, 1.07 LEVEL 1 Maternal Age 15-19 years 1.00 0.94, 1.07 20-24 years 1.02 0.98, 1.06 25-29 years 1.00 1.00, 1.00 30-34 years 1.09 1.05, 1.14 > 35 years 1.39 1.32, 1.46 Maternal Education < High School 1.11 1.07, 1.14 = High School 1.00 1.00, 1.00 > High School 0.82 0.80, 0.85 Marital Status Unmarried 1.18 1.14, 1.21 Married 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.00 0.98, 1.02 Cigarette Use During Pregnancy No 1.00 1.00, 1.00 Yes 2.06 1.95, 2.18 Ethnicity Black 2.19 2.07, 2.33 White 1.00 1.00, 1.00 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 130

118 Findings for the main effects of reside ntial segregation on adverse perinatal outcomes were not supported in multilevel models tested; there were no significant effects (p > .05) of segregation on dependent variables after indi vidual and contextual covariates were included in models. Direct effects of contextual and individual effects were weak to moderate. The strongest c ontextual effect was found for the positive relationship between the proportion of high sc hool graduates and each adverse perinatal outcome. The individual risk factors associ ated with the greatest risk for adverse perinatal outcomes were maternal ages of 35 years or older, smoking during pregnancy and Black ethnicity. Hypothesis Two: The relationship between residential segregation and the increased likelihood of adverse perinatal outc omes is moderated by ethnicity. Hypothesis Two was tested to examine wh ether residential se gregation variables differentially impacted adverse perinatal outcomes conditional by ethnicity. A random intercept model was developed to allow the slope for ethnicity (Level 1) to randomly vary across all structural variables (Level 3) to examine the impa ct of residential segregation and ethnicity on perinatal outcomes (i.e., Level 3 variables served as predictors of the slope of ethnicity3). Significant results were found for the moderating effect of ethnicity on the relationship between neighborhood Isola tion and LBW. Statistically significant support was also found for crosslevel interactions between ethnicity an d statistical area size in the preterm delivery model 3 In a three-level multilevel model, the slope of an independent variable is set by the researcher to vary randomly across Level II and III variables in order to assess the extent to which the magnitude of the relationship between the independent variable and the dependent variable varies across neighborhoods.

PAGE 131

119 Low Birth Weight Table 18 provides the parameter estimates for the moderating effects of ethnicity. For Black women, decreased LBW was associated with each unit increa se in the Index of Isolation (OR 0.36, 95% CI: 0.15, 0.87, p=0.02). Cross-level interactions between ethnicity and Area Size, Dissimilarity a nd Proportion Black, respectively, were nonsignificant (p>.05). Table 18 Parameter Estimates for Random Effects of Et hnicity: Multilevel Analysis of Moderating Effects on the Relationship Between Resident ial Segregation and Low Birth Weight for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 LOW BIRTH WEIGHT OR 95% CI LEVEL 3 Intercept 0.01 0.01, 0.02 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.01 0.89, 1.15 *Dissimilarity 0.77 0.36, 1.65 *Isolation 1.49 0.58, 3.84 *Proportion Black 1.04 0.42, 2.55 LEVEL 2 *Proportion Black 1.02 0.92, 1.13 Median Income 0-$24999 1.12 1.05, 1.20 $25,000-$59,999 1.00 1.00, 1.00 > $60,000 1.13 1.08, 1.19 *Proportion of Owner-Occupied Homes 1.14 1.04, 1.25 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 1.01 0.98, 1.05 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 132

120 Table 18 (Continued) LOW BIRTH WEIGHT OR 95% CI LEVEL 3 *Age of Housing 1.00 0.99, 1.01 Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 1.01 0.97, 1.06 Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.00 0.96, 1.04 Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.04 0.99, 1.09 *Proportion of Residents with High School Diploma 1.35 1.11, 1.65 Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.02 0.99, 1.06 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.06 1.03, 1.10 Maternal Age 15-19 years 0.90 0.84, 0.96 20-24 years 0.89 0.84, 0.95 25-29 years 1.00 1.00, 1.00 30-34 years 1.19 1.13, 1.25 > 35 years 1.65 1.55, 1.76 Maternal Education < High School 1.18 1.12, 1.23 = High School 1.00 1.00, 1.00 > High School 0.82 0.80, 0.84 Marital Status Unmarried 1.24 1.20, 1.28 Married 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.34 1.27, 1.41 Cigarette Use During Pregnancy No 1.00 1.00, 1.00 Yes 1.72 1.63, 1.81 Ethnicity Black 2.41 2.04, 2.85 White 1.00 1.00, 1.00 MODERATING EFFECTS OF ETHNICITY Ethnicity x Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.07 0.92, 1.24 Ethnicity x Dissimilarity 1.71 0.86, 3.37 Ethnicity x Isolation 0.36 0.15, 0.87 Ethnicity x Proportion Black 1.85 0.85, 4.04 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 133

121 Preterm Delivery The preterm model random effect mode l indicated that while residential segregation indicators did not have a significant effect, one st ructural, level3 variable did demonstrate a significant infl uence on preterm delivery (Table 19). These findings were illustrated by non-significant (p > .05) crosslevel interactions be tween each residential segregation variable (Dissim ilarity, p=0.12; Isolation, p=0.30) and preterm delivery. Area size was significantly associated with the likelihood of preterm delivery such that residence in micropolitan statistical areas was associated with increased risk of preterm delivery among Black women when compared to White women (OR 0.87, 95% CI: 0.77, 0.99, p<.01). The proportion of Blacks within cen sus tracts was non-significant (p=0.93).

PAGE 134

122 Table 19 Parameter Estimates for Random Effects of Et hnicity: Multilevel Analysis of Moderating Effects on the Relationship Between Reside ntial Segregation and Preterm Delivery for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 PRETERM DELIVERY OR 95% CI LEVEL 3 Intercept 0.01 0.01, 0.02 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 0.87 0.77, 0.99 *Dissimilarity 0.43 0.17, 1.11 *Isolation 1.95 0.62, 6.12 *Proportion Black 0.94 0.28, 3.12 LEVEL 2 *Proportion Black 0.98 0.85, 1.13 Median Income 0-$24,999 1.03 0.95, 1.12 $25,000-$59,999 1.00 1.00, 1.00 > $60,000 1.04 1.00, 1.08 *Proportion of Owner-Occupied Homes 1.08 1.00, 1.16 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 1.02 1.00, 1.05 *Age of Housing 0.99 0.98, 1.00 Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 1.04 1.00, 1.08 Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.01 0.96, 1.06 LEVEL 1 Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.03 0.99, 1.08 *Proportion of Residents with High School Diploma 1.55 1.22, 1.97 Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.01 0.98, 1.04 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.02 0.98, 1.05 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 135

123 Table 19 (Continued) PRETERM DELIVERY OR 95% CI Maternal Age 15-19 years 0.93 0.89, 0.99 20-24 years 0.89 0.86, 0.92 25-29 years 1.00 1.00, 1.00 30-34 years 1.19 1.15, 1.24 > 35 years 1.45 1.39, 1.51 Maternal Education < High School 1.18 1.11, 1.25 = High School 1.00 1.00, 1.00 > High School 0.90 0.87, 0.93 Marital Status Unmarried 1.17 1.12, 1.22 Married 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.59 1.48, 1.71 Cigarette Use During Pregnancy No 1.00 1.00, 1.00 Yes 1.14 1.06, 1.22 Ethnicity Black 1.33 1.06, 1.67 White 1.00 1.00, 1.00 MODERATING EFFECTS OF ETHNICITY Ethnicity x Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.23 1.08, 1.41 Ethnicity x Dissimilarity 1.94 0.84, 4.51 Ethnicity x Isolation 0.47 0.17, 1.33 Ethnicity x Proportion Black 1.94 0.68, 5.57 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable Small for Gestational Age Birth Table 20 illustrates parameter estimated de veloped to test moderating effects of ethnicity on the relationship be tween residential segregation and SGA births. Results did not support Hypotheses Two. Residential segreg ation (Dissimilarity, p=0.66; Isolation, p=0.10) and other structural variables (Area Size, p= 0.46; Proportion Black, p=0.70) were not significantly associated with ethnicity in cross-le vel interactions.

PAGE 136

124 Table 20 Parameter Estimates for Random Effects of Et hnicity: Multilevel Analysis of Moderating Effects on the Relationship Between Resident ial Segregation and Sm all for Gestational Age Births for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 SMALL FOR GESTATIONAL AGE BIRTH OR 95% CI LEVEL 3 Intercept 0.05 0.04, 0.06 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.04 0.95, 1.15 *Dissimilarity 1.22 0.66, 2.27 *Isolation 0.90 0.44, 1.83 *Proportion Black 1.67 0.80, 3.52 LEVEL 2 *Proportion Black 1.00 0.95, 1.06 Median Income 0-$24999 1.13 1.05, 1.20 $25,000-$59,999 1.00 1.00, 1.00 > $60,000 1.08 1.04, 1.12 *Proportion of Owner-Occupied Homes 1.09 1.02, 1.15 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 0.99 0.97, 1.02 *Age of Housing 1.01 1.00, 1.03 LEVEL 3 Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 0.97 0.93, 1.02 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 137

125 Table 20 (Continued) SMALL FOR GESTATIONAL AGE BIRTH Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.00 0.97, 1.03 Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.02 0.98, 1.07 *Proportion of Residents with High School Diploma 1.39 1.15, 1.68 *Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.00 0.97, 1.04 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.04 1.00, 1.00 LEVEL 1 Maternal Age 15-19 years 1.00 0.94, 1.07 20-24 years 1.02 0.98, 1.06 25-29 years 1.00 1.00, 1.00 30-34 years 1.10 1.05, 1.14 > 35 years 1.39 1.32, 1.46 Maternal Education < High School 1.10 1.07, 1.14 = High School 1.00 1.00, 1.00 > High School 0.83 0.80, 0.85 Marital Status Unmarried 1.18 1.15, 1.21 Married 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.00 0.98, 1.03 Cigarette Use During Pregnancy No 1.00 1.00, 1.00 Yes 2.07 1.96, 2.18 Ethnicity Black 2.82 2.24, 3.55 White 1.00 1.00, 1.00 MODERATING EFFECTS OF ETHNICITY Ethnicity x Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 0.95 0.84, 1.08 Ethnicity x Dissimilarity 1.21 0.53, 2.74 Ethnicity x Isolation 0.49 0.21, 1.15 Ethnicity x Proportion Black 1.16 0.55, 2.43 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 138

126 Findings indicated mixed findings for th e moderating effects of ethnicity on relationships between resident ial segregation and adverse pe rinatal outcomes in crosslevel interactions. Signifi cant effects were found for the negative association between Isolation and LBW among Black women. A pos itive relationship was found for the cross level interactions be tween Areas Size and ethnicity among Black women. These results indicate that the moderating effects of ethni city may be specific to birth outcome and ethnic group. Hypothesis Three: The relationship between residen tial segregation and the increased likelihood of adverse perinatal outcom es is moderated by median income. To examine Hypothesis Three, a random intercept model was developed to allow the slope for median income (Level 2) to randomly vary across all structural variables (i.e., Level 3 variables served as predictors of th e slope of median income). Cross-level interactions examined for each adverse bi rth outcome model were non-significant (p >.05). Low Birth Weight Median income did not moderate the rela tionship between residential segregation and LBW after controlling for other factors in models. Table 21 provides parameter estimates for this model. Median income was not significantly associat ed with residential segregation (Dissimilarity, p= 0.67; Isolation, p=0.75) and stru ctural variables (Area Size, p=0.07; Proportion Black, p=0.94) in cross-level interactions.

PAGE 139

127 Table 21 Parameter Estimates for Random Effects of Median Income: Multilevel Analysis of the Moderating Effect on the Relationship between Residential Segregation and Low Birth Weight for Sample of Florida, Georgia and Louisiana Mothers 1999-2001 LOW BIRTH WEIGHT OR 95% CI LEVEL 3 Intercept 0.01 0.01, 0.02 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.12 0.98, 1.28 *Dissimilarity 0.88 0.44, 1.78 *Isolation 1.29 0.51, 3.27 *Proportion Black 1.13 0.49, 2.60 LEVEL 2 *Proportion Black 1.00 0.91, 1.11 *Median Income 0-$24,999 1.24 0.94, 1.64 $25,000 or more 1.00 1.00, 1.00 *Proportion of Owner-Occupied Homes 1.07 0.98, 1.16 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 1.01 0.98, 1.05 *Age of Housing 1.01 1.00, 1.02 Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 1.02 0.98, 1.07 Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.00 0.96, 1.05 Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.04 0.99, 1.09 *Proportion of Residents with High School Diploma 1.63 1.33, 2.00 Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.03 1.00, 1.06 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.06 1.02, 1.10 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 140

128 Table 20 (Continued) LOW BIRTH WEIGHT OR 95% CI LEVEL 1 Maternal Age 15-19 years 0.90 0.84, 0.96 20-24 years 0.90 0.84, 0.95 25-29 years 1.00 1.00, 1.00 30-34 years 1.19 1.13, 1.25 > 35 years 1.65 1.55, 1.75 Maternal Education < High School 1.18 1.12, 1.24 = High School 1.00 1.00, 1.00 > High School 0.82 0.79, 0.84 Marital Status Unmarried 1.24 1.20, 1.28 Married 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.34 1.27, 1.41 Cigarette Use During Pregnancy No 1.00 1.00, 1.00 Yes 1.71 1.63, 1.80 Ethnicity Black 2.20 2.11, 2.30 White 1.00 1.00, 1.00 MODERATING EFFECTS OF MEDIAN INCOME Median Income x Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 0.86 0.73, 1.01 Median Income x Dissimilarity 0.79 0.28, 2.28 Median Income x Isolation 0.83 0.25, 2.75 Median Income x Proportion Black 1.04 0.38, 2.83 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 141

129 Preterm Delivery Findings indicated that median income di d not moderate the relationships between residential segregation indicators and pr eterm delivery (Table 22). Cross-level interactions were non-significan t (p >.05) for associations between median income and Dissimilarity (OR 1.19, 95% CI: 0.54, 2.67) Isolation (OR 0.50, 95% CI: 0.22, 1.15), Area Size (OR 0.96, 95% CI: 0.85, 1.09), and th e Proportion Black (OR 1.09, 95% CI: 0.32, 3.69), respectively. Table 22 Parameter Estimates for Random Effects of Median Income: Multilevel Analysis of the Moderating Effect on the Relationship betw een Residential Segregation and Preterm Delivery for Sample of Florida, Ge orgia and Louisiana Mothers 1999-2001 PRETERM DELIVERY OR 95% CI LEVEL 3 Intercept 0.01 0.01, 0.02 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.00 0.86, 1.16 *Dissimilarity 0.55 0.23, 1.32 *Isolation 1.73 0.56, 5.36 *Proportion Black 1.10 0.35, 3.42 LEVEL 2 *Proportion Black 0.98 0.85, 1.13 Median Income 0-$24,999 1.20 0.92, 1.56 > $25,000 1.00 1.00, 1.00 *Proportion of Owner-Occupied Homes 1.06 0.99, 1.13 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 1.02 1.00, 1.05 *Age of Housing 0.99 0.98, 1.00 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 142

130 Table 22 (Continued) PRETERM DELIVERY OR 95% CI Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 1.05 1.00, 1.09 Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.01 0.96, 1.07 Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.03 0.99, 1.08 *Proportion of Residents with High School Diploma 1.65 1.27, 2.14 Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.01 0.98, 1.04 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.01 0.98, 1.05 LEVEL 1 Maternal Age 15-19 years 0.93 0.89, 0.99 20-24 years 0.89 0.86, 0.92 25-29 years 1.00 1.00, 1.00 30-34 years 1.19 1.15, 1.24 > 35 years 1.45 1.39, 1.51 Maternal Education < High School 1.18 1.11, 1.25 = High School 1.00 1.00, 1.00 > High School 0.90 0.87, 0.93 Marital Status Unmarried 1.17 1.12, 1.22 Married 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.59 1.48, 1.71 Cigarette Smoking During Pregnancy No 1.00 1.00, 1.00 Yes 1.14 1.06, 1.22 Ethnicity Black 1.53 1.47, 1.59 White 1.00 1.00, 1.00 MODERATING EFFECTS OF MEDIAN INCOME Median Income x Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 0.96 0.85, 1.09 Median Income x Dissimilarity 1.19 0.54, 2.67 Median Income x Isolation 0.50 0.22, 1.15 Median Income x Proportion Black 1.13 0.54, 2.37 Note. OR=Odds Ratio. *Continuous Variable

PAGE 143

131 Small for Gestational Age Birth Findings for the random effects of medi an income on the relationship between residential segregation and preterm delivery were similar to the LBW model for the overall hypothesis, after controlling for other fa ctors in models, including the cross level interaction between residential segregation and median income. Cross level interactions were non-signficant for associations betw een median income and Area Size (OR 0.89, 95% CI: 0.79, 1.01), Dissimilarity (OR 1.09, 95% CI: 0.51, 2.33), Isolation (OR 0.73, 95% CI: 0.33, 1.61), and the Proportion Black (OR 1.28, 95% CI: 0.66, 2.47), respectively (Table 23).

PAGE 144

132 Table 23 Parameter Estimates for Random Effects of Median Income: Multilevel Analysis of the Moderating Effect on the Relationship betw een Residential Segregation and Small Gestational Age Birth for Sample of Fl orida, Georgia and Louisiana Mothers 1999-2001 SMALL FOR GESTATIONAL AGE BIRTH OR 95% CI LEVEL 3 Intercept 0.06 0.05, 0.07 Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 1.09 0.98, 1.20 *Dissimilarity 1.06 0.62, 1.81 *Isolation 0.96 0.51, 1.81 *Proportion Black 1.34 0.71, 2.55 LEVEL 2 *Proportion Black 0.97 0.91, 1.03 Median Income 0-$24,999 1.12 0.90, 1.38 > $25,000 1.00 1.00, 1.00 *Proportion of Owner-Occupied Homes 1.05 0.99, 1.12 Household Vacancy < 10% 1.00 1.00, 1.00 > 10% 0.99 0.96, 1.02 *Age of Housing 1.02 1.01, 1.03 Proportion of Female-Headed Households < 10% 1.00 1.00, 1.00 > 10% 0.98 0.94, 1.03 Proportion in Correctional Institutions > 10% 1.00 1.00, 1.00 > 10% 1.00 0.98, 1.03 Proportion of Resident Civilians Unemployed < 5% 1.00 1.00, 1.00 > 5% 1.02 0.98, 1.07 *Proportion of Residents with High School Diploma 1.55 1.33, 1.79 Proportion of Blacks Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.01 0.97, 1.05 Proportion of Whites Below Poverty < 10% 1.00 1.00, 1.00 > 10% 1.04 1.01, 1.07 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 145

133 Table 23 (Continued) SMALL FOR GESTATIONAL AGE BIRTH OR 95% CI LEVEL 1 Maternal Age 15-19 years 1.00 0.94, 1.07 20-24 years 1.02 0.98, 1.06 25-29 years 1.00 1.00, 1.00 30-34 years 1.09 1.05, 1.14 > 35 years 1.39 1.32, 1.46 Maternal Education < High School 1.11 1.07, 1.14 = High School 1.00 1.00, 1.00 > High School 0.82 0.80, 0.85 Marital Status Unmarried 1.18 1.14, 1.21 Married 1.00 1.00, 1.00 *Prenatal Care Adequacy 1.00 0.98, 1.02 Cigarette Smoking During Pregnancy No 1.00 1.00, 1.00 Yes 2.06 1.95, 2.18 Ethnicity Black 2.19 2.07, 2.33 White 1.00 1.00, 1.00 SLOPE OF MEDIAN INCOME Median Income x Area Size Metropolitan 1.00 1.00, 1.00 Micropolitan 0.89 0.79, 1.01 Median Income x Dissimilarity 1.09 0.51, 2.33 Median Income x Isolation 0.73 0.33, 1.61 Median Income x Proportion Black 1.28 0.66, 2.46 Note. OR=Odds Ratio. CI=Confidence Interval. *Continuous Variable

PAGE 146

134 Summary of Findings The results of the study did not support the role of reside ntial segregation on adverse perinatal outcomes (Hypothesis On e ). While findings did not support the moderating role of median income (Hypothesi s Three), findings for the moderating role of ethnicity (Hypothesis Two) were mixe d, signaling ethnic group and outcome-specific associations. Black women experience a decreas ed risk of LBW with increased isolation and increased risk of preterm delivery when they lived in micropolitan areas. In summation, the direct effects of residentia l segregation on adverse perinatal outcomes were not illustrated after accounting for other variables in models, including the cross level interactions between re sidential segregation and ethni city or median income. Several individual and contex tual variables emerged as significantly associated with all birth outcomes although they were w eak and modest findings. The proportion of individuals with high school diplomas and the proportion of owner-occupied homes were positively associated with all adverse perinatal outcomes such that each unit increase in contextual variables resulted in an increas e in LBW, preterm and SGA births. Median income and the proportion of Whites in povert y were significantly associated with SGA birth and LBW. Single, outcome-specific findi ngs showed the significant influence of the proportion of incarcerated indivi duals on preterm delivery and the impact of the age of houses in census tracts and SGA birth.

PAGE 147

135 CHAPTER FIVE: DISCUSSION Major Research Findings The purpose of this study was to inves tigate the direct relationship between residential segregation (the physical separation of Black s and Whites in residential contexts) and adverse perinatal outcomes (LBW, preterm delivery and SGA births) in Florida, Georgia and Louisian a. This study was specifically designed to determine the independent effect of the level of reside ntial segregation on th e likelihood of adverse perinatal outcomes after controlling for contex tual (i.e. poverty level, unemployment) and individual factors (i.e. prenat al care, substance use). The study also examined whether the direct relationship between the level of resi dential segregation and each adverse perinatal outcome was moderated by ethnicity and me dian income, respectively. Multilevel modeling techniques were employed to examine these associations. This study did not support the major ity of postulated hypotheses for the relationships between resident ial segregation and adverse pe rinatal outcomes. Residential segregation was not associat ion with LBW, preterm deliv ery or SGA births, after controlling for factors representing the physic al, social and economic environments in communities and the sociodemographic and behavioral risk factors of individuals (Hypothesis 1). Given the large sample size, there was adequate statistical power to detect small effects. The non-significant resu lts indicate that the effect sizes for these independent variables were relatively smaller than could be detected and likely to be negligible. One significant cross-level interaction was found between ethnicity (Level 1) and segregation (Hypothesis 2). Isolation was asso ciated with decreased risk of LBW among

PAGE 148

136 Black women. Further, Black women in smaller, micropolitan statistical areas (population size between 10,000 and 50,000) were le ss likely to have preterm infants. Cross-level interactions between ethnic ity and other Level 3 variables were nonsignificant. Non-significant re sults were also found for the moderating effect of median income on the segregation-a dverse perinatal outcome rela tionship (Hypothesis 3). Several contextual and individual factor s were significantly associated with adverse perinatal outcomes with primarily w eak or modest associations. These findings imply that macro level, structural factors may not be important predictors of adverse perinatal outcomes, but that proximate, nei ghborhood contexts and indi vidual risk factors are critical precursors of perinatal outcomes. The proportion of individuals with high school diplomas and proportion of owneroccupied homes in census tracts had positive and statistically significant associations with each adverse perinatal outcome in this study. Home-ownership has been positively associated with life satisfaction, psychologi cal health and well-being due to improved social status and personal freedom (Danes & Morris, 1986; Fanie Mae, 1999; Rossi & Weber, 1996). These benefits may be challe nged or negated by in creased numbers of homeowners who are at risk of losing thei r homes due to mortgage foreclosure (Rohe, Van Zandt, & McCarthy, 2001). The threats associ ated with losing one’s home have been associated with lower self perceived wellbeing among persons who are unable to pay their mortgages (Nettleton & Burrows, 1998). Th ese findings may help to explain why an increased proportion of homeowners in censu s tracts were associated with increased adverse perinatal outcome risk in this study. An explanation for the positive association between the proportion of residents with a high school diploma and the likelihood of

PAGE 149

137 adverse perinatal outcomes may be due to the heterogeneity of educational attainment in census tracts and other commorbi dities that were not accoun ted for. For instance, there was no distinction between women of advanced maternal age who postponed childbearing to complete their graduate e ducation and high school graduates who had other sociodemographic or be havioral risk factors. The majority of individual–level factors were significantly associated with each perinatal outcome in multilevel models Women who were Black, unmarried, smoked, older and less educated (less than high school diploma) were more likely to experience all adverse perinatal outcomes. These findings ar e supported by the major ity of studies on the role of individual risk f actors on perinatal health (Ana nth, Balsubramanian, Demissie, & Kinzler, 2004; Goldenberg et al., 1 996; Guyer, Martin, MacDorman, Anderson, & Strobino, 1997; Martin, Hamilton, Sutton, Ventura, Menacker, & Munson, 2003; Schiono, Rauh et al., 1997). The inverse significant association betw een prenatal care adequacy, LBW and preterm delivery may seem counterintuitive upo n initial examination. However, intensive prenatal care services targeted to women at heightened risk for poor outcomes implies that low-risk women had lowe r prenatal care adequacy, but better perinatal outcomes. There was a positive, significant relationship between small for gestational age births and prenatal care adequacy. The results of this study offer several imp lications for future research and practice focused on reducing disparitie s in health. Findings point to the need for further specification of a comprehensive constellati on of variables starting from the onset of pregnancy through to labor and delivery. Due to an increased number of Blacks in

PAGE 150

138 poverty, the additional threat of economic segregation (i.e. income inequality) may further attenuate the overall well-being and life experience of Blacks. However, social capital and cohesion, composed of a number of factors, such as how connected groups feel, share resources and perceive the availa bility of moral support may help to explain for why residential segregation was not a ssociated with an in creased likelihood of adverse perinatal outcomes. Possible Explanations for Research Findings A possible explanation of current study fi ndings is that ethni c stratification of residential segregation in the United States is overestimated. While residential segregation has decreased both nationally and regionally, the average Index of Dissimilarity across the nation is 0.652, indi cating hypersegregation (Glaeser & Vigdor, 2001). Residential segregation may not have significantly affects on adverse perinatal outcomes because segregation is still re latively high across the United States. The state of residential segregation in the South and residential segregation indices in this study may also shed light on findings in the study. With residential indices greater than 0.60 considered hypersegrega ted (Glaeser & Vigdor, 2001), the mean Dissimilarity Index for the study sample wa s 0.47 (scores ranged from 0.20-0.69) and the mean Isolation Index for the study samp le was 0.45 (scores ranged from 0.04-0.74). Further, studies indicate that the South is now the second most integrated region, experiencing the largest regional redu ction in segregation between 1990 and 2000 (Glaeser & Vigdor, 2001). Generally low residen tial segregation levels in areas studied may explain no significant associations w ith adverse perinatal outcomes due to a

PAGE 151

139 potentially deleterious effects of segr egation when areas are hypersegregated. A second interpretation of findings centers on the distinction between historic and current influences of residential segregation. Previous studies indicate that, historically, residential segregation had effects on Black mobility, creating ghettos and segregated communities (Massey & Denton, 1993; Sugure, 1996). Since the Housing Act of 1968, residential segregation does not legall y constrain Blacks from making housing and neighborhood choices. This explanation points to the greater important of choice, rather than administrative policies, in current disc ussion of the role of neighborhood residence. These factors were not measured in this st udy. As described in directions for future research, choice of residence may be te mpered by preferences, by both Blacks and Whites, to live among those who are ethnically similar to themselves in order to reduce perceived stresses that may be inherent of living in more integrated neighborhoods. Evidence of both non-significant and pr otective effects of segregation in ethnically homogeneous neighborho ods indicate the complexity of macro and contextual factors (Johnson, Drisko, Ga llagher & Barela, 1999; O’Ca mpo et al., 1997; Pearl, Braveman, & Abrams, 2001; Rauh, Andrews et al., 2001) and may help to explain differing associations between segregation indicators and ea ch perinatal outcome studied. While many studies on the infl uence of residential segregation on infant outcomes have demonstrated deleterious effects (Elle n, 2000; LaVeist, 1989; Polednak, 1991), other evidence has shown that segregation may have either protective or non harmful effects in more ethnically homogeneous neighborhoods (Laveist, 1993, 1999; Pickett, Collins, Masi, & Wilkinson, 2005; Roberts, 1997). Pickett and colleague s (2005) found that among Black women in affluent census trac ts, those in predominantly Black areas

PAGE 152

140 experienced decreased risk of LBW and pr eterm delivery. The effects of affluence on birth outcomes among Black women were non-significant among those living in ethnically heterogeneous census tracts. Eviden ce that increased Isolation was associated with decreased likelihood of LBW among Black women (Hypot hesis 2) may be due to perceived social support and cap ital which may buffer agains t the stressors associated with integration. The residential segregation level of aggregation employed in this study may represent a distinct effect with magnitude and directionality that differs from dimensions that could be measured at the neighbor hood level (census tract). Birth certificate measures of adverse perinatal outcomes were linked to segregation indices at the macro (metropolitan or micropolitan sta tistical area) level in this st udy. This level of causation is associated with social, political and ec onomic structures th at may represent institutionalized racism. Segreg ation measured at the contextual level could represent the individual experience of discri mination in daily interactions Measurement of residential segregation at the contextual level may be more important to individual health outcomes than measures at the macro level. Inability to fully measure the degree to which income and education may moderate the degree of choice and realized preferences in neighborhood selection is am important methodological explanation for research findings. This study was limited by measures of socioeconomic status at the c ontextual level (census tract) because this variable was not included on birt h certificates. Previous resear ch signals that residential segregation influences communities through depr ivation of physical environments as well as limited employment opportunities. These e ffects are most evident in impoverished,

PAGE 153

141 segregated neighborhoods (Jargowsky, 1994, 1996, 1997; Strait, 2006). The degree to which income and education mediate or mode rate the associations between residential segregation and individual health may help to more appropriately explicate the contextindividual health outcome relationship. Multilevel model sample size was based on the number of groups and units at the highest level of aggregation te sted in hypotheses, the macro le vel. Published studies that center on sample size sensitivity in multilevel an alysis indicate that there should be at least 50 clusters at Level 3 (MSAs) with an average of 10 units per cluster (Hox, 2001). Level 3 consisted of 63 statistical areas, with an average of 69.21 ce nsus tracts per area, a sample size, exceeding the threshold required to achieve statistical power of 0.80 (Cohen, 1988). The potential for spuri ous or unimportant findings mu st be noted given the large number of Level 1 observations and that in dividual observations were critical to examining compositional effects. The identifi ed associations may not have existed, but yielded statistics that were expected to occur a certain proportion (alpha) of the time when null hypotheses were true. Further, ps eudo-associations may have occurred, making it difficult to detect intermediate variable s. These variables could have been excluded from analyses or may have been intermediati ng in ways that were not tested (Raferty, 1986, 1995). Finally, findings may indicate that ther e is no relationship between residential segregation and adverse peri natal outcomes, notwithsta nding previously described explanations. However, results provide importa nt insight into the influence of personal and contextual factors on adverse perinatal outcomes utilizing a technique that most comprehensively accounts for multiple levels of exposure. Results indicate that more

PAGE 154

142 proximal, downstream factors (individual and co ntextual) should be c onsidered in efforts to reduce ethnic disparities in perinatal outcomes. Study Strengths Multilevel modeling of the independent e ffects of residentia l segregation on the likelihood of adverse perinatal outcomes c ontributes to better understanding of the determinants of health disparities. Most studies examining the impact of residential segregation on population health have been limited by cross-sectional and ecological study designs analyzed at individual or aggregate levels. Further, contextual factors that may be controlled for in the relationship betw een structural factors and health outcomes are often excluded in single level studies. Evidence indicates that the relationships between residential contexts, structures and health outcomes are not homogeneous and require more specificity in investigation. None of the identifie d studies researching community levels of residential segregation have utilized this analytical approach (Collins et al., 1997; Po lednak et al., 1991, 1996). This research expanded upon existing studies of the associations between place on health outcomes through the simultaneous i nvestigation of thr ee adverse perinatal outcomes associated with increased infant mortality and morbidity. Studies previously investigating area effects on health outcom es broadly include se lf-rated health, life expectancy, all-cause mortality, and infant mortality. Virtually unexplored is how the effects of place on health may vary by reside ntial segregation dimensions measured and perinatal outcome examined (LBW, preterm delivery and SGA). This study also sheds light on the unique experience of pregnancy and perinatal

PAGE 155

143 outcomes. Study results indicate d that contextual and indivi dual factors were the chief contributors to increased risk for adverse pe rinatal outcomes, rather than macro-level, structural factors that have been significant predictors of mortality and other health outcomes. Findings infer that the effects of residential segregation are unique to the outcome studied. Study Limitations Limitations of the data sources utilized in this study were discussed earlier in this study and are summarized in this section. E rrors in coding and underestimation of key labor and delivery outcomes in birth certifi cate data are among noteworthy challenges (Brunskill, 1990; Green et al., 1998; Hamvas, Kwong, DeBaun, Schramm, & Cole, 1998; Watkins et al., 1996). The specificity, sensitivit y and predictive value of birth certificates across sites vary in the estimation of maternal medical risk factors, complications of labor and delivery, number of interv entions or procedures, congenital anomalies, and measures of prenatal care. Explanations for lower sensitivity among certain birth certificate elements are vast and include issues surroundi ng site variations in data management and self-report bias. Birth certific ate data may also be subject to non-sampling error occurring during initial data entry (at the hospital level) and in processing birth certificates at health departments. The lack of information on psychosoc ial, attitudinal, or stress measures that have been previously associated with pe rinatal outcomes (Rich-Edwards et al., 2001; Stancil et al., 2000) are a recognized lim itations of birth certificate data. Individual preferences and values associated with wh ere one chooses to reside were not accounted for in this study. For ex ample, a person may perceive benefits to living in an ethnically homogeneous nei ghborhood which are not captured by the data

PAGE 156

144 employed in this study. It is anticipated that th e contribution of this in itial investigation is a foundation for future longitudinal examina tion of the meanings and perceptions of residential structures and cont exts to more fully understand th e risk or protective effects of residential segregatio n on perinatal outcomes. While the decision to employ an obser vational study design that was crosssectional and utilized secondary data was c onstrained by the availability and nature of data at each level of analysis, inherent limitations must be acknowledged. Pocock and Elbourne (2002) indicated that ob servational studies are not e xperimental and, as a result, do not involve randomization, the use of entry criteria, and the rigor ous use of standard definitions of conditions. The primary limitations of cross-sectional designs are that they preclude any formal conclusions about causali ty and directionality in the association between residential segregati on and adverse perinatal outc omes. The primary challenge associated with the use of secondary data in this study was potentially differing levels of the accuracy of live birth certificate data entry across the three states that were examined. Like many other studies investigating ar ea variations in he alth, this study was restricted to the choice of spatial unit (censu s tract). The census trac t was the selected unit of analysis in this study becau se it is a small, relatively pe rmanent statistical subdivision. Designed to be relatively ho mogeneous with respect to population characteristics, economic status and living conditions, census tr acts capture within c ity variation and the immediacy of social context that is lost to cross-metro analyses (Guest et al., 1998). Census tract identifiers on birth certificates were used in orde r to link residential segregation indices and contextu al covariates calculated by the U.S. Census Bureau to create a three-level profile for each woma n in the study. Altering the geographical unit

PAGE 157

145 may have lead to different conclusions re garding the importance of areas on health (Boyle & Willms, 1999). The data in this research were limited by residential segregation indicators associated with the year of labor and deliv ery. Nothing is known about whether a woman was new to the census tract indicated on the live birth certificate or whether she was a resident within this community for several years. This limitation of the data may be interpreted as a methodological artifact of the live birth certificate da taset utilized and may have influenced research findings. Findings from this study offer many opportunities for pub lic health professionals and clinicians to 1) gain increased understanding of pa st and current biological, psychosocial, and geographical influences on health through bot h qualitative and quantitative methods, 2) address chronic and emerging health issues and 3) investigate both clinical and social risk factors over time. Directions for future research in this area among members of this important public hea lth population are described in the section that follows. Directions for Future Research The ethnic disparities that persist for the majority of health outcomes in the United States demonstrate the need for expa nded investigation of a more comprehensive array of variables that influence birth outco mes at multiple levels. While administrative and geopolitical boundaries most frequently gover n the assessment of structural forces (e.g. Isolation Index, Dissimilarity), the inte rnalization and meani ngs associated with these social, cultural, and political influences are critical to complete assessment of the

PAGE 158

146 determinants of and pathways to adverse pe rinatal outcomes. Pyschosocial and biological assessments, context and duration of current and previous residential exposure, and individuals’ perceptions of their neighborhoods are among the measures that should be analyzed in each study of associations betw een space, place and perinatal outcomes. Future research should include an asse ssment of perceived stress to examine how interactions between psychological and struct ural factors may influence birth outcomes. Previous research suggests that, for some subgroups, psychosocial stress may be a byproduct of perceptions of discriminati on (Farmer, 1997; Krieger, & Sidney, 1996; Livingston, 1994). Higher stress levels among African American and Hispanic women in studies of ethnically and socioeconomically di verse samples have led some to conclude that perceived racial and gender discrimination and their associated stressors contribute to an increased risk for preterm birth (RichEdwards et al., 2001; Stancil et al., 2000). Second, chronic stressors experienced by women of lower SES may also lead to adverse intrapsychic processes (i.e. low levels of personal resilience, self esteem/optimism, mastery belief/sense of personal control) medi ated by perceived stress. Further, living in chronically stressful environments appear to erode personal re silience, which may heighten perceived stress, anxiety, a sense of helplessness, a l ack of optimism, and depression increasing the risk of poorer birth outcomes. The biological manifestations of stress a nd their associations with contextual and structural factors should also be assessed in multilevel studies. Studies supporting the influence of psychosocial factors on fetal development and pregnancy outcomes suggest that the consequences of stress include increased corticotrophin-releasing hormone (CRH) a neuropeptide synthesized primarily in the paraventricu lar nucleus of the

PAGE 159

147 hypothalamus. Higher CRH levels have b een found in women who have preterm (Wadhwa et al., 1998). Adverse effects of chroni c stress exposure can be seen as early as the prenatal environments. Mate rnal stress at 18 to 20 weeks' gestation has been found to significantly predict cortic otrophin-releasing hormone (CRH) at 28 to 30 weeks gestation, even after controlling for CRH at 18 to 20 weeks (Hobel et al., 1999). Stress may also increase proinflammatory cytoki ne production, which leads to increased prostaglandin production, increased uterine contractility and pret erm labor (Turnbull & Rivier, 1999). Stress may be measured as an objective stimuli, an emotional/biological response, and a perception of control or a combination of these factors. A number of theoretical issues that may provide fresh avenues of research include determining what constitutes an integrated neighborhood in “real life”. The quantitative residential segregation indicators employed in this study are based on ideal values and thresholds of integration. A question that may be investigated is whether having an “equal” number of Blacks and Whites in nei ghborhoods will necessarily improve health. The degree of political power and governance in residentially segregated neighborhoods may be more significantly associated with birth outcomes than previously utilized indicators (Hart et al., 1998; Williams & Co llins, 2001). Investigation of voting records and city government initiatives that are focused on improving neighborhoods will further provide information on the poli tical power of a community. The previous associations between measures of social capital, cohesi on, integration and health (House et al., 1988, Kawachi et al., 1996; Wilkinson; 1996) call for assessment and incorporation of these contextual measures in order to stre ngthen studies in this research field. The length of time that an individual has liv ed in a community is critical to better

PAGE 160

148 understanding of the impact of residential segregation and should also be assessed in future studies. Similar to the majority of re search on the factors asso ciated with perinatal outcomes, the data in this research was limited by residential segregation indicators associated with the year of labor and deliv ery. Nothing is known about whether a woman was new to the census tract reported on the li ve birth certificate or whether she was a resident within this community for several years. A longitudinal survey of expectant mothers, for the duration of pregnancy, woul d allow better understanding of previous place and duration of residence. Perhaps a lif etime of living in neighborhoods that are residentially segregated may be less detrimental to health than more recent residence in a segregated neighborhood due to changes is so cioeconomic status or other stressful life events. Other studies in this area coul d focus on duration and level of residence segregation in the current nei ghborhood relative to the previous residence. No identified studies on how transitional effects associated wi th place of residence effect health status have been identified in the public health literature. Preference for residence in neighborhoods with higher concentra tions of residents that are ethnically similar should also be di stinguished from housing choice. For instance, Blacks report preferences for neighborhoods that are about half Black and Half White (Farley, et al., 1994; Timberlake, 2000; Z ubrinsky & Bobo, 1996). However, surveys do not typically ask respondents about perceived costs of livi ng in integrated neighborhoods. Blacks might prefer to trade residential se gregation for increase d proximity to close friends and family members (Clark, 1986). Becau se Blacks tend to be less affluent than Whites on many measures of socioeconomic st atus, the choice to live among those who are ethnically similar may more often mean that Blacks live among a greater percentages

PAGE 161

149 of poorer individuals than Whites do. However, the de gree to which social support may buffer against the deleterious effects of resi dential segregat ion may help to explain the current research findings. Blacks or Whites who choos e to live in ethnically homogeneous neighborhoods may, in essence, be choosing, residential solidarity over residential integration. Residential segregati on may not be protective of health, but these key, health enhancing moderators may buffer ag ainst its anticipated, deleterious effects. Variations in protocol for collecting liv e birth certificate da ta across sites, nonclinical personnel recording data and little or no auditing to ensure data reliability in hospitals are among explanations for lower sensitivity of specific birth certificate elements. However, little or no data on per ception of neighborhoods, racism, and stress to complement studies that utili ze birth certificate data do not bring researchers much closer to understanding and addressing disparities in health. The collection of psychosocial data could be accomplished by administering questionnaires among pregnant women at multiple points during pregnancy, beginning with the first prenatal visit. Cortisol readings would provide insight into th e self–reported effect of community level variables on stress. Areas selected for inclusion in thes e studies could represent a continuum on the residential segregation scale in order to de termine the relative influence of residential segregation after controlling for biological samples, survey and administrative data. Ethnic group heterogeneity in health and patter ns of segregation has been identified in research studies and should receive increased attention in multilevel examinations. Several previously described st udies have identified variatio ns in health outcomes among Black women of Hispanic, Caribbean and Afri can descent with findi ngs suggesting that the burden of ethnic disparities is not e qually shared among Blacks (Cabral, Fried,

PAGE 162

150 Levenson, Amaro, & Zuckerman, 1990; Collins, and David, 2000; Cooper et al., 2003; Fang, Madhavan & Alderman, 1997; Palloto, Friedm an et al., 1993). Furt her, patterns of segregation vary as much within as acro ss ethnic subgroups, implying the importance of closer scrutiny of resi dential patterns across and within groups to asse ss their varying associations with health (White, Kim, & Glick, 2005). Public health funding and interventions that are developed with incr eased understanding of both the protective and deleterious effects of cultural contexts and residential mobility are critical to more targeted and effective interventions. Implications for Public Health Locally implemented studies are needed in response to this research and others that are limited by state and national datase ts. National data collection mechanisms may not provide the full picture of the relationshi ps between adverse perinatal outcomes and place due to the effect of regression to th e mean, which may obscure what occurs in neighborhoods. Cultural and re gional norms may further color the definitions and experience of residential segreg ation and other structural f actors that are theorized to influence health. Health departments may be the most appropriate vehicle for mandated collection of this data due to the fact that they are publicly governed, and are characterized by more frequent contact betw een pregnant women and providers than the limited interaction mothers have with hospita l personnel who complete birth certificates after labor and delivery. Despite the emphasis of public health re search to avoid victim blaming due the

PAGE 163

151 long established influence of fo rces outside the realm of indi vidual control, most prenatal education and interventions are still primarily tailored in respons e to individual risk. Findings of this study indicated the importance of a number of cont extual factors that were significant predictors of adverse perina tal outcomes. A cognitive-shift among public health practitioners is needed to respond to research study findings in order to arm expectant mothers with tools needed respond to and cope with the resi dential contexts in which they live. The future research directi ons described above must be translated to practice through practitio ners informed and trained to implement interventions that address the multilevel strengths and challenges faced by the women that they serve. Conclusion The complex web of causation associated withvariations in population health requires identification of the spiders, or funda mental causes, that instigate risk for risk factors (Krieger, 1994). In order to draw cl oser to identifying the determinants and pathways to adverse perinatal outcomes, co mprehensive data on biological, social and spatial markers is critical. Further, studies of the association between place and health should incorporate a thorough a ssessment of residential pr eferences and experiences. Until more is known about motivations for resi dential perceptions, our understanding of how place and space influence on health is incomplete.

PAGE 164

152 REFERENCES Abrams, B., & Newman, V. (1991) Small-for-ges tational-age birth: Ma ternal predictors and comparison with risk factors of s pontaneous preterm delivery in the same cohort. American Journal of Obstetrics and Gynecology, 164 (3), 785-790. Acevedo-Garcia, D. (2001). Zip code le vel factors or tube rculosis: Neighborhood environment and residential segregation in New Jersey 1985-1992. American Journal of Public Health 81, 1480-1482. Acevedo-Garcia, D., & Lochner, K. (2003) Re sidential segregation and health. In I. Kawachi, & L. Berkman (Eds.) Neighborhoods and Healt (pp. 265-281). New York, NY: Oxford University Press, 265-281. Acevedo-Garcia, D., Lochner, K., Osypuk, T., & Subramanian, S. (2003). Future directions in residential segregation and health res earch: A multilevel approach. American Journal of Public Health 93(2), 215-221. Adler, N., Boyce, T., & Chesney, M. ( 1994). Socioeconomic status and health: The challenge of the gradient. American Psychologist, 49 (1), 15-24. Aiken, L. S. (1991). Multiple regression: Testi ng and interpreting interactions Newbury Park, CA: Sage Alba, R., & Logan, J. (1992). Assimilation and stratification in the homeownership patterns of racial and ethnic-groups. International Migration Review, 26 (4), 13141341. Alexander, G. (1992). Validity of postnatal as sessments of gestational age: A comparison of Ballard et al. a nd early ultrasonography. American Journal of Obstetrics and Gynecology 166, 891-895.

PAGE 165

153 Alexander, G., Hines, J., Kaufman, R., Mo r, J., & Kogan, M. (1996). A United States national reference for fetal growth. Obstetrics and Gynecology 87(2), 87-163. Alexander, G., Kogan, M., Himes, J., Mo r, J., & Goldenberg, R. (1999). Racial differences in birthweight for gestationa l age and infant mortality in extremelylow-risk US populations. Paediatric and Perinatal Epidemiology 13, 205-217. Alexander, G., Kogan, M., & Nabukera, S. (2002) Racial differences in prenatal care use in the United States: Ar e disparities decreasing? American Journal of Public Health, 92 (12), 1970-1975. Alexander, G., & Kotelchuck, M. (1996). Quanti fying the adequacy of prenatal care: A comparison of indices. Public Health Reports, 111 (5), 408-416. Alexander, G., & Petersen, D. (1997). Threat s to accurate interpretation of secondary data. In J. Kotch (Ed.), (pp. 395-403). Gaithersburg, MD: Aspen. Alexander, G., Petersen, D., Powell, E., & Tompkins, M. (1989). A comparison of gestational age reporting methods based on physician estimate and date of last normal menses from fetal death reports, American Journal of Public Health, 79 (5), 600-602. Alexander, G., Tompkins, M., Altekruse, J., Hornung, C. (1985). Racial differences in the relation of birth weight and gesta tional age to neonatal mortality. Public Health Reports, 100 (5), 539-47. Alexander, G., Tompkins, M., & Cornely, D. (1990). Factors influencing the relationship between a newborn assessment of gestat ional maturity and the gestational interval. Paediatric and Perinatal Epidemiology 4, 133-146.

PAGE 166

154 Alexander, G., Tompkins, M., Petersen, M., Hulsey, T., & Mor, J. (1995). Discordance between LMP-based and clini cally estimated gestationa l age: Implications for research, programs and policy. Public Health Reports, 110 (4), 395-402. Alker, H. (1969). A typology of ecological fall acies. In L Dogan, M., Rokkam, S., Ed.s. Social ecology. Boston: The MIT Press, 1969, 69-86. Altshuler, A., Morrill, W., Wolman, H., Mitc hell, F., & The Committee on Improving the Future of U.S. Cities through Improved Metropolitan Governances. (1999). Governance and opportunity in metropolit an America. (p. ix; 347). Washington, D.C. National Academy Press. American Anthropological Association (1997) American Anthropolog ical Association: Response to OMB Directive 15: Race and ethnic standards for federal statistics and administrative reporting. Available at: http://www.aaanet/gvt/ombdraft.htm Accessed May 1, 2005. Ananth, C., Balasubramanian, B., Demissie, K., & Kinzler, W. (2004). Small-forgestational-age births in the United St ates: An age-period cohort analysis. Epidemiology, 15 (1), 28-35. Ananth, C., Demissie, K., Kramer, M., & Vintz ileos. A. (2003). Small for gestational age births among black and white women: Te mporal trends over a recent decade in the United States. American Journal of Public Health 93, 577-579. Ananth, C., Joseph, K.A., Oyelese, Y., Demissi e, K., & Vintzileos. A. (2005). Trends in preterm birth and perinatal mortality among singletons: United States, 1989 through 2000. Obstetrics and Gynecology., 105 (5), 1084-1091.

PAGE 167

155 Ananth, C., & Platt, R. (2004). Reexamining th e effects of gestation age, fetal growth, and maternal smoking on neonatal mortality. BMC Pregnancy and Childbirth 4, 22. Atkinson, J., MacDorman, M., & Parker, J. (2001 ). Trends in births to parents of two different races in the United States: 1971-1995. Ethnicity Disease 11, 273-285. Bandura, A. (1986). Social Foundations of thought and action: A social cognitive theory Englewood Cliffs, N.J.: Prentice-Hall. Battaglia, F., Frazier, T., & Hellegers, A. (1966). Birth weight, gestational age and pregnancy outcome with special references to high birth weig ht-low gestational age infants. Pediatrics 37, 417-422. Beers, M., & Berkow, R. (1999). The Merck manual of diagnosis and therapy: 17th Edition. John Wiley & Sons; 17th edition. Ben-Shlomo, Y., White, I., & Marmot, M. (1996). Does the variation in the socioeconomic characteristics of an area affect mortality? British Medical Journal 312, 1013-1014. Berkowitz G., & Papernik E. (1993 ). Epidemiological preterm birth. Epidemiologic Reviews, 15, 414–443. Beuscher, P., Taylor, K., Davis, M., & Bo wling J. (1993). The quality of the new certificate data: A validation study in North Carolina. American Journal of Public Health 83, 1163-1165. Biggeri, A., Marchi, M., Lagazio, C., Ma rtuzzi, M., & Bohning, D. (2000). Nonparametric maximum likelihood estimators for disease mapping. Statistics in Medicine. 19 (17-18) 2539-54.

PAGE 168

156 Bishop, J. (1977). Jobs, cash transfers, and marital inst ability: A review of the evidence. Madison: Institute for Research on Poverty, University of Wisconsin. Blakely, T., Lochner, K., & Kawachi, I. ( 2002). Metropolitan area in come inequality and self-rated health-A multilevel study. Social Science and Medicine, 54 (1), 65-77. Blondel, B., Morin, I., Platt, R., Kramer, M ., Usher, R., & Breart, G. (2002). Algorithms for combining menstrual and ultraso und estimates of gestational age: Consequences for rates of preterm and post term birth. BJOG: An International Journal of Obstetrics and Gynecology 109, 718-720. Brodish, P., Massing, M., & Tyroler, H. ( 2000). Income inequality and all-cause mortality in the 100 counties of North Carolina. Southern Medical Journal 93 (4), 386-391. Brooks-Gunn, J., Duncan, G., & Aber, J. (Eds.) (1997). Neighborhood poverty: Context and consequences for children, Vol. 1. New York: Russell Sage Foundation. Brosco, J. (1999). The early history of infant mortality rate in Amer ica: A reflection upon the past and prophecy of the future. Pediatrics 103, 478-485. Brunskill, A. (1990). Some sources of error in the coding of birth weight. American Journal of Public Health, 80 (1), 72-73. Buekens, P., Delvoye, P., & Robyn, C. (1984). Epidemiology of pregnancies with unknown last menstrual period. Journal of Epidemiology and Community Health, 38 79-80. Buka, S., Brennan, R., Rick-Edwards, J., Raudenbush, S., & Earls, F. (2002) Neighborhoods support and the birth weight of urban infants. American Journal of Epidemiology 157, 1-8.

PAGE 169

157 Burton, P., Gurrin, L., & Sly, P. (1998). Tutorial in biostatistics: Extending the simple linear regression model to account for co rrelated responses: An introduction to generalized estimating equations and multi-level mixed modeling. Statistics in Medicine, 12 1261-1292. Cabral, H., Freid, L., Levenson, S., Amaro, H., & Zuckerman, B. (1990). Foreign -born and Us-born black women: Differences in health behaviors and birth outcomes. American Journal of Public Health, 80 (1), 70-71. Centers for Disease Contro l and Prevention (2005). Racial/ethnic disparities in infant mortality-United States, 1995-2002 Morbidity and Mortality Weekly Report, 54 (22), 553-556. Clark, K., Chun-Mei, F., & Burn ett, C. (1997). Accuracy of birth certificate data regarding the amount, timing, and adequacy of prenatal care using prenatal clinic medical records as referents. American Journal of Epidemiology, 145 68-71. Clark, P. (1998). Programming of the hypthalamo-pituitary-adr enal axis and the fetal origins of adult di sease hypothesis. European Journal of Pediatrics 157 (Supplement 1), s7-s10. Cohen, J. (1988). Statistical power analysis for the behavioral sciences Hillsdale, NJ: Lawrence Erlbaum. Collins, C. (1985). Four sociological traditions. New York, NY: Oxford University Press. Collins, C., & David, R. (1997). Urban vi olence and African-American pregnancy outcome: An ecologic study. Ethnicity and Disease, 7 (3), 184-190. Collins, C., & Williams, D. (1995). Examining the black-white adult mortality: The role

PAGE 170

158 of residential segreg ation. Abstr. For 25th Public Health Conference on Records and Statistics and the National Committee on Vital and Health Statistics 45thh Anniversary Symposium, July 17-19, Washington, D.C. Collins, C., & Williams, D. (1999). Segregati on and mortality: The deadly effects of racism? Sociological Forum, 14 (3), 495-523. Control, C. (2000). Healthy People 2000 review. National Center for Health Statistics 16, 3-9. Cooper, R., Kaufman, J., & Ward., R. (2003) Race and genomics: Comment. New England Journal of Medicine, 348 (12), 1166-1170. Coulton, C., Chou, J., Wang, E., & Su, M. (1996) Geographic concentration of affluence and poverty in 100 metropolitan areas, 1990. Urban Affairs Review, 32 (2), 186216. Cutler, D., Glaeser, E ., & Vigdor, J. (1999) The Rise and Decline of the American Ghetto." Journal of Political Economy, 107, 455-506 Daly, M., Duncan, G., Kaplan, G., & Lynch, J. (1998). Macro to micro links in the relation between income inequality and mortality. Milbank Quarterly 76, 315339. Daniels, N., Kennedy, B., & Kawachi, I. (2000). Is inequality bad for our health ? Boston: Beacon Press. Darden, J. (1987). Socioeconomic Status and Racial Residential Se gregation: Blacks and Hispanics in Chicago. International Journal of Comparative Sociology. 281 (2), 1–13. Darroch, A., & Marston, W. (1971). The soci al class basis of ethnic residential

PAGE 171

159 segregation: The Canadian Case. American Journal of Sociology, 77 (3), 491-510. Davey Smith, G. (1996). Income inequality and mortality: Why are they related? British Medical Journal 313, 987-988. Davey Smith, G., & Eggers, M. (1996). Commentary: Understa nding it all—Health, meta-theories and mortality trends. BMJ 313(7072), 1584-1585. David, R. (1980). The quality and completeness of birthweight and gestational age data in computerized birth files. American Journal of Public Health 70, 964-973. Deaton, A. (2001). Health inequality and economic development. Prepared for Work Group 1 WHO committee. Ma croeconomic Health. New Jersey: Princeton University. Deaton, A., & Lubotsky, D. (2003). Mortality, in equality and race in American cities and states. Social Science and Medicine, 56 (6), 1139-1153. Denton, N., & Massey, D. (1988). Residential segregation of Blacks, Hispanics, and Asians by socioeconomic status and generation. Social Science Quarterly 69, 797-817. Diez,-Roux, A. (1998). Bringing context back in to epidemiology: variables and fallacies in multilevel analysis. American Journal of Public Health 88, 216-222. Diez-Roux, A. (2002). A glossary for multilevel analysis. Journal of Epidemiology and Community Health 56, 588-594. Diez, Roux, A. (2003). A glossary of multilevel analysis. Epidemiological Bulletin, 24 (3), 11.13. Diez-Roux, A. (2004). Estimating neighborhood hea lth effects: The challenges of causal inference in a complex world. Social Science and Medicine 58, 1953-1960.

PAGE 172

160 Diez-Roux A., Link B., & Northridge M. (2000). A multilevel analysis of income inequality and cardiovascula r disease risk factors. Social Science & Medicine, 50 (5), 673-87. Diez-Roux A., Nieto, F., Muntaner, C., Tyroler, H., Comstock, G., Shahar, E., Cooper L., Watson R., & Szklo, M. (1997). Neighborhood environments and coronary heart disease: a multilevel analysis. Ameri can Journal of Epidemiology. 146:48-63. DiGuiseppe, D., Aron, D., Ranbom L, David, L., Harper, D., & Rosenthal, G. (2002). Reliability of birth cer tificate data: a multi-hospita l comparison to medical ... Maternal and Child Health Journal 6, 169-179. Dobie, S., Baldwin, L., Rosenblatt, R., For dyce, M., Androlla, C., & Hart, L. (1998). How well do birth certificates describe d the pregnancies they report? The Washington State experience wi th low-risk pregnancies. Maternal Child Health Journal 2, 145-154. Doctor, B., O’Riordan, M., Kirchner, L., Shah, D., & Hack, M. (2001) Perinatal correlates and neonatal outcomes of small for gestational age infants born at term gestation. American Journal of Obstetrics and Gynecology, 185 (3), 652-659. Duncan, G., Connell, J., & Klebanov, P. (1997) Conceptual and met hodological issues in estimating causal effects on neighborhoods and family conditions on individual development. In J. Brooks-Gunn, G. Duncan & J. L. Aber (Eds.), Neighborhood poverty: Context and consequences for children. (Volume I) (pp. 219-250). New York: Russell Sage Press Durch, J., & Madans, J. (2001). Methodologi cal issues for vita l rates and population estimates: 1997 OMB standards for data on race and ethnicity. Vital Health

PAGE 173

161 Statistics 31, 1-30. Durlauf, S. (1996). A theory of persistent income inequality. Journal of Economic Growth, 1, 75-93. Edwards, L., Coulter, C., Symonds, M., & McMillem, I. (2001). Prenatal undernutrition, glucocorticoids, and the program ming of adult hypertension. Clinical Exp. Pharmacology, Physiology 28, 938-941. Ellen, I. (2000). Is segregation bad for your health? The case of lowbirth weight. In W. Gale, & J. Pack (Eds.), Papers in urban affairs (pp. 203-238). Washington, D.C.: Brookings Institution Press. English, P., Kharrazi, M., Davies, S., Scalf, R., Waller, L., & Neutra, R. (2003). Changes in the spatial pattern of low birth weight in a southern California county: the role of individual and neighborhood level factors. Social Science and Medicine 56, 2073-2088. Eshleman, J. (2003). The family. 10th Ed. Boston, MA: Allyn and Bacon. Fang, J., Madhaven, S., & Alderman, M. (1997) Low birth weight: Race and maternal nativity-impact of community income. Pediatrics 103, e5. Fang, J., Madhaven, S., Bosworth, W., & Alderm an, M. (1998). Residential segregation and mortality in New York City. Social Science and Medicine 47, 469-476. Fannie Mae (1999). Fannie Mae national Housing Survey 1999. Washington, D.C: Fannie Mae. Farley, R. (1970). The changing distribution of Negroes within metr opolitan areas: The emergences of the black suburbs. American Journal of Sociology 75, 512-529. Farmer, M., & Ferraro, K. (1997). Distress & perceived health: Mech anisms of health

PAGE 174

162 decline. The Journal of Health and Social Behavior 39, 298-311. Figueroa, J., & Breen, N. (1995). Significan ce of underclass reside nce on the stage of breast or cervical cancer diagnosis. AEA Pa. Proc., 85 (2), 112-116. Fiscella, K., & Franks, P. (1997). Poverty or income inequality as a predictor of mortality: Longitudinal cohort study. British Medical Journal 314, 1724-1727. Foster, H., Wu, L., Bracken, M., Semenya K., Thomas, J., & Thomas, J. (2000). Intergenerational effects of high socioeconomic stat us on low birthweight and preterm birth in African Americans. Journal of the National Medical Association, 92 (5), 213-221. Franzini, L., Ribble, J., & Spears, W. ( 2001). The effects of income inequality and income level on mortality vary by po pulation's size in Texas counties. Journal of Health and Social Behavior 42(4), 373-387. Frazier, E. (1957). The Black Bourgeoisie. New York, NY: Simon and Schuster, 15-22. Fretts, R., Schmittdiel, J., McLean, F., Usher, R., & Goldman, M. (1995). Increased maternal age and the risk of fetal death. The New England Journal of Medicine, 333 (15), 955-957. Friedman D., Cohen, B., Mahan, C., Lede rman, R ,Vezina, R., & Dunn, V. (1993). Maternal ethnicity and birthweight among blacks. Ethnicity & Disease 3 (3), 255269. Frisbie, W., Parker, Forbes, D., & Pullum, S. (1996). Compromised birth outcomes and infant mortality among racial and ethnic groups. Demography 33, 469-481. Galster, G., & Killen, S. (1995). The geography of metropolitan opportunity: A reconnaissance and conceptual framework. Housing Policy Debate, 6 (1), 7-43.

PAGE 175

163 Galster, G., & Mikelsons, M. (1995). The ge ography of metropolitan opportunity: A case study of neighborhood conditions confr onting youth in Washington, D.C. Housing Policy Debate, 6 (1), 73-102. Gee, G. (2002). A multilevel analysis of th e relationship between institutional racial discrimination and health status. American Journal of Public Health 92, 615– 623. GeoLytics (2004). Neighborhood Change Data base (1970-2000). Retrieved November 10, 2004, from http://www.geolytics.com Gerber, E., & de la Puenta, M. (1996) The development and cognitive testing of race and ethnic origin questions for the Year 2000 Decennial Census. Proceedings of the Bureau of the Census 1996 Annual Conference Rosslyn, VA. Geyer, S., & Peter, R. (2000). Income, occ upational position qualification and health inequalities—competing risks? (Compa ring indicators of social status). Journal of Epidemiology and Community Health 54, 299-305. Glaeser, E., & Vigdor, J. (2001) Racial Segregation in the 2000 Census. Promising News. The Brookings Institution Survey Series Washington, D.C. Goldenberg, R., Cliver, S., Mulvihill, F., Hickey, C., Hoffman, H., Lorraine V., Klerman, Y., & Johnson, M. (1996). Medical, psychol ogical, and behavioral risk factors do not explain the increased risk for low birth weight among black women. American Journal of Obstetrics and Gynecology 175, 1317-1324. Goldenberg, R., Davis, R., Cliver, S., Cutter, G., Hoffman, G., Dubard, M., & Copper, R., (1993). Maternal risk factors and their influence on fetal anthropometric measurements. American Journal of Obst etrics and Gynecology, 168 (4), 1197-

PAGE 176

164 1205. Goldenberg, R., Davis, R., Cutter, G., Hoffm an, H., Brumfield, C., & Foster, J. (1989). Prematurity, postdates and growth reta rdation: The influence of use of ultrasonography on reported gestational age. American Journal of Obstetrics of Gynecology 160, 462-470. Gorman, B. (1999). Racial and ethnic variati on in low birthweight in the United States: Individual and contextual determinants. Health and Place 5, 195-207. Gravelle, H. (1998). How much of the rela tion between population mortality and unequal distribution of income is st atistical or artifactual. British Medical Journal, 316 382-385. Green, D., Moore, J., Adams, M., Berg, C ., Wilcox, L., & McCarthy, B. (1998). Are we underestimating rates of vagi nal birth after previous ce sarean birth? The validity of delivery methods from birth certificates. American Journal of Epidemiology, 147 (5), 581-586. Greenberg, M., & Schneider, D. (1994). Violen ce in American cities: Young black males are the answer, but what is the question? Social Science and Medicine 39, 179187. Guest, A., Almgren, G., & Hussey, J. (1998). The ecology of race and socioeconomic distress: Infant and worki ng-age mortality in Chicago. Demography 35, 23-34. Guyer, B., Martin, J., MacDorman, M., Anderson, R., & Strobino, D. (1997). Annual summary of vital statistics-1996. Pediatrics, 100, 905-919.

PAGE 177

165 Haan, M., Kaplan, G., & Camacho, T. (1987). Poverty and health: Prospective evidence from the Alameda County study. American Journal of Epidemiology, 125 (6), 989998. Hall, M., Carr-Hill, R., Fraser C., Campbell, D., & Samphi er, M (1985). The extene and antecedents of uncertain gestation. British Journal of Obstetrics and Gynaecology 92, 445-451. Hamilton, B., Martin, J., & Sutton, P. (2004). Births: Preliminary Data for 2003. National Vital Statistics Reports. National Vital Statistics Reports, 53(9 ). Hamvas, A., Kwong, P., DeBaun, M., Schram m, M., & Cole, F. (1998). Hyaline membrane disease is underreported in a linked birth-infant death certificate database. American Journal of Public Health, 88 (9), 1387-1389. Hart, K., Kunitz, S., Sell, R., & Muka mel, D. (1998). Metropolitan governance, residential segregation, and mort ality among African Americans. American Journal of Public Health, 88 (3), 434-438. Hatcher, L., & Stepanski, E. (1994). A step-bystep approach to using the SAS system for univariate and multivariate statistics. SAS Institue, Inc., Cary, North Carolina. Hattersley, A., & Tooke, J. (1999). The fe tal insulin hypothesis: an alternative explanation of the associa tion of low birth weight w ith diabetes and vascular disease. The Lancet 353, 1789-1792. Hayes-Bautista,D.E. & Chapa, J. (1987). Lation terminology: conceptual bases for standardized terminology. American Journal of Public Health, 77 (1), 61-68. Hedeker, D., Gibbons, R., & Flay, B. (1994) Random effects regression models for clustered data with an example from smoking prevention research. Journal of

PAGE 178

166 Consulting and Clinical Psychology, 62 757-765. Hessol, N., Fuentes-Afflick, E., & Bacchetti, P. (1998). Risk of low birthweight among black and white parents. Obstetrics and Gynecology, 92, 814-822. Hillemeier, M., Lynch, J., Harper, T., Raghunath an, T., & Kaplan, G. (2003). Relative or absolute standards for child poverty: A st ate-level analysis of infant and child mortality. American Journal of Public Health, 93 (4), 652-657. Himes, J., & Hoaglin, D. (1989). Resist ant cross-age smoothing of age-specific percentiles for growth reference data American Journal of Human Biology, 1, 165-173. Hobel, C., Dunkel-Shetter, C., Roesch, S., Ca stro, L., Arora, C. (1999). Maternal plasma cotricotropin-releasing hormone associated with stress at 20 weeks' gestation in pregnancies ending in preterm delivery. American Journal of Obstetrics and Gynecology, 180 (1), S257-S263. Hogberg, U., & Larsson, N. (1997). Early da ting by ultrasound and perinatal outcome. A cohort study. Acta Obstetrica Gyne cologica Scandinavica, 76, 907-912. House, J., Landis, K., & Umberson, D. ( 1988) Social relationships and health. Science 214, 540-545. Hox, J. (1998). Multilevel mode ling: When and why? In I. Balderjahn, R. Mathar & M. Schader (Eds.), Classification, data analysis and data highways (pp. 147-154). New York: Springer Verlag. Hox, J. & Maas, C. (2001). The accuracy of multilevel structural eq uation modeling with psuedobalanced groups and small samples. Structrual Equation Modeling, 8 (2), 157-174.

PAGE 179

167 Huxley, R., Neil, A., & Collins, R. (2002). Un raveling the fetal origins hypothesis: Is there really and inverse association be tween birthweight and subsequent blood pressure? Lancet 360, 659-665. Iceland, J., Weinberg, D., & Steinmetz, E. (2002). Raci al and ethnic residential segregation in the United States: 19 80-2000. U.S. Census Bureau: Washington, D.C. Jackson, S., Anderson, R., Johnson, J., & Sorlie P. (2000). The relation of residential segregation to all-cause mortal ity: A study in black and white American Journal of Public Health, 90 (4), 615-617. Jargowsky, P. (1996). Take the money a nd run: Economic segregation in U.S. metropolitan areas. American Sociological Review 62, 984-998. Jargowsky, P. (1997). Poverty and place: Ghettos, ba rrios, and the American city. New York: Russell Sage Foundation. Jargowsky, P., & Bane, M. (1991). Ghetto poverty in the United States, 1970-1980. In C. Jencks & P. Peterson (Eds.), The urban underclass Washington, DC: The Brookings Institution. Jencks, C., & Mayer, S. (1990). The social consequences of growing up in a poor neighborhood In Inner-city poverty in Am erica, edited by Laurence E. Lynn, Jr., & Michael, G.H. McGreary. Washington, D.C.: National Academy Press. Jiobu, R.M. (1972). Urban Determinants of R acial Differentiation in Infant Mortality. Demography, 9, 603-615. Jones, C. (2001). Invites commentary: Race, racism and the practice of epidemiology:

PAGE 180

168 Comment. American Journal of Epidemiology, 154 (4), 299-304; discussion 305306. Johnson, T., Drisko, K., Gallagher, J., & Ba rela, C. ( 1999). Low birth weight: A women's health issue. Women's Health Issues 9, 223–230. Joseph, K., Kramer, M., Marcoux, S., Ohlss on, A., Wen, S., & Allen, A., & Platt R. (1998). Determinants of preterm deliver ies rates in Canada from 1981 through 1983 and from 1992 through 1994. New England Journal of Medicine 339, 14341439. Judge, K. (1999). Income distribution and life expectancy: A critical appraisal. In I. Kawachi, B. Kennedy, & R. Wilkinson (E ds.), The society and population health reader, Volume 1: Income inequality and Health. New Yor k, NY: New Press. Judge, K., & Paterson, I., (2001). Poverty, income, inequality and health Treasury Working Paper. Retrieved October 6, 2004, from http://www.treasury.govt.nz/wor kingpapers/2001/01-29.asp. Kahn, R., Wise, P., Kennedy, B., & Kawachi, I. (2000). State income inequality, household income, and maternal mental and physical health : Cross-sectional national survey. British Medical Journal 321, 1311-1315. Kaplan, G., Palmuk, E., Lynch, J., Cohen, R ., & Balfour, J. (1996). Income inequality and mortality in the United States: Analys is of mortality and potential pathways. British Medical Journal 312, 999-1003. Kaplan, G., Pamuk, E., Lynch, J., Cohen, R. & Balfour, J. (1999). Inequality in income and mortality in the United States: Analys is of mortality and potential pathways. In I. Kawachi, B. Kennedy, R. Wilkinson (Eds.), The Society and population

PAGE 181

169 health reader, Volume I: In come Inequality and Health. New York, NY: New Press. Kasarda, J. (1989). Urban industr ial transition and the underclass. Annals of the American Academy of Political Social Science 501, 26-47. Kaufman, J., Copper, R., & McGee, D. (1997). Socioeconomic status and health in blacks and whites: The problem of residual c onfounding and the resiliency of race. Epidemiology, 8 (6), 621-628. Kaufman, J., Dole, N., Savitz, D., & He rring, A., (2003). Modeling community-level effects on preterm birth. Annals of Epidemiology 13, 377-384. Kawachi, I., Colditz, G., Ascherio, A., Rimm E., Giovannucci, E., Stamfer, M., Willet, W.C. (1996). A prospective study of social ne tworks in relations to total mortality and cardiovascular disease in men in the U.S. Journal of Epidemiology Community Health 50, 245-251. Kawachi, I., & Kennedy, B. (1997). The relations hips of income inequality to mortality: Does the choice of indicator matter? Social Science and Medicine, 45 1121-1127. Kawachi, I., & Kennedy, B. (1999). Relationship of income inequality to mortality: Does the choice of indicator matter? In I. Ka wachi, B. Kennedy, & R. Wilkinson (Eds.), The society and population health Reader. Volume 1: Income inequality and Health New York, N.Y.: New Press, 1999. Kawachi, I., & Kennedy, B. (2002). Health of nations. Why ine quality is harmful to your health. New York: New Press. Kawachi, I., Kennedy, B., Lochner, K., & Prothrow-Stith, D. (1997). Social capital income inequality and mortality. American Journal of Public Health 87, 1491-

PAGE 182

170 1499. Kawachi, I., Kennedy, B., Lochner, K., & Pr othrow-Stith, D. (1999). Social capital, income inequality and mortality. In I. Kawachi, B. Kennedy, & R. Wilkinson (Eds.), The society and population health Reader. Volume 1: Income inequality and Health New York, N.Y.: New Press, 1999. Kawachi, I., Kennedy, B., & Wilkinson, R. (Eds.) (1999). The society and population health reader-Income inequality and health. New York: The New press, 506 pp. Kawachi, I., Subramanian, S., & AlmeidaFilho, N. (2002). A glossary for health. Journal of Epidemiology and Community Health, 56 (9), 647-652. Kennedy, B., Kawachi, I., Glass, R., & Prot hrow-Stith, D. (1998). Income distribution, socioeconomic status, and self rated health in the United States: Multilevel analysis. British Medical Journal 317, 917-921. Kennedy, B., Kawachi, I, & Prothrow-Stith, D. (1999). Income distribution and mortality: Cross sectional ecological st udy of the Robin Hood Index in the United States. In I. Kawachi, B. Kennedy, & R. Wilkinson (Eds.), The Society and Population Health Reader, Volume I: Income Inequality and Health. New, NY: New Press. Kiely, J., Brett, K.,Yu, S., & Rowley, D. (1994). Low birth weight and intrauterine growth retardation. From data to action. CDC's public health surveillance for women, infants and children. U.S. Department of Health and Human services: Atlanta, GA Kissim, E., Herrera, E., & Nakamoto, J. (1993). Hispanic Responses to Census Enumeration Forms and Procedures. Report prepared for the Bureau of the

PAGE 183

171 Census. Suitland, MD. Knapp, T. R. (1990). Treating oridinal scales as interval scales: An attempt to resolve the controversy. Nursing Research, 39 (2), 121-123. Kotch, J. (1997 ). Maternal and child health: Progr ams, problems and policy in public health Gaithersburg, MD: Aspen Publishers. Kramer, M. (1987). Determinants of low bi rth weight: Methodological assessment and meta-analysis. Bulletin of the World Health Organization 65, 663-737. Kramer, M., McLean, F., Boyd, M., & Usher, R. (1988). The validity of gestational age estimation by menstrual dating in term preterm, and postterm gestations. Journal of the American Medical Association 260, 3306-3308 Krieger, N. (1992) Overcoming the absence of socioeconomic data in medical records: Validation and application of a census-based methodology. American Journal of Public Health, 82 (5), 703-710. Krieger, N. (1994). Epidemiology and the web of causation: Has anyone seen the spider? Social Science and Medicine, 39 (7), 887-903. Krieger, N. (1999). Embodying inequality: A re view of concepts, measures, and methods for studying health conseque nces of discrimination. International Journal of Health Services, 29 (2), 295-352. Krieger, N. (2001). Theories for social epidemiology in the 21st century: An ecosocial perspective. International Journal of Epidemiology 30(4), 608-677. Krieger, N., Rowley, D., Herman, A., Avery, B., & Phillips, M. (1993). Racism, sexism, and social class: Implications for stud ies of health, diseas e, and well-being. American Journal of Preventive Medicine 9, 82–122.

PAGE 184

172 Krieger, N., & Sidney, S. (1996) racial di scrimination and blood pressure: The cardiac study of young black and white adults. American Journal of Public Health, 86 (10), 1370-1378. Lang, J., Lieberman, E., & Cohen, A., (1996). A comparison of risk for preterm labor and term small-for-gestational-age birth. Epidemiology, 7 (4), 369-376. Laporte, A. (2002). A note on the use of single in equality index in te sting the effect of income distribution on mortality. Social Science and Medicine, 55 (9), 1561-1570. Last, J..M. (1995). Dictionary of epidemiology. (1995). 3rd ed., New York, NY: Oxford University Press Laveist, T. (1989). Linking residential segregat ion and infant mortalit y disparity in U.S. cities. Social Science Research 73, 90-94. Laveist, T. (1992). The political empowerment and health status of African-Americans: Mapping a new territory. American Journal of Sociology, 97 (4), 1080-1095. LaVeist, T. (1993). Segregation, poverty, a nd empowerment: Health consequences for African Americans. The Milbank Quarterly, 71 (1), 41-66. LaVeist, T. (1995). The color line and the health of African-Americans. Humboldt Journal of Social Relations, 21 (1), 119-137. LaVeist, T. (1999). Book review: Anthony P. Polednak. Segregation, poverty and mortality in urban African Americans. Journal of Health Po litics, Policy and Law, 24 (6): 1410-1412. LeClere, F., Rogers, R., & Peters, K. (1997). Ethnicity and mortality in the United States. Individual and community correlates. Social Forces 76, 169-198. LeClere, F., & Soobader, M. (2000). The effect of income inequality on the health of

PAGE 185

173 selected U.S. demographic groups. American Journal of Public Health, 90 (12), 1892-1897. Lee, P., Chernausek, S., Hokkem-Koela, A ., & Czernichow, P. (2003). International small for gestational age advisory bo ard consensus development conference statement: Management of short children born small for gestational age, April 24October 1, 2001. Pediatrics, 111 (6), 1253-1261. Lillie-Blanton, M., & LaVeist, T. (1996). R ace/ethnicity, the social environment and health. Social Science and Medicine, 43 (1), 83-91. Link, B., & Phelan, J. (1995). Social conditi ons as fundamental causes of disease. The Journal of Health and Social Behavior, Spec No, 80-94. Littrell, R., Milliken, G., & Str oup, W., & Wolfinger, R. (1996). SAS System for mixed models Cary, NC: SAS Institute, Inc. Livingston, I. (1994). Social st atus, stress and health: Black Americans at risk. In: I. Livingston, (Ed.), Handbook of Black American health (pp. 236-252). Westport, Connecticut: Greenwood Press. Lobmayer, P., & Wilkinson, R. (2000). Income, inequality and mortality in 14 developed countries. Sociology of Health & Illness, 22 (4), 401-414. Lochner, K., Pamuk, E., Makuc, D., Kennedy, B., & Kawachi, I. (2001). State-level income inequality and individual mort ality risk: A prospective multilevel study. American Journal of Public Health 91, 385-391. Logan, J., & Alba, R. (1993). Locational re turns to human capital: Minority access to suburban community resources. Demography 30, 243-268. Logan, J., & Messner, D. (1987) Racial residential segreg ation and suburban violent

PAGE 186

174 crime. Social Science Quarterly 68, 510-527. Lumley, J. (2003). Recent work on the epidemiology of preterm birth. Acta Obstetricia et Gy necologica Scandinavica 84 (6), 541-54. Lynch, J. (2000). Income inequality and health: Expanding the debate. Social Science and Medicine 51, 1001-1005. Lynch, J., & Davey Smith, M. (2003). Rates a nd states: Reflections on the health of Nations. International Journal of Epidemiology, 32 (4), 663-670. Lynch, J., Davey Smith, M., Harper, S., Hillemeier, M., & Wolfson, M. (2004). Is income inequality a determinant of p opulation health? Part 1. A systematic review. The Milbank Quarterly, 82 (1), 5-99. Lynch, J., Davey Smith, M., Hillemeier, Shaw, T., Raghunthan, & Kaplan, G. (2001). Income inequality, the psychosocial e nvironment and health: Comparisons of wealthy nations. Lancet, 358, 194-200. Lynch, J., Davey Smith, M., Kaplan, G., & H ouse, J. (2000). Income inequality and mortality: Importance to health of indi vidual income, psychosocial environment or material conditions. British Medical Journal, 320 (7243), 1200-1204. Lynch, J., Kaplan, G. Pamuk, E., Cohen, R., & Heck, K. (1998). Income inequality and mortality: Importance to health of indi vidual income, psychosocial environment, or material conditions. British Medical Journal 320, 1200-1204. Lynch, J., Kaplan, G., Pamuk, E., Cohen, R., H eck, K., Balfour, J., et al. (1999). Income inequality and mortality in metropolitan areas of the United States. In I. Kawachi, B. Kennedy, & R. Wilkinson (Eds.), The society and population health Reader. Volume 1: Income inequality and Health New York, N.Y.: New Press, 1999.

PAGE 187

175 MacCallum, R., Zhang, S., Preacher, K., & Ruck er, D. (2002). On the dichotomization of quantitative variables. Psychological Methods, 7 (1), 19-40. Macinko, J., Shi, L., & Starfield, B. (2004). Wage inequality, the health system and infant mortality in wealthy indust rialized countries, 1970-1996 Social Science and Medicine, 58 (2), 279-292. Macintyre, S. (1994). Understand ing the social patterning of h ealth: The role of the social sciences. Journal of Public Health Medicine 16, 53-59. Malstrom, M., Sandquist, J., & Johansson, S. (1999). Neighborhood environment and self-reported health status: A multilevel analysis. American Journal of Public Health, 89 (8), 1181-1186. Marmot, M. (1999). Epidemiology of socioecono mic status and health: Are determinants within countries the same as between countries? Annals of New York Academy of Science 896, 16-29. Marmot, M., Bobak, M., & Davey Smith, G. (19 95). Explanations for social inequalities in health. In B.C. Amick, S. Le vine, A. Tarlov, D.C. Walsh (Eds.), Society and Health (pp. 172-210). Oxford University Press. Marmot, M., & Wilkinson, R. Eds. (1999) Social determinants of health New York: Oxford University Press. Marmot, M. & Wilkinson, R. (2001). Psychosocia l and material pathwa ys in the relation between income and health: A response to Lynch et al. BMJ, 322 (7296), 11241128. Martin, J., Hamilton, B., Sutton, P., Ventura, S., Menacker, F., & Munson, M. (2003). Births: Final data for 2002. National Vital Statistics Reports, 52(10). Hyattsville,:

PAGE 188

176 MD. National Center for Health Statistics. Massey, D. (1990). American Apartheid: Segr egation and the maki ng of the underclass. American Sociological Review 61, 984-998. Massey, D. (2001). Residential segregati on and neighborhoods conditions in U.S. metropolitan areas. In N. Smelser, W. Wilson, & F. Mitchell (Eds.), American becoming: Racial trends and their consequences Vol I (pp. 391-434). Washington, D.C.: National Academy Press. Massey, D. (1996). The age of extremes: C oncentrated affluence and poverty in the twenty-first century. Demography 33, 395-412. Massey, D., Condran, G., & Denton, N. (1987). Th e effect of residential segregation on black social and economic well-being. Social Forces, 66 (1), 29-56. Massey, D., & Denton, N. (1988). The dime nsions of residential segregation. Social Forces, 76 (2), 281-315. Massey, D., & Denton, N. (1993). American apartheid: Segre gation and the making of the underclass. Cambridge, MA: Harvard University Press. Massey, D., & Eggers, M. (1990). The ecol ogy of inequality: Minorities and the concentration of poverty, 1970-1980. American Journal of Sociology, 95 (5), 1153-1188. Massey, D., & Eggers, M. (1993). The spatia l concentration of affluence and poverty during the 1970s. Urban Affairs Quarterly, 29 (2), 299-315. Massey, D., & Fischer., M. (1999) Does rising income bring in tegration? New results for Blacks, Hispanics, and Asians in 1990. Social Science Research 28, 316-326. Massey, D., & Fischer, M. (2000). Ho w segregation concentrates poverty. Ethnic and

PAGE 189

177 Racial Studies, 23 (4), 670-691. Massey, D., White, M., & Phua, V. (1996). The dimensions of segregation revisited. Sociological Methods and Research 25, 172-206. Matthews, T., Menacker, F., & Macdorman, M. (2002). Infant mortality statistics from the 2000 period linked birth/Infant data set. National Vital Statistics Report, 50 (12). Hyattsville, MD: National Ce nter for Health Statistics. Mayer, S. (2001). How the growth in income inequa lity increased econom ic segregation. Northwestern University of Chicago Jo int Center for Poverty Research in Working Paper #230. Mayer, S., & Sarin, A. (2005). Some mech anisms linking economic inequality and infant mortality. Social Science and Medicine 60, 439-455. McLaughlin, D., Stokes, C., & Nonoyama A. (2002). Residence and income inequality: Effects on mortality among U.S. Counties. Rural Sociology 66 (4), 579-598. McLean, D., Hatfield-Timajchy, K., Wi ngo, P., & Floyd, R. (1993). Psychosocial Measurements: Implications for the Study of Preterm Delivery in Black Women American Journal of Preventive Medicine 9(6), 39-81. Mellor, J., & Milyo, J. (2003). Is exposur e to income inequality a public health concern? Lagged effects of income inequality on individual and population health. Health Services Research 38, 137-152. Messner, S., & Tardiff, K. (1986). Economic-Inequality and levels of homicide-An Analysis of urban neighborhoods. Criminology 24 (2), 297-317.

PAGE 190

178 Milner, R., & Richards, B. (1974). An analysis of birth weight a nd gestational age of infants born in England and Wales, 1967 to 1971. Journal of Obstetrics and Gynecology of the British Commonwealth 81, 956-967. Mokdad, A., Ford, E., Bowman, B., Dietz, W., Vi nicor, F., Bales, V ., & Marks, J. (2003). Prevalence of obesity, diabetes, and obes ity-related health ri sk factors, 2001. Journal of the American Medical Association 289, 76-79. Muller, A. (2002) Education, income inequa lity and mortality: A multiple regression analysis. British Medical Journal, 324(7328), 23-25. Muthen, B., & Muthen, L. (2002). How to us e a Monte Carlo study to decide on sample size and determine power. Structural Equation Modeling, 4 599-620. Naeye, R. (1983). Maternal age, obstetric complications, and the outcome of pregnancy. Obstetrics and Gynecology, 61, 3135-3139. Martin, J., Hamilton, B., Ventura, S., Menacker F., & Park, M. (2002). Births: Final data for 2002. National Center for Health Statistics. Hyattsville, M.D. National Center for Health Statistics (2004) Health, United States, 2004 with chartbook on trends in the health of Am ericans. Hyattsville, M.D. Nettleton, S, & Burrows, R. (1998). Hous ing norms, housing satisfaction and the propensity to move. Journal of Marriage and the Family 38, 309-320. Newburn-Cook, C., White, D., Svenson, L., De mianczuk, N., Bott, N., & Edwards, J. (2002). Where and to what extent is pr evention of low birth weight possible? Western Journal of Nursing Research, 24 (8), 887-904. Nguyen, V., & Preschard, K. (2003). Anthropol ogy, inequality, and disease: A review. Annual Review of Anthropology 32, 447-474.

PAGE 191

179 O'Campo, P., Xue, X., & Wang, M. ( 1997). Neighborhood risk factors for low birthweight in Baltimore multilevel analysis. American Journal of Public Health 87, 1113-1118. Office of Management and Budget. Race and ethnic standards for federal statistics and administrative reporting. Statistical Polic y Directive 15, 1977. Office of Management and Budge t Federal Register Notice. Revisions to the standards for the classification of Fede ral data on race and ethnicity. Federal Register 62FR58781-58790, October 30, 1997. Available at: http://www.whitehouse.gov/wh/eop/omb/ html/fedreg.html. Accessed May 1, 2005. Oliver, M., & Shapiro, T. (1997). Black wealth/White wealth: A new perspective on racial inequality New York: Routledge. Parker, J. (2001). Birthweight trends among interracial black and white infants. Epidemiology 11, 242-248. Parker, J., Schenker, N., Ingram, D., Weed, J, Heck, K., & Madans, J. (2004). Bridging between two standards for collecting information on race and ethnicity: An application to Census 2000 and vital rates. Public Health Reports, 119 192-205. Pearl, M., Braveman, P., & Abrams, B. (2001). The city. Chicago: University of Chicago Press. Peterson, R., & Krivo, L. (1993). Racial segr egation and black urban homicide. Social Forces, 71, 1001-1026. Peterson, R., & Krivo, L. (1999). Racial segregation, the conc entration of disadvantage, and Black and White homicide victimizations. Social Forum, 14 (3), 465-493.

PAGE 192

180 Pettigrew, T., & Meertens, R. (1995). Subtle and blatant prejudice in Western Europe. European Journal of Social Psychology 25, 57-75. Phillips, D. (2001). Programming of adrenocortia l function and the fetal origins of adults disease. Journal of Endocri nological Investment 24, 742-746. Pickett, K., Collins, J., Masi, C., & Wilkins on, R. (2005). The effects of racial density and income incongruity on pregnancy out comes. Social Science and Medicine, 60(10), 2229-2238. Piper, J., Mitchel, E., Snowden, M., Hall, C ., Adams, M., & Taylor P. (1993). Validation of 1989 Tennessee birth certificate using ma ternal and newborn hospital records. American Journal of Epidemiology 137, 758-768. Piper, J., Xenakis, E., McFarland, M., Ellio tt, B., Berkus, M., & Langer, O. (1996). Do growth retarded premature infants have di fferent rates of peri natal morbidity and mortality than appropriately grown premature infants? Obstetrics and Gynecology, 87( 2), 169-174. Polednak, A. (1991). Black-white differences in infant mortality in 38 standard metropolitan statistical areas. American Journal of Public Health, 86 (5), 723-726. Poldenak, A. (1993). Poverty, residential segreg ation and black/white mortality ratios in urban areas. Journal of Health Care of the Poor and Underserved 4, 363-373. Poldenak, A. (1996). Segregation, discrimination and mortality in U.S. blacks. Ethnicity and Disease 6, 99-108. Polednak, A. (1997). Segregation, poverty and mortality in urban African Americans. New York, NY: Oxford University Press. Pollato, E., Collins, J., & David, R. (2000) Enigma of maternal race and infant

PAGE 193

181 birthweight: A population-base d study of U.S. born bl ack and Caribbean born black women. American Journal of Epidemiology 151, 1080-1085. Potter, L. (1991). Socioeconomic determinan ts of white and black life expectancy differentials. Demography 28, 303-320. Preston, S.H. (1975). The changing relation between mortality and level of economic development. Population Studies 29, 231-248. Raferty, A.E. (1986) Choosing models for cr oss-classifications: comment on Grusky and Hauser. American Sociological Review, 51, 145-146. Raferty, A.E. (1995) Bayesian mode l selection in social research. Sociological Methodology 25 111-163. Rauh, V., Howard, F., Andrews, H., & Garfinke l, R. (2001). The contribution of maternal age to racial disparities in birthweight: A multilevel perspective. American Journal of Public Health 91, 1815-1824. Regev, R., Lusky, A., Dolfin, T., Litmanov itz, I. Arnon, S., & Reichmank, B. (2003). Excess mortality and morbidity among small-for-gestational-age premature infants: A population-based study. Journal of Pediatrics 143, 186-191. Reichman, N., & Hade, E. (2001) Validation of birth certificate da ta: A study of women in New Jersey’s HealthStart Program. Annals of Epidemiology 11, 186-193. Reloar, A., Behn, B., & Cowan, D. (1967). Analysis of gestational interval. American Journal of Obstetrics and Gynecology 99, 34-45. Ren, X., Amick, B., & Williams, D. (1999). R acial/ethnic dispari ties in health: The interplay between discrimination and socioeconomic status. Ethnicity and Disease 9, 151-165.

PAGE 194

182 Rich-Edwards, J., Krieger, N., Majzoub, J., Sierler, S., Lieberman, E., & Gillman, M. (2001). Maternal experiences of racism and violence as predictors of preterm birth: Rationale and study design. Paediatric and Perinatal Epidemiology 15, S124-S135. Roberts, E. (1997) Neighborhood social en vironments and the distribution of low birthweight in Chicago. American Journal of Public Health 87, 597-603. Rodgers, G. (1979). Income and inequality as de terminants of mortality: An international cross-sectional analysis. Populations Studies 33, 343-351. Rodriguez, C.E., & Cordero-Guzman, J.M. ( 1992). Place, race in context. Ethnic racial studies. Vol. 15. Pp. 523-543. Rohe, W., Van Zandt, V., & McCarthy, G. ( 2001). The social benefits and costs of homwownershp: a critical assessment of the research Joint Center for Housing Studies of Harvard Univer sity: Chapel Hill: NC. Roisman, F. (1995). The lessons of American apartheid: The necessity and means of promoting residential racial integration. Iowa Law Review 81, 479-525. Rosenberg H., Maurer, J., Sorlie, P., Johns on, N., MacDorman, M., Hoyert D., Spitler, J., & Scott, C. (1999). Quality of d eath rates by race and Hispanic origin: A summary of current research, 1999. National Center for Health Statistics. Vital Health Stat 2 (128). 1999 Rosenfield, R. (1986). Urban crime rates: Effects of inequality, welfare dependency, region and race. In J. Bryne and R. Sampson (Eds.), The Social Ecology of Crime : (pp. 116-130). New York: Springer-Verlag. Rossi, P., & Weber, E. (The social benef its of homeownership: Empirical evidence from

PAGE 195

183 national nurveys. Housing Policy Debate, 7 (1), 1-35. Roohan, P., Josberger, R., Acar, J., Dabi r, P., Feder, H., & Gagliano, P. (2003). Validation of birth certificat e data in New York State. Journal of Community Health, 28 (5), 335-346 Rowley, D. (1994). Research issues in the study of very low birt hweight in AfricanAmericans. Journal of the National Medical Association, 91 (12), 663-667. Ryan, B., Joiner, B., & Ryan, T. (1985). Minitab handbook. 2nd Ed. Boston: Duxbury Press. Sampson, R. (1985). Race and criminal vi olence: A demographically disaggregated analysis of urban homicide. Crime and Delinquency 31, 47-82. Sampson, R. (1987). Urban black violence: The effect of male jobl essness and family disruption. American Journal of Sociology 93, 348-382. Sanders, M., Allen, M., Alexander, G., Yankow itz, J., Graeber, J., Johnson, T., & Repka, M. (1991). Gestational age assessment in preterm infants weighing less than 1500 grams. Pediatrics, 88 (3), 542-545. Sastry, N., Ghosh-Dastidar, B., Adam s, J., & Pebley, A. (Eds.). (2003). The design of a multilevel survey of children, families, and communities: The Los Angeles Family and Neighborhood Survey Santa Monica: RAND. Scheuch, E. (1969). Social context and indivi dual behavior. In: Dogan, M. Rokkam, S. Eds., Social ecology. Boston: The MIT Press, 133-155. Schiono, P., Rauk, V., Park, M., Lederman, S ., & Zuskar, D. (1997). Ethnic differences in birth weight: The role of lifestyle and other factors. American Journal of Public Health 87, 787-793.

PAGE 196

184 Shihadeh, E., & Flynn, N. (1996). Segregation an d crime: The effect of black isolation on the rates of black urban violence. Social Forces 74, 1325-1352. Siegrist, J., & Marmot, M. (2004). Health in equalities and the psychosocial environmenttwo scientific challenges. Social Science and Medicine 58, 1463-1473. Singer, J. (1998). Using SAS PROC MIXED to fit multilevel models, hierarchical models and individual growth models. Journal of Educational and Behavioral Statistics, 24 (4), 323-355. Singer, B., & Ryff, C. (Eds.) (2001). New hor izons in health: An integrative approach. Washington, D.C.: National Academies Press. Singer, J., & Willett, J.. (2003). Applied longitudinal data ana lysis: Modeling change and even occurrence New York: Oxford University Press. Sloggett, A., & Joshi, H. (1994). Higher mo rtality in deprived areas: Community or personal disadvantage? British Medical Journal 309, 1470-1474. Snijders, T., & Bosker, R.(1993). Standard erro rs and sample sizes for two-level research. Journal of Educational Statistics, 18 (3), 237-259. Soobader, M, & LeClere, F. (1999). Aggr egation and the measurement of income inequality: Effects on morbidity. Social Science and Medicine, 48 (6), 733-744. South, S., & Crowder. K. (2000). The decl ining significance of neighborhoods? marital transitions in community context. Social Forces 78, 1067-1099 Spinnato, J., Sibai, B., Shaver, D., & Anderson, G. (1984) Inaccuracy of Dubowitz gestational age in low birth weight infants. Obstetrics and Gynecology 63, 491495. Stancil, T., Hertz-Picciotto, I., & Schramm, M., & Watt-Morse, M. (2000). Stress and

PAGE 197

185 pregnancy among African American women. Paediatric and Perinatal Epidemiology, 14, 127-135. Strait, J. (2006). An epidemiology of nei ghborhood poverty: Causal factors of infant mortality among Blacks and Whites in metropolitan United States. The Professional Geographer, 58 (1), 39-53. Subramanian, S., Blakely, T., & Kawachi, I. (2 003). Income inequality as a public health concern: Where do we stand? Commentary on “Is exposure to income inequality a public health concern?” Health Services Research 38, 153-167. Subramanian, S., Acevedo-Garcia, D., & Osypuk, T. (2005). Racial residential segregation and geographic heterogeneity in black-white disparity in poor selfrates health in the U.S.: A multilevel statistical analysis. Social Science and Medicine 60 (8), 1667-1679. Subramanian, S., Kawachi, I., & Kennedy, B. (2001). Does the state you live in make a difference? Multilevel analysis of self-rated health in the U.S. Social Science and Medicine, 53, 9-19. Sugure, T. (1996). The Origins of the Urban Crisis: Race and Inequality in Postwar Detroit. Princeton, NJ: Princeton University Press, 1996). Sutton, P., & Matthews, M. (2004). Trends in ch aracteristics of birt hs by state: United States, 1990, 1995, and 2000-2002. National Vital Statistics Reports, 52 (19). Hyattsville, Maryland: National Ce nter for Health Statistics. Taeuber, K., & Taeuber, A. (1964). The Negro as an immigrant group: Recent trends in racial and ethnic segregation in Chicago. American Journal of Sociology 69, 374382.

PAGE 198

186 Taffel, S., Johnson, D., & Heuser, R. (1982). A method of imputing length of gestation on birth certificates. Vital and Health Statistics, 2 (93), 1-11. Taffel, S., Ventura, S., & Gay. G. (1989). Revised U.S. certificate of birth--new opportunities for research on birth outcome. Birth 16(4), 188-93. Testa, M., Astone, N., Krogh, M., & Necker man, K. (1993). Employment and marriage among inner-city fathers. In W. Wilson (Ed.), The ghetto underclass Newberry Park, CA: Sage Publications. Tien, J., Sandler, I., MacKinnon, D., & Wolc hik, S. (2004). How did it work? Mediation in the context of a moderated prevention effect for children of divorce. Journal of Consulting and Clinical Psychology, 72 (4), 617-624. Turnbull, A., & Rivier, C. ( 1999). Regulation of the hypothala mic-pituitary-adrenal axis by cytokines: Actions an d mechanisms of action. Physiological Reviews 79, 1-71. U.S. Department of Health and Human Services (2000). Healthy People 2010 2nd Ed. With Understanding and Improving Health and Objectives for Improving Health. 2 vols. Washington, DC: U. S. Government Printing Office. United States Department of Hea lth and Human Services (2000). Healthy People 2010: Understanding and Improving Health. Washington: GPO. Velleman, P. (1980). Definition and compar ison of robust nonlinear data smoothing algorithms. Journal of the American Statistical Association 75, 609-615. Wadhwa, P., Culhane, J., Rauh, V., Barve, S., Hogan, V., Sandman, C., Hobel, C.J., Chicz-DeMent, A., Dunkel-Schetter, C., Garite, T.J., & Glynn, L. (2001). Stresses, infection and preterm birt h: A biobehavioral perspective. Paediatric and Perinatal Epidemiology 15, 17-29.

PAGE 199

187 Wadhwa P., Porto, M., Chicz-DeMet, A., & Sandman, C. (1998). Maternal CRH levels in the early third trimester predict leng th of gestation in human pregnancy. Ameican Journal of Obstetrics and Gynecology, 179, 1079-1085. Wagstaff, A., & van Doorslaer (2000). Inco me inequality and health: What does the literature tell us? Annual Review of Public Health, 21, 543-567. Waitzman, N., & Smith, K. (1998). Separate but lethal: the e ffects of economic segregation on mortality in metropolitan America. The Milbank Quarterly, 76 (3), 341-373. Waldmann, R. (1999). Income distribution and infant mortality. In I. Kawachi, B. Kennedy, & R. Wilkinson (Eds.), The Society and Popula tion Health Reader, Volume I: Income Inequality and Health. New, NY: New Press. Wallace, R. (1990). Urban desertification, public health and public order: Planned shrinkage, violent death, substa nce abuse and AIDS in the Bronx. Social Science and Medicine 31, 801-813. Wallace, R. (1991). Expanding couples shock fronts of urban decay and criminal behavior: How U.S. cities are becoming hollowed out. Journal of Quantitative Criminology 7, 333-356. Wang, X., Guyer, B., & Paige, D. (1994). Differences in gestational age-specific birthweight among Chines, Japa nese and white Americans. International Journal of Epidemiology 23, 119-128. Watkins, M., Edmonds, L, McClearn, A., Mullins, L., Mulinare, J., & Khoury, M. (1996). The surveillance of birth defects: The usef ulness of the revised U.S. standard birth certificate. American Journal of Public Health, 86 (5), 731-734.

PAGE 200

188 Warren, W., Patrick, S., & Goland, R. (1992). El evated maternal plasma corticotropinreleasing hormone levels in pregna ncies complicated by preterm labor. American Journal of Obstetrics and Gynecology, 166, 1198–204. Wen, W., Goldenberg, R., Cutter, G., Hoffman R., & Cliver, S. (1990). Intrauterine growth retardation and preterm delivery: Prenatal risk factors in an indigent population. American Journal of Ob stetrics and Gynecology 162, 213-218. Wenner, W., & Young, E. (1974). Nonspeci fic date of last menstrual period. American Journal of Obstetrics and Gynecology 120, 1071-1079. White, M., Kim, A., Glick, J. (2005). Mappi ng social distance: Ethnic residential segregation in a multiethnic metro. Sociological Methods and Research, 43 (2), 173-203. Wilcox, L., & Marks, J. (Eds). (1994). From data to action: CDC's public health surveillance for women, infants and children. Atlanta, GA: Center for Disease Control and Prevention. Wilkinson, R. (1992). Income distribution and life expectancy. British Medical Journal, 304, 165-168. Wilkinson, R. (1996). Unhealthy societies: The afflictions of inequality London: Routledge. Wilkinson, R. (1997). Health inequalities: Re lative or absolute material standards? British Medical Journal, 314, 591-595. Wilkinson, R. (1999). Health hierarchy and social anxiety. Annals of New York Academy of Science, 896 48-63. Wilkinson, R. (2000). Mind the Gap London: Weidenfeld & Nicolson.

PAGE 201

189 Williams, D. (1994). The concept of race in Health Services Research: 1996 to 1990. Health Services Research, 29 (3), 261-274. Williams, D. (1996). Racism and health: A research agenda. Ethnicity and Disease 6, 16. Williams, D. (1997). Race and health: Ba sis questions, emerging directions. Annals of Epidemiology, 7, 322-333. Williams, D. (2001). Race and health: Trends an d policy implications. In J.A., Auerbach, & B. Krimgold (Eds.), Income, socioeconomic status and health: Exploring the relationships. Washington, D.C.: National Policy Association and the Academy for Health Services Research and Policy. Williams, D. (2002). Racial/ethnic varia tions in women's health: The social embeddedness of health. American Journal of Public Health, 92 (4), 588-597. Williams, D., & Collins, C. (1995). US socioeco nomic and racial differences in health: Patterns and explanations. Annual Review of Sociology 21, 349-386. Williams, D., & Collins, C. (2001). Racial resi dential segregation: A fundamental cause of racial disparities in health. Public Health Reports 116, 404-416. Wilson, W. (1980). The declining significance of race. Chicago, IL: The University of Chicago Press. Wilson, W. (1987). The truly disadvantaged: The inne r city, the underclass, and public policy in Chicago University of Chicago. Yang, H., Kramer, M., Platt, R., Blondel, B., Breart, G., Morin, I., Wilkins, R., & Usher, R. (2002). How does early ultrasound scan estimation of gestational age lead to higher rates of preterm birth? American Journal of Obst etrics and Gynecology,

PAGE 202

190 186, 433-437. Yankauer, A. (1950). The relationship of fe tal and infant mortality to residential segregation. American Sociological Review 15, 644-648. Zambonato, A., Pinherio, R., Horta, B., & To masi, E. (2004). Risk factors for small for gestational age births among infants in Brazil. Revista de Saude Publica, 38 (1), 14. Zeitlin, J., Ancel, P., Surel-Cubizolles, M., & Papiernik, E. (2001). Are risk factors the same for small for gestationa l age versus preterm births? American Journal of Obstetrics and Gynecology, 185 (1), 208-215

PAGE 203

191 APPENDIX A: CONCEPTUAL MODEL OF THE RELATIONSHIP BETWEEN RESIDENTIAL SEGREGATION AND PERINATAL OUTCOMES

PAGE 204

192 CONCEPTUAL MODEL OF THE RELATIONSHIP BETW EEN RESIDENTIAL SEGREGATION AND PERINATAL OUTCOMES CONTROL VARIABLES CONTROL VARIABLES MACRO LEVEL STRUCTURAL FACTORS Residential Segregation Evenness -Index of Dissimilarity Exposure -Isolation Index Area Size -Metropolitan -Micropolitan Proportion Black MEZZO LEVEL CONTEXTUAL FACTORS Physical Environment -Age of Housing -Housing Unit Vacancy -Owner Occupied Homes Social Environment -Female-headed Households -Proportion Institutionalized -Proportion Black MICRO LEVEL INDIVIDUAL FACTORS Maternal Characteristics -Age -Education -Marital Status -Prenatal Care -Tobacco Use MICRO LEVEL INDIVIDUAL PERINATAL OUTCOMES -Low Birth Weight -Preterm Delivery -Small for Gestational Age Births Economic Environment -Proportion with High School Diploma Race-specific Poverty Rate -Unem p lo y ment Rate Ethnicity Median Income

PAGE 205

ABOUT THE AUTHOR Tabia Henry Akintobi received a Bachelor’s Degree in English from the University of Miami in 1997 and a Master’s of Public Hea lth from the College of Public Health, at University of South Florida in 1999. Tabia lives in Conley, Georgia with her husband, Adebayo, and daughter, Ifelola. She will expand her role as Senior Research er at the Morehouse School of Medicine Prevention Research Center, in A tlanta, Georgia, after graduation.