USF Libraries
USF Digital Collections

Biomedical and psychosocial determinants of problematic birth outcomes

MISSING IMAGE

Material Information

Title:
Biomedical and psychosocial determinants of problematic birth outcomes
Physical Description:
Book
Language:
English
Creator:
Kroelinger, Charlan Day
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
high birth weight
psychosocial stressors
pregnancy screening tests
pregnancy stressors
reproductive epidemiology
Dissertations, Academic -- Public Health -- Doctoral -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: The primary objective of this study was to evaluate the associations between psychosocial stressors, urine sugar levels, and subsequent birth outcomes, specifically high birth weight babies and Caesarean section births. In a prospective cohort study, 506 Black and White women of childbearing age were followed for the duration of one pregnancy in Tuscaloosa and Mobile counties in Alabama from 1990 to 2001. Participants were interviewed twice throughout pregnancy, during the first and third trimesters, respectively, and birth outcome data were collected via medical chart reviews. Six percent (6.1%) of the women in the sample had a high birth weight baby, and 18.4% received a C-section during childbirth. Adjusted logistic regression results indicate that urine sugar levels are predictive of high-weight births, with women who have higher urine sugar levels were more than three times likely to birth a high weight baby compared with women who have no detectable urine sugar spill (OR 3.25; 95% CI 1.30, 8.10). In addition, the interaction of familial social support throughout pregnancy, physical or verbal abuse during the second and third trimesters, and ethnicity is significantly associated with increased risk of having a high birth weight baby. For C-section, single participants are over two times less likely to receive a C-section during childbirth compared with currently married participants (OR 0.46; 95% CI 0.21-1.00). Examining structural equation modeling results; pathways leading from urine sugar levels, physical or verbal abuse during the latter half of the pregnancy, and a mother's social support among White participants are indicative of high weight births (R² = 0.65). White abused women who receive their mother's social support are more likely to have a high birth weight baby compared with both White and Black women who are not abused and receive the same amount of social support. Recommendations to public health practitioners include primary prevention through promotion of familial support during pregnancy, secondary prevention through urine sugar screening at every prenatal visit, and direct intervention by identifying and inquiring about instances of suspected abuse during pregnancy.
Thesis:
Thesis (Ph.D.)--University of South Florida, 2004.
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 Charlan Day Kroelinger.
General Note:
Includes vita.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 356 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 56546874
notis - AJS2461
usfldc doi - E14-SFE0000413
usfldc handle - e14.413
System ID:
SFS0025105:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Biomedical and Psychosocial Determinant s of Problematic Birth Outcomes by Charlan Day Kroelinger A dissertation submitted in partial fulfillment of the requirement s for the degree of Doctor of Philosophy Department of Epidemiology and Biostatistics College of Public Health University of South Florida Major Professor: Thom as J. Mason, Ph.D. Heather G. Stockwell, Sc.D. Getachew A. Dagne, Ph.D. William W. Dressler, Ph.D. Date of Approval: May 20, 2004 Keywords: Reproductive Epidemiology, Pregnancy Stressors, High Birth Weight, Psychosocial Stressors, Pregnancy Screening Tests Copyright 2004, Charlan Day Kroelinger

PAGE 2

DEDICATION The current research, analysis, and dissertation are dedicated to my daughter, Ember Kimmerbryce Crutchfield. She had no choice in supporting me throughout the doctoral process, and means more to me than any other person I have known. When she reads this dedica tion one day, I hope she will be proud.

PAGE 3

ACKNOWLEDGEMENTS I would like to acknowledge my disse rtation committee, first and foremost: Thomas J. Mason, PhD; Heather G. Sto ckwell, ScD; Getachew A. Dagne, PhD; and William W. Dressler, PhD. Secondly, I would like to thank Kath ryn S. Oths, PhD, for her invaluable input throughout my graduate career, and Kevi n Kip, PhD, for his support during the proposal phase of this dissertation. Thirdly, I would like to thank my family and friends for their support over the past five years. Specifically, C harles and Lu Kroelinger; Keara, Cory, and Kyle Kroelinger; William and Bobbi Kroel inger; Mel and Betty Enyart; Jan Stanners; Jennifer Hudson; Jennifer Chiprich ; Janelle Novak; Kristine Shanteau; Erin Hughey; and Zachary Thompson.

PAGE 4

i TABLE OF CONTENTS LIST OF TABLES iv LIST OF FIGURES xxv ABSTRACT xxxii CHAPTER ONE: INTRODUCTION 1 1.0 Introduction 1 CHAPTER TWO: LITERATURE REVIEW 9 2.0 Introduction 9 2.1 Socio-cultural Factors and Pregnancy 10 2.2 Outcome Measures 18 2.2.1 Urine Sugar Levels 18 2.2.2 High Birth Weight 25 2.2.3 Caesarean Section 29 2.3 Theoretical Framework 33 2.4 Application of the Framework 35 CHAPTER THREE: METHODOLOGY 39 3.0 Introduction 39 3.1 Issues of External Validity 43 3.1.1 Comparability of Both Cohorts 43 3.1.2 Generalizability of Findings 46 3.2 Issues of Internal Validity 47 3.2.1 Variables in Analysis 47 Psychosocial and Physical Risk Factors 50 Intermediate Outcome 55 Main Outcomes 56 3.2.2 Statistical Tests 56 Hypothesis 1 57 Hypothesis 2 58 Hypothesis 3 58 Hypothesis 4 59 Hypothesis 5 59 Comprehensive Modeling 59 3.3 Power Analysis 61

PAGE 5

ii 3.3.1 Hypothesis 1 63 3.3.2 Hypothesis 2 65 3.3.3 Hypothesis 3 73 3.3.4 Hypotheses 4 and 5 81 3.4 Limitations 81 CHAPTER 4: RESULTS 86 4.0 Introduction 86 4.1 Descriptive Statistics 86 4.1.1 Demographic Characteristics 87 4.1.2 Predictor Variables 89 4.1.3 Outcome Variables 97 4.1.4 Confoundi ng Factors 99 4.1.5 Transformation of Non-normally Distributed Outcomes 101 4.1.6 Reliability Analysis of Scales 102 4.1.7 Exclusion of Specific Factors 106 4.1.8 Uncollected Data 107 4.2 Inferential Statistics 109 4.2.1 Evaluation of Confounding Factors 110 Urine Sugar Levels 112 High Birth Weight 117 Caesarean Section 120 4.2.2 Evaluation of Multicollinearity 123 Multicollinearity Am ong Confounding Factors 123 Multicollinearit y Among Predictors 124 4.2.3 Separate Analysis of Predictors and Outcomes 125 Independent Associations Between Each Predictor and Urine Sugar Levels 125 Independent Associations Between Each Predictor and Problemat ic Birth Outcome 131 4.2.4 Combined Anal ysis of Predictors and Outcomes 133 Hypothesis 1 134 Hypothesis 2 138 Hypothesis 3 149 Hypothesis 4 168 Hypothesis 5 173 CHAPTER 5: STRUCTURAL EQUATION MODELING 182 5.0 Introduction 182 5.1 Structural Equatio n Modeling Methodology 182 5.2 Structural Equation Mode ling for Overall Findings 188 5.2.1 Evaluation of Hypothesis 2 190

PAGE 6

iii 5.2.2 Evaluation of Hypotheses 3 and 1 198 5.3 Structural Equation Mode ling for Interaction Terms 214 5.3.1 Evaluation of Hypothesis 4 215 5.3.2 Evaluation of Hypothesis 5 218 CHAPTER 6: DISCUSSION 236 6.0 Introduction 236 6.1 Major Findings 236 6.1.1 Inferential Results 237 6.1.2 Structural Equat ion Modeling Results 240 6.2 Application of the Theoretical Framework 244 6.3 Study Limitations 251 6.4 Study Strengths 256 6.5 Consistency with Current Literature 257 6.6 Public Health Implications 259 6.7 Further Research 261 CHAPTER SEVEN: CONCLUSION 264 7.0 Introduction 264 REFERENCES 268 APPENDICES 288 Appendix A: Description of Predictor Variables 289 Appendix B: Description of Potentially Confounding Factors 303 Appendix C: Description of Outcome Measures 310 Appendix D: Model Fit Results for Analyses 311 Appendix E: Residual Plots of Multiple Regression Analyses 312 Appendix F: Re-analysis Ex cluding Low Birth Weight Infants 316 ABOUT THE AUTHOR End Page

PAGE 7

iv LIST OF TABLES Table 1 Original and current threshold limits for glucose tolerance testing diagnosis of gestational diabetes 24 Table 2 Literature review of epidemiologic studies of risk factors of and diseases associated with being born high birth weight 27 Table 3 Literature review of epidemiologic studies of risk factors of and problematic outcomes associated with Caesarean section 31 Table 4 Distribution of age and et hnicity by site of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 42 Table 5 Distribution of reproduct ive characteristics for the state of Alabama, Tu scaloosa, and Mobile Counties in 1999, U.S. Census Bureau, 2003 45 Table 6 Power calculation of urine sugar levels and estimated high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 64 Table 7 Power calculation of urine sugar levels and estimated Caesarean sections of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 65 Table 8 Power calculation of estimated physical work strain and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa a nd Mobile Counties, AL 1990-2001 67

PAGE 8

v Table 9 Power calculation of estimated depression score and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 68 Table 10 Power calculation of estimated physical abuse and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 69 Table 11 Power calculation of estimated lack of social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 70 Table 12 Power calculation of estimated lack of autonomy and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 71 Table 13 Power calculation of estimated pregnancy wantedness and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 72 Table 14 Power calculation of estimated marital status and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 73 Table 15 Power calculation of estimated autonomy, physical work strain, depression, lack of social support, and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 74 Table 16 Power calculation of physical abuse and estimated high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 75

PAGE 9

vi Table 17 Power calculation of pregnancy wantedness and estimated high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 75 Table 18 Power calculation of marital status and estimated high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa a nd Mobile Counties, AL 1990-2001 76 Table 19 Power calculation of estimated physical work strain and Caesarean sect ion of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 77 Table 20 Power calculation of estimated depression and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 78 Table 21 Power calculation of estimated physical abuse and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 78 Table 22 Power calculation of estimated lack of social support and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 79 Table 23 Power calculation of estimated lack of autonomy and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 79 Table 24 Power calculation of estimated pregnancy wantedness and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 80

PAGE 10

vii Table 25 Power calculation of estimated marital status and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 81 Table 26 Categorical demographic c haracteristics including ethnicity and educational level attained by the initial interview of pr egnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 89 Table 27 Continuous demographic c haracteristics including age, pre-pregnant weigh t, and height of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 89 Table 28 Categorical descripti ve statistics of predictor variables from the initial interview during the first trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 91 Table 29 Continuous descripti ve statistics of predictor variables from the initial interview during the first trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 92 Table 30 Categorical descripti ve statistics of predictor variables from the final in terview during the third trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 93 Table 31 Continuous descripti ve statistics of predictor variables from the final in terview during the third trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 94

PAGE 11

viii Table 32 Categorical descripti ve statistics of the percentage of predictor change between initial and final interviews during the first and third trimesters of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 95 Table 33 Continuous descripti ve statistics of the percentage of predictor change between initial and final interviews during the first and third trimesters of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 95 Table 34 Categorical descripti ve statistics of predictor variables categorized as present or absent during the course of a pregnancy of women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 97 Table 35 Descriptive statis tics of outcome variables including urine sugar levels, birth weight of infants, and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 99 Table 36 Descriptive statisti cs of potentially confounding categorical factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 100 Table 37 Descriptive statisti cs of potentially confounding continuous factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 100 Table 38 Reliability analysis of the autonomy scale of pregnant women attending the County Health Department Prenatal Cli nic in Tuscaloosa and Mobile Counties, AL 1990-2001 103

PAGE 12

ix Table 39 Reliability analysis of the physical work strain scale at the initial interview during the first trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 104 Table 40 Reliability analysis of the physical work strain scale at the final interview during the third trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 104 Table 41 Reliability analysis of the depression scale at the initial interview during the first trimester of pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 104 Table 42 Reliability analysis of the depression scale at the final interview during the third trimester of pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 105 Table 43 Reliability analysis of the social support scale of the participant’s partner of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 105 Table 44 Reliability analysis of the social support scale of the participant’s mother of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 106 Table 45 Assessment of conf ounding factors for partner social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 113

PAGE 13

x Table 46 Assessment of confounding factors for a mother’s social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 114 Table 47 Selected confounding factors for all other predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 116 Table 48 Selected confounding fact ors of all categorical predictors and high birth we ight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 118 Table 49 Selected confounding fact ors of all cont inuous predictors and high birth we ight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 119 Table 50 Selected confounding factor s of urine sugar levels and high birth weight in fants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 120 Table 51 Selected confounding fact ors of all categorical predictors and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 121 Table 52 Selected confounding fact ors of all cont inuous predictors and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 122

PAGE 14

xi Table 53 Selected confounding factor s of urine sugar levels and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 123 Table 54 Multinomial logistic regression model of physical work strain during the sec ond and third trimesters of pregnancy and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 128 Table 55 Multinomial logist ic regression model of the mother’s total social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 129 Table 56 Multinomial logist ic regression model of the mother’s emotional soci al support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 130 Table 57 Multinomial logist ic regression model of the mother’s instrumental so cial support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 130 Table 58 Multiple regression model of urine sugar level as a dichotomous measure and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 137 Table 59 Multiple regression m odel of urine sugar level as an ordinal measure and the bi rth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 137

PAGE 15

xii Table 60 Logistic regression model of urine sugar level as an ordinal measure and Caesarean section of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 138 Table 61 All predictors in one l ogistic regression model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 140 Table 62 Logistic regression model of predictors assessed as present or absent dur ing pregnancy and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 141 Table 63 Logistic regression model of predictors assessed from the initial interview during the first trimester and urine sugar leve ls of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 142 Table 64 Logistic regression model of predictors assessed from the initial interview during the first trimester and urine sugar leve ls of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 143 Table 65 Logistic regression model of predictors assessed from the final interview dur ing the third trimester and urine sugar leve ls of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 144 Table 66 Logistic regression model of predictors assessed from the final interview during the third trimester and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 145

PAGE 16

xiii Table 67 Final logistic regressi on predictor model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 147 Table 68 Final logistic regressi on predictor model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 148 Table 69 Final logistic regressi on predictor model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 148 Table 70 Final logistic regressi on predictor model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 149 Table 71 All predictors in one l ogistic regression model and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 151 Table 72 Logistic regression model of predictors assessed as present or absent during pregnancy and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 152 Table 73 Logistic regression model of predictors assessed from the initial interview during the first trimester and high birth weight in fants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 153

PAGE 17

xiv Table 74 Logistic regression model of predictors assessed from the initial interview during the first trimester and high birth weight in fants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 153 Table 75 Logistic regression model of predictors assessed from the final interview dur ing the third trimester and high birth weight in fants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 154 Table 76 Logistic regression model of predictors assessed from the final interview dur ing the third trimester and high birth weight in fants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 155 Table 77 All predictors in one multiple regression model and the birth weight of in fants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 157 Table 78 Multiple regression model of predictors assessed as present or absent during pregnancy and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 157 Table 79 Multiple regression model of predictors assessed from the initial interview during the first trimester and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 159

PAGE 18

xv Table 80 Multiple regression model of predictors assessed from the initial interview during the first trimester and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 159 Table 81 Multiple regressi on model of the predictors assessed from the final in terview during the third trimester and the birth we ight of infants born to pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 160 Table 82 Multiple regression model of predictors assessed from the final interview dur ing the third trimester and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 160 Table 83 Multiple regression model of predictor difference scores between the initial and final interviews and the birth weight of inf ants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 161 Table 84 Final multiple regr ession predictor model and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 161 Table 85 All predictors in one l ogistic regression model and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 163 Table 86 Logistic regression model of predictors assessed as present or absent during pregnancy and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 164

PAGE 19

xvi Table 87 Logistic regression model of predictors assessed from the initial interview during the first trimester and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 164 Table 88 Logistic regression model of predictors assessed from the initial interview during the first trimester and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 165 Table 89 Logistic regression model of predictors assessed from the final interview dur ing the third trimester and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 166 Table 90 Logistic regression model of predictors assessed from the final interview dur ing the third trimester and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 167 Table 91 Final logistic regr ession predictor model and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 168 Table 92 Logistic regression model of the interaction between marital status from the initial interview during the first trimester and ethnicity with urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 170

PAGE 20

xvii Table 93 Logistic regression model of the interaction between marital status from the final interview during the third trimeste r and ethnicity with urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 170 Table 94 Logistic regression model of the interaction between partner social support and ethnicity with urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 171 Table 95 Logistic regression model of the interaction between physical work strain during the second and third trimesters and ethnicity with urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 172 Table 96 Logistic regression model of the interaction between history of physica l or verbal abuse and ethnicity with high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 174 Table 97 Logistic regression model of the interaction between physical or verbal abuse during the second and third trimeste rs and ethnicity with high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 175 Table 98 Logistic regression model of the interaction between the mother’s social support scale and ethnicity with high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 176

PAGE 21

xviii Table 99 Multiple regressi on model of the interaction between marital status from the initial interview during the first trimester and ethnicity on the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 178 Table 100 Multiple regressi on model of the interaction between marital status from the final interview during the third trimeste r and ethnicity on the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 178 Table 101 Multiple regressi on model of the interaction between the partner social support scale and ethnicity on the birth we ight of infants born to pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 179 Table 102 Logistic regression model of the interaction between marital status from the initial interview during the first trimester and ethnicity with Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 180 Table 103 Logistic regression model of the interaction between the mother’s social support scale and ethnicity with Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 181 Table 104 Statistically signific ant associations between predictors and outcomes fr om hypotheses 1-3 in the results chapter fo r structural equation modeling of pregnant wom en attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 189

PAGE 22

xix Table 105 Goodness-of-fit indices for hypothesis 2 assessing associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 192 Table 106 Model 4 statistics of associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 193 Table 107 Correlation matr ix for Model 4 assessing associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 194 Table 108 Goodness-of-fit indices for hypothesis 2 assessing associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 195 Table 109 Model 3 statistics of associations between predictors and urine sugar le vels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 196 Table 110 Correlation matr ix for Model 3 assessing associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 198 Table 111 Model statistics of as sociations for predictors of urine sugar levels and high bi rth weight infants of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 200

PAGE 23

xx Table 112 Correlation matrix for predictors of urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 201 Table 113 Correlation matrix for predictors of high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 202 Table 114 Model statistics of associations for predictors of urine sugar levels and high birth weight infants of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 203 Table 115 Correlation matrix for predictors of urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 204 Table 116 Correlation matrix for predictors of high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 205 Table 117 Model statistics of as sociations for predictors of urine sugar levels and the bi rth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 207 Table 118 Correlation matrix for predictors of urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 208 Table 119 Correlation matrix for predictors of t he birth weight of infants born to preg nant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 209

PAGE 24

xxi Table 120 Model statistics of as sociations for predictors of urine sugar levels and the bi rth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 210 Table 121 Correlation matrix for predictors of urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 211 Table 122 Correlation matrix for predictors of t he birth weight of infants born to preg nant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 211 Table 123 Model statistics of asso ciations for predictors of Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 213 Table 124 Correlation matrix for predictors of Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 214 Table 125 Statistically significant associations between predictors and outcomes from hypotheses 4 and 5 in the results chapter for structural equation modeling of pregnant wom en attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 215 Table 126 Goodness-of-fit indices for hypothesis 4 assessing the interaction between ethnicity and predictors on urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 216

PAGE 25

xxii Table 127 Model 1 statistics of the interaction between ethnicity and predictors on urine sugar levels of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 217 Table 128 Correlation matrix for Model 1 assessing the interaction between et hnicity and predictors on urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 218 Table 129 Goodness-of-fit indices for hypotheses 4 and 5 assessing the interaction between ethnicity and predictors on urine sugar levels and high birth weight infants of preg nant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 220 Table 130 Model 2 statistics of the interaction between ethnicity and predictors on urine sugar levels and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 221 Table 131 Correlation matrix for the interaction between ethnicity and predictors of urine sugar levels of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 222 Table 132 Correlation matrix for the interaction between ethnicity and predictors of high birth weight infants of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 223

PAGE 26

xxiii Table 133 Logistic regression model of the interaction between physical or verbal abuse during the second and third trimes ters and the mother’s social support scale with high birth weight infants of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 224 Table 134 Logistic regression model of the three-way interaction between physical or verbal abuse during the second and thir d trimesters, ethnicity, and the mother’s social support scale with high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 227 Table 135 Goodness-of-fit indices for hypotheses 4 and 5 assessing the interaction between ethnicity and predictors on urine sugar levels and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 229 Table 136 Model 2 statistics of the interaction between ethnicity and predictors on urine sugar levels and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 230 Table 137 Correlation matrix for the interaction between ethnicity and predictors of urine sugar levels of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 231 Table 138 Correlation matrix for the interaction between ethnicity and predictors of the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 232

PAGE 27

xxiv Table 139 Model 2 statistics of the interaction between ethnicity and predictors on Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 234 Table 140 Correlation matrix for the interaction between ethnicity and predictors C aesarean section births to pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 235 Table 141 Model fit statisti cs for each logistic regression model table from Chapter 4, the Results Chapter including Chi-square goodness of fit statistics and statistical significance 311 Table 142 Hypothesis 3 compar ison of original odds ratios and re-analyzed odds ratios excluding low-weight births of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 317 Table 143 Hypothesis 5 and s upplemental analysis of the original interaction of ethnicity on predictors and high-weight births and analyses excluding lowweight births among pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 319

PAGE 28

xxv LIST OF FIGURES Figure 1 Map of Alabama for location of Tuscaloosa and Mobile Counties from the U.S. Census Bureau, 2002 5 Figure 2 Proposed causal pathway model of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 7 Figure 3 Proposed pathway to insulin resistance during pregnancy with unidentifi ed hormonal influence with solid lines repres enting known effects and dotted lines representing suspected factors 21 Figure 4 Proposed pathway to development of gestational diabetes among pregnant women and resulting hyperinsulinemia among newborn infants 21 Figure 5 Integration of framewor k, theory, and m odel based on Kreiger, Berkman, and Glass’ current work in social epidemiology 36 Figure 6 Outline of the inte rviewing procedures for the combined data set of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 41 Figure 7 Proposed causal pathway model of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 61 Figure 8 Organization of the de scriptive section for the Results Chapter 87 Figure 9 Organization of inferent ial statistics section for the Results Chapter 110

PAGE 29

xxvi Figure 10 Strategy for assessing confounding factors for the Results Chapter 112 Figure 11 Analysis strategies for hypotheses 2 and 3 for the Results Chapter 134 Figure 12 Logistic regression model of the interaction between ethnicity and physical work strain in the second and third trimesters on urine sugar levels of pregnant women attendi ng the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 (All possible scores; Black/0 score as reference), with triangles representing White women and squares representing Black women 173 Figure 13 Logistic regression model of the interaction between ethnicity and the mother’s social support scale on high birth wei ght of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 (All possible scores; Black/0 score as reference), with triangl es representing White women and squares repres enting Black women 176 Figure 14 Proposed causal pathway model of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 190 Figure 15 Structural equati on modeling procedure for Hypothesis 2 assessment of predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 191 Figure 16 Comprehensive modeli ng for predictors of urine sugar levels excluding physical work strain of pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 194

PAGE 30

xxvii Figure 17 Comprehensive modeli ng for predictors of urine sugar levels including physical work strain of pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 197 Figure 18 Structural equati on modeling procedure for Hypotheses 1 and 3 assessm ent for predictors of urine sugar levels and high birth weight infants of pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 199 Figure 19 Comprehensive modeli ng for predictors of urine sugar levels and high birth weight excluding physical work strain of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 201 Figure 20 Comprehensive modeli ng for predictors of urine sugar levels and high birth weight infants including physical work strain of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 204 Figure 21 Structural equati on modeling procedure for Hypotheses 1 and 3 assessm ent for predictors of urine sugar levels and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 206 Figure 22 Comprehensive modeli ng for predictors of urine sugar levels and the birth weight of infants excluding physical work strain of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 208

PAGE 31

xxviii Figure 23 Comprehensive modeling for predictors of urine sugar levels and the birth weight of infants including physical work strain of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 210 Figure 24 Structural equati on modeling procedure for Hypothesis 3 assessment of predictors on Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 212 Figure 25 Comprehensive mode ling for predictors of Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 213 Figure 26 Structural equati on modeling procedure for Hypothesis 4 assessment of the interaction between ethnicity and predi ctors on urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 216 Figure 27 Comprehensive mode ling for the interaction between ethnicity and predi ctors of urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 218 Figure 28 Structural equati on modeling procedure for Hypothesis 5 assessment of the interaction between ethnicity and predi ctors of urine sugar levels and high birth we ight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 219

PAGE 32

xxix Figure 29 Comprehensive mode ling for the interaction between ethnicity and predi ctors of urine sugar levels and high birth we ight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 222 Figure 30 Logistic regression model of the interaction between the mother’s social support scale and physical or verbal abu se during the second and third trimesters on high bi rth weight infants of pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 (All possible scores; non-abused/0 score as reference), with triangles representing non-abused women and squares representi ng abused women 225 Figure 31 Logistic regression model of the three-way interaction between the mother’s social support scale, ethnicity, and physical or verbal abuse during the second and third trimesters on high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 (All possible scores; Black/non-abused/0 score as reference), with the stra ight line representing Black, non-abused women; squares representing Black, abused women; circles representing White, non-abused women; and triangles representing White, abused women 227 Figure 32 Structural equati on modeling procedure for Hypothesis 5 assessment of the interaction between ethnicity and predi ctors of urine sugar levels and the birth wei ght of infants born to pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 228

PAGE 33

xxx Figure 33 Comprehensive mode ling for the interaction between ethnicity and predi ctors of urine sugar levels and the birth wei ght of infants born to pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 231 Figure 34 Structural equati on modeling procedure for Hypothesis 5 assessment of the interaction between ethnicity and predictors on Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 233 Figure 35 Comprehensive mode ling for the interaction between ethnicity and pr edictors of Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 234 Figure 36 Revised theoretic al model and framework based on synthesis of results 250 Figure 37 The residual scatte rplot of birth weight and ethnicity controlling fo r confounding factors of pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 313 Figure 38 The residual scatte rplot of birth weight and ethnicity excluding low birth weight infants and controlling for confounding factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 313 Figure 39 The residual scatterp lot of birth weight and urine sugar levels controlling for confounding factors of pregnant women attending the County Health Department Prenatal Clini c in Tuscaloosa and Mobile Counties, AL 1990-2001 314

PAGE 34

xxxi Figure 40 The residual scatterp lot of birth weight and urine sugar levels excluding low birth weight infants and controlling for confounding factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 315

PAGE 35

xxxii BIOMEDICAL AND PSYCHOSOCIAL DE TERMINANTS OF PROBLEMATIC BIRTH OUTCOMES Charlan Day Kroelinger ABSTRACT The primary objective of this stud y was to evaluate the associations between psychosocial stressors, urine sugar levels, and subsequent birth outcomes, specifically high birth weight babies and Caes arean section births. In a prospective cohort study, 506 Black and White women of childbearing age were followed for the duration of one pregnancy in Tuscaloosa and Mobile counties in Alabama from 1990 to 2001. Participants were interviewed twice throughout pregnancy, during the first and thir d trimesters, respectively, and birth outcome data were collected via medical c hart reviews. Six percent (6.1%) of the women in the sample had a high bi rth weight baby, and 18.4% received a Csection during childbirth. Adjusted logist ic regression results indicate that urine sugar levels are predictive of high-wei ght births, with women who have higher urine sugar levels were more than three times likely to birth a high weight baby compared with women who have no detect able urine sugar spill (OR 3.25; 95% CI 1.30, 8.10). In addition, the interact ion of familial social support throughout pregnancy, physical or verbal abuse duri ng the second and third trimesters, and ethnicity is significantly associated wit h increased risk of having a high birth

PAGE 36

xxxiii weight baby. For C-section, single parti cipants are over two times less likely to receive a C-section during childbirth com pared with currently married participants (OR 0.46; 95% CI 0.21-1. 00). Examining structural equation modeling results; pathways leading from urine sugar levels physical or verbal abuse during the latter half of the pregnancy, and a mo ther’s social support among White participants are indicative of high weight births ( R2 = 0.65). White abused women who receive their mother’s social support are more likely to have a high birth weight baby compared with both White and Black women who are not abused and receive the same amount of so cial support. Recommendations to public health practitioners include prim ary prevention thr ough promotion of familial support during pregnancy, sec ondary prevention through urine sugar screening at every prenatal visit, and di rect intervention by identifying and inquiring about instances of suspected abuse during pregnancy.

PAGE 37

1 CHAPTER 1 INTRODUCTION 1.0 Introduction Birth complications such as high birth weight and Caesarean section are health problems specific to post-industr ial nations. Although both problems are most likely prevalent in all post-industria l nations, data are only available for the United States, the Un ited Kingdom, and Canada1. In the U.S. for the year 1999, approximately 9.9% of all live births we re greater than 4000 gr ams, while 7.6% were considered low birth weight (>2500 grams) [1]. Likewise, for 2000-2001, 10% of all births in the U.K. were greater than 4000 grams, and 6% were less than 2500 grams [2]. Data were only avai lable for low birth weight births in Canada (5.6% in 1999) [3]. The World Health Organization mandates that not more than 15-20% of all birt hs should be by Caesarean section in any region of the world [4]. The C-section rate in the United States was 22.9% in 2000 [5], 22% in the U.K. [2], and 19.9% in Canad a in 2002 [6]. Further, in an early study by Nortzon et al., C-section rates in the U.S., Scotland, and Norway all increased between the years 1970 and 1985 [7]. The listed proportions for the 21st century are as high or higher than the WHO deems ju stifiable. High-weight births occur 1 Data from other WHO member countries were unavailable, not translatable, or not collected for the outcome variables.

PAGE 38

2 in greater proportions in both the United States and Great Britain, and the prevalence rate of Caesarean sections exceed the percentage recommended by public health agencies. These factors int ensified by modifiable risk factors during pregnancy are appropriate targets of public health prevention, intervention, and research programs. Routine measures taken at each prenatal visit are potential early indicators of high birth weight babies and the need for subsequent C-sections. These markers, including weight, blood pressure, and urine sugar levels, are integral to estimating the probability of ear ly identification of such complications. The objectives of the current study are to investigate the extent to which sociocultural factors (both p sychosocial and physical) infl uence the development of higher urine sugar levels, high birth we ight babies, and Caesarean sections; the potential of urine sugar level readings as useful indicators of specific birth complications; and the effect s of ethnicity on both psychosocial and physical factors, urine sugar leve ls, and birth complications. In essence, the first objective ai ms to establish a relationship between specific high-risk predictors with hi gher urine sugar levels and birth complications. The second study objec tive addresses the question of urine sugar levels predicting poor pregnancy out comes. Finally, the third study objective is an evaluation of whether relationships between psychosocial and physical factors and subsequent developm ent of high urine sugar levels, and these same factors and birth complic ations are modified by ethnicity.

PAGE 39

3 The data used to achieve the research objectives are from two prospective cohorts of pregnant women in Tuscaloosa County, Alabama (N= 397) and Mobile, Alabama (N = 109) followed from baseline (1-20 weeks gestation) through childbirth. Evaluation includes measurement of predictors at baseline or initial interview (time 1 or t1), and subsequent measurement during the third trimester of pregnancy (final interv iew; time 2 or t2). Risk factors are assessed overall for significance. Signi ficant risk factors are then assessed to determine when during pregnancy t hey are most influential (e.g., first trimester or third trimester), and finally are modeled th rough the use of causal pathways to further explain the associations. The two samples, taken in similar geographic regions, are both located in the state of Alabama (Figure 1) [8]. Both counties contain urban and rural environments, with the majority of women in the sample residing in metropolitan, low-income areas. The state ethnic dist ribution for the year 2000 is 71.1% White and 29.9% Black/other, with a median inco me of $34,135 (1999), and 16.1% of the population below poverty level ( 1999). Approximately 75.3% of the population has obtained a high school degree, and 19.0% have at least a bachelor’s degree. The land area of the entir e state in square miles is 50,744 with 87.6 persons per square mile [9]. For Tuscaloosa County, 68.1% of the population is White with 31.9% Black/ot her (2000), and the median income is $34,436 per year (1999). Seventeen percent (17%) live below the poverty level (1999), 78.8% have high school degrees (2000), 24.0% have at least a

PAGE 40

4 bachelor’s degree (2000), and the size of the county is 1,324 square miles with 124.5 persons per square mile [10]. Sixt y-three percent (63. 1%) of those in Mobile County are White with 36.9% Bla ck/other (2000). The median income is $17,178 per year (1999), and 18.5% of the population live below the poverty level (1999). Seventy-seven percent (76. 7%) are high school degreed (2000) with 18.6% degreed at the bachelor ’s level (2000). The county is 1,233 square miles with 324.3 persons per square mile [11].

PAGE 41

5Figure 1 Map of Alabama for location of Tuscaloosa and Mobile Counties from the U.S. Census Bureau, 2002 Source: Alabama County Selection Map 2002, U.S. Census Bureau. http://quickfacts.census.gov/ qfd/maps/alabama_map.html

PAGE 42

6 The two ethnic groups of interest in the study are Black and White, as all other ethnicities combined in both cohorts represent less than four percent of the total sample. Ages for participation range fr om fourteen to thirty -four. All other ages are excluded as part of initial study protocol due to high-risk of pregnancy complications based on various age-relat ed risk factors. Since low income, Medicaid waiver women are traditionally underserved in terms of medical care; they have been selected as the soci o-economic group for this study. Five specific research hypotheses will be tested: (1) Urine sugar levels during pregnan cy are positively associated with development of pregnancy complicati ons (e.g., high birth weight and Cesarean section). (2) Psychosocial and physical factors (e.g., physical stress, lack of social support, depression, autonomy, pregnan cy wantedness, and physical and verbal abuse) during pregnancy are asso ciated with higher urine sugar levels. (3) Psychosocial and physical factors dur ing pregnancy are associated with pregnancy complications. (4) The associations between psychosocia l and physical factors and urine sugar levels differ among Black wom en compared with White women. (5) The associations between psychos ocial and physical factors and pregnancy complications differ among Black women compared with White women.

PAGE 43

7 By utilizing the ecosocial framework, a model incorporating all five hypotheses will be conceptualized in order to characterize a prototypical pathway for disease development among women in the target population (Figure 2). Upon completion, t he study results will be potentially generalizable to ethnicities of Bla ck and White women among lower income groups in the southern United States. Figure 2 Proposed causal pathway model of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 In summary, the study objectives are to identify interrelationships between psychosocial and physical factors, development of high urine sugar levels, and subsequent pregnancy complications. S pecifically, the focus on Medicaid recipients highlights the affects of thes e factors on a defined population typically less educated, lower wage receiving, and less likely to receive the same Ethnicity Depression Marital Status and Autonomy Social Support Physical Work Stress Physical and Verbal Abuse Urine Sugar Reading High Birth Weight Cesarean Section H1 H2 H3 H4 and H5 Pregnancy Wantedness

PAGE 44

8 treatment options and benefits of those patients who rece ive private-pay care. Results are then generalizable to this underserved population, where factors such as access to care and the decision-ma king process of treatment play less of a role in preventing problematic birth outcomes. The public health impact of these study results will aid in identification of risks specific to this group of women, hi ghlight strategies in prevention of poor pregnancy outcomes, and present modifications of current protocols to assist practitioners in treatment options for poor er groups of women. In addition to aiding practitioners and health care clinics in identifying wom en at high risk of pregnancy complications and in developi ng effective interventions (e.g., programs that reduce psychosocial and phy sical strain); ideally, results will promote an understanding of pot entially modifiable factors that influence the development of poor birth outcomes, help to lower the current rate of disease, and assist in raising the awareness of post-industrial specific diseases among both practitioners and the lay public.

PAGE 45

9 CHAPTER 2 LITERATURE REVIEW 2.0 Introduction Most research to date has focused on predictors of and disease resulting from low birth weight. However, Colhoun and Charturvedi indicate that the same co-morbid events that occur in children and adults born with low birth weight may also occur in children and adults born with hi gh birth weight [12]. In fact, they posit that there may be a U-shaped curv e in terms of development of glucose intolerance in adulthood and birth weigh t. Any deviation from the norm, or normal birth weight (e.g., low or high bi rth weight), may involve increased risk of specific morbidities of adulthood. To support this supposition, Egeland, Skjaerven, and Irgens state that both low and high birt h weight contributed to adult development of gestational diabet es among Norwegian women [13]. McCance et al. reported similar findings with an increase in overall diabetes prevalence among Pima Indians born low and high birth weight (prevalence among low birth weight = 38%; prev alence among normal weight = 20%; prevalence among high birth weight 35%) [14]. Dabelea et al., confirm these findings of a U-shaped curve between birt h weight and develop ment of Type II Diabetes among Pima Indian children aged 10 to 14 [15]. In addition to glucose intolerance or diagnosed diabetes, it is bi ologically plausible that morbidities

PAGE 46

10 associated with being low-weight at birt h may also be associated with being born high-weight. The purpose of this research is to cl arify the association of specific sociocultural factors affecting pregnancy and resulting high birth weight and Caesarean section. A causal pathway beginning with thes e factors including increased urine sugar levels as indicative of developing glucose intolerance and ending with high birth weight and s ubsequent C-section is proposed and supported by current research. The follo wing chapter reviews the literature on socio-cultural factors and pregnancy, t he research history of each outcome measure (i.e. urine sugar levels, high bi rth weight, and Caesarean section), and the selected framework and theory ut ilized in this dissertation. 2.1 Socio-cultural Factors and Pregnancy Although exposure and disease ar e affected by the environment and biologic pathways, individual societal and cultural factors also influence the amount and type of exposures, and the subs equent development and severity of disease. Robert Hahn describes three modes of socio-cultural influence on exposure and disease [16]. He posits that socio-cultural factors construct, mediate, and aide in the production of conc epts of disease. The construction of disease definitions is influenced by soci o-cultural factors of individuals and society and is perpetuated by generati onal use and re-use through constant social interaction. Mediation is the influence of cultural values, ideas, and concepts on exposure and disease assessment and interpretation. Production of

PAGE 47

11 disease is the direct exposure to viru ses or other pathogens by contact through defined social relationships or institutions such as clinics or hospitals. The mediation of socio-cultur al influences is considered by Hahn as the most measurable in the context of medical re search. Measures of these mediating socio-cultural influences include such factors as socio-economic status, education level, income level, and marital status. Current international research on the impact of these soci o-cultural factors on pregnancy is focused on adverse outcome s including mater nal death during pregnancy or childbirth [17-19], complicat ions of home labor and delivery [20], access to family planning [21], elec tive Caesarean section [22], early age childbearing [23], in fant mortality [24], and postpartum illnesses [25]. The purpose of this dissertation is to assess t he impact of a series of socio-cultural factors on the development of problematic birth outcomes in a low income population in the United States The socio-cultural fact ors of concern are marital status and feelings of autonomy, pregnancy wantednes s, depression, physical work strain, physical or verbal abuse and social support both partner and familial. Other socio-cultural influ ences such as socio-economic status, education level, and income level are contro lled in analysis in order to assess the direct impact of the selected predictors on birth outcomes. The concept of autonomy, or independe nce, is associated with the healthy functioning of a marriage pr imarily for women [26, 27]. Marital status and autonomy are directly related based on the associations between higher mortality

PAGE 48

12 and widowhood [28, 29], or cultural practice s of certain ethnic groups such as single Black women with children produced by multiple male partners [30, 31]; lack of autonomy regardless of marital st atus is associated with poorer health among women [32]. In He lsing, Szklo, and Comstock’s studies, mortality occurred within one year after the deat h of a spouse after controlling for confounding of socio-economic status and heal th care. These findings indicate that sudden independence follo wing the end of a long-term relationship has a negative effect on mortality. In co ntrast, Boone asserts that in Black communities, living independent from a partner is highly adaptive for women even in the presence of multiple childr en. Here, among single women, autonomy or independence is positively associat ed with decreasing abuse, and increased social and financial status. In a study by Lou et al., common law marriages were associated with the worst pregnancy outcome s (e.g., pre-term birth, low birth weight, or stillbirth) compared to married and single women [33]. For analyses of the effects of marital status on Caesarean section, Kabir et al. found that married women most often delivered via C-secti on, and had higher rates of repeat Csection deliveries [34]. This dissertation focuses on the affects of marital status, specifically being non-married, and la ck of autonomy, or independence, on increased urine sugar spill, birthing highweight infants, and Caesarean section during childbirth. Pregnancy wantedness as defined by Mille r is whether or not a woman, upon finding out she is pregnant, wants her baby [35]. Wanting a pregnancy is

PAGE 49

13 different from intending to get pre gnant or planning a pregnancy, however, the three terms are often used interchangeably Studies of unwanted pregnancy as a socio-cultural factor and resulti ng adverse outcomes include psychiatric morbidity of the infant, educational attain ment of the child, and neonatal mortality [36-38]. Predictors of probl ematic birth outcomes in conjunction with wantedness are receiving adequate prenatal care and weight gain during the pregnancy [3941] To date, only two studies have examined the possible association of wantedness and birth weight. Morris, Udry and Chase examined the affects of unwanted pregnancy on resulting low birth weight infant s [42]. They found no association between the st ate of not wanting a pr egnancy and reduction in birth weight. Likewise, Sable et al. ex amine the association between having an unwanted pregnancy and birth wei ght [43]. In contrast, they found that women who had low birth weight babies were more likely to have experienced feelings of unwantedness during pregnancy compared with women who had normal weight babies. The focus of this study is to explore the association between having an unwanted pregnancy and resulting problematic birth outcomes such as high birth weight and C-section; since a paucity of research exists on this topic, all analyses are considered exploratory. Depression as a socio-cultural factor and pregnancy studies include associations between maternal depression2 and increased fetal heart rate [44], malnourishment of the in fant [45, 46], alcohol and drug use during pregnancy 2 Maternal depression during pregnancy only is included in this literature review. Studies of post-partum depression are excluded from this review.

PAGE 50

14 [47], and inconsolable infants with exce ssive crying [48]. In terms of the outcomes of interest in this study, P aarlberg et al. examined the association between maternal depression in the first tr imester and its affect on birth weight [49]. Depression was associated with havin g a low birth weight baby. In support, Copper at al. determined that an associ ation between maternal depression and resulting low-weight births was statisti cally significant [50]. In addition, depression was also associated with prematur e birth. In cont rast, Hedegaard et al. and Brooke et al. found no associati on between maternal depression and birth weight [51, 52]. When exam ining the affects of depressi on on C-section delivery, Wu et al. reported no association bet ween maternal depression and subsequent Caesarean section delivery [53]. Referring to Colho un and Charturvedi’s review, since having an unwanted pregnancy is sign ificantly associated with birthing a low-weight baby, it is plausible that unwanted pregnancy may also influence the birth of high-weight babies. Poerksen and Petitti researched the impact of physical strain in the workplace during pregnancy and problematic birth outcom es [54]. They found no association between work strain and birthi ng low-weight infants. Nurminen et al. concluded that physical work stra in during pregnancy was predictive of spontaneous abortion only if the work involved a lar ge amount of standing, and observed an increase in maternal hy pertension with work that included heavy loads [55]. However, Hansteen, Kjuus and Fandrem found that psychological and physical strain at work increased t he risk of spontaneous abortion [56].

PAGE 51

15 Klonoff-Cohen, Cross, and Pieper found that increased physical strain increased the risk of development of pre-eclamp sia (e.g., hypertension during pregnancy) compared with low strain among working wo men [57]. Further they found that the risk increased when all working women were compared with non-working women during pregnancy. In research regarding birth weight, Henrikson, Hedegaard, and Secher found a trend betwe en increased job strain and small for gestational age and premature delivery, although none of the findings were statistically significant [58]. While much of the research on physical work strain focuses on pregnancy complications typically associated with low birth weight and pre-term births, no res earch has addressed these issues in terms of high birth weight or Caesarean section. Current research in the area of ver bal and physical abuse includes studies associating these factors with low birt h weight and lack of prenatal care, especially of adolescent pregnancies [59-62]. Am ong women of childbearing age, Neggers et al. report ed significant associations between physical abuse and low birth weight and premature birth am ong a cohort of Afric an American women [63], as did Valladares et al. in a cohor t of Nicaraguan women [64]. In a review of current research and a meta-analysis of findings from 14 studies associating abuse and low birth weight, Murphy et al. found a significant association between physical and psychological abuse duri ng pregnancy and resulting low-weight births [65]. However, Kearney et al. found no association between abuse and low birth weight [66]. In support, Alta rac and Strobino found no association

PAGE 52

16 between abuse and low birth weight, but did report an association between the stress caused by abuse and low birth weight [67]. They posit that stress is a mediating factor between abuse and low bi rth weight and has a much greater impact on low birth weight than the abuse itself. Rachana et al. found an association between abuse and abdominal in jury, placental abruption, premature birth, and subsequent C-sect ion [68]. In contrast, Berenson et al. reported no association between abuse and C-section delivery [69]. Current research supports the association between abuse and C-section, but an association between abuse and high birth weight is unexplored. Social support during pregnancy is pr ovided by both the current partner and other family members. In resear ch on partner and familial support and adverse outcomes, Norbeck and Anders on found that support decreased the risk of gestation complications, prolonged labor, and C-section among African American women; but increased the risk of problematic outcomes and substance abuse during pregnancy among Whit e women [70]. They indicate that instead of providing protection against adverse outcomes, support among White women reinforces negative behavior during pregnan cy. Lespinasse et al. state that support in the delivery room significantly reduced low-weight a nd very low-weight births among African American women [ 71]. Norbeck, DeJoseph, and Smith report that increased social support si gnificantly increased birth weight among African American women [72]. Da Costa et al. and Feldman et al. support this finding in their studies of social suppor t and birth weight among diverse ethnic

PAGE 53

17 groups [73, 74]. Jesse, Wallace, and Seav er found that the impact of increased partner support significantly increased inf ant birth weight [75]. For studies of small for gestational age infants, Dejin-Karlsson et al. found a significant association between decreased social support and small for gestational age infant births [76]. In contrast, Pryor et al. reported no association between social support during pregnancy and small for gestational age births [77], as did Sheehan [78]. In summary, Hahn defines care in pregnancy and childbirth or obstetrics as a discipline that, in principle, a ssumes pregnancy is an experience but in practice observes the pregnancy process as a series of biological changes within a woman [79]. Such a viewpoint precludes the associations of any socio-cultural factors and biologic pregnancy outcomes, ex cluding demographic characteristics. Hahn supports his supposition with a revi ew of obstetric textbooks from 1903 to 1989 and the changing definitions and descripti ons of obstetrics as a practicing discipline. It is the intent of this dissertation to in clude these socio-cultural factors using a theoretical framework to analyze the effect, if any, on biologically measurable problematic pregnancy and chil dbirth outcomes while controlling for other biologic risk factors that are asso ciated with these outcomes of interest: increased urine sugar levels, high birth weight, and Caesarean section. Each set of biologic risk factors controlled in analysi s is addressed in terms of the specific outcome in the following sections.

PAGE 54

18 2.2 Outcome Measures The following sections are compos ed of an in-depth review of the epidemiologic literature for t he outcomes of urine sugar levels, high birth weight, and Caesarean section. The biologic basis for using urine sugar levels as a mediating outcome is provided, as well as its link to high birth weight infants and C-section. Also, the pat hway beginning with increased urine sugar levels and ending in C-section is reviewed in the context of this analysis. 2.2.1 Urine Sugar Levels Increased sugar in the urine is not independently problematic, but is indicative of underlying morbidity in the mother and the fetus. Current epidemiologic research associates incr eased urine sugar levels with known exposure to arsenic [80], diagnosis of r enal glycosuria [81] and chorioamnionitis [82]. Urine sugar levels are also st udied as screening and monitoring tools for diabetes in human and animal studies [83-86], multiple sclerosis in humans [87], and renal impairment in animal studies [88 ]. To date, research on urine sugar screening does not include analysis of sociocultural factors and their impact on increased sugar in the urine. Urine sugar screening is literally a monitoring of the sugar in a pregnant woman’s urine. The American Diabet es Association (ADA) recommends monitoring throughout pregnancy of sugar in the urine, and highly recommends a diagnostic test, glucose tolerance test ing, identifying gestational diabetes between 24 and 28 weeks gestation [89]. The urine sugar screening test is

PAGE 55

19 routinely performed at each prenatal vi sit by placing a dipstick in a urine specimen3. If high spill is recorded on multiple occasions, a diagnostic test for gestational diabetes may be perform ed based on physician preference. Since the bulk of the reviewed current research utilizes urine sugar testing as a screening and monitoring tool for diabe tes, a description of the biologic mechanisms involved in the development of diabetes is necessary. During pregnancy, carbohydrates consumed by wo men are processed and converted to glucose that circulates in the blood or glycogen that is stored in the liver until needed. Glycogen is then converted to glucose and released in the blood stream. The body requires a minimal am ount of glucose to function properly; less than 30 milligrams per deciliter are c onsidered too low for proper functioning and may result in disorientation. In contrast, greater than 300 mg/dl is considered too high and if chronic, lead to diagnosis of diabetes or other health complications [90]. Referring to Figure 3, increased produc tion of glucose in the blood leads to an increase of insulin production in the pancreas of the beta cells. During the course of the pregnancy, due to unidentified factors, insulin resistance may occur. Hypothesized factors for de creased insulin production include genetic strings and increased hormones produced by the placenta such as progesterone, estrogen, human placental lactogen, or human chorionic somatotropin. As a result, the subsequent glucose intolerance mimics a state of Type II diabetes. A 3 See the Methodology Chapter’s description of the urine sugar scr eening instrument for further definition of the tool in this dissertation. Screening practices may vary depending on phys ician practice, clinic, or hospital protocol. Detailed descriptions of the protocol for clinics participating in this study are provided.

PAGE 56

20 physiologic response of hyperinsulinemia occurs and glucose circulation increases in the maternal blood stream (Figure 4). The glucose crosses through the placenta into the chord blood and to the fetus [91-93]. The fetal response to this increase in sugar is an overproduction of insulin to proce ss the glucose, and resulting over-nutrient transfer causes an increase in fetal growth and ultimately, birth weight [94, 95]. Routine testing fo r gestational diabetes occurs during the beginning of the third trimester unless otherwise indicated by screening tests (e.g., urine sugar spill, other high-risk indicators, etc.), and the mother is diagnosed and treated through diet modification or insulin injection [96]. If diagnosis occurs too far along in the pr egnancy, the woman is unwilling to follow treatment protocols, or the problem remains unrecognized, the resulting high birth weight infant precipitates operati ve deliveries through the use of Caesarean section [97].

PAGE 57

21Figure 3 Proposed pathway to insulin resistance during pregnancy with unidentified hormonal influence with solid lines representing known effects and dotted lines representing suspected factors Figure 4 Proposed pathway to development of gestational diabetes among pregnant women and resulting hyperinsulinemia among newborn infants Pancreasbeta cells Insulin produced Insulin resistance occurs Reasons unknown Insulin production increases Body no longer processes insulin State mimics Type II Diabetes Genetic factors Hormones Produced by Placenta -Progesterone -Estrogen -Human placental lactogen -Human chorionic somatotropin Normally Hyperinsulinemia Poor insulin response Glucose circulation increases Glucose crosses the placenta to the fetus Mother’s pancreas cannot produce enough insulin MomHyperglycemia Gestational diabetes FetusPancreas over-produces insulin

PAGE 58

22 Gestational diabetes is defined as “any degree of clinical glucose intolerance with onset or fi rst recognition during pregnancy” [89, 93]. Complications of pregnancy that may occu r are altered gestation of the fetus (e.g., shorter or longer gestation), placental failure, pre-eclampsia, or high birth weight of the infant. Ge stational diabetes may be pre liminarily identified using the urine sugar screening tool previously discussed, but is diagnosed using the glucose tolerance test performed at approximately 28 weeks gestation. Although Danforth’s Obstetrics and Gynecology identifies gestational diabetes as a medium-risk pregnancy comp lication, it recommends screening at least once during pregnancy [98]. Siccardi defines traditional high-risk factors for gestational diabetes as including age, pr e-pregnancy weight, fa mily history of diabetes in a first degree relative, pr evious high-weight baby, and previous perinatal loss [99]. Specif ic ethnic groups are also identified as more likely to have gestational diabetes including His panics and specific Native American groups. Protocols for nurse practitioners and nurse-midwives identify previous history of gestational diabetes, previ ous macrosomic infant, stillborn or malformed infant, previous polyhydramnios, obesity, high urine sugar spill, family history of diabetes in a first degree rela tive, and age greater than 25 years as high-risk factors for gestati onal diabetes [100]. Clinical practice guidelines for midwives list similar protocols, but limit age to greater than 35, and add preeclampsia or hypertension to the list [ 101]. The high-risk factors from these clinical sources are similar, but not uni form. The ADA also lists a group of

PAGE 59

23 characteristics for women considered ‘low-risk’ for developing gestational diabetes, and presumably to lower obstetr ic costs, recommends no diagnostic screening for this group of women: le ss than 25 years of age, normal prepregnant weight, member of an ethnic group with low rates of gestational diabetes mellitus (e.g., Black and White women), and no known hi story of familial diabetes, previous glucose impairmen t, or poor birth outcome [89]. Therefore, diagnosis of gestational diabetes may not be uniform across medical practice. Only using these high-risk factors to screen pregnant women will identify approximat ely 50% of those with gestati onal diabetes [99]. It is plausible to conclude that since scr eening mechanisms are not uniform, more studies linking urine sugar screening to diagnosis of gestational diabetes and subsequent problematic birth outcomes are necessary. Although a urine sugar level testi ng is used as a screening and monitoring tool for gestational diabetes, it is not recommended for diagnostic use. Reasons supporting its use are expense, its rapid response, and the fact that it is noninvasive. More importantly, it is not recommended for diagnosis due to its falsepositive results (low sensitivity) through ox idation or lack of fasting by the patient, and its false negative results through reduction of ascorbic acid in the urine [93]. Therefore, the ADA and the World Health Organiza tion (WHO) recommend using the glucose tolerance test for diagnostic purposes [89, 93]. The gold standard, or glucose toler ance testing threshold limits were established in 1974 by O’Sullivan and these standards were adopted and

PAGE 60

24 modified by the National Diabetes Data Group and then the American Diabetes Association [89, 102]. Original thresholds and current thresholds are located in Table 1. The fasting level is the gluc ose amount prior to the test following 24 hours of fasting. The hour tests indicate the glucose levels in blood drawn at intervals following oral administr ation of the glucose solution. Table 1 Original and current threshold limits for glucose tolerance testing diagnosis of gestational diabetes Glucose Tolerance Testing Thresholds Glucose Tolerance Result Level O’SullivanNDDG ADA Fasting 90 g/L 105 g/L 95 g/L 1 Hour 165 g/L 190 g/L 180 g/L 2 Hour 145 g/L 165 g/L 155 g/L 3 Hour 125 g/L 145 g/L 140 g/L Sources: Siccardi, D.C. Obstet rics and Gynecology: Gestational Diabetes. 2004 Medstudents. http://www.medstudents.com.br/ginob/ginob4.htm Gestational Diabetes Mellitus. Diabetes Care, 2004. 27 (Suppl 1): S88-S90. Although current research does not examine associations between urine glucose or urine sugar and pregnancy outcomes such as high birth weight and Csection, the link between diagnosis of gestational diabetes and these outcomes is clear [103, 104], although the link betwe en diabetes and C-section is mediated by high birth weight [105, 106]. Since ur ine sugar levels are used as a screening tool for later diagnosis of ges tational diabetes, it is bi ologically plausible that an association should exist between elevated urine sugar levels and subsequent high-weight births; presuming the caus al relationship between urine sugar and high birth weight is mediated by gestational diabetes. Scholl et al., found a significant association between higher ma ternal glucose levels and increased

PAGE 61

25 fetal growth and resulting large for gesta tional age births. In addition, they found an association between increased mate rnal glucose and C-section birth [82]. To support the main hypothesis for this dissert ation, urine sugar screening is the test given prior to the glucose tolerance te st; its results are correlated with the tolerance test results. Therefore, te sting the association between urine sugar screening and the subsequent outcomes of hi gh birth weight and C-section is logical based on the concept that t he causal pathway is defined as: Urine sugar screening glucose tolerance testing gestational diabetes diagnosis high-weight birth C-section A main purpose of this dissertation is to identify an association between high urine sugar spill or glycosuria and high-weig ht births or C-section deliveries while bypassing the diagnosis of gestational diabetes which is considered in the causal pathway. The following literature review se ctions highlight the impact of focusing on these two problematic outcomes in terms of their impact on the morbidity and mortality of both mothers and infants, and as a result of impaired glucose functioning. 2.2.2 High Birth Weight The impact of having a high birth weight baby is problematic to both the mother and infant. High birt h weight is defined as an infant weighing more than 4000 grams or 4 kilograms or an infant in the 90th percentile on the intrauterine growth curve [107]; however, this defin ition varies across countries with some definitions of high birth weight beginning at 3800 gr ams or as high as 4500 grams. Recommendations for such la rge for gestational age infants include

PAGE 62

26 monitoring for hypoglycemia in the first fe w hours after childbirth as these infants typically result from maternal complicati ons such as gestational diabetes [108]. Current research on morbidities that result from being born high birth weight includes infant, childhood, and adul t onset diseases (Table 2). Most epidemiologic studies exami ne high birth weight as a predictor for childhood and adult obesity; childhood and adult cancers; development of diabetes; and birth complications such as gestational di abetes among adult women who were born high birth weight, Caesarean section, and inf ant injury and fever at birth. For example, He et al. and Gallaher et al both found associations between being born high birth weight and childhood obesity [109, 110]. In a review article by Ekbom, the main adult onset cancers a ssociated with being a high birth weight infant are breast, prostate, and testicular [111]. Many studies also associate high birth weight with the onset of childhood canc ers, Kaatsch et al. and Robison et al. indicate associations between being bor n high birth weight and development of childhood leukemia [112, 113].

PAGE 63

27Table 2 Literature review of epidemiologic studies of risk factors of and diseases associated with being born high birth weight Disease Year of Publication Author/s Study Design Measure of Association 95% Confidence Interval* Obesity Childhood 2003 1999 Gillman et al. [114] Dabelea et al. [15] Cross-sectional Retrospective Cohort OR 1.3 N/A 1.1-1.5 N/A Adult 2000 1996 1996 Mikulandra et al. [115] Curhan et al. [116] Curhan et al. [117] Case-control Cross-sectional Cross-sectional N/A OR 2.08 OR 1.62 N/A 1.7-2.5 1.4-1.9 Cancer: Childhood Leukemia 2002 2002 2002 1997 1995 Okcu et al. [118] Ou et al. [119] Murray et al. [120] Yeazel et al. [121] Cnattinguis et al. [122] Case-control Case-control Retrospective Cohort Case-control Nest Casecontrol OR 2.2 OR 1.4 OR 1.7 OR 1.8 OR 1.7 1.2-4.1 1.4-1.8 1.2-2.3 1.2-2.5 1.1-2.7 Nephroblastoma/ Wilm’s Tumor 2001 1997 Schuz et al. [123] Yeazel et al. [121] Case-control Case-control OR 1.6 OR 2.1 1.0-2.5 1.2-3.8 Cancer: Adult Breast 2000 Innes, Byers, and Schymura [124] Case-control OR 3.1 1.2-8.0 Colorectal 2002 Sandhu et al. [ 125] Cohort HR 2.6 1.2-5.7 Type II Diabetes 2003 Wei et al. [126] Case-control OR 1.8 1.0-3.1 Birth Complications: Maternal Diabetes (among women born hbw) 2003 Savonna-Ventura and Chircop [127] Case-control OR 2.7 N/A Cesarean section 1992 Webster et al [128] Cross-sectional N/A N/A Birth Complications: Infant Injury 1988 Wikstrom et al. [ 129] Cross-sectional 12.2 3.3-44.4 Fever 2003 Maayan-Metzger, Mazkereth, and Kuint [130] Case-control OR 3.38 1.4-8.2 Respiratory Distress 2001 Sutton et al. [131] Case-control OR 1.8 1.0-3.2 Total number of studies: 20 *NA represents not applicable and refers to a study that used incidence, prevalence, or univariate analyses only. Other areas of current research in high birth weight include brain tumors [132], back pain [133], coronary artery disease [134], Rickets [135],

PAGE 64

28 Erb/Duchenne’s palsy [136], and schizophrenia [137]. These studies, however, either found no significant association bet ween high birth weight and disease, or in every case, only one study has been conducted for each topic. Further research includes examinati on of factors predictive of high birth weight that impact both the mother and fetus such as impaired glucose tolerance and low alpha-fetoprotein, elevated amni otic insulin, fetal hyperinsulinism, gestational diabetes mellitus, gestational hypertension, body mass index or prepregnant weight of the mother, mater nal weight gain during pregnancy, and gestational age of the infant at birth [98, 104, 138-143]. As discussed in the urine sugar level section, glucose intolerance affected by alpha-fetoprotein and insulin levels, and maternal weight gain during pregnancy are part of the pathway to diagnosis of gestational diabetes mellitus. No current studies to date, how ever, address the possible link between urine sugar spill and subsequent high birth weight independent of the diagnosis of gestational diabetes. In addition, t he majority of published research is retrospective with the exception of t he cross-sectional studies. Two major strengths of this study ar e the testing of the associ ation between urine sugar screening and high birth weight, and prospect ive data collection. The reason for selecting urine sugar screening is that women who have higher urine sugar levels inconsistently, or do not present with levels high enough during the third trimester to be diagnosed with gestati onal diabetes, may still be at risk for birthing a high-weight infant. The resu lts of having a high birth weight baby

PAGE 65

29 include adverse events during the birthi ng process and potential increased risk of disease development in bot h childhood and adulthood. As with gestational diabetes and resulti ng high birth weight, similar sociocultural factors contribute to low birth we ight such as maternal age and ethnicity. Other socio-cultural factors associated with premature and lo w-weight births include marital status, education and inco me level, lack of social support, and depression; all factors under investigation in this re search [144, 145]. This dissertation examines the plausibility of a U-shaped curve in terms of morbidities associated with both low and high-weight birth outcomes by assessing the associations between these socio-cult ural factors and high birth weight. 2.2.3 Caesarean Section Caesarean section births are defined as a ‘delivery of the fetus by means of an incision into the uterus’ [107]. Emergency C-sections are most often performed when there is fetal distress during labor and delivery. The most common types of incisions are horizont al through the lower uterine segment, although during profound fetal distress, t he vertical midline incision may be performed. C-section birth is consider ed safe by current medical standards and elective based on physician decision in the absence of fetal distress [108]. Due to the large body of literature on elective C-section, this review focuses on literature directly relevant to this dissertation, such as resulting Csection for women whose pr egnancies are complicated by glucose intolerance or gestational diabetes, large for gestational a ge or high birth weight infants, small

PAGE 66

30 for gestational age or low birth weight in fants, pre-existing conditions of the mother such as obesity, and fe tal distress of the infant. Table 3 displays results from epidem iologic studies in all five areas previously listed. In terms of the effects of glucose intolerance or gestational diabetes and resulting large for gestational age infants or high birth weight infants on C-section, Haram, Pir honen, and Bergsjo reviewed current literature, and based on results recommended that C-sect ion only be performed in infants suspected of weighing more than 5000 grams, or very high birth weight infants [146]. Persson and Hanson report a 60% increase in Caesarean section births for women diagnosed with gestational diabet es [147], and Maymon et al. also recommend C-sections for multiparous wom en with more than six previous births [148]. In terms of low birth weight infants, C-section increases survival, especially for breech infants and is re commended by Jain, Ferre, and Vidyasagar [149]. Finally, maternal obesity and fe tal complications during childbirth are positively associated with increas ed C-section and infant survival.

PAGE 67

31Table 3 Literature review of epidemiologic studies of risk factors of and problematic outcomes associated with Caesarean section Disease/Condition Date of Study Author/s Study Design Measure of Association 95% Confidence Interval+ Glucose Intolerance 2003 Ostlund et al. [150] Case-control OR 1.9 1.2-2.9 2001 Xiong et al. [104] Retrospective Cohort OR 1.1 1.1-1.2 Large for Gestational Age or High Birth Weight 2003 Ostlund et al. [150] Case-control OR 7.3 4.1-12.7 2001 Xiong et al. [104] Retrospective Cohort OR 1.1 1.1-1.2 1997 Jardim et al. [151] Case-control N/A N/A 1994 Aucott et al. [152] Case-control N/A N/A 1993 Lawoyin [153] Prospective Cohort N/A N/A Small for Gestational Age or Low Birth Weight 2001 Teberg et al. [154] Retrospective Cohort N//A N/A 1998 Jain, Ferre, and Vidyasagar [149] Retrospective Cohort OR 5.1 4.0-6.4 Maternal Conditions Obesity 2003 Bo et al. [142] Prospective Cohort OR 1.5 1.0-2.2 2001 Lu et al. [143] Retrospective Cohort OR 1.6* 1.4-1.8 Fetal Distress 2003 Maayan-Metzger, Mazkereth, and Kuint [130] Case-control OR 4.9 1.7-13.8 2002 Kjos, Berkowitz, and Kung [155] Case-control OR 2.2 2.0-2.3 2001 Sutton et al. [131] Case-control OR 3.7 2.0-6.5 1998 Albrechtsen et al. [156] Retrospective Cohort OR 5.9** 5.6-6.2 Total Number of Studies: 15 *Odds ratio for the study years 1995-1999. **Odds ratio for the study year 1994. +N/A refers to not applicable and refers to studies using incidence, prevalence, or univariate analyses. In addition to the predictors of C aesarean section births previously discussed, other factors precipitate a ph ysician’s decision to perform a C-section

PAGE 68

32 delivery. Tong et al. examined the perinatal outcomes among physicians who had high, medium, and low C-section deliv ery rates [157]. Among low rate practitioners, the risk of intracranial hemorrhage in infants was higher compared with medium rate practitioners (OR 1. 53; 95% CI 1.07-2.19). High rate practitioners were more likely to perform a C-section for all major indications, and infants delivered by this group had an over all lower risk of mortality [158-160]. The reason cited for the difference in rates between practitioners is physician style of practice; a predictor that is extremely difficult to quantify. As evidenced in this literature review, an association exists between having a high birth weight infant and r equiring a C-section during delivery. Caesarean section births are associated, however, with glucose intolerance both in the presence and absence of a diagnosis of gestational diabetes. Since glucose intolerance is associated with Csection births, it is plausible that increased urine sugar levels which occur prior to and during hyperinsulinemia are associated with Caesarean section deliveries. It follows that C-section deliveries are also associated with high-weight bi rths. Ideally, a single causal pathway beginning with the proposed socio-cultur al factors and ending with C-section should be supported by this research. The pathway is described below: Socio-cultural factors Increased urine sugar levels High birth weight infants C-section births The Results Chapter addresses the applic ability of this causal pathway, and findings are supported by the Stru ctural Equation Modeling Chapter.

PAGE 69

33 2.3 Theoretical Framework Current theory in epidemiology includes a synthesis of economic and social perspectives. Much of the current work in epidemiologic theory has been developed by Krieger, Kawachi, and Berk man. Krieger outlines three major frameworks for these perspectives, t he psychosocial, social production of disease/political economy, and ecosocial [161]. She discusses the psychosocial framework in terms of it s history, beginning with ear ly associations between stress and disease, stress and the environment leading to disease susceptibility, and ending with the merging of this pers pective and political economy. The political economy framework, as outlined by Krieger, introduces the concept of an individual’s physiologic movement from exposure to disease to exposure (i.e., stress to disease to stress and vice/versa), and adds the layer of ecologic factors affecting individuals such as social ineq uality (e.g., lack of access to resources and health care) and politics. The ecosoc ial framework is a synthesis of the former frameworks with the addition of mu ltiple layers mediating health and disease states among individuals and soci al groups. Krieger metaphorically alludes to the image of a “fractal bus h” with each branch intertwined with all others. Krieger’s ecosocial framework is composed of four main parts: embodiment; pathways of embodiment; cumu lative interplay between exposure, susceptibility, and resistance; and acc ountability and agency. Embodiment is defined as the inseparable natur e of the biologic, social, and material within each

PAGE 70

34 individual. States of health and disease are directly tied to ecologic and social factors affecting individuals, social groups and societies. Embodiment occurs at the individual level, or at the micro (individual) level. The pathways of embodiment are composed of personal an d social history, politics, and evolutionary history. Each pathway to disease or health is mediated by these sociologic and ecologic factor s. Again, any pathway to disease may occur at the individual or micro level. The cumulati ve interplay between health and disease is expressed and measured through each pathway of embodiment at all levels, the individual or micro, the group or mezzo and the societal or macro. The cumulative interplay, therefore, is an ex tension of measuring embodiment at the individual level. By identifying the e ffects of embodiment at group and societal levels, the impact of disease and pathways to disease can be measured in the context of the population. T herefore, ecologic effects of disease are measurable. Accountability and agency involves both the epidemiologist becoming aware of epidemiologic limitations, and understanding the limitations of interpretation of each measure through bio-medically defined definitions of exposure and disease. The ecosocial framework is an attempt to address the current limitations in explanation of epidemiologic research. While epidemiologic studies follow a rigorous methodology, many do not attemp t to hypothesize about associations outside of the exposure disease pathway. Nor do they measure factors indirectly influential in disease coping mechanisms and susceptibility. The ecosocial framework allows the epidemio logist to consider secondary factors related to

PAGE 71

35 societal concepts and influences as nec essary in the exposure and disease pathway. As part of this dissertation, the ecosocial framework is utilized to explain the relationships bet ween each component of the model4, interaction between ethnicity and specific parts of t he model, and the physiologic effects of defined psychosocial stressors on pregnancy and birth outcomes. 2.4 Application of the Framework Berkman and Glass present a theoret ical model robust enough to apply to all four components of the ecosocial fr amework [162, 163]. The model is a synthesis of current social structural influence, and support theories. Figure 5 outlines the direct application of the fr amework and portions of the theoretical model to the hypothetical model in this study. 4 See the Introduction for a description of the model.

PAGE 72

36Figure 5 Integration of framework, theory, and model based on Kreiger, Berkman, and Glass’ current work in social epidemiology The macro and micro levels depic ted in the figure are ecosocial measurements of embodiment of disease at the societal and individual level. The macro level measures include assessm ent of the societal structure through data collected on ethnicity as a crude m easure of culture, and analysis of the socioeconomic structure by focus on postindustrial nations only. Specifically, the study focuses on lower income women in the post-industria l era. The micro levels are measurements of social s upport and influence on low-income pregnant women. Perceived social support is ke y to buffering stressors in the embodiment pathway, while social influences such as marital status and autonomy may cause increased psychosocial stress during pregna ncy leading to birth complications. Macro Level Micro Level Pathways Outcomes Social Structural ConditionsCulture -Interaction of ethnicity Socioeconomic Structure -Focus on Post-Industrial NationsPsychosocial MechanismsSocial Support -Perceived Support Social Influence -Marital Status and AutonomyPsychological-Depression -Verbal Abuse -Pregnancy WantednessPhysiologic-Work Stress -Physical AbuseHigh Urine Sugar Levels High Birth Weight Babies Cesarean SectionAccountability and Agency EmbodimentPathway To Embodiment Cumulative Interplay (all levels)

PAGE 73

37 The pathways to embodiment include both the psychological and physiologic. The psychological pathway focuses on increased depression, the presence of verbal abuse, and the wantedn ess of a pregnancy. The physiologic pathway contains measures of increas ed physical work stress and the presence of physical abuse. All of these factors produce a pathway that in conjunction with the micro and macro level measures of embodiment, increase or decrease risk of problematic outcomes. The cumulative interplay between the exposures or predictors, susceptibility, and resistance is assess ed when measuring the influence of the pathways on the outcome measures of higher urine sugar levels, high birth weight babies, and increased probability of Cesarean sect ion. Also, the micro and macro levels of embodiment are in cluded in an analysis of the outcome measures to examine which specific level factors of embodiment are most influential. Accountability and agency, the f ourth part of t he framework will be commented on throughout the analysis and interpre tation of the results. It is the epidemiologist’s responsibility to recognize both the methodological and ecosocial limitations of research. As such, accountability will be fully addressed in the discussion of results. The theoretical model out lining the integration of the ecosocial framework, social structure, integration, and support theory, and this study’s model encompasses only those data selected for study. While all four major

PAGE 74

38 components of the ecosocia l model are addressed, components of Berkman’s and Glass’ theoretical model not relev ant to this research are excluded.

PAGE 75

39 CHAPTER 3 METHODOLOGY 3.0 Introduction This study assesses the affects of psychosocial and physical factors on pregnancy in relation to an intermediate out come (i.e., high urine sugar levels) and problematic birth outcomes (i.e., hi gh birth weight and C-section). The prospective cohort data set is used to assess the five primary hypotheses5. This chapter discusses the external validity of the sample, describes the variables for the proposed model, defines possible conf ounding factors, and includes power estimation and possible limit ations of the study. The data set consists of 506 pregnant women prospectively interviewed during pregnancy, by trimester, with bi rth outcome data collected from medical charts. The interviews were conduct ed for an NIH study on “Psychosocial and Physical Stressors in Low Birth Weight ” (Grant # 5-R29-HD29559, Tuscaloosa, AL), and for an evaluation of a Human Resources and Service Administration (HRSA) funded Healthy Start site (Grant # 5MJC-0186 32-02-0, Mobile, AL). Both samples followed similar data colle ction protocol. Each participant was interviewed during her initial prenatal vi sit at the local health department or Medicaid waiver clinic (ranging from 120 weeks gestation). A second interview 5 See the Introduction chapter fo r the five primary hypotheses.

PAGE 76

40 was conducted during the third trimester (between 28-40 weeks gestation), and birth data were collected within two to six m onths after childbirth. All participants were Medicaid waiver recipients with a cumulative total household income below $30,000 (contingent upon total number of h ousehold members and total number of legal-aged working adults), and no hist ory of prior high-risk medical factors such as Type II Diabetes, chronic hypert ension, heart disease, or genetic diseases. Women who were pregnant with twins or experienced spontaneous abortion prior to 21 weeks gestation were excluded from each study. Each data set had low attrition rates (Birth Weight Study <3%; Healthy Start Evaluation <10%). Recruitment of participants fo r the two samples was similar based on location of interview (local Health D epartment Prenatal and Family Planning Clinic). In addition, all va riables of interest were collected using the same questions and scales in each sample during the same scheduled interview (1st trimester and 3rd trimester)6. Specifically, recruitment of partici pants occurred in the clinic for both samples. While multiple clinics were used in the Birth Weight Study, only the Mobile County Health Department was utiliz ed for recruitment of participants. In both samples, women were approached afte r check-in, but prior to examination by the medical resident or nurse prac titioner. After reviewing the potential participants’ charts in detail for inclusion criteria and receiving informed consent, participants were interviewed while waiting for laboratory result s, counseling, or 6 Excluding assessment of autonomy, pregnancy wantedness, and physical work stress, which were collected at varying times depending on site.

PAGE 77

41 physical examination. For both samples, if an interview was incomplete, it was finished at the following prenatal visit or at the participant’s convenience in the home, at a different location, or over the phone. The same methodology was employed for the third trimester intervie w (final) in both samples. Figure 6 outlines how the data were collected over the course of the pregnancy for each participant. Figure 6 Outline of the interviewing procedures for the combined data set of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 The primary differences between the two combined samples were the distribution of age and ethnicity (see Table 4). Age ranges for the Birth Weight study included all women between the ages of twenty to th irty-four with an average age of twenty-four. For the Healthy Start Eval uation, participants’ ages ranged from fourteen to twenty with an average age of seventeen. These Conception 14 weeks gestation 28 weeks gestation Childbirth Initial Interview (first prenatal visit) Final Interview Outcome Data

PAGE 78

42 differences in eligibility criteria were grant-specific, with a focus on either financially independent women or t eenage mothers. Combined, both represented all women of childbearing age excludi ng early pubescence (ages nine to thirteen). As a result, both samples combined represent almost all Women of Childbearing Age (WCBA, Health y Start Sample = ages 14 – 20, and Birth Weight Study = ages 20 – 34). Ethni city was available for the NIH sample from Tuscaloosa County only. Demogr aphically, both samples consisted of a higher proportion of Black com pared with White participants. Table 4 Distribution of age and ethnicity by site of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Primary Hypotheses Sample Variable Name Birth Weight Sample (N = 397)* Healthy Start Evaluation Sample (N = 109)* Combined Sample for Proposed Study (N = 506)* Age Range 20-34 (years) 14-20 (years) 14-34 (years) Mean 23.96 (years) 17.02( years) 22.5 (years) Ethnicity** Black 57.2 NA 57.2 White 42.8 NA 42.8 *Presented as a proportion unless otherwise specified. **Excluding the proportion missing. Based on the small number of participant s listing an ethnicity of other, only Black and White participants were ut ilized in analysis. As a result, generalizability is limited to Black or White women of childbearing age who receive Medicaid medical coverage during pregnancy, and subsequently have a cumulative income not exceeding $30,000 or 185% of poverty level.

PAGE 79

43 3.1 Issues of external validity External validity, or generalizability of results is discussed in relation to the demographic similarities between the tw o counties where women were sampled, and in relation to the state of Alabama. Results of this dissertation are addressed in terms of the populations to wh ich findings are relevant; specifically, to Black and White women residing in the Deep South. 3.1.1 Comparability of both cohorts Both Tuscaloosa and Mobile counties are located in different areas of Alabama, and demographically, Mobile Coun ty contains a proportionately larger population than Tuscaloosa Coun ty. The population of M obile County includes approximately 9% of the total population of Alabama, while Tuscaloosa County comprises 4%. Tuscaloosa County consis ts of approximately 4% of the total number of live births for Alabama, whil e Mobile County comprises more than double that amount at 10%. As exemplified in Table 5 [9], the proportional distributio ns of specific reproductive characteristics by ethnici ty clearly differ from the state. Demographically, Alabama is roughly 71% White and 29% Black/other. In comparison, Mobile County is 63% Wh ite and 37% Black/other, and Tuscaloosa County is 68% White and 32% Black/other [164]. Compared with the state, the distributions of live births and live birt hs by birth weight are higher in both counties among Black women only. The in fant mortality rate among Blacks is

PAGE 80

44 slightly higher in both countie s in comparison to the state, while it varies slightly among White women between counties. While both counties vary by birth out comes and ethnicity from the state, they are similar to each other. The proportion of women who receive less than adequate prenatal care is slightly larger among Black women in Mobile County compared with Black women in Tuscaloos a County, and both are larger than the state proportion. In cont rast to the state, both counties contain a higher proportion of Black/other ethnicities, and as exemplified by the study sample, an even higher proportion is Medicaid elig ible. Therefore, though both counties differ in the distribution of ethnicity co mpared with the state, together, they are demographically comparable.

PAGE 81

45Table 5 Distribution of reproductive characteri stics for the state of Alabama, Tuscaloosa, and Mobile Counties in 1999, U.S. Census Bureau, 2003 All Women of Childbearing Age by County for the Year 1999* Reproductive Characteristics Measures by Ethnicity Alabama Population: 4,447,100 (2000 census) Tuscaloosa County Population: 164,875 (2000 census) Mobile County Population: 399,843 (2000 census) Total Live Births All62,061 2,296 6,216 White41,681 (67.2%)** 1, 393 (60.7%) 3,592 (57.8%) Black/Other20,380 (32.8%) 903 (39.3%) 2,624 (42.2%) Average to High Birth Weight ( 2500 grams) All56,231 2,040 5,558 White38,610 (68.7) 1, 275 (62.5) 3,332 (60.0) Black/Other17,621 (31.3) 765 (37.5) 2,226 (40.0) Low Birth Weight (< 2500 grams) All5,799 256 658 Black/Other3,049 (52.6) 118 (46.1) 260 (39.5) White2,750 (47.4) 138 (53.9) 398 (60.5) Infant Mortality Rate 1997-1999*** All9.8 10.1 12.0 White7.3 5.4 8.2 Black/Other14.8 17.6 17.1 Adequate Prenatal Care (Kessner Index) All48,109 1,569 4,432 White35,107 (73.0) 1, 083 (69.0) 2,913 (65.7) Black/Other13,002 (27.0) 486 (31.0) 1,519 (34.3) Less than Adequate Prenatal Care (Kessner Index) All13,692 721 1,768 White6,397 (46.7) 307 (42.6) 671 (38.0) Black/Other7,295 (53.3) 414 (57.4) 1,097 (62.0) *Data are reported for the year 1999 unless otherwise noted **Proportions are presented in parentheses unless otherwise noted ***Rates are per 1,000 live births Source: State and County Quick Facts: Alabama 2003, U.S. Census Bureau. http://quickfacts.census. gov/qfd/states/01000.html Selected Maternal and Child Health Statistics: Alabama 1999 2001, Montgomery: Alabama Department of Public Health Center for Health Statistics. However, though both are comparable at the county level, due to the specific focus at each site, they may diffe r in terms of distributions of outcome measures or possible predi ctors. The focus of par ticipant selection at the

PAGE 82

46 Tuscaloosa site (Low Birth Weight Study ) was to collect data from a communitybased sample of women; at the Mobile si te (Healthy Start), the focus was on high risk pregnant teenagers. Therefore, due to the differences in ages and risk between the two samples, responses to certain questions may differ (e.g., sensitive issues such as physical abuse). To account for these differences, a variable has been created to identify at whic h site the data were collected. The site variable is used in the asse ssment of confounding, and when found significant, controlled in analysis7. 3.1.2 Generalizability of findings As stated in the introduction, resu lts of the current study are only applicable to southern Black and White women of childbearing age who have a cumulative income of less than $30,000 per y ear. In terms of generalizing to all poverty-level women in the South, the issue of selection bias must be addressed. In an effort to recruit as many wo men as possible into both studies, a nonprobabilistic consecutive samp ling scheme was used. In essence, each time a woman entered the County Health Departm ent or local participating Medicaidwaiver clinic, her chart was first reviewed for eligibility crit eria, that is, age and fetal gestational age. If she met the criteria for participation the interviewer approached her to receive informed c onsent. Therefore, only women who sought prenatal care in the first half of their pregnancies ar e included in the study. However, as exemplified in t he table above, the proportion of women 7 See the results chapter for instances when site is included as a confounding factor in analysis.

PAGE 83

47 receiving adequate (>8 prenatal visits) or less than adequate prenatal care (1-8 prenatal visits), is similar to the stat e proportions. Therefor e, results are only generalizable to women who seek prenatal care prior to the mid-point of pregnancy. As cultural practices, although simila r in the United States among specific ethnic groups, differ from regi on to region, generalizability is severely limited. For example, dietary practices differ from geogr aphic region to region as well as from ethnic group to ethnic gr oup. Since both samples are no more than 450 miles apart and located in the same regi on (e.g., western Alabama), though comparable to each other, ar e not comparable to the entire United States. At best, results are applicable to states includ ed in the grouping of the “deep south” such as Alabama, Mississippi, Georgia, Tennessee, and Louisiana. At worst, results will be generalizable only to Bla ck and White women in Alabama. 3.2 Issues of internal validity Internal validity is addressed in t he following section with a discussion and description of the variables in analysi s including predictors and outcomes, the specific tests used in assessing each hypothesis, and the structural equation modeling proposed to compliment inferential findings. 3.2.1 Variables in Analysis The primary focus of the study is on psychosocial factors in relation to development of birth complications among Black and White ethnicities. Specifically, social support, depression, ma rital status and issues of autonomy,

PAGE 84

48 pregnancy wantedness, and verbal abuse are included in the analysis. Physical factors such as work strain, and physical abuse are also included (see Appendix A). Most factors hypothesiz ed to predict high birth weight and Caesarean section are assessed during the first half of the pregnancy and again during the third trimester of pregnancy. Specific risk fa ctors such as social support, abuse and pregnancy wantedness are measur ed as either present or absent throughout the course of the pregnancy for t heir affect on the outcome me asures. All factors are then analyzed at both time periods separately during pregnancy (1st and 3rd trimesters) to determine when throughout pregnancy they are most influential. Finally, scale measures such as depr ession and physical work strain are assessed for change between trimesters and analyzed for affects on the outcomes. Also, possible confounding fa ctors such as age, body mass index, educational level attained, pre-pregnant we ight, interview site, total number of pregnancies, total number of live births previous Caesarean section, total number of abortions and miscarriages, to tal number of pr emature births, gestational age of the infant at birth, excessive weight gain during pregnancy, lack of prenatal care, and substance abus e are controlled in analysis (see Appendix B). The primar y outcome measures of in terest include urine sugar level, high birth weight, and Caesarean section (see Appendix C). While the secondary objective of the study is to assess when during pregnancy associated risk factors most influence birth outcomes, missing data affects such an analysis. In order to analyze risk factors throughout pregnancy,

PAGE 85

49 they must have been measured during both ti me points. Approximately 14% of women were missed during the third tr imester of pregnancy (71/506) reducing the sample size for the third trimester portion of the analysis. In addition, specific measures are available only in the third tr imester such as the scale of autonomy and pregnancy wantedness. The analysis of the above variables is limited by data not collected at both time points. A second major concern in relation to the impact of when and how the risk factors affect the problematic outcomes is the actual stru cture and measurement of each measure. Each pr edictor is defined as either continuous or categorical, and the composition of each is located in each respective appendix (A, B, or C as listed above). The structure of the c ontinuous predictor variables, excluding other potential confounders, is debat able based on changes made during the piloting phases of the study8. The established scales used in this dissertation were modified based on pilot study results of participant interpretation and content analysis. Such methods for modifying psychometrically sound scales to fit a specifically defined population in order to evaluat e more culturally relevant predictors have been suggested by Dressler et al [165]. Prior to inferential analyses, all scales are tested for reli ability and Cronbach’s alpha used as a criterion for a more extensive evaluation of the effectiveness of each modified scale9. 8 Further discussion is located throughout the chapter and in the Discussion Chapter. 9 See the Reliability of Scales section of the Results Chapter for further discussion.

PAGE 86

50 Psychosocial and physical risk factors A six-question scale assessing both perceived and received support from the current male partner or the mother of the participant defines social support for the proposed study10. The received support from the current partner is called the partner support scale, and the support received from the mother of the participant is called the mother’s support scale or t he surrogate measure for familial support. As perceived support emotionally impacts women more than received support [162], the scale was modified to asse ss only perceived support in the Healthy Start Evaluation. However, the semant ic difference between the two studies is minimal as evidenced in Appendix A. T he scale is divided into emotional and instrumental support, three questions assessing emotional support and three questions assessing instrum ental support, respectfully. Social support is measured both at the first interview and seco nd interview (N = 506 first trimester; N = 432 third trimester). Depression is assessed in the st udy utilizing the CESD (Center for Epidemiologic Study Depre ssion scale; see Appendix A for full description). When pilot tested prior to use in either study, it became apparen t that there were similarities between the physical sym ptomology of depression and sickness associated with pregnancy. More than 60% of the women interviewed were characterized as clinically depressed bas ed on preliminary results. The scale was then ethnographically modified to fit the population under study. Words 10 Scales are presented prior to testing of reliability.

PAGE 87

51 were replaced with terms understandable at the third grade level, and the ten items physically assessing depression we re omitted (e.g., insomnia, stomach cramps, headache, etc). In stead, the ten items asse ssing emotional state and depression were used. Depression was assessed during both interviews in the first and third trimesters of pregnancy (N = 506 first trimester; N = 432 third trimester). Personal autonomy is assessed only in the Birth Weight Sample during the third trimester interview. Autonomy is not measured in the Healthy Start Evaluation sample. As a result, the tota l sample size in analysis is 397 rather than 506. The scales consis ts of eight questions fram ed to measure a woman’s sense of independence, four questions to measure a woman’s feelings about authoritarianism, and is only assessed durin g the third trimester of pregnancy. Marital status is available for all wom en in the data set. Marital status is assessed in two manners, from the medi cal record of each participant, and during the interviewing process. The in terview responses are compared with the medical records (used as a tool for ve rification) and any differences reconciled prior to analysis. Women in the study are categorized as never-married, evermarried single (e.g., separated, divorc ed, widowed), or married (N = 506). Pregnancy wantedness is defined as whet her a woman want s, plans, or intends to get pregnant. Due to the multip le meanings of wanting, planning, or intending, wanting will be used specifica lly in the current study. Wanting a pregnancy is assessed during t he third trimester of pregnancy (N = 432) as both

PAGE 88

52 a current question, “Do you want this baby now?” and as a reflective question, “When you first found out you were pregnant, did you want the baby?” Abuse is categorized as verbal or physical within the sample (see Appendix A). Verbal abus e is defined as anyone saying hurtful things to the participant prior to and during pregnancy. Physical abuse is assessed by a question addressing whether t he participant has ever been hit or hit during her current pregnancy. Further, if a parti cipant admits to receiving physical abuse during pregnancy, she is asked to show the area of abuse on a body map. Both types of abuse are combined to create a measurement of comprehensive abuse among all participants (N = 506 first trimester; N = 43 2 third trimester). Physical strain is assessed in terms of work strain at a job for pay. The Karasek scale consists of seven questions covering topics such as physical and emotional strain, and rest breaks at work Questions categorizing type of work and number of hours worked per week ar e also included in work stress assessment. Work stress is assessed dur ing both interviews only for women who worked for pay during the pregnancy (N = 230 first trimester; N = 109 third trimester). Confounding factors include age11, body mass index, education level attained, pre-pregnant weight, interview site, total number of pregnancies, total number of live births, previous C-se ction, total number of abortions and miscarriages, total number of premature bi rths, gestational age of the infant at 11 See Demographics section for description of age in the samp le. Due to data constraints, specific confounding factors assessed during data collection are controlled in analysis.

PAGE 89

53 birth, excessive weight gain durin g pregnancy, lack of prenatal care, and substance abuse during pregnancy. All t he former characteristics are possible confounders of increased urine sugar leve ls, high birth weight, or Caesarean section. Age, a factor that can a dd to a woman’s risk during pregnancy, is calculated by subtracting the participant’s date of birth from the initial interview date. Body mass index, a factor that also affects birth weight and a physician’s decision to perform a Caesar ean section, is determined using each participant’s pre-pregnant weight and height. Educatio nal status affects how participants understand information provided at each prenatal visit and the type of job for pay available to participants during pregnancy. Educational level attained is categorized as less than a high school education (middle school or equivalent), a high school education with no degree, a high school education with a degree/GED, and any college. Pre-pregnant weig ht is defined as either weight at first prenatal visit or a ve rbal account from the pati ent during the first prenatal visit. As previously noted in the external validity section, the site at which the participants were enrolled may confound the associations between the predictors and outcomes due to the focus of each i ndependent grant objective. Site is defined as either Tuscaloosa, AL (Birth Weight Study) or Mobile, AL (Healthy Start Study). The total number of pregnancies may affect analysis due to the physiologic differences between women who are experiencing their first pregnancy compared to women who have had multiple pregnancies. The same

PAGE 90

54 logic follows for the number of previous live births. In addition, previous Csections play a role in a physician’s decision to perform a subsequent C-section and may also confound associations involv ing current C-secti on as the outcome measure. The number of previous premature births and either elective or spontaneous abortions impact the current pregnancy as well. Finally, current analyses of birth weight are affected by t he gestational age of the infant at the time of birth. Therefore, gestational age is assessed for its confounding effects. The prior number of pregnancies carried to term may affect both a woman’s pre-pregnant and her baby’s birt h weight. Women with multiple pregnancies are at higher ri sk of being overweight, and t herefore, predisposed to development of pregnancy complicati ons. Average weight gain during a pregnancy is estimated at twenty-five to thirty-five pounds for women of average pre-pregnant weight [107]. For overweight women, weight gain is typically less as determined by the attending physician. Excessive weight gain is defined as the weight gained between the first and last prenatal visit, and is contingent upon pre-pregnant weight. If the first prenatal visit occurs at twenty weeks gestation, the pregnancy is already half completed. Wi th only the patient’s estimate of prepregnant weight available (r equiring those patients at twenty weeks gestation to recall their weight almost five months prio r), it is not reliable to estimate probable weight gained during the firs t half of the pregnancy. As a result, only the amount of verifiable weight gain will be used in analysis. Adequate prenatal care is defined as six or more prenatal visi ts during pregnancy. Any participant

PAGE 91

55 attending less than six visits may have pregnancy complications unknown to the attending physician or nurse practitioner. Substance abuse is defined as alcohol consumption and drug use dur ing pregnancy, as well as use of marijuana and cocaine/crack. Smoking is not included in this analysis due to its strong association with the birth of low-wei ght babies and lack of evidence for the association with high-weight babies. Intermediate outcome Recorded urine sugar level, the intermediate outcome variable, is predicted to be in the pathway between the specified risk factors and the primary outcomes. Urine sugar levels are reco rded at each prenatal visit. The highest urine sugar reading of each participant is included in the data set. A binomial measure of urine sugar spill is used to init ially assess the effects of any sugar in the urine. Subsequent analysis includes creating an ordinal measure of urine sugar based on the dipstick test perform ed at each clinic, and an analysis of higher urine sugar spills only. Those par ticipants with readings of 1+ or higher are at higher risk for development of ges tational diabetes or other complications, while those with trace readings are monitored throughout pregnancy for any increase in urine sugar spill. Theref ore, participants are grouped as no spill (‘none’, ‘no detectable level’), trace (‘low urin e sugar spill’), or 1+ or greater (‘high or higher urine sugar spill’). Throughout the study, women with no urine sugar spill are referred to as “No spill” or “N o Detectable Level” of spill, women with trace readings are “Possible cases” or “Low urine sugar spill,” and women with

PAGE 92

56 readings of 1+ or higher are “Probable case s” complicated by urine sugar levels or “High or Higher urine s ugar spill”. Also, time of baseline measurement and highest sugar reading are measured to ensure assessment of exposure (predictors) prior to outcome12. Main outcomes The main birth outcome variables of in terest include high birth weight and Caesarean section. High birth weight is defined as any birth weighing more than 4000 grams [108]. Birth weight is measur ed using two strategies. First, it is dichotomized to high birth weight and other birth weight to assess the effects on high birth weight infants alone. Second, the continuous measur e of birth weight is used as the outcome measure to more succinctly measure the effect of each factor on incremental change in birth weight. Caesarean section is defined as the surgical removal of a fetus from t he uterus and is a binom ial measure in the study (compared with vaginal births). 3.2.2 Statistical Tests The overall analysis for the study is the testing of the five hypotheses and the organization of significant results in prototypical order of occurrence and impact using structural equati on modeling. Descriptive statistics are reported for each variable within the study, and based on the specific statistical test assumptions of each mode ling procedure, all variab les are assessed. The primary analysis is addressed in order of hypotheses below. Specifically, based 12 See Descriptive statistics for more details.

PAGE 93

57 on distribution, chi-square, t-tests, linear and logist ic regression are used to assess Hypothesis 1, logistic regression is used for Hypothesis 2, multiple and logistic regression for Hypothesis 3, l ogistic regression with interaction for Hypothesis 4, and in Hypothesis 5, multip le and logistic regression models with an interaction term are calculated. Confounding factors are independently assessed in the descriptive statistics section, and based on the strength of the association with the outcomes and pr edictors, added to each model for adjustment. Hypothesis 1 The outcomes of interest in Hypot hesis 1 are high birth weight and Caesarean section. The predictor variable is urine sugar level. Excluding birth weight, all variables in the primary analysis are categorical. A t-test is used for initial assessment of birth weight and urine sugar levels, and a chi-square for the categorical measurement of high birth weight. Mult iple regression models are utilized for birth weight, the predict or urine sugar, and possible confounding factors. For the binomial measure of high birth weight, a logistic regression model is used including all confounding fact ors. A logistic regression model is used in assessment of Caesarean sect ion and urine sugar levels including possible confounding factors. As previously stated, it is hypothesized that higher urine sugar levels increase the risk of hi gh birth weight babies and the likelihood of C-section.

PAGE 94

58 Hypothesis 2 Urine sugar levels are the outcome of interest for Hypothesis 2. Hypothetically, psychosocial and physical factors such as social support, marital status and autonomy, pr egnancy wantedness, depression, abuse, and physical work strain physiologically affect women during pregnancy and may impair glucose processing. When urine sugar is measured as present or absent, a binary logistic regression model is used. Polytomous logistic regression models (specifically multinomial and ordinal) are used to assess whether psychosocial or physical factors increase urine sugar leve ls over the course of a pregnancy. Confounding factors are c ontrolled in each analysis. Hypothesis 3 Multiple outcome measures are utiliz ed in assessment of Hypothesis 3. The likelihood of birth outcomes such as high birth weight and Caesarean section are hypothesized to increase among wo men experiencing psychosocial and physical strain during pregnancy. Again, these factors are composed of social support, marital status, autonomy, pr egnancy wantedness, depression, abuse, and physical work strain. For the measures of birth weight and high birth weight multiple and logistic regression model s are used. For the assessment of Caesarean section, a logistic regre ssion model is utilized controlling for confounding factors.

PAGE 95

59 Hypothesis 4 Hypothesis 4 is similar to Hypothesis 2 with the inclusion of the interaction of ethnicity. Binary and polytomous logi stic regression models are used during analysis to test whether psychosocial and physical factors associated with high urine sugar levels differ among White and Black women. The relationship between all predictive factors and ethnicity is also assessed. A modified alpha level (0.20) is used to assess the asso ciation between the interaction term and the outcome due to lack of power for the analysis13. Hypothesis 5 The interaction of ethnicity is assessed for Hypothesis 3 in Hypothesis 5. Multiple and logistic regression models ar e used to determine if ethnicity interacts with psychosocial and physica l factors in the birth of high-weight babies. A logistic regression model is used to det ermine whether ethnicity interacts with stressors and Caesarean section at birt h. The relationship between all psychosocial and physical factors and ethnici ty are assessed, and due to lack of power, a modified alpha-level is used. Comprehensive Modeling After inferential testing of Hypothes es 1 through 5, a model is created using structural equation mode ling techniques to display the prototypical order of negative events throughout pregnancy and birt h beginning with psychosocial and physical factors, testing of sugar spill in urine, and ending with problematic birth 13 See power analysis section for further detail.

PAGE 96

60 outcomes. Since structural equation m odels do not support interaction, analysis of ethnicity includes the creation of s eparate models to represent the experiences of both Black and White women. Only fa ctors significantly associated with the outcomes of interest from the inferential analysis are included in the structural equation models. As in the introduction, the proposed model is represented in Figure 7 below. The locations of s pecific hypotheses are denoted by the reference ‘H.14’ The specific type of structural equation modeling used in analysis is confirmatory factor analysis followed by path analysis of latent variables. Confirmatory factor analysis encompasse s the evaluation of the relationships between predictors, and how those predict ors correlate to produce a ‘latent’ construct. Latent constructs are vari ables created during analysis by groups of variables collected for data analysis. In this dissertation, referring to Figure 5 from Chapter 2, both the psychologica l and physiologic pathways are latent constructs composed of predictors co llected for study. The psychological pathway contains depression duri ng pregnancy, pregnancy wantedness, and verbal abuse during pregnancy; while the physiologic pathway contains physical work strain and physical abuse during pr egnancy. The confirmatory factor analysis aids in evaluation of the adequacy of the latent constructs. If the associations between the collected variables and latent constructs are significant, the next step is a path analysis of latent variables. The path analysis is an 14 See aims section for full description.

PAGE 97

61 evaluation of the causal pathways bet ween the collected variables, latent variables, and outcome variabl es (e.g., urine sugar leve ls, high birth weight, and Caesarean section) of interest. Chapter 5 outlines the strategy for the structural equation models in greater detail. Figure 7 Proposed causal pathway model of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 3.3 Power Analysis To ensure the statistical validity of results, the ability of each test to adequately assess an association must be determined. When using a fixed sample size, power is calculated based on the type of test (e.g., alpha level, effect size, standard deviation, etc.) and type of variable (e.g., categorical, continuous, etc.). The following power anal yses are described in terms of worst to best-case scenario, and the subsections are discussed by order of each study Ethnicity Depression Marital Status and Autonomy Social Support Physical Work Stress Physical and Verbal Abuse Urine Sugar Reading High Birth Weight Cesarean Section H1 H2 H3 H4 and H5 Pregnancy Wantedness

PAGE 98

62 hypothesis. Each calculation is based solely on the sample size, relevant national prevalence data, and distributions of the sample (N = 506; unless otherwise indicated), and are cr eated using nQuery 3.0 [166]. In epidemiology, in order to assess t he strength of an association, chronic disease epidemiologists plan studies adequately powered to assess a minimum of a two-fold increase in risk. Historically, more emphasis has been placed on study results that indicate a 200% increase in risk both statistically and clinically [167]. Using the current sample size, pow er calculations for a two-fold increase in risk are shown with the minimal difference in risk between groups at approximately 80% power. It is antici pated that a two-fold (200%) increase in risk between the exposed groups is un likely, and therefore, the minimum difference provides a realistic reference for increasing risk with exposure. When national prevalence rates are available, they are used as baseline measures for the non-exposed group, and proportions of the baseline measure are used to calculate adequate power. When no prevalen ce data are available, the standard deviation of the mean of the specific m easure is used to calculate differences between groups. Due to the impact on analysis for some of the outcome variables (e.g., birth weight as a cont inuous measure and ordinal urine sugar measures), those outcome measures r equiring the most power are shown as well those measures requiring the l east power. Power analyses for the binary outcome measure of urine sugar le vels are omitted for brevity.

PAGE 99

63 3.3.1 Hypothesis 1 The purpose of Hypothesis 1 is to analyze the association between urine sugar levels and the outcomes of high birt h weight and Caesarean section. The power of the polytomous logistic regre ssion (outcome birth weight) is discussed below in Table 6, and Table 7 represents of the logistic regression (outcome Caesarean section). Power for birth weight is based on predicted proportional differences between birth weight and urine sugar level. The 1999 national high birth weight prevalence rate is cons idered to be the minimum rate expected among otherwise healthy women, and is therefore used as the rate of outcome among women with no detectable urine suga r spill (9.9% births >= 4000 grams [1]). Power is assessed by taking 62% and 200% of the baseline measure to create the levels of possible and probable ca ses. Sixty-two percent (62%) of the baseline measure is the minimal differenc e required for the polytomous logistic regression to be adequately pow ered (62% = 6.14). For each urine sugar level, 6.14 is added to create an incremental incr ease in risk (No spill = 9.9%; low = 9.9 + 6.14 = 16.0%; high = 16.0 + 6.14 = 22.2% ). As shown, a 62% difference is sufficient for detecting a difference between the three urine sugar level groups in terms of high birth weight.

PAGE 100

64Table 6 Power calculation of urine sugar levels and estimated high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Analysis of Variance – High Birth Weight Estimated High Birth Weight Urine sugar levels* N 62% increase from baseline 200% increase from baseline None (0)390 9.9% 9.9% Low (Trace)47 16.0% 19.8% High (1+ or higher)69 22.2% 29.7% Power 81% 98% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). The power calculation for Caesarean sections is based on the national prevalence rate for 1999 (22% Caesarean se ctions [5]). Again, the rate is assumed to be the minimum expected among otherwise healthy women, and is used for the women with no detectable urin e sugar spill. Differences among the three groups are initially calculated using the baseline measure of 22% and creating the two higher risk gr oups. When calculating powe r, it is assumed that at a minimum, urine sugar le vels will impact the more at risk groups by a factor of 41% of the baseline measure. In column 1 Table 7, 36% of is a dded to the higher risk groups (7.9%) in order to predict the minimum impact of urine sugar spill (e.g., No detectable sugar spill = 22%, low = 22 + 7.9 = 29.9%, high = 29.9 + 7.9 = 37.8%). In the following column, the high-risk group outco mes are calculated by doubling the risk from the baseline percentage to meas ure a larger impact on urine sugar spill (e.g., No sugar spill = 22%, low = 22 + 22 = 44.0%, higher = 44. 0 + 22 = 66.0%).

PAGE 101

65 In order to achieve sufficient power (~80%) for logistic regression (Caesarean section), each group must vary at least 36% from the baseline rate (national prevalence rate). To detect a signific ant difference between urine sugar level groups, there must be a minimum of a 7. 9% difference between each successive risk group, assuming the sample size is equal to 506. Table 7 Power calculation of urine sugar levels and estimated Caesarean sections of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Logistic Regression – Es timated Proportion of Caesarean Sections Estimated Proportion of Caesarean Sections Urine sugar levels* N 36% increase from baseline 200% increase from baseline None390 22.0% 22.0% Low47 29.9% 44.0% High69 37.8% 66.0% Power 81% >99% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). 3.3.2 Hypothesis 2 Physical stress, depression, physical or verbal abuse, social support, autonomy, pregnancy wantedness, and mari tal status are associated with increased urine sugar levels as described in Hypothesis 2. The power of the polytomous logistic regression in asse ssment of the hypothesis is based on an alpha of .05, and an increas e of the baseline rate (if available) for each successive sugar level (e.g., physical stress, depression, social support, and autonomy). Calculations for abuse questions are based on the lifetime prevalence of domestic violence among women of childbearing years [168].

PAGE 102

66 Pregnancy wantedness calculations are based on the preval ence of unintended pregnancies for 1998 in the stat e of North Carolina [169]. For marital status, power is based on the nati onal proportion of women wh o report themselves as never-married as of 2002 [170]. Increased physical work strain duri ng pregnancy may increase the risk of urine sugar spill. Power for the polytomous logistic regression is estimated in Table 8 below using the mean score on the wo rk strain scale (2.83) as a baseline measure, or the score expe cted in the group wit h no detectable urine sugar spill. In the third and fourth columns, 19% and 200% of the standard deviation (sd = 1.71) from the mean are calculated. In column 3, a 19% increase (1.71*.19 = 0.325) of the standard deviation from t he mean score (none = 2.83; low = 2.83 + 0.325 = 3.15; high = 3.15 + 0.325 = 3.48) is the minimal difference (between groups) adequately powered to test the associ ation. Since the scale is a sevenpoint scale, a difference in 1/3 of a point between groups is more realistic than assuming the highest risk group mean will fall at the upper end of the scale (6.25) as when the risk is doubled.

PAGE 103

67Table 8 Power calculation of estimated physical work strain and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Polytomous Logistic Regression – Physical Work Strain Estimated Physical Work Strain Urine sugar levels* N 19% increase from baseline 200% increase from baseline None390 2.83 2.83 Low47 3.15 4.54 High69 3.48 6.25 Power 82% >99% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). As with physical stress, increased depr ession over the course of a pregnancy may be associated with increased urine sugar levels. In order to estimate power, the 1993 prevalence of depressive disorders among women (aged five and older) is used as the baseline measure (12.0% [171]). The minimal difference between groups that is adequately powered is 58% (Table 9). Since depressive disorders include ma jor depression, dysthymia, and bipolar disorder, 12% is likely an over-estimate. However, a 7% difference between groups is not unreasonable given the link between hormone levels and major depression [171].

PAGE 104

68 Table 9 Power calculation of estimated depression score and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Polytomous Logistic Regression Depression Estimated Score on the CESD Urine sugar levels* N 58% increase from baseline 200% increase from baseline None390 12% 12% Low47 19% 24% High69 25% 36% Power 80% >99% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). In Table 10, the power needed to detect a significant association between physical abuse and urine sugar level is shown. T he lifetime prevalence of domestic violence among women of childbearing age (WCBA) is 25% [168]. Therefore, the nati onal prevalence rate is used as the baseline measure in determining power. The mi nimal difference between gr oups that is adequately powered for analysis is 32%.

PAGE 105

69Table 10 Power calculation of estimated physical abuse and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Polytomous Logistic Regression – Physical Abuse Estimated Physical Abuse Urine sugar levels* N 32% increase from baseline 200% increase from baseline None390 25% 25% Low47 33% 50% High69 41% 75% Power 80% >99% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). In contrast, a decrease in social s upport over the course of the pregnancy may be predictive of higher urine sugar le vels. In order to predict power, the mean score of social support (3.69, st andard deviation = 2.43) is used as the average score of women with no detectabl e sugar spill during pregnancy. The first column in Table 11 is calculated as a decrease in social support of 19% of the standard deviation for each successive higher risk group, and displays the minimal difference between groups in terms of score. A difference of 19% is unlikely, as the standard deviation bet ween groups will most likely exceed 0.462 of a point. Since a 200% increase in t he standard deviation at all three levels would include a score of <0 for the highest risk group, a 200% increase is calculated between the no detectable s ugar spill and highest risk group only. The middle risk group is calculated as a 50% increase in risk.

PAGE 106

70Table 11 Power calculation of estimated lack of social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Polytomous Logistic Regression – Social Support Estimated Lack of Social Support Urine sugar levels* N 19% decrease from baseline 200% decrease from baseline None390 3.69 3.69 Low47 3.23 2.46 High69 2.77 1.26 Power 82% >99% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). A decrease in personal autonomy ov er the course of a pregnancy may also be predictive of higher urine sugar levels. Table 12 displays the power required to detect a significant association between urine sugar levels and decreasing autonomy. The standard devia tion of the baseline measure of autonomy (sd = 1.36) is used to calculat e the differences between urine sugar level groups. If scores differ by at l east 22% of the baseline, the test is adequately powered.

PAGE 107

71Table 12 Power calculation of estimated lack of autonomy and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Polytomous Logistic Regression – Autonomy Estimated Lack of Autonomy Urine sugar levels* N 22% decrease from baseline 200% decrease from baseline None261 4.46 4.46 Low37 4.16 3.10 High50 3.86 1.74 Power 81% >99% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). An unwanted pregnancy may more likel y affect the pregnancy experience for both the mother and fetus. The Pregnancy Risk Assessment Monitoring System for the state of No rth Carolina in 1998 estimat ed that 44% of the term pregnancies were unintended by the Medica id recipient population [169]. No baseline data on pregnancy wantedness exis t for any surveillance system, and therefore, intention will be used as the bas eline measure. In Table 13, for a 21% increase in unwanted pregnancies from baseline the test is adequately powered. Since a true increase of 200% would pr oduce a proportion greater than 0.99 for the highest risk group, a total increase of 200% is calculated between the lowest and highest risk groups only.

PAGE 108

72Table 13 Power calculation of estimated pregnancy wantedness and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Polytomous Logistic Regression – Pregnancy Wantedness Estimated Pregnancy Wantedness Urine sugar levels* N 21% increase from baseline 200% increase from baseline None351 44% 44% Low42 53% 66% High65 63% 88% Power 81% >99% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). Like decreasing autonomy and an unw anted pregnancy, the absence of a stable relationship may affect the out come of a pregnancy and contribute to complications. The 2002 National Survey of Family Growth fr om the Department of Health and Human Services indicates that 28% of all wo men of childbearing age are never-married single women [170]. Therefore, 28% is used as the minimum percentage of women expected to be single in the cohort. As shown in Table 14, the difference between each su ccessive group must be a minimum of 30% to achieve at least 80% power.

PAGE 109

73Table 14 Power calculation of estimated marital status and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Polytomous Logistic Regression – Marital Status Estimated Marital Status – Never-Married Women Urine sugar levels* N 30% increase from baseline 200% increase from baseline None390 28% 28% Low47 36% 56% High69 45% 84% Power 81% >99% *Urine sugar levels are grouped as No detectable readi ng, Low (possible case), and High (probable case). 3.3.3 Hypothesis 3 Hypothesis 3, or the association of psychosocial and physical stressors and birth complications, is assessed by calculating power for each of the outcome measures of high birth weight and Caesarean section. It is predicted that increased physical stress, depre ssion, and abuse, while decreased social support and autonomy, unwanted pregnancy, and never-married marital status are associated with higher birth we ight babies and a higher proportion of Caesarean sections. The unadjusted Pearson correlation is used to determine the power of continuous predictors and birth weight. In Table 15, each of the scales used in analysis and birth weight are corr elated and the estimated power shown (correlations are based on data from the birt h weight sample only [N = 397] as the second sample of birth weight dat a had not been standardized by the time the power was assessed [N = 109]). As i ndicated, the testing of Hypothesis 3 is

PAGE 110

74 underpowered for the continuous variables Since the correlation between the predictors and outcome are weak (ranging from -.042 to .007), the fixed sample size is too small to power the hypothes is test. Regardless of statistical significance, examining the impact of t hese factors on birth weight aids in defining the relationship between soci o-cultural measures and pregnancy, and explains a minimal amount of t he variance in the final model. Table 15 Power calculation of estimated autonomy, physical work strain, depression, lack of social support, and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Pearson Correlation between Continu ous Predictors and Birth Weight Autonomy Work Stress CESD Social Support Birth Weight 0.007 -0.042 -0.005 -0.009 Power 5% 13% 5% 5% Table 16 outlines the power requir ed for testing the association between physical abuse and high birth weight. Us ing a baseline rate of 9.9% of highweight births (3), a 92% differ ence between abused and non-abused women is required for adequate power. The 92% increas e in risk is proportionately close to a two-fold increase in risk as shown below.

PAGE 111

75Table 16 Power calculation of physical abuse and estimated high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Multiple Regression – Physical Abuse and High Birth Weight Estimated High Birth Weight Physical Abuse N 92% increase from baseline 200% increase from baseline No330 9.9% 9.9% Yes176 19.0% 19.8% Power 80% 85% The association between pregnancy want edness and high-weight births is adequately powered when the two groups (w omen who want versus women who do not want their pregnancies) differ by no less than 96% (Table 17). Table 17 Power calculation of pregnancy wantedness and estimated high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Multiple Regression – Pregnancy Wantedness and Birth High Weight Estimated High Birth Weight Pregnancy Wantedness N 96% increase from baseline 200% increase from baseline Yes225 9.9% 9.9% No210 19.4% 19.8% Power 80% 83% A 51% difference between never-married, ever-married, and married women provides adequate power to test the association between marital status and high-weight births (Table 18).

PAGE 112

76Table 18 Power calculation of marital status and estimated high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Multiple Regression – Marital Status and High Birth Weight Estimated High Birth Weight Marital Status N 51% increase from baseline 200% increase from baseline Never-Married155 9.9% 9.9% Single Ever-Married45 14.95% 19.8% Married or Living with Partner 306 20.00% 29.70% Power 81% >99% For the outcome of Caesarean section, power is assessed for each predictor in turn. Table 19 displays the results of power calculations for the association between physical work strain and Caesarean section. The average score on the physical work strain sca le is 3.39, and adequate power for the hypothesis test is reached by adding 33% of the standard deviation (sd = 1.71) to the mean score of women who delivered va ginally. A two-fold increase in risk almost doubles the differences between the two groups.

PAGE 113

77Table 19 Power calculation of estimated physical work strain and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dichotomous Logistic R egression – Physical Work Strain Estimated Physical Work Strain Caesarean Section N 33% increase from baseline 200% increase from baseline No413 2.83 2.83 Yes93 3.39 4.54 Power 81% >99% Increased depression may affect the onset of labor and thus, Caesarean versus vaginal birth. Table 20 displays t he power required to test the association between depression score and C-section. As in Hypothesis 2, the national prevalence rate of depressive disorders for 1999 (12%) is used as the baseline. There must be a minimum of a 99% difference in depression score between women who do and do not receive a C-section for an adequately powered analysis (80%).

PAGE 114

78Table 20 Power calculation of estimated depression and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dichotomous Logistic R egression – Depression Estimated Score in the CESD Caesarean Section N 99% increase from baseline 200% increase from baseline No413 12% 12% Yes93 23.9% 24% Power 80% 81% Table 21 indicates that in order to be adequately powered (80%), women who received a C-section must have been abused by at least 62% of the baseline measure (25%). That is, almo st half the women with a C-section (41%) must have been abused compared with a quarter of the women without a Csection. Table 21 Power calculation of estimated physical abuse and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dichotomous Logistic R egression – Physical Abuse Estimated Physical Abuse Caesarean Section N 62% increase from baseline 200% increase from baseline No413 25% 25% Yes93 40.5% 50% Power 80% 99% In terms of social support and C-se ction, Table 22 disp lays the required differences between support scale scores based on the standard deviation (2.43) subtracted from the mean bas eline score (3.69). As shown, there must be a

PAGE 115

79 difference of at least 62% of the standard error betw een the two groups in order for the test to be adequately powered (81%). Table 22 Power calculation of estimated lack of social support and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dichotomous Logistic Regression – Social Support Estimated Lack of Social Support Caesarean Section N 33% decrease from baseline 200% decrease from baseline No413 3.69 3.69 Yes93 2.89 1.26 Power 81% >99% Power calculations of autonom y are again based on the standard deviation (1.36) of the mean score (4.46; Ta ble 23). In order fo r the test statistic to be adequately powered, t here must be at least a 33% decrease in the standard deviation subtracted fr om the baseline mean score in order to detect a difference between the two groups (81% power). Table 23 Power calculation of estimated lack of autonomy and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dichotomous Logistic Regression – Autonomy Estimated Lack of Autonomy Caesarean Section N 33% decrease from baseline 200% decrease from baseline No413 4.46 4.46 Yes93 4.01 3.10 Power 81% >99%

PAGE 116

80 Table 24 displays the minimal differences required in proportions between those who receive a C-section and those who do not in terms of pregnancy wantedness. In order to be adequately powered (81%), the groups must differ by at least 37% of the bas eline measure (16.28%). Table 24 Power calculation of estimated pregnancy wantedness and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dichotomous Logistic R egression – Pregnancy Wantedness Estimated Pregnancy Wantedness Caesarean Section N 37% increase from baseline 200% increase from baseline No 44% 44% Yes 60.3% 88% Power 81% 99% In examining the relationship between marital status and C-section, the differences in proportions are based on the prevalence of single, never-married women of childbearing age. The data in Table 25 indicate that a 57% increase from baseline adequately powers the test of the association.

PAGE 117

81Table 25 Power calculation of estimated marital status and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dichotomous Logist ic Regression – Marital Status Estimated Marital Status – NeverMarried Women Caesarean Section N 57% increase from baseline 200% increase from baseline No413 28% 28% Yes93 44% 56% Power 80% 99% 3.3.4 Hypotheses 4 and 5 Hypotheses 4 and 5 directly mirror Hypotheses 2 and 3 with the addition of the interaction of ethnicity. As indi cated earlier, the sample is too small for adequately powered analysis of interaction. Ho wever, Hypotheses 4 and 5 are considered exploratory, and if the probability value of the interaction term is less than 0.20, interaction will be consi dered present and further explored. 3.4 Limitations As with all epidemiologic studies, lim itations must be addressed. The “systematic error in the co llection and analysis of data” [172] must be discussed in terms of study design and data collecti on. The major probl ems in conducting a prospective cohort study include disease ascertainment and loss of participants during the study period. In a prospective cohort, the major biases include selection, attrition, interviewer, observa tion/information, to a limited extent recall, and possible misclassification [172]. T he issues of confounding and effect

PAGE 118

82 modification are addressed throughout the chapter with potential confounders included and controlled in analyses, and inte raction specified in two of the five primary hypotheses of study. Selection bias, or the syst ematic, differential enrollm ent of participants in a study, may be identified dur ing the enrollment process or during the analysis phase of the study. Selection bias is comp rised of more specific biases including exclusion, referral, diagnostic, and non-res ponse. In the current study, exclusion bias affected women asked to participate. Since participation is restricted to women at an initial prenatal visit in the first half of their pregnancies, any women beginning prenatal care in the second hal f of the pregnancy or not receiving prenatal care during the pregnancy are excl uded from the sample. Referral bias, or selecting a group of participants based on the reference of an individual or clinic, affected the study. Specific c linics were selected for interviewing. Therefore, only women who received care at clinics chosen for study were approached for participation. In terms of diagnosis, medical records were reviewed for exclusion criteria If errors occurred in the assessment of due date, for example, eligible women may have been excluded. Finally, while there were no non-respondents, there was a small proporti on of refusals (>2%). No data were collected on women who refused to participate in the study. Selection bias may also have occurred in the differing methodologies implemented at the two sites. As stated earlier in the chapter, the Birth Weight Study sample was community-based, em ploying a catch-all methodology.

PAGE 119

83 Women from all area clinics were review ed for participation, and those excluded did not meet study requirements. Howeve r, for the Healthy Start sample, only high-risk teenage mothers were recruited in to the sample. Therefore, the data collected for the high-risk group may be affect ed simply by their increased risk of adverse outcomes. Selection of this s pecific sub-group impacts both external and internal validity. As mentioned earlier in the chapter, attr ition rates were very small for both samples. Minimal data were obtained fo r women who became lost to follow up during the course of the study. Howeve r, no outcome data were available for those participants who became lost. T hey were excluded from analyses. Interviewer bias is of major concer n in the study. All data excluding outcome measures were collected via interv iew. All interviewers participated in a two-day training class, observed more experienced interviewers, audio-taped their first few interviews, and were observed by other staff. Each new interviewer then met with the project director who reviewed technique, gave a list of suggestions for improvement, and obser ved subsequent interviews. The mentioned precautions did not eliminate interviewer bias, but minimized the effect of multiple interviewers during data collection. Information or observation biases include surveillance, recall, and reporting biases. Surveillance bias, or m onitoring a specific group over time to ascertain disease status, did not occur in the current study given the design and methodology. Recall bias, however, ma y affect the analysis and results.

PAGE 120

84 Although participants were interviewed at two distinct time points during their pregnancies, in both interviews they were asked to recall events prior to pregnancy, during early pregnancy, and between the two interviews. As a result, certain measurements of stressors ma y be biased. Even though recall was affected, the participants we re unaware of the outcomes of interest: urine sugar readings, C-section, and high birth weigh t, and therefore, could not bias their responses in favor of a specific outcome Reporting bias may also affect study results. Specific measures such as abuse may be under-reported due to the stigma and possible legal ramifications of admitting to living in an abusive household. All such reporting biases would lead to an underestimate of the measure of association, and t herefore, are conservative. Misclassification may have occurred in terms of exposure (predictors). Most questions were asked during both interv iews for verification. Interviewers checked initial responses during the second interview, and questioned participants if their answers changed. If any misclassification occurred, it was most likely non-differential as the spec ific predictors and outcome measures were not known for any participant unt il after the interviewing process was complete. Misclassification of outcome measures is unlikely unless there was an error in recording data into participants’ medical records. Although such errors occur, it is not in a proportion la rge enough to induce misclassification. In conclusion, attrition bias most like ly played a minimal role in the current study. Selection, interviewer, and obser vation biases affected the study to a

PAGE 121

85 greater extent, and if miscla ssification occurred, it was conservatively biased or non-differential.

PAGE 122

86 CHAPTER 4 RESULTS 4.0 Introduction The following chapter is divided into se ctions of data results consisting of descriptive and inferential statistical su mmaries. The descriptive section is composed of a summary of available demog raphic characteristics of participants in the study; frequency and distribution of predictors, outcomes, and other possible confounders; transformation of non-normally distributed outcomes; reliability testing of all scales used in analysis; exclusion of variables; and discussion of uncollected data. The inferentia l section is divided into multivariate assessment of confounding and multicolline arity, analysis of pa irs of predictors and outcomes separately, and testing of eac h primary hypothesis. A descriptive schematic appears at the beginning of each major section outlining the organization of that particular section. 4.1 Descriptive Statistics This section is composed of the bas ic demographic characteristics of the sample followed by descriptive statistics of all predictors, outcomes, and possible confounding factors. A brief discussion of variable transformation and reliability testing follows with an explanation of variables excluded from analysis due to missing data. Figure 8 describes the layout of the section in detail.

PAGE 123

87Figure 8 Organization of the descripti ve section for the Results Chapter 4.1.1 Demographic Characteristics Demographic characteristics collect ed within the study include ethnicity, age, educational level attained at the time of the initial interv iew, the pre-pregnant weight of the participan t, and each participant’s height15. Marital status, body mass index, and site location were colle cted, but are included in the section describing the distribut ion of the predictors16. As shown in Table 26, 57% of the participants in the sample are Black, and 43% are White excluding the proportion missing17. Educational level attained is defined as grade level comp leted at the init ial interview. Approximately 10% of the sample completed grades 6th through 8th regardless of 15 All demographic characteristics and additional potentially c onfounding factors are defined in the methodology chapter. 16 Normality is assessed for each cont inuous variable based on descriptive ev aluation of skewness and kurtosis. 17 A detailed discussion of the missing values is located under the Uncollected Data section. • Basic Demographic Characteristics-Categorical Followed by Continuous Measures• Distributions and Frequencies(Categorical Followed by Continuous Measures)-PredictorsInitial Interview Distributions Final Interview Distributions Change in Distribution between Initial and Final Interviews-Outcomes -Possible Confounding Factors• Data Transformations-Birth Weight as a Continuous Measure• Reliability Testing-Autonomy Scale -KarasekPhysical Work Strain Scale -CESD (Depression Scale) -NorbeckSocial Support Scale (Partner and Mother)• Variable Exclusion

PAGE 124

88 age, 53% completed grades 9th through 12th but did not graduate from high school or receive a GED, 32% completed high school or received a GED, and 5% completed some college. In summary approximately 63% of the participants had less than a completed high school education at the time of the initial interview. Table 27 presents continuous demogr aphic characteristics. The mean age of participants within the sample is 22; 22% of the women in the sample are teenagers (aged < 20 years), 70% are bet ween the ages of 20 and 29, and 8% are 30 or older. A participant’s prepregnant weight ranges from 82 to 411 pounds with a mean of 152 pounds. Dividing the sample into quartiles, 25% of the sample weighed 120 pounds or less prior to pregnancy, 25-50% weighed between 121 and 139 pounds, 50-75% weighed between 140 and 175 pounds, and the upper quartile ranged between 176 and 411 pounds. In fact, the two largest weights, 355 and 411 appear to be outliers in the sample. Height of each participant ranges from four f eet nine inches to six feet four inches. The mean height is five feet f our and half inches.

PAGE 125

89Table 26 Categorical demographic characterist ics including ethnicity and educational level attained by the initial interview of pr egnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Categorical Demographic Characteristic Distributions* Characteristic Frequency Percentage Ethnicity Black 227 44.9 White170 33.6 Missing109 21.5 Educational Level Less than 9th grade (middle school)53 10.5 Less than high school/GED, but greater than middle school267 52.8 Less than college, but greater than high school/GED 160 31.6 College or greater26 5.1 *Marital status is excluded due to its incl usion as a predictor in the model. It is included in the section on descriptive analysis of the predictors. Table 27 Continuous demographic characteristics including age, pre-pregnant weight, and height of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Continuous Demographic Characteristic Distributions Characteristic Range Mean Standard Deviation Missing Age 14-35 22.5 4.3872 Pre-pregnant weight of the participant (pounds/lbs.) 82.0-411.0 151.5 46.0709 Height of the participant (inches) 57-78 64.5 2.8163 2 4.1.2 Predictor variables The plausible predictors of high ur ine sugar levels, C-section and high birth weight are social support, depressi on, marital status, autonomy, physical work strain, abuse, and pregnancy w antedness. As described in the methodology chapter, predictors were collect ed at both interviews from the initial interview, or from the final interview18. The following section displays the 18 See the methodology chapter for further information on sample size for each measure.

PAGE 126

90 distribution of variables colle cted from the initial interview, final interview, and the differences in scores or the change in st atus of variables collected during both interviews. The frequencies and distributions of each predictor fr om the initial interview are in Tables 28 and 29. Social support is assessed in four ways for both current partner and mother of the parti cipant (e.g., the presence or absence of support, total support, emotional suppor t, and instrumental/mat erial support). Seventy-seven percent (77%) of the sample list current partner as supportive for at least one question in the social suppor t scale, while 33% omitted current partner from the scale. The mean score s for emotional, ma terial, and total partner support are 2.0, 1.7, and 3.7 res pectively. Eighty-six percent (86%) of participants listed social support from t heir mother as present, compared with 14% who did not list their mother as s upportive. The mean scores for the three scales assessing the mother of the parti cipant’s social support are 2.2 for emotional support, 2.0 for material suppor t, and 4.2 for total support. The scores for the depression scale range from 1 to 40 with a mean score during the initial interview of 18.8. Fifty-nine percent ( 59%) of the sample are single, and have never been married as of the initial interview. Nine percent (9%) are single and have been married, and 32% are married or living with a partner. The mean score on the physical work strain scale is 3.4. Thirty-five per cent (35%) of the sample reported some form of abuse (either verbal or physical) prior to the initial interview and of those women, 26% r eported verbal abuse and 21% reported

PAGE 127

91 physical abuse. Of wom en who indicated whether or not they wanted their pregnancy, 35% said they did not want the pregnancy from the initial interview (excluding missing data). Table 28 Categorical descriptive statistics of predictor variables from the initial interview during the first trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Categorical Predictor Variable Distributions Predictor Frequency Percentage Social support partner No116 22.9 Yes390 77.1 Social support mother No73 14.4 Yes433 85.6 Marital status Single/never married296 58.5 Single/ever married45 8.9 Married/living with partner165 32.6 Overall Abuse Yes176 34.8 No330 65.2 Verbal Abuse Yes129 26.1 No374 73.9 Missing3 Physical Abuse Yes107 21.1 No399 78.9 Pregnancy Wantedness No71 14.0 Yes38 7.5 Missing397 78.5

PAGE 128

92Table 29 Continuous descriptive statistics of predictor variables from the initial interview during the first trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Continuous Predictor Variable Distributions Predictor Range Mean Standard Deviation Missing Social support scale partner 0-6 3.7 2.41 Emotional social support scale – partner 0-3 2.0 1.29 Material social support scale – partner 0-3 1.7 1.25 Social support scale mother 0-6 4.2 2.16 Emotional social support – mother 0-3 2.2 1.16 Material social support – mother 0-3 2.0 1.16 Depression scale 1-40 18.8 6.13 Physical work strain scale 1-6 3.4 1.45 276 Tables 30 and 31 present the distributi ons of data collected from the final interview. For variables collected from the initial interview, a discussion is included on the change in frequency, per centage, or mean from Table 29. Seventy-three participants ( 14%) did not participate in th e second interview. The mean score for the social support measur es in Table 31 indicate that minimal change occurred in support, and, therefore, support rema ined the same for the final interview. Partner social support changed from 23% of the participants claiming no support to 25%. In addition, the mean of each partner support scale increased from the final interview. Social suppor t of the participant’s mother, however, remained virtually the same as the init ial interview. The mean depression score increased, while more participants became either married or began living with their partners by the final in terview (33% initial interview; 41% final interview). The mean score for the physical work stra in scale also increased by the final

PAGE 129

93 interview. The autonomy scale, not assess ed from the initial interview, ranged from 1 to 10 and had an average score of 4. 5. Total cumulative abuse increased from 35% to 43% from the final intervie w with verbal abuse increasing from 26% to 38% and physical abuse decreasing fr om 21% to 16%. The percentage of women who stated they did not want the pregnancy when they first found out they were pregnant also increased from 35% to 48% by the final interview. Table 30 Categorical descriptive statistics of predictor variables from the final interview during the third trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Categorical Predictor Variable Distributions* Predictor Frequency Percentage Social support partner No128 25.3 Yes378 74.7 Social support mother No71 14.0 Yes435 86.0 Marital status Single/never married223 51.5 Single/ever married33 7.6 Married/living with partner177 40.9 Missing73 Abuse Yes186 43.0 No247 57.0 Missing73 Verbal Yes164 37.9 No269 62.1 Missing73 Physical Yes69 15.9 No364 84.1 Missing73 Pregnancy Wantedness No210 48.3 Yes225 51.7 Missing71 *Seventy-one final interviews are missing for all observat ions (approximately 14%). All percentages will be reported excluding this missing number.

PAGE 130

94Table 31 Continuous descriptive statistics of predictor variables from the final interview during the third trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Continuous Predictor Variable Distributions Predictor Range Mean Standard Deviation Missing Social support scale partner 0-6 5.4 1.56 493 Emotional social support scale – partner 0-3 2.7 0.75 493 Material social support scale – partner 0-3 2.7 0.85 493 Social support scale mother 0-6 4.3 2.87 502 Emotional social support – mother 0-3 2.3 1.50 502 Material social support – mother 0-3 2.0 1.41 502 Depression scale 0-40 19.4 5.96 73 Physical work strain scale 0-6 4.3 2.05 362 Autonomy scale 1-10 4.5 1.36 158 Tables 32 and 33 present the per centage change of all variables measured at both the in itial and final interviews. Th irty participants changed their opinions on whether or not they perc eived they had partner support, whereas only four changed their opinions on mother ’s support. The mean depression score change, that is, final interview score subtracted from initial interview score was 0.55 with a range of (-22) to 18. One hundred and twent y-one participants (121) stated they had changed their marita l status since the initial interview (28%). The change in the physical work st rain scale ranged from a score of (-4) to 5 with a mean score of 1. 22. Twelve percent (12%) of the participants in the sample reported a change in overall abuse between the initial and final interviews, and 20% expressed a change in their opinions of pregnancy wantedness.

PAGE 131

95 Table 32 Categorical descriptive statistics of the percentage of predictor change between initial and final interviews during the first and third trimesters of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Predictor Variable Distributions Predictor Frequency Percentage Social support – partner Change30 8.4 No change328 91.6 Missing148 Social support – mother Change4 1.2 No change343 98.8 Missing159 Marital status Change121 28.1 No change309 71.9 Missing76 Abuse Change52 12.0 No change381 88.0 Missing73 Pregnancy wantedness Change17 19.8 No change69 80.2 Missing420 Table 33 Continuous descriptive statistics of the percentage of predictor change between initial and final interviews during the first and third trimesters of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Predictor Variable Distributions Predictor Range Mean Change Standard Deviation Missing Depression scale (-22) to 18 0.55 5.96 73 Physical work strain scale (-4) to 5 1.22 2.09 406 All continuous variables were a ssessed for symmetry. All variables appeared normally distributed except for t he pre-pregnant weight of participants and the change in partner and mother s upport between the initial and final

PAGE 132

96 interviews. Pre-pregnant weight was not transformed as it is considered a potentially confounding factor and peripheral in analysis. The change in support variables were not transformed for analysis19. Finally, predictors measured at both time points during pregnancy were assessed for their affect on outcome measures based on whether they were present or absent throughout t he duration of the pregnancy. That is, the affect of having a predictor occur during pregnancy on the outcome measures regardless of time. Table 34 presents the percentages of these dichotomous variables. The majority of participants perceived they had support from both their partners and mothers. Thirty-seven percent (37%) sa id they were verbally or physically abused during or prior to their pregnancy; also, almost 45% stated they did not want their pregnancy when they firs t found out they were pregnant. 19 See the Exclusion of Specific Fact ors section for further explanation.

PAGE 133

97Table 34 Categorical descriptive statistics of predictor variables categorized as present or absent during the course of a pregnancy of women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Categorical Predictor Variable Distributions* Predictor Frequency Percentage Social support partner Absent116 22.9 Present390 77.1 Social support – mother Absent73 14.4 Present433 85.6 Abuse Absent247 48.8 Present186 36.8 Missing73 14.4 Pregnancy Wantedness (Unwanted) Absent210 41.5 Present225 44.5 Missing71 14.0 *Seventy-one final interviews are missing for all observat ions (approximately 14%). All percentages will be reported excluding this missing number. 4.1.3 Outcome Variables The study is composed of three outcome variables, an intermediate outcome (urine sugar levels), and two fi nal outcomes (high birth weight and Csection). Table 9 lists the frequency and distribution of all three outcome variables. Approximately 14% of the sa mple had urine sugar readings indicative of susceptibility to glucose intoler ance or a physiologic change occurring during pregnancy (e.g., high urine sugar spill). Ni ne percent (9%) presented trace urine sugar level readings (e.g., low urine s ugar spill), while t he remainder of the sample had no detectable sugar in their urine. The birth weight distribution ranged from 318 to 4570 grams with a mean weight of 3143 grams. Birth weight was not normally distributed within the samp le (skewness -1.5; kurtosis 4.39). When dichotomized into high and other birth weight categories, 6% of the

PAGE 134

98 participants’ babies weighed greater than or equal to 4000 grams. Eighteen percent (18%) of the sample had a Csection birth, while 82% gave birth vaginally. Additional descriptive outcome data include the timing of each urine sugar screening and the distribution of low birth weight infant s in the sample. Although the highest urine sugar level is the only level collected for analysis, when that screening test occurred has an impact on the monitoring of each participant. The average week gestation in which the highes t urine sugar level was recorded is 29 weeks (range 6 to 42 weeks gestation). The majority of the high readings occurred during the second trimester with 24 or less weeks gestation encompassing the first quartile (25%), and 25 to 31 encompassing the second quartile (50%). The third quartile r anged from 32 to 35 weeks gestation. Approximately 10% of the sample gave birth to low-weight babies (>= 2500 grams) in comparison to t he six percent who gave birth to high-weight babies in Table 35. The percentage of low birth we ight babies is high for this sample; however, referring back to the original foci of each grant, low birth weight was the outcome of interest. Therefore, such a large percentage is expected in these high risk populations.

PAGE 135

99Table 35 Descriptive statistics of outcome variables including urine sugar levels, birth weight, and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Outcome Variable Distributions Outcome Frequency* Percentage Urine sugar level (binomial) Any urine sugar116 22.9 No detectable urine sugar390 77.1 Urine sugar level (ordinal) High (1+ or higher) urine sugar69 13.6 Low (trace) urine sugar47 9.3 No detectable urine sugar390 77.1 High birth weight >= 4000 grams31 6.1 < 4000 grams475 93.9 Birth weight Range 318-4570 Mean 3143.0 Caesarean section Yes93 18.4 No413 81.6 *Frequencies are shown unless otherwise specified within the table. 4.1.4 Potentially Confounding Factors In addition to the demographic charac teristics of age, educational level attained, pre-pregnant weight, and height, ot her possible confounding factors, or factors physiologically a ssociated with the outcomes of interest in the study include parity, number of miscarriages/abortions, number of live births, number of previous C-sections, gestational age of t he infant at birth, weight gain during pregnancy, participants’ body mass index, to tal number of prenatal visits, alcohol consumption throughout pregnancy, and illicit drug use during pregnancy (specifically marijuana, cocaine, other narcotics). The frequencies and distributions of each potentially confoundi ng factor are shown in Tables 36 and 37.

PAGE 136

100Table 36 Descriptive statistics of potentially confounding categorical factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Confounding Factors (N = 506) Confounding Factors Frequency Percentage Previous Caesarean section Yes61 12.3 No436 87.7 Missing9 Alcohol and drug use during pregnancy Yes143 28.3 No363 71.7 Alcohol use during pregnancy Yes137 27.1 No369 72.9 Drug use during pregnancy Yes25 4.9 No481 95.1 Study site Tuscaloosa397 78.5 Mobile109 21.5 Table 37 Descriptive statistics of potentially confounding continuous factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Confounding Factors (N = 506) Confounding Factors Range Mean Standard Deviation Missing Parity (Total number of pregnancies) 1-8 2.2 1.37 1 Number of miscarriages/abortions 0-4 0.4 0.72 2 Number of live births 0-5 0.8 1.04 2 Number of pre-term births 0-3 0.2 0.49 2 Gestational age of infant at birth 20.0-44.0 38.9 3.088 1 Weight gain during pregnancy (-32.0)-84.0 27.5 15.590 4 Body Mass Index 14.0-60.7 25.5 7.028 2 Number of prenatal visits 2-30 13 3.76 In addition to the demographic variables previously discussed20, 12% of the women in the sample had a history of a previous C-section (procedures for elective abortions were excluded). Fo r parity history, t he range of previous 20 See the demographic distribution section at the beginning of the chapter for further information.

PAGE 137

101 pregnancies was one to eight (including the current pregnancy) with a mean of two pregnancies (not necessarily carried to term). The number of previous live births ranged from zero to five with a m ode of zero, while the number of pre-term births ranged from zero to three with a mode of zero, and the number of previous elective abortions and miscarriages ranged from zero to four with a mode of zero. The average gestational age at birth for t he current pregnancy of each participant was 38.9 weeks with a range of 20 to 44 weeks. In terms of participant characteristics of pregnancy, the av erage weight gain was 28 pounds with a range of 32 pounds lost to 84 pounds gai ned throughout the course of the pregnancy. Body mass index, calcul ated using pre-pregnant weight and height, has a mean value of 25.5 with a range of 14 to 60.7. The number of prenatal visits of participants ranged from two to 30 with an average of thirteen visits. Twenty-eight percent (28%) of the partici pants used either alcohol or drugs or both during pregnancy, with 27% consuming alcohol at least one time during pregnancy and 5% consuming drugs excludi ng smoking or illegal inhalants. For evaluation of confounding by study site a variable identifying where each participant’s data were collected is inclu ded (i.e., clinics in either Tuscaloosa County [78.5%] or Mob ile County [21.5%]). 4.1.5 Transformation of Nonnormally Distributed Outcomes As described in the outcome distribut ion section, birth weight was not normally distributed within the sample. In order to change the distribution, ten different transformations were empl oyed. First, the logarithm (log10) of birth

PAGE 138

102 weight was taken, however, the distri bution remained non-normal (skewness 4.0; kurtosis 20.1) with a m ean score of 3.5 and a range of 2.5 to 3.7. Next, birth weight was standardized by taking each measure and subtracting the mean then dividing by the standard deviation. The sample still seemed non-symmetric. The mean was subtracted alone fr om each measure for the third transformation, and the square root, inverse, and natural l og of each measure was taken for the fourth, fifth, and sixth atte mpts. For the seventh a ttempt, measures were squared. The distribution appeared to be symmetric (skewness -0.17; kurtosis 0.62), though the numbers were too la rge for subsequent analysis. The measures were converted to pounds, but the interpretation of the measures was extremely complicated. Ther efore, grams were converte d to kilograms for clarity of presentation. The final transformati on of birth weight was kilograms squared with a skewness of -0.17 and kurtosis of 0.62. In subsequent analyses birth weight was squared, and then backtransformed for interpretation. 4.1.6 Reliability Analysis of Scales All scales used in the analysis were te sted for reliability prior to hypothesis testing [173]. The autonomy, physical wo rk strain, depression, and both partner and mother social support scales were assessed using the Cronbach’s alpha to indicate scale cohesion. Multiple iterat ions were conducted if results indicated the alpha would increase with removal of certain questions from the scale, and if scales were assessed at both interviews, reliability testing was conducted for each interview.

PAGE 139

103 Table 38 presents the iterations of the autonomy scale, collected only from the final interview. After removal of si x of the ten variables within the scale, the standardized item alpha remained 0.40. Th e factors within the scale were not moderately or highly correlated regardless of the iteration. Table 38 Reliability analysis of the aut onomy scale of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Autonomy Scale (N = 348) Scale Item Deleted Alpha Mean Variance Final Alpha No items deleted 0.20 4.46 1.85 Men should spend the same amount of time as women in caring for children and the home. 0.24 3.65 1.76 It would feel strange if your boss at work was a woman. 0.32 3.63 1.77 Women should be paid the same as men for doing the same job. 0.37 2.64 1.76 You should treat experts with respect even if you do not think much of them. 0.39 1.82 1.55 A trusted person in authority tells you to do something. You should do it even if you do not see the reason for it. 0.39 1.27 1.19 It is the natural duty of the woman to provide the love and caring for the family. 0.39 0.75 0.84 0.39 The physical work strain scale consisted of seven questions and was calculated for both interviews for a portion of the sample. Table 39 shows the alpha score of the final scale for the initial interview. After remo val of four of the seven questions, the alpha level was 0.72 with all remaining variables moderately correlated. From the final interview, however, only one item was deleted with a final alpha of 0.76 and a mean score of 3.78 (Table 40). All remaining variables were minimally, moderately and highly correlated within the scale.

PAGE 140

104Table 39 Reliability analysis of the physical wo rk strain scale from the initial interview during the first trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Physical Work Strain Scale (N = 230) Scale Item Deleted Alpha Mean Variance Final Alpha No items deleted 0.28 3.39 2.10 Does the work you do on the job cause you to worry a lot? 0.37 3.13 2.03 Is your work physically difficult? 0.47 2.84 1.91 At your job, are you alwa ys on the move? 0.69 2.05 1.98 Do you get enough breaks during work hours? 0.72 1.37 1.41 0.72 Table 40 Reliability analysis of the physical wo rk strain scale from the final interview during the third trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Physical Work Strain Scale (N = 144) Scale Item Deleted Alpha Mean Variance Final Alpha No items deleted 0.72 4.31 4.21 At your job, are you alwa ys on the move? 0.76 3.85 3.78 0.76 The depression scale or modified C ESD was composed of ten questions assessing participants’ perceived level of depression. One item was removed from the scale for the responses from t he initial interview with a final alpha of 0.82 and a mean score of 16.49 (Table 41). For the final interview, two items were removed for an alpha of 0.84 and a mean score of 15.14 (Table 42). Table 41 Reliability analysis of the depression scale from the initial interview during the first trimester of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Depression Scale (N = 506) Scale Item Deleted Alpha Mean Variance Final Alpha No items deleted 0.80 18.78 37.60 You felt hopeful about the future. 0.82 16.49 34.79 0.82

PAGE 141

105Table 42 Reliability analysis of the depression scale from the final interview during the third trimester of pregnant women attending the County Health Departme nt Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Depression Scale (N = 433) Scale Item Deleted Alpha Mean Variance Final Alpha No items deleted 0.81 19.36 39.03 You felt hopeful about the future. 0.82 17.15 34.75 You enjoyed life. 0.84 15.14 31.27 0.84 The social support scales consisted of the same series of six questions however, the participant selected the indi vidual for the series. In this study, support of the mother was used to a ssess familial support and support of the current partner was used to assess part ner support. Due to the small magnitude of change in mother and part ner support between the init ial and final interviews (refer to Table 32), only the initial intervie w scales were assessed for reliability. Tables 43 and 44 present the partner and mother support scale Cronbach’s alpha scores. No items were deleted from either scale due to the high reliability of each (partner alpha 0.92; mother of participant alpha 0.89). The partner support scale had a mean score of 3.74, wh ile the mother support scale had a mean score of 4.16. Table 43 Reliability analysis of the social support scale of the participant’s partner of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Social Support Scale (N = 506) Scale Item Deleted Alpha Mean Variance Final Alpha No items deleted 0.92 3.74 5.80 0.92

PAGE 142

106Table 44 Reliability analysis of the social support scale of the participant’s mother of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Social Support Scale (N = 506) Scale Item Deleted Alpha Mean Variance Final Alpha No items deleted 0.89 4.16 4.66 0.89 4.1.7 Exclusion of Specific Factors All variables discussed in the previous section of this chapter were initially to be included in analysis. Based on the di stribution of specific variables, lack of change between measurements during the in itial and final interviews, and low Cronbach alpha scores specific measures we re removed from analysis. These variables include social support from the final intervie w, autonomy, pregnancy wantedness from the initial interview, and change between interviews of abuse and physical work strain. Specifically due to the small percentage of change in social support (8% for partner support and 1% for a mother’s support) during the final interview, support as measured from the initial interview was used throughout the analysis to assess the impac t of social support over the entire pregnancy. In the reliability section the autonomy scale was listed as having a final Cronbach’s alpha of 0.39. Such a low alpha indicated that the scale was not cohesive and should not be used; therefore, it was omitt ed from the analysis. In addition, due to a high proportion of unc ollected data, pr egnancy wantedness

PAGE 143

107 taken from the initial interview, c hange in pregnancy wante dness, and change in abuse and physical work strain were omitted from the final analysis21. 4.1.8 Uncollected Data Uncollected data affect t he reliability and validity of study results. Three types of uncollected data impact t he analysis and interpretation of this dissertation: data uncollected due to the original grant pr otocols; data collected at during only one interview and uncollected du ring the other interview; and data missing due to interviewer e rror, participant’s giving birt h, or attrition. This section provides a detailed di scussion of these limitations. Both sites had specific protocols; the Tuscaloosa site focused on psychosocial and physical stressors and low birth weight babies, while the Mobile site focused on high-risk teenage mothers. Since the current study centers on high birth weight infants and Caesarean sect ion births, certain data predictive of these outcomes may not have been collected in the original interview schedules. For example, data were not collected i dentifying diagnoses in the pathway to high-weight births such as ketone-level m onitoring, glucose tolerance testing, or identification of previous or current gesta tional diabetes. The highest urine sugar level is the only recorded level for analysis. Inclusion of all urine sugar testing results would have aided in a time-related analysis of sugar spill. These additional data would provide invaluable information in the assessment of the pathway from socio-cultural factors to hi gh birth weight infant s. In addition, 21 See the Uncollected Data section for a further explanation.

PAGE 144

108 multiple factors affect a physician’s dec ision to perform a Caesarean section during childbirth. No data were collect ed including operating facilities, physician preparedness or preference, hospi tal protocol, or Medicaid policy. All of these factors contribute to the decision to operate during l abor. As a result, the associations between the identified psyc hosocial and physical predictors may not be as strong with the exclusion of these necessary factors. Due to time constraints, certain da ta were collected only once during the interviewing process, and specific measures were collected at different times for the two samples. The autonomy scale was only evaluated during the final interview for the Tuscaloosa sample, as well as pregnancy wantedness. The Karasek physical work strain scale was collected during both interviews for the Tuscaloosa sample, but after extensive analysis it was conc luded its greatest impact was during the latter half of the pregn ancy. Therefore, the protocol for the Mobile sample included only collecting the data during the final interview. Ethnicity was not collected for the Mobile sample based on the characteristics of the high-risk group from the pilot study wh ich was 95% Black and only 5% White. Ethnicity is only coded for the Tuscaloos a sample and analysis of Hypotheses 4 and 5 are based on that site only. Modified protocols for the Mobile site affected the analytic capabilities of this disse rtation, and the m easurement of the predictors and interaction terms. Finally, data were uncollected due to la ck of attention dur ing the interview process, pre-term births and miscarri ages, and loss to follow-up of a small

PAGE 145

109 proportion of participants. The intervie w guides consisted of complicated skip patterns based on participant response. Du ring the interview process, sections may have been omitted by interviewers who could not initially follow the questioning sequence. The amount of data d ue to interviewer e rror is minimal, although up to 5% of the proportion of missing for some predictors may be attributed to this error. Approximately 73 participants gave birth prior to finishing the entire final interview, or 14.4% (a ll miscarriages were excluded completely from the sample). These data were unobtainable as each grant protocol scheduled the final interview between 28 and 40 weeks gestation and any birth prior to that time removed the respondent from participation. Some participants, however, were partially interviewed prior to their pre-term birth and so partial data were collected for a small proportion of thos e women. Attrition consisted of less than 3% for the Tuscaloosa site and 10% fo r the Mobile site; since outcome data were not available for those participants lost to follow-up, they were excluded from the analysis completely. Although unc ollected data affected results, only a small proportion were considered trul y missing based on interviewer error or attrition, and therefore, the bias attributed to these data is minimal22. 4.2 Inferential Statistics As previously described at the begi nning of the chapter, the Inferential Statistics section is composed of asse ssment of potentially confounding factors along with multicollinear e ffects. Next, assessment of each predictor and each 22 See the Discussion Chapter for further interpreta tion of the impact of missing or uncollected data.

PAGE 146

110 outcome is performed (Hypotheses 2 and 3) in separate models. The subsequent sections include a complete ana lysis of all five hypotheses (Figure 9). Figure 9 Organization of inferential stat istics section for the Results Chapter 4.2.1 Evaluation of Confounding Factors Figure 10 describes the methodology for the evaluati on of potentially confounding factors in the final analysis. As shown, predictors and outcomes individually are modeled including all fourteen potentially confounding factors (e.g., each predictor is modeled with each outcome and all confounding factors for a total of 45 models). With alpha set at 0.10, all signif icant confounders are next included in a partial model with the predictor and outcome of interest [174]. The change in beta between the predictor and outcome is compared between the • Assessment of Confounding– See Figure 3 for Further Detail• Evaluation of Multicollinearity– Potential Confounders – Predictors – Outcomes• Analysis– Each Predictor and Outcome Separately • Hypothesis 2 • Hypothesis 3– Hypothesis 1 – Hypothesis 2 (See Figure 4 for Further Detail) – Hypothesis 3 – Hypothesis 4 – Hypothesis 5

PAGE 147

111 full and partial models for each set of predi ctors and outcomes. If there is less than a 10% change in the beta, the partial model is re-run removing each potentially confounding factor in turn to assess the magnitude of confounding. Each confounder is considered weak if t he change in beta (between the predictor and outcome) is less than 5%, moderate if the change in beta is 5 – 10%, and strong if the change in beta is greater t han 10%. A final model is run including only moderate and strong conf ounders. The predictor-outcome beta is compared to the full model beta, and if the change is less than 10%, then analysis stops and those factors in the model are c onsidered the most influential confounders and included in all subsequent analyses of t hat predictor-outcome combination. If the change is greater than 10%, the weak confounders are added back to the model one by one to assess the change in the predictor-outcome beta. If the predictor-outcome beta falls within a 10% range of the fu ll model predictoroutcome beta, analysis stops. If not, then all non-signific ant confounders are added back to the model one by one to i ndividually assess the change in the predictor-outcome beta. Once ther e is a less than 10% change between the partial and full model predictor-outcome betas, analysis stops. Further modeling follows the initial pathway on the left side of Figure 10.

PAGE 148

112Figure 10 Strategy for assessing confounding factors for the Results Chapter Urine Sugar Levels Potentially confounding factors are assessed for all fourteen measurements of predictors, the interact ion term of ethnicity, and urine sugar levels. The predictors include categoric al measurements of social support (partner and mother), abuse (total, initial and final interviews), marital status (initial and final interviews), and pr egnancy wantedness. C ontinuous predictors include social support scale s (partner and mother), depre ssion scale (initial, final, and change in score), and physical work stra in scale (initial and final). For brevity, only the pathway of confounding ana lysis is reiterated in detail for the first two predictors; subsequent predict ors are described in Table 47. Full model: Each predictor and outcome including all 14 confounders; alpha < 0.10 Partial model: Each predictor and outcome including all significant confounders Beta between full and partial model changes by <10% Beta between full and partial model changes by >10% Run the model removing each potential confounder in turn; assess whether each is a weak, moderate, or strong confounder (change in beta) Begin with full model and reassess confounding Run final model including only moderate and strong confounders

PAGE 149

113 In Table 45, the final model includi ng moderate and strong confounders is shown. The full model yielded a predi ctor-outcome beta of 0.672. Four potentially confounding factors contained an alpha < 0.10 (weight gain during pregnancy p < 0.001; pre-pregnant weight p < 0.10; body mass index p > 0.10; education level attained by the initial interview p < 0.10). A partial model was constructed including the pr edictor and all four potentially confounding factors. The partial model yielded a beta of 0.681. Next, each factor was removed from the model sequentially to assess the af fect of its remova l on the predictoroutcome beta. Results are located in the t able next to the column labeled “partial model”. When examining the predict or-outcome beta changes, pre-pregnant weight gain and body mass index were determined to be weak confounders and removed from the model. The final m odel included the predictor and only the moderate confounders of weight gai n during pregnancy and education level attained by the initial interview, and yiel ded a beta of 0.695 (wit hin 10% of the full model beta of 0.672 [range 0.605 – 0.739]) indicating the adequate control of confounding. Table 45 Assessment of confounding factors for partner social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding (N = 502) Model Confounder Predictor-Outcome Beta Full All fourteen 0.672 Partial Weight gain 0.724 Pre-pregnant weight 0.670 BMI 0.680 Education level 0.623 Final Predictor-Outcome Beta 0.695

PAGE 150

114 For the assessment of mother’s social support and urine sugar levels, all fourteen potentially confounding factors were included in the full model with a beta of 1.375 (predictor-out come). Table 46 present s modeling results. The partial model contained the potentially c onfounding factors of educational level attained by the initial interview, pre-pregnant weight, weight gain during pregnancy, and body mass index (beta = 1.333). Education level was determined to be a weak confounder and was removed from the final model (change in beta > 10% compared with the full model). The final predictoroutcome beta was 1.333. Table 46 Assessment of confounding factors for a mother’s social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding (N = 500) Model Confounder Predictor-Outcome Beta Full All fourteen 1.375 Partial Education level 1.333 Pre-pregnant weight 1.484 Weight gain 1.537 BMI 1.488 Final Predictor-Outcome Beta 1.333 Table 47 presents all of the final m odels for the remaining predictors and urine sugar levels. As shown, none of t he potential confounders had an affect on the social support, depression, and physical wo rk strain (initial interview only) scales (43% of the predict ors). Age (75%), pre-pr egnant weight (50%), and weight gain during pregnancy (50%) confounded most of the remaining predictors and urine sugar levels. Pregnancy history (total number of

PAGE 151

115 pregnancies), education level attained by t he initial interview, and alcohol and drug use during pregnancy also confounded the predictors of marital status, abuse, and pregnancy wantedness.

PAGE 152

116Table 47 Selected confounding factors for all other predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding Predictor Confounder N Exponent of Beta Marital Status t1* (1)** Age 506 0.943 Marital Status t1 (2)*** 0.771 Marital Status t2+ (1) Age 424 1.218 Marital Status t2 (2) Alcohol and Drug Use 0.685 Past Premature Births Previous C-section Total Number of Live Births Total Number of Pregnancies Weight Gain Overall Abuse Age 430 0.601 Alcohol and Drug Use Body Mass Index Education Level Past Premature Births Pre-pregnant Weight Abuse t1 Age 491 1.011 Body Mass Index Education Level Past Premature Births Pre-pregnant Weight Previous C-section Weight Gain Abuse t2 Alcohol and Drug Use 425 0.496 Body Mass Index Education Level Pre-pregnant Weight Previous C-section Previous Premature Births Partner Social Support Scale None 506 1.030 Mother Social Support Scale None 506 0.942 Depression Scale t1 None 506 1.026 Depression Scale t2 None 433 1.032 Change in Depression Score None 433 0.990 Physical Work Strain Scale t1 None 230 1.094 Physical Work Strain Scale t2 Age 144 1.046 Total Number of Abortions Total Number of Live Births Total Number of Pregnancies Continued on the next page

PAGE 153

117Table 47 (Continued) Pregnancy Wantedness Age 431 0.659 Alcohol and Drug Use Body Mass Index Pre-pregnant Weight Total Number of Prenatal Visits Weight Gain Ethnicity++ Weight Gain 394 1.238 t1 refers to data collected during the initial interview. ** 1 is the comparison of single participants wi th married/living with partner participants. *** 2 is the comparison of single, ev er-married participants with married/ living with partner participants. + t2 refers to data collected during the final interview. ++ Interaction term. High Birth Weight Confounding factors for categorical predictors of high birth weight are listed in Table 48. Each predictor is c onfounded by at least two or more factors. Education level attained by the initial interview confounds 78% of the predictors, while body mass index, pre-pregnant weigh t, total number of premature births, and weight gain during pregnancy each conf ound 56% of the predictors. Other confounders include alcohol and drug use during pregnancy, gestational age of the infant at birth, interview site and total number of prenatal visits.

PAGE 154

118Table 48 Selected confounding factors of all categorical predictors and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding Predictor Confounder N Exponent of Beta Marital Status t1* (1)** Alcohol and Drug Use 497 0.820 Marital Status t1 (2)*** Body Mass Index 1.775 Education Level Gestational Age of Infant Interview Site Pre-pregnant Weight Total Number of Premature Births Total Number of Prenatal Visits Weight Gain Marital Status t2+ (1) Body Mass Index 427 1.141 Marital Status t2 (2) Education Level 1.530 Gestational Age of Infant Pre-pregnant Weight Total Number of Premature Births Total Number of Prenatal Visits Weight Gain Overall Abuse Education Level 431 0.464 Weight Gain Abuse t1 Body Mass Index 502 0.600 Education Level Pre-pregnant Weight Total Number of Premature Births Abuse t2 Alcohol and Drug Use 428 0.471 Body Mass Index Education Level Gestational Age of Infant Pre-pregnant Weight Weight Gain Pregnancy Wantedness Education Level 431 0.768 Gestational Age of Infant Total Number of Premature Births Weight Gain Partner Social Support Education Level 502 0.972 Total Number of Prenatal Visits Weight Gain Mother Social Support Body Mass Index 501 0.144 Gestational Age of Infant Pre-pregnant Weight Total Number of Premature Births Total Number of Prenatal Visits Continued on the next page

PAGE 155

119Table 48 (Continued) Ethnicity++ Education Level 397 2.447 Gestational Age of Infant Total Number of Prenatal Visits t1 refers to data collected during the initial interview. ** 1 is the comparison of single participants wi th married/living with partner participants. *** 2 is the comparison of single, ev er-married participants with married/ living with partner participants. + t2 refers to data collected during the final interview. ++ Interaction term. Confounders for continuous variables are located in Table 49. None of the continuous predictors are confounded by identified factors except for the physical work strain scales measured at the initial and final interviews. Both scales are confounded by gestational age of t he infant at birth, while the total number of previous C-sect ions and premature births confounds only the final interview scale. Table 49 Selected confounding factors of all continuous predictors and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding Predictor Confounder N Exponent of Beta Partner Social Support None 506 1.103 Mother Social Support None 506 1.311 Depression t1* None 506 0.994 Depression t2** None 433 1.024 Change in Depression Score None 433 1.022 Physical Work Strain t1 Gestational Age of Infant 230 1.055 Physical Work Strain t2 Gestational Age of Infant 140 1.394 Previous C-section Total Number of Premature Births t1 represents data collected during the initial interview ** t2 represents data collected during the final interview In order to assess the association between urine sugar spill and high birth weight, possible confounders must also be assessed. Table 50 indicates that education level attained by the initial in terview, gestational age of the infant at

PAGE 156

120 birth, and total number of prenatal visi ts attended confounds the association between urine sugar testing and high birth weight. Table 50 Selected confounding factors of urine sugar levels and high birth weight infants of pregnant women attending the County Heal th Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding Predictor Confounder N Exponent of Beta Urine Sugar Level Education Level 505 0.388 Gestational Age of Infant Total Number of Prenatal Visits Caesarean Section A list of confounding factors is show n for each categorical predictor and the interaction term ethnicity. Each of the nine categorical predictors has moderate and strong confounders associated with Caesarean section. Alcohol and drug use during pregnancy (56%), body mass index (67%), previous Csection (89%), and total num ber of prenatal visits (78%) confound most of the predictor-outcome associations. Other confounders include age, education level attained, total number of live births, pre-pregnant weight, interview site, and gestational age of the infant at birth.

PAGE 157

121Table 51 Selected confounding factors of all categorical predictors and Caesarean section of pregnant women attending the County Heal th Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding Predictor Confounder N Exponent of Beta Marital Status t1* (1)** Age 495 0.613 Marital Status t1 (2)*** Alcohol and Drug Use 1.155 Body Mass Index Education Level Previous C-section Total Number of Live Births Total Number of Prenatal Visits Marital Status t2+ (1) Age 424 0.409 Marital Status t2 (2) Alcohol and Drug Use 1.075 Body Mass Index Gestational Age of Infant at Birth Pre-pregnant Weight Previous C-section Total Number of Live Births Total Number of Prenatal Visits Overall Abuse Alcohol and Drug Use 426 0.487 Interview Site Previous C-section Total Number of Live Births Abuse t1 Alcohol and Drug Use 506 0.787 Education Level Total Number of Prenatal Visits Abuse t2 Age 425 0.591 Alcohol and Drug Use Body Mass Index Education Level Gestational Age of Infant at Birth Pre-pregnant Weight Previous C-section Total Number of Prenatal Visits Pregnancy Wantedness B ody Mass Index 428 0.982 Education Level Pre-pregnant Weight Previous C-section Total Number of Prenatal Visits Partner Social Support Body Mass Index 496 0.826 Education Level Previous C-section Total Number of Prenatal Visits Mother Social Support Body Mass Index 496 1.036 Education Level Previous C-section Continued on the next page

PAGE 158

122Table 51 (Continued) Ethnicity++ Education Level 396 2.179 Previous C-section Total Number of Prenatal Visits t1 refers to data collected during the initial interview. ** 1 is the comparison of single participants wi th married/living with partner participants. *** 2 is the comparison of single, ev er-married participants with married/ living with partner participants. + t2 refers to data collected during the final interview. ++ Interaction term. Table 52 presents confounders fo r the continuous predictors and Caesarean section. Of a ll continuous predictors, only physical work strain measured from the final interview is c onfounded by educational level attained. Associations between C-section and all other continuous pr edictors are not impacted by the confounders identified for this study. Table 52 Selected confounding factors of all continuous predictors and Caesarean section of pregnant women attending the County Heal th Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding Predictor Confounder N Exponent of Beta Partner Social Support None 506 1.063 Mother Social Support None 506 1.071 Depression t1 None 506 1.027 Depression t2 None 433 1.025 Change in Depression Score None 433 0.981 Physical Work Strain t1 None 230 0.941 Physical Work Strain t2 Education Level 144 0.830 t1 represents data collected during the initial interview ** t2 represents data collected during the final interview As with high birth weight, in order to assess the association between urine sugar levels and C-section, possible conf ounders must be identified. Table 53 presents confounders of the predictor-outcome association. As shown, education level attained, alcohol and drug use during pregnancy, pre-pregnant

PAGE 159

123 weight, and total number of prenatal vi sits attended confounds the association between urine sugar levels and C-section. Table 53 Selected confounding factors of urine sugar levels and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Assessment of Confounding Predictor Confounder N Exponent of Beta Urine Sugar Level Alcohol and Drug Use 506 0.791 Education Level Pre-pregnant Weight Total Number of Prenatal Visits 4.2.2 Evaluation of Multicollinearity Multicollinearity is defined as a high correlation between two factors (>0.70) that may affect the association of either or both factors on an outcome of interest [175]. In this study, multicollinearity is assessed for all potentially confounding factors and predi ctors. The following section is divided into a discussion of multicollinearity among confounding factors and predictors. Multicollinearity Among Confounding Factors Of all fourteen potentially confoundi ng factors, two sets are highly correlated. Total number of live birt hs and total number of pregnancies are correlated as expected (r = 0.85) given t hey are almost a m easurement of the same reproductive phenomenon. In addition, pre-pregnant weight and body mass index are high correlated (r = 0.96). Again, such a correlation is expected, since body mass index is a calculation based on pre-pregnant weight and height. In analysis for confounding, models were calculated with both sets of highly

PAGE 160

124 correlated variables, with one present and the other absent, and with neither to ensure multicollinearity did not affect the overall assessment23. In addition, in hypothesis testing procedures, total num ber of live births and pregnancies were not included simultaneously in analysis. However, body mass index and prepregnant weight appeared to confound associ ations often. When testing models with both factors and with each factor remo ved, it was determined that only one was required to account for the confounding effect of both. Body mass index was left in all models as it accounted fo r not only pre-pregnant weight, but height as well. Multicollinearity Among Predictors As discussed in the section on multic ollinearity of co nfounding factors above, multiple sets of predictors are highly correlated as expected based on multiple assessments of the same behavior. To exemplify, presence or absence of partner social support is highly correlated with the measurement scale assessing partner social supp ort (r = 0.85), in turn, sup port of the participant’s mother is highly correlated with the scale assessing a mother’s support (r = 0.79). Measurement of marital status from the final interview is highly correlated with the initial assessment of ma rital status (r = 0.78), both measurements of abuse are highly correlated (r = 0.76), and the presence or absence of abuse is highly correlated with the assessment of abuse from the final interview (r = 0.71). All of the listed multicollinear relationships are expected based on their conceptual 23 See the confounding section for models including and excluding the multicolli near sets of variables.

PAGE 161

125 relationships (i.e., history of abuse s hould be correlated with current abuse). All subsequent analyses excluding the initial m odels comprised of a ll predictors, do not include multicollinear pr edictors in the same model. 4.2.3 Separate Analysis of Predictors and Outcomes The analysis of each predictor and out come association including only confounding factors exemplif ies the independent effect of each predictor on each outcome. Prior to testing the five prim ary hypotheses, the indep endent effect of each predictor on each outcome is necessary The following section consists of assessment of each predictor and ur ine sugar separately including only confounding factors; the following secti on is composed of analysis of predictors and birth weight and then Caesarean secti on again only including confounding factors. Independent Associations Between Each Predictor and Urine Sugar Levels Hypothesis 2 is stated as follows: P sychosocial and physical factors (e.g., physical strain, lack of social su pport, depression, autonomy, pregnancy wantedness, and physical and verbal abuse) during pregnancy are associated with higher urine sugar levels (See A ppendix D for model fit statistics for all logistic regression analysis t ables presented in this chapter24). A participant’s history of physical or verbal abus e and physical or verbal abuse during pregnancy are significantly associated wit h urine sugar spill when examining 24 All smaller order models presented a good fit; however, many of the larger models did not present a good fit, but are included as they follow the original study protocol. All models including interaction terms are exploratory only and therefore may not present a good fit but are interpreted in the context of the dissertation.

PAGE 162

126 each predictor including all identified confounding factors25. Binomial logistic regression analyses were assessed for t he association between each predictoroutcome combination including confounding factors. Essentially, women in the study with a history of physical or ver bal abuse are 66% more likely to have elevated urine sugar levels during the c ourse of the pregnancy (low or high levels) compared with women with no histor y of abuse (OR = 1.66; 95% CI 1.042.66)26. Women in the study who are ph ysically or verbally abused during pregnancy are 47% more likely to have el evated urine sugar levels compared with women who are not abused during pregnancy (OR = 1.47; 95% CI 1.193.28)27. Secondary analysis involves an examin ation of urine sugar spill as both an ordinal and multinomial meas ure. Specifically, urine sugar is grouped as ‘no detectable levels of sugar’, ‘low or trac e amounts of sugar’, and ‘high sugar spill’ (1+ or higher). Analysis using ordinal l ogistic regression yielded no significant results as the models were almost all underpowered. For example, when assessing model fit -2 log likelihood measures and Pearson goodness-of-fitstatistics, both were significant for se ven out of fifteen model s. Of the models where goodness of fit statistics indi cated a good fit, no predictors were significantly associated with urine sugar spill. As an alternative, multinomial logistic regression was used for predict or-outcome assessment. Multinomial 25 Urine sugar levels are dichotomized for the prim ary analysis and are ordinal for secondary analysis. 26 Analysis adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, prepregnant weight, and total number of premature births. 27 Analysis adjusted for alcohol and dr ug use during pregnancy, body mass index, education level attained, pre-pregnant weight, previous C-section, and to tal number of premature births.

PAGE 163

127 modeling assumes each level of the outcome (urine sugar) is mutually exclusive, and, therefore, does not take into a ccount cumulative probabilities between levels [176]. As a result, significant predictors are only compared with one other group. That is, odds ratios are compar ing women with trace or low amounts of sugar spill to the no detectable levels of urine sugar group (reference), high sugar spill with the no detectable levels of sugar spill group, and a separate analysis compares the low or trace am ounts to the high group. Comparisons between the no detectable levels, low or trace, and high sugar spills are not analyzed taking into account the e ffects of the other groupings. When examining the same set of predictors and the multinomial measurement of urine sugar levels, physi cal work strain during the second and third trimesters of pregnancy is signific antly associated with sugar readings, as well as social support of the participant ’s mother (total support, emotional, and instrumental support). Tables 54 thr ough 57 present the results. Table 54 presents the results for physical work stra in and urine sugar levels. There is no significant association between trace urine sugar levels and physical work strain. However, for every one point increase in the work strain scale during the second and third trimesters of pregnancy, women are 32% more likely to high have urine sugar spill compared with women who hav e decreasing physical work strain scores (OR = 1.32; 95% CI 1.00-1.75). In addition, when comparing high to low groups, for every one point increase in the work strain scale, women are 61% more likely to have high urine sugar spills compared with women who have

PAGE 164

128 decreasing physical work strain indicat ed by decreasing scores (OR = 1.61; 95% CI 1.07-2.43). Table 54 Multinomial logistic regression model of physical work strain during the second and third trimesters of pregnancy and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Logistic Regression Model (N = 144) -2 Log Likelihood = 167.52; p > 0.05* Independent Variables Compared with No Sugar Level Odds Ratio 95% Confidence Intervals No Detectable Levels 1.000 Low Urine Sugar Levels Physical Work Strain 0.822 0.592 1.143 High Urine Sugar Levels Physical Work Strain 1.322 1.001 1.745 Independent Variables Compared with Low Levels Low Urine Sugar Levels 1.000 High Urine Sugar Levels Physical Work Strain 1.608 1.065 2.427 *Adjusted for age, total number of abortions, and total number of pregnancies. The Norbeck social support scale is analyzed in its entirety, and with its two components of emotional and instru mental (material) support. Table 55 indicates an association between low uri ne sugar levels and total social support from the participant’s mother For every decrease in t he level of social support from a participant’s mother, participant s are 16% more likely to have low amounts of sugar in their urine compar ed with participants who have increasing social support from thei r mothers (OR = 0.86; 95% CI 0.76-0.98). High urine sugar levels (>= 1+) are not associat ed with no detectable urine sugar, and high levels are not associated with low amounts of spill. In terms of emotional support from a participant’s mother (Table 56), again as support decreases, participants

PAGE 165

129 are 30% more likely to have low amounts of sugar in their urine compared with participants who have increasing support (O R = 0.77; 95% CI 0.61-0.98). High urine sugar levels are not significantly associated with emotional social support compared with no detectable sugar or low levels. Results for instrumental (material) support are sim ilar to total and emotional support (Tables 57). Participants who have decreasing support from their mothers are 30% more likely to have low amounts of sugar in t heir urine compared with participants who have increasing support (OR = 0.77; 95% CI 0.60-0.99). Table 55 Multinomial logistic regression model of the mother’s total social support and urine sugar levels of pregnant women attendi ng the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Logistic Regression Model (N = 506)* -2 Log Likelihood = 59.03; p > 0.05 Independent Variables Compared with No Sugar Level Odds Ratio 95% Confidence Intervals No Detectable Levels 1.000 Low Urine Sugar Levels Mother Social Support Scale 0.864 0.760 0.982 High Urine Sugar Levels Mother Social Support Scale 1.007 0.892 1.138 Independent Variables Compared with Low Levels Low Urine Sugar Levels 1.000 High Urine Sugar Levels Mother Social Support Scale 1.166 0.990 1.373 No significant confounding factors were used in adjustment of a mother’s social support and the ordinal measurement of urine sugar levels. See the assessment of c onfounding factors section for further discussion.

PAGE 166

130Table 56 Multinomial logistic regression model of the mother’s emotional social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Logistic Regression Model (N = 506) -2 Log Likelihood = 38.97; p > 0.05 Independent Variables Compared with No Sugar Level Odds Ratio 95% Confidence Intervals No Detectable Sugar Levels 1.000 Low Urine Sugar Levels Mother Emotional Social Support Scale 0.772 0.608 0.978 High Urine Sugar Levels Mother Emotional Social Support Scale 1.009 0.806 1.263 Independent Variables Compared with Low Levels Low Urine Sugar Levels 1.000 High Urine Sugar Levels Mother Emotional Social Support Scale 1.307 0.965 1.771 Table 57 Multinomial logistic regression model of the mother’s instrumental social support and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Logistic Regression Model (N = 506) -2 Log Likelihood = 38.62; p > 0.05 Independent Variables Compared with No Sugar Level Odds Ratio 95% Confidence Intervals No Detectable Sugar Levels 1.000 Low Urine Sugar Levels Mother Instrumental Social Support Scale 0.771 0.602 0.989 High Urine Sugar Levels Mother Instrumental Social Support Scale 1.017 0.811 1.274 Independent Variables Compared with Low Levels Low Urine Sugar Levels 1.000 High Urine Sugar Levels Mother Instrumental Social Support Scale 1.318 0.964 1.801

PAGE 167

131 Independent Associations Between E ach Predictor and Problematic Birth Outcome That psychosocial and physical fa ctors during pregnancy are associated with pregnancy complications is proposed for Hypothesis 3. First, predictor associations with high birth weight are assessed. Next, associations between the continuous measure of birth weight and each pr edictor are examined. Finally, Caesarean section and each predictor are exami ned. As stated in the Methodology chapter, binary logistic regr ession is used when birth weight is categorized as high or not high and with Ca esarean section, and multiple logistic regression is used when birth we ight is treated as continuous. As with urine sugar levels the social support of a participant’s mother and ethnicity are also associated with highweight births. However, unlike the protective effect of social support on ur ine sugar spill, a mother’s social support increases the likelihood of a high birth we ight infant. In fact, as a mother’s support increases on the scale, participants are 31% more likely to have a high birth weight baby compared with participant s who have decreasing social support from their mothers (OR = 1.31; 95% CI 1.03-1.67). Specifically, emotional support affects high birth weight. For every one point increase in a mother’s social support, participants are 74% more likely to have a high birth weight baby compared with participants lacking social support (OR = 1.74; 95% CI 1.0820.79)28. Instrumental (material) social supp ort of the participant’s mother is not significantly associated with high-weight bi rths (OR = 1.43; 95% CI 0.07-1.43). In 28 The upper end of the 95% Confidence Interval is high due to t he small size of the scale ( scores range from 0-3) and the lack of variability between partici pant responses and the dichotomou s outcome of high birth weight.

PAGE 168

132 addition, ethnicity is independently associ ated with high-weight births. White women in the study are over two and half ti mes more likely to have a high birth weight baby compared with Black women in the study (OR = 2.45; 95% CI 1.035.79)29. When examining birth weight as a c ontinuous measure, only ethnicity is significantly associated. The model has a good fit (F-test 84. 864; p < 0.001), and 46% of the variance is explained by et hnicity controlling for education level attained, gestational age of the infant at birth, and total number of prenatal visits30. The regression equation is as follows: birth weight ( ) = -4.478 + 0.956(ethnici ty) + 0.493(education) + 0.855(gestation) + 0.944(prenatal) As stated above, White women are more likely to have heavier babies (standard error ethnicity 0.280; t stat istic 3.236; p < 0.05). Similarly to birth weight, ethnicity is associated with the proportion of Caesarean sections performed in the study However, unlike high birth weight results, history of abuse and current marital status are associated with C-section. White women in the sample are over two times more likely to have had a Csection compared to Black women in t he study (OR = 2.18; 95% CI 1.27-3.75)31. In terms of abuse and marital status, participants with a history of abuse are two times more likely to have a C-se ction compared with pa rticipants with no 29 Analysis is adjusted for education level attained, gestational age of the infant at birth, and the total number of prenatal visits attended. 30 For the residual plots of this analysis, see Appendix E. 31 Analysis is adjusted for education level attained, prev ious C-section, and total number of prenatal visits.

PAGE 169

133 history of abuse (OR = 2.05; 95% CI 1.16-3.66)32. In addition, single nevermarried participants are almost two and hal f times more likely to have a Csection birth compared with married partici pants or those currently living with a partner (OR = 2.44; 95% CI 1.32-4.51)33. Being previously married but currently single during the initial interview was not statistically associated with Caesarean section (OR = 0.89; 95% CI 0.39-2.68)34. 4.2.4 Combined Analysis of Predictors and Outcomes In the following section, all five hypotheses are evaluated in sequential order. Hypothesis 1 includes univaria te and multivariate assessments, while Hypotheses 2 and 4 include multivariate an alysis using binomial, multinomial, and ordinal measurements of ur ine sugar levels. High birth weight is analyzed as both categorical and continuous for Hypotheses 3 and 5. Interaction is examined by the use of interaction terms incl uded in analysis. Figure 11 displays the specific organization of analysis for Hypotheses 2 and 3. Hypothesis 1 is excluded as it is a straightforward asse ssment of urine sugar as a predictor for high birth weight and C-section. Hypot heses 4 and 5 follow the methodology of the separate analysis of predi ctors section; each predictor and outcome pair are assessed separately using the interact ion term and confounding factors only. 32 Analysis is adjusted for alcohol and drug use during pregnancy, interview site, previous C-section, and total number of live births. 33 Analysis is adjusted for age, alcohol and drug use during pregnancy, body mass index, gestational age of the infant at birth, previous C-section, total number of live birt hs, and the total number of prenatal visits attended. 34 Analysis is adjusted for the same confounders as listed above.

PAGE 170

134Figure 11 Analysis strategies for Hypotheses 2 and 3 for the Results Chapter Hypothesis 1 Urine sugar levels during pregnan cy are positively associated with development of pregnancy complications (e.g., high birth weight and Caesarean section) as stated in Hypothesis 1. Univ ariate and multivariate analyses indicate an association between urine sugar spill and birth weight and high birth weight, but no association between sugar leve ls and Caesarean section. When examining the association bet ween urine sugar spill and birth weight, both presence or absence of sugar and varying levels of sugar in the urine are significantly asso ciated with increasing birth weight. When urine sugar is measured as present or absent and birt h weight is categorized as high or other, having any amount of sugar in ur ine is more likely among women who have high birth weight babies ( 2 = 6.754, p < 0.01). Fort y-two percent (42%) of • Analysis Strategy– Model of All Predictors with each outcome measure (urine sugar levels, high birth weight, and C-section) – Model of the Presence/Absence of specifically measured predictors and all outcomes (i.e., for a total of 3 models; outcomes are each assessed separately) – Model of predictors assessed at the initial interview (time 1 or x1) and all outcomes – Model of predictors assessed at the final interview (time 2 or x2) and all outcomes – Model of predictors that measure a change from the initial to final interviews and all outcomes – Model of all significant predictors regardless of time and all outcomes – Final model of all significant predictors and all outcomes

PAGE 171

135 women sampled who have any amount of sugar in their urine birthed babies weighing greater than 4000 gr ams compared with 22% who birthed babies less than 4000 grams. Among parti cipants, the odds of havi ng a high birth weight baby are 2.61 times higher with any amount of sugar in their urine compared with participants who lack sugar in their uri ne (OR = 2.61; 95% CI 1.24-5.50). However, when controlling for confounding factors, adjusted logistic regression indicates no statistically significant a ssociation between urine sugar spill and high birth weight (OR = 0. 49; 95% CI 0.22-1.10)35. When urine sugar spill is divided into levels of no detectable spill, low amounts of sugar, or high amounts; high ur ine sugar spills are associated with high-weight births ( 2 = 9.951, p < 0.01). Essentia lly, 32% of women with high amounts of sugar in their urine had a high birth weight baby compared with 12% who birthed a normal weight baby and had high amounts of sugar in their urine. Equal proportions of women with low am ounts of sugar in their urine had both normal and high birth weight babies, and only 5% of women with no detectable sugar spill gave birth to high-weight babies Participants with a high amount of urine sugar are three and a half times more likely to have a high birth weight baby compared with participants with no detec table levels of sugar (OR = 3.50; 95% CI 1.54-7.96). These re sults of ordinal urine suga r levels and birth weight indicate a positive association. A fter adjusting for confounding, there is a significant association between the high urine sugar spill group and high birth 35 Analysis is adjusted for education level attained, gestational age of the infant at birth, and the total number of prenatal visits attended.

PAGE 172

136 weight, but not between low sp ill and high birth weight (l ow spill OR = 0.89; 95% CI 0.23-3.43). Participant s with a high spill are more than three times likely to have a high birth weight baby com pared with participants who have no detectable sugar spill in their uri ne (OR = 3.25; 95% CI 1.30-8.10)36. Since the high urine sugar spill group alone is signifi cantly associated with having high birth weight babies, dichotomizi ng urine sugar spill to hi gh and low to none exemplifies the relationship between this specific group and high-weight infant births. Participants with high urine sugar spill are 3.30 times more likely to have a high birth weight baby compared with participant s with low to no detectable levels of spill after adjusting for confounding fa ctors (OR = 3.30; 95% CI 1.35-8.08)37. When birth weight is treated as continuous, the positive association between increasing urine sugar and increasing birth wei ght remains. The mean difference in birth weight (kilograms) bet ween the two groups of participants is 1.024 (no detectable sugar spill = 3.175; sugar spill = 3.336; t = -2.745; p < 0.01). However, when urine sugar levels are m easured ordinally, only trace amounts of spill are significantly associ ated with increasing birth wei ght (t = 3.59; p < 0.001; high spill t = 0.87; p > 0.05) Controlling for the confounding factors, any urine sugar spill is associated with high birth weight (Table 58)38. The regression equation for any spill versus no detectable spill is listed: 36 Analysis is adjusted for education level attained, gestational age of the infant at birth, and the total number of prenatal visits attended. 37 Analysis is adjusted for the same se t of confounding factors as listed above. 38 For the residual plots, see Appendix E.

PAGE 173

137 birth weight ( ) = -4.450 + 0.785(sugar sp ill) + 0.415(education) + 0.855(gestation) + 0.979(prenatal) Low amounts of sugar are associated with increasing birth weight as a continuous measure (Table 59). T he regression equation for low and high amounts of sugar compared with no detect able levels is located below: birth weight ( ) = -4.426 + 1.073(lowspi ll) + 0.510(highspill) + 0.396(education) + 0.851(gestation) + 0.992(prenatal) Table 58 Multiple regression model of urine sugar level as a dichotomous measure and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Multiple Regression Model (N = 505) F-test 107.786; p < 0.001 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -4.450 1.532 -12.928 Urine Sugar Level 0.785* 0.286 2.158 Education Level 0.415 0.167 1.033 Gestational Age of In fant 0.855* 0.042 17.335 Total Number of Prenatal Visits 0.979* 0.035 2.739 *p<0.05. Table 59 Multiple regression model of urine sugar level as an ordinal measure and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Multiple Regression Model (N = 505) F-test 87.208; p < 0.001 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -4.426 1.534 -12.769 Urine Sugar Level Low Level 1.073* 0.417 2.762 High Level0.510 0.350 0.745 Education Level 0.396 0.167 0.942 Gestational Age of In fant 0.851* 0.042 17.179 Total Number of Prenatal Visits 0.992* 0.035 2.814 *p<0.01.

PAGE 174

138 Caesarean section and urine sugar leve ls are not significantly associated (OR = 0.49; 95% CI 0.22-1.10)39. Among participants, pr esence or absence of sugar in the urine makes no difference in the number of C-sections performed ( 2 = 1.633; p > 0.05). There is no trend in the proportion of C-sections as sugar in the urine increases ( 2 = 1.701; p > 0.05; Table 60). In addition, birth weight and Caesarean section are not associated ( 2 = 4.298; p > 0.05). Although, a higher proportion of women who delivered by Csection had high birth weight babies compared with women who had normal birth weight babies (hbw = 29%; normal = 18%). Table 60 Logistic regression model of urine sugar level as an ordinal measure and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model (N = 506)* Independent Variables Odds Ratio 95% Confidence Interval Urine Sugar Level No Detectable Level1.000 Low Level1.165 0.541 2.509 High Level1.334 0.709 2.510 *Adjusted for alcohol and drug use duri ng pregnancy, education level attained, pre-pregnant weight, and total number of prenatal visits. Hypothesis 2 As hypothesized, specific socio-cultur al factors (e.g., physical strain, lack of social support, depression, autonom y, pregnancy wantedness, and physical and verbal abuse) during pr egnancy were associated with urine sugar levels. According to initial methodology, urine sugar was analyzed as a binomial 39 Analysis is adjusted for alcohol and drug use during pregnancy, education level attained, pre-pregnant weight, and the total number of prenatal visits attended.

PAGE 175

139 (presence and absence), multinomial (no det ectable sugar, low sugar levels, high sugar levels), and ordinal measure. Initially, each predictor was modeled separately, together in one model, and temp orally. As previously described in the separate analysis of predictors and outcomes section, binomial or dichotomous measurement of urine sugar spill was adequately powered for all analyses (>/= 80%). However, multi nomial and ordinal analyses were both underpowered and yielded no statisti cally significant results40. Only the results of t he binomial analysis are de scribed in the assessment of Hypothesis 2. For binom ial analysis, predictors ar e assessed as present or absent during the course of the entire pr egnancy, for their effect s from the initial interview during the first trimester, for their impact on the second and third trimesters (final intervie w), and for any temporal effe ct between interviews (for scale measures assessed during both interv iews). All significant predictors are combined and included in final models. Binomial analysis of all predictors including confounding factors yielded one significant predictor, mari tal status (Table 61). Due to the reduction of sample size by examining participants who work outside the home for pay to assess physical work strain (eliminated mo re than three-fourths of the sample); significant results are applicable to a specif ic subset of participants only. Further analyses are, therefore, required to expl ore the associations of predictors with urine sugar levels beyond assessing the e ffects of all predictors simultaneously. 40 Refer to Figure 4 for a reminder of the analysis strategy.

PAGE 176

140Table 61 All predictors in one logistic regres sion model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200141 Full Model of All Predictors (N = 100)* Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Yes1.000 No1.847 0.037 92.945 Partner Social Support Scale 1.155 0.575 2.322 Mother Social Support Yes1.000 No0.027 0.000 3.469 Mother Social Support Scale 0.758 0.341 1.686 Marital Status t1 Married1.000 Single/Never Married0.110 0.004 3.085 Single/Ever Married 102.24 1.696 6161.63 Marital Status t2 Married1.000 Single/Never Married5.965 0.257 139.399 Single/Ever Married0.020 0.000 2.142 Total Abuse No1.000 Yes1.403 0.034 57.325 Abuse t1 No1.000 Yes0.904 0.057 14.368 Abuse t2 No1.000 Yes0.063 0.002 1.899 Depression Scale t1 1.073 0.695 1.655 Depression Scale t2 1.053 0.690 1.607 Difference in Depression Score 0.939 0.574 1.535 Physical Work Strain t1 1.658 0.782 3.516 Physical Work Strain t2 1.013 0.619 1.658 Pregnancy Wantedness Yes1.000 No1.134 0.224 5.747 *Adjusted for age, alcohol and drug us e during pregnancy, body mass index, educ ation level attained, previous Csections, total number of abortions, total number of live birt hs, total number of pregnancies, total number of premature births, total number of prenatal visits and weight gain during pregnancy. 41 The upper ends of the 95% Confidence Intervals for predict ors such as partner social support, single/ever married status during both the initial and final interviews, and history of physical or verbal abuse are high due to a lack of variabil ity in responses for both the dichotomous out come of urine sugar and the categoric al measures of each predictor. The reduced sample size also affected the range of the 95% CI ’s, and the model does not present a good fit (see Appendix D).

PAGE 177

141 As discussed in the descriptive pr edictor variables section, binary variables were created to assess t he presence or absence of predictors measured at two time points in the study Table 62 presents results of these predictors on urine sugar levels (pres ence was defined as occurrence at any point during the pregnancy regardless of ti me; these predictors were measured at both interviews). Although the sample size includes over 80% of the sample, none of the predictors that were measured as pres ent or absent during the pregnancy are associated with presenc e of sugar in the urine. Table 62 Logistic regression model of predictors assessed as present or absent during pregnancy and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Present or Abs ent Predictors (N = 428)* Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support 0.671 0.358 1.258 Mother Social Support 1.733 0.917 3.276 Total Abuse No1.000 Yes0.664 0.407 1.083 Pregnancy Wantedness Yes1.000 No0.663 0.402 1.092 *Adjusted for age, alcohol and drug us e during pregnancy, body mass index, education level attained, total number of premature births, total number of prenatal visits, and weight gain during pregnancy. Further analysis includes temporally assessing the impact of specific predictors and their effect on urine sugar levels. Measurements from both the initial and final interviews are modeled to determine the impact of predictors on urine sugar levels. Models including and excluding physical work strain scale measurements are analyzed due to the high proportion of women in the sample who did not work during their pregnancy.

PAGE 178

142 Table 63 presents results from the init ial interview excluding physical work strain scale measurements. Only depression is signi ficantly associated with urine sugar spill. For every one point in crease in reported depressive symptoms (scale 0 – 38), participants are 4% more likely to have sugar spill in their urine compared with participants who have decre asing depression scores (OR = 1.04; 95% CI 1.00-1.08). Using the same model and including physical work strain scores, the depression scale remains si gnificantly associated with urine sugar levels (Table 64). Among this group, participants with in creasing depression scores are 7% more likely to have ur ine sugar spill compared with participants who have decreasing depression scores (OR = 1.06; 95% CI 1.01-1.13). Table 63 Logistic regression model of predictors assessed from the initial interview during the first trimester and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Initial Interview Predictors (N = 491)* Excluding Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.032 0.932 1.144 Mother Social Support Scale 0.930 0.839 1.031 Marital Status Married1.000 Single/Never Married1.046 0.625 1.748 Single/Ever Married0.805 0.339 1.909 Abuse t1** No 1.000 Yes 1.214 0.733 2.012 Depression Scale t1 1.039 1.000 1.079 *Adjusted for age, body mass index, education level attained, pr evious of C-sections, total number of premature births, and weight gain during pregnancy. ** t1 refers to data collected during the initial interview.

PAGE 179

143Table 64 Logistic regression model of predictors assessed from the initial interview during the first trimester and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Initial Interview Predictors (N = 230)* Including Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.148 0.974 1.353 Mother Social Support Scale 0.999 0.841 1.187 Marital Status Married1.000 Single/Never Married1.125 0.538 2.352 Single/Ever Married1.597 0.476 5.360 Abuse t1** No1.000 Yes1.387 0.617 3.119 Depression Scale t1 1.068 1.010 1.129 Physical Work Strain Scale 1.061 0.840 1.340 *Adjusted for age, body mass index, education level attained, pr evious of C-sections, total number of premature births, and weight gain during pregnancy. ** t1 refers to data collected during the initial interview. Analysis from the final interview (e .g., the second and third trimesters of pregnancy) includes all predictors measured at the end of the pr egnancy. In the model excluding the assessment of physical work strain, social support by the mother of the participant (scale) is signi ficantly associated with urine sugar spill (Table 65). A mother’s support decreases the risk of sugar in the urine. Participants with increasing social support from their mothers are 13% less likely to develop sugar spills in their urine (f or each increase in the social support scale) compared with participants who have dec reasing social support from their mothers (OR = 0.87; 95% CI 0.79-0.99) Depression is also significantly associated with urine sugar spill. For ev ery one point increase in the depression scale, participants are 5% more likely to have urine sugar spill compared with participants who have decreasing depression scores (OR = 1.05; 95% CI 1.00-

PAGE 180

144 1.09). When examining t he same model including physical work strain, no predictors are significantly associat ed with urine sugar spill (Table 66). Table 65 Logistic regression model of predictors assessed from the final interview during the third trimester and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 423)* Excluding Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.014 0.904 1.136 Mother Social Support Scale 0.886 0.793 0.990 Marital Status t2** Married1.000 Single/Never Married1.086 0.623 1.895 Single Ever Married0.543 0.178 1.657 Abuse t2 No1.000 Yes0.742 0.410 1.343 Depression Scale t2 1.045 1.000 1.092 Pregnancy Wantedness Yes1.000 No0.758 0.446 1.289 *Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, previous C-section, total number of live births, total number of pregnancies, total number of premature births, and weight gain during pregnancy. ** t2 refers to data collected at the final interview.

PAGE 181

145Table 66 Logistic regression model of predictors assessed from the final interview during the third trimester and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 203)* Including Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.028 0.840 1.258 Mother Social Support Scale 0.871 0.712 1.065 Marital Status t2** Married1.000 Single/Never Married1.015 0.345 2.989 Single Ever Married0.295 0.019 4.694 Abuse t2 No 1.000 Yes 0.381 0.111 1.307 Depression Scale t2 1.086 0.997 1.183 Pregnancy Wantedness Yes1.000 No0.696 0.250 1.941 Physical Work Strain Scale 1.215 0.933 1.583 *Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, previous C-section, total number of abortions, total number of live births, tota l number of pregnancies, total number of premature births, and weight gain during pregnancy. ** t2 refers to data collected during the final interview. The only scale with a large enough sample to assess the difference in scores is the depression scale. The differ ence in scores between the initial and final interviews is modeled with urine s ugar and yielded no significant association (OR = 0.31; 95% CI 0.95-1.03)42. Final analyses included combining pr edictors found to be significant in the separate analysis with significantly asso ciated predictors in the multivariate modeling section. Initially, since hist ory of abuse and abuse during the second and third trimesters of pregnancy were si gnificantly associated with urine sugar levels, a model was analyzed combining both predictors. However, due to 42 No confounding factors were used in adjustment of the difference in depression scores and urine sugar levels. For further explanation see the assessment of confounding factors section.

PAGE 182

146 multicollinear effects, nei ther predictor was significantly associated with urine sugar readings. Therefore, when combined with the predict ors of the multivariate analysis, two separate models incorporat ing history of abuse and abuse during the latter half of pregnancy were constructed. The same premise follows for meas uring depression during the initial interview and depression during the sec ond and third trimesters. Four final models are presented in Tables 67 thr ough 70. Table 67 pres ents results from the model including history of abuse, marita l status, mother’s total social support scale, and depression from the initial in terview. There are no statistically significant predictors of urine sugar spill in this model. Table 68 summarizes the same analysis with depression during the second half of the pregnancy, excluding depression from the initial inte rview. Here, as social support of participants’ mothers decreases, they are 12% more likely to have sugar in their urine compared with participants who have in creasing social support from their mother (OR = 0.90; 95% CI 0.81-0.99). Also, for every one point increase in depression, participants are 5% more likely to have sugar present in their urine compared with participants who have decre asing depression scores in the latter half of the pregnancy (OR = 1.05; 95% CI 1.00-1.09). Tables 69 and 70 comprise the same analysis, except abuse during the second and third trimesters is included instead of hist ory of abuse. Results from Table 69 indicate no associations between the predictors and uri ne sugar spill. However, both social support of a participant’s mother and depression during the latter half of the

PAGE 183

147 pregnancy are significantly associated wit h urine sugar spill in Table 70. Participants who believe they lack the support of their mother are 11% more likely to have sugar in their urine ( dose-response), compared with women who believe they have increasing social suppor t (OR = 0.90; 95% CI 0.81-1.00). For every point increase in depression, partici pants are 4% more likely to have sugar spill compared with participants who have decreasing depression scores (OR = 1.04; 95% CI 1.00-1.09). Table 67 Final logistic regression predictor model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 430)* Excluding Abuse and Depression During Second and Third Trimesters Independent Variables Odds Ratio 95% Confidence Interval Marital Status Married1.000 Single/Never Married1.055 0.626 1.780 Single/Ever Married1.043 0.427 2.549 Mother Social Support Scale 0.902 0.811 1.003 History of Abuse No1.000 Yes0.771 0.462 1.287 Depression Scale t1 1.040 0.999 1.082 *Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, and total number of premature births.

PAGE 184

148Table 68 Final logistic regression predictor model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 430)* Excluding Abuse During Second and Third Trimesters; Depression in the First Trimester Independent Variables Odds Ratio 95% Confidence Interval Marital Status Married1.000 Single/Never Married1.071 0.633 1.811 Single/Ever Married1.047 0.430 2.550 Mother Social Support Scale 0.895 0.805 0.994 History of Abuse No1.000 Yes0.807 0.478 1.363 Depression Scale t2 1.045 1.003 1.088 *Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, and total number of premature births. Table 69 Final logistic regression predictor model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 430)* Excluding History of Abuse Independent Variables Odds Ratio 95% Confidence Interval Marital Status Married1.000 Single/Never Married1.056 0.626 1.781 Single/Ever Married1.077 0.440 2.632 Mother Social Support Scale 0.904 0.813 1.007 Abuse During Second and Third Trimesters of Pregnancy No1.000 Yes0.729 0.417 1.277 Depression Scale t1 1.039 0.998 1.081 *Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, previous C-section, and total number of premature births.

PAGE 185

149Table 70 Final logistic regression predictor model and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 430)* Excluding History of Abuse Independent Variables Odds Ratio 95% Confidence Interval Marital Status Married1.000 Single/Never Married1.071 0.634 1.811 Single/Ever Married1.073 0.439 2.618 Mother Social Support Scale 0.898 0.807 0.999 Abuse During Second and Third Trimesters of Pregnancy No1.000 Yes0.757 0.430 1.333 Depression Scale t2 1.043 1.001 1.087 *Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, previous C-section, and total number of premature births. Hypothesis 3 Hypothesis 3 consists of an examinat ion of the impact of psychosocial and physical factors on high birth weight and Caesarean section births. Similar methodology is used in analysis as was wit h Hypothesis 2. All predictor models are used in examining associations bet ween these factors and birth outcomes; then temporally related models are asse ssed. Specifically, predictors are modeled as present or absent during the c ourse of the pregnancy, present in the initial interview, present in the final interview, and changing between the initial and final interviews if measured at both time points43. For high birth weight, no predictors in the all inclusive model are significantly associated with high birth we ight (Table 71). Due to the lack of power for such an analysis, physical work st rain from the final interview is omitted 43 Refer to Figure 4 for a reminder of the analysis strategy.

PAGE 186

150 from analysis as the sample size signifi cantly decreased when it was initially included. Similarly, when predictors are assessed as present or absent throughout the pregnancy, none are associated with high birth weight (Table 72).

PAGE 187

151Table 71 All predictors in one logistic regr ession model and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200144 Full Model of All Predictors (N = 203)* Independent Variables** Odds Ratio 95% Confidence Interval Partner Social Support Yes1.000 No13.510 0.059 3100.53 Partner Social Support Scale 1.797 0.729 4.431 Mother Social Support Yes1.000 No0.003 0.000 2.37E+18 Mother Social Support Scale 2.384 0.908 6.263 Marital Status t1 Married1.000 Single/Never Married0.067 0.003 1.696 Single/Ever Married0.104 0.001 7.282 Marital Status t2 Married1.000 Single/Never Married6.354 0.305 132.387 Single/Ever Married42.184 0.928 1917.039 Total Abuse No1.000 Yes0.125 0.002 6.979 Abuse t1 No1.000 Yes9.108 0.480 172.910 Abuse t2 No1.000 Yes0.673 0.028 16.142 Depression Scale t1 1.007 0.655 1.548 Depression Scale t2 1.221 0.816 1.825 Difference in Depression Score 0.865 0.533 1.404 Physical Work Strain t1 1.006 0.530 1.907 Pregnancy Wantedness Yes1.000 No2.276 0.394 13.150 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant, previous C-section, total number of premature births, tota l number of prenatal visits, and weight gain during pregnancy. **Physical work strain from the final interview is e xcluded due to the reduction of sample size and adequate power for analysis (N = 100). 44 The large range of the 95% Confidence Intervals of specific predictors such as partner so cial support, mother’s social support, marital status measured at the final interview, physical or verbal abuse measured at both interviews, and pregnancy wantedness are due to a lack of variability in par ticipant response to these ca tegorical predictors and the dichotomous measure of high and other birth weight. The r eduction in sample size also affected the variability, and the model does not present a good fit (see Appendix D).

PAGE 188

152Table 72 Logistic regression model of predictors assessed as present or absent during pregnancy and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Present/Absent Predictors (N = 427)* Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support 0.808 0.249 2.617 Mother Social Support 0.153 0.017 1.420 Total Abuse No1.000 Yes0.459 0.189 1.114 Pregnancy Wantedness Yes1.000 No0.834 0.344 2.021 *Adjusted for body mass index, education le vel attained, gestational age of infant, total number of premature births, total number of prenatal visits, and weight gain during pregnancy. When both the initial and final interv iews are modeled, the social support of the participant’s mother is significant ly associated with high birth weight. In Table 73, as participants’ perceptions of their mother’s social support increases, they are 47% more likely to have a high birth weight baby compared with participants who lack support (OR = 1.47; 95% CI 1.10-1.96). Including physical work strain in the model, for every incr ease in the mother’s social support scale, participants are 78% more likely to have a high birth weight infant compared with participants who have decreasing social su pport from their mo thers (OR = 1.78; 95% CI 1.09-2.90; Table 74).

PAGE 189

153Table 73 Logistic regression model of predictors assessed from the initial interview during the first trimester and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Initial Interview Predictors (N = 497)* Excluding Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.077 0.881 1.316 Mother Social Support Scale 1.466 1.095 1.963 Marital Status Married1.000 Single/Never Married0.827 0.317 2.161 Single/Ever Married3.599 0.805 16.088 Abuse t1 No1.000 Yes0.379 0.142 1.012 Depression Scale t1 0.957 0.883 1.036 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant, interview site, total number of premature births, total number of prenatal visits, and weight gain. Table 74 Logistic regression model of predictors assessed from the initial interview during the first trimester and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200145 Full Model of Initial Interview Predictors (N = 230)* Including Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.097 0.817 1.474 Mother Social Support Scale 1.777 1.091 2.895 Marital Status Married1.000 Single/Never Married0.469 0.120 1.829 Single/Ever Married3.085 0.393 24.195 Abuse t1 No1.000 Yes0.999 0.234 4.266 Depression Scale t1 1.081 0.968 1.209 Physical Work Strain Scale 0.996 0.379 4.942 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant, interview site, total number of premature births, total number of prenatal visits, and weight gain. 45 The wide 95% Confidence Interval for ever married wom en is due to the lack of variability between the dichotomous predictor and outcome. Also, the reduction in sample size fr om including the work strain scale affected the variability.

PAGE 190

154 Similar to results of the initial intervie w, from the final interview, the social support of the participant’s mother is associated with high bi rth weight. When physical work strain is excluded fr om analysis (Table 75), participants with increasing support from thei r mothers are 51% more lik ely to have a high birth weight baby compared with women who have decreasing social support (OR 1.51; 95% CI 1.092.09). When physical work stra in is included in the analysis, there are no significant predictors of hi gh birth weight (Table 76). In addition, there is no association between the change in depression throughout the pregnancy and resulting high birth weight in fants (OR = 1.02; 95% CI 0.95-1.10). Table 75 Logistic regression model of predictors assessed from the final interview during the third trimester and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 427)* Excluding Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.148 0.923 1.428 Mother Social Support Scale 1.507 1.086 2.091 Marital Status Married1.000 Single/Never Married1.251 0.471 3.320 Single/Ever Married3.080 0.611 15.520 Abuse t2 No1.000 Yes0.397 0.141 1.119 Depression Scale t2 1.010 0.933 1.092 Pregnancy Wantedness Yes1.000 No0.839 0.329 2.136 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant at birth, total number of premature births, to tal number of prenatal visits, and weight gain.

PAGE 191

155Table 76 Logistic regression model of predictors assessed from the final interview during the third trimester and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200146 Full Model of Final Interview Predictors (N = 139)* Including Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.184 0.739 1.898 Mother Social Support Scale 1.118 0.684 1.829 Marital Status Married1.000 Single/Never Married0.633 0.104 3.845 Single/Ever Married5.834 0.277 122.888 Abuse t2 No1.000 Yes0.239 0.024 2.394 Depression Scale t2 1.042 0.898 1.210 Pregnancy Wantedness Yes1.000 No1.753 0.245 12.543 Physical Work Strain Scale 1.294 0.737 2.275 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant at birth, previous C-section, total number of premature births, total number of prenatal visits, and weight gain. The final models include all signif icant predictors. When examined separately, ethnicity and a mother’s so cial support, specifically emotional support, are statistically associated wit h high birth weight. When modeled together, only a mother’s social support is associated with high birth weight. Both social support and ethnicity remain significant when modeled together. Again, women with increasing social supp ort are 56% more likely to have a high birth weight baby compared with wom en who have decreasing support (OR = 1.56; 95% CI 1.13-2. 17); and White women are almost three times more likely to have a high birth weight baby compared with Black women (O R = 2.81; 95% CI 46 The wide 95% Confidence Intervals for ever married wom en and pregnancy wantedness is due to the lack of variability between the predictors and outcome. Also, the reduction in samp le size due to the inclusion of the work strain scale affected the variability.

PAGE 192

156 1.16-6.82)47. When focusing on emotional social support only, women with increasing emotional support ar e over two times more likely to have a high birth weight baby compared with women who have decreasing emotional social support (2.03; 95% CI 1.13-3.63); and Whit e women are 2.74 times more likely to have a high birth weight baby compared with Black women (O R = 2.74; 95% CI 1.14-6.63)48. The same methodology is employed fo r the continuous measure of birth weight. A model of all pr edictors is shown in Tabl e 77. No predictors are significantly associated with birth weight in the all inclusive model. When examining presence or absence of s pecific predictors, again, none are significantly associated with birth weight (Table 78). 47 Analysis is adjusted for education level attained, gestational age of the infant at birth, and the total number of prenatal visits attended. 48 Analysis is adjusted for education level attained, gestational age of the infant at birth, and the total number of prenatal visits attended.

PAGE 193

157Table 77 All predictors in one multiple regressi on model and the birth weight of infant born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of All Predictors (N = 100)* F-test 4.058; p < 0.001; R2 = 0.404 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -5.196 7.834 -4.468 Partner Social Support -1.132 1.611 -0.795 Partner Social Support Scale 0.580 0.283 1.186 Mother Social Support 1.673 1.996 1.403 Mother Social Support Scale -0.436 0.348 -0.548 Marital Status t1 Single/Never Married-0.877 1.053 -0.730 Single/Ever Married-0.142 1.484 -0.014 Marital Status t2 Single/Never Married0.769 0.959 0.616 Single/Ever Married1.065 1.590 0.714 Total Abuse -0.648 1.551 -0.271 Abuse t1 0.605 1.228 0.298 Abuse t2 1.507 1.519 1.495 Depression Scale t1 -0.321 0.201 -0.514 Depression Scale t2 0.252 0.205 0.309 Difference in Depression Score -0.349 0.228 -0.533 Physical Work Strain t1 0.489 0.242 0.988 Physical Work Strain t2 -0.423 0.187 -0.956 Pregnancy Wantednes s -0.358 0.696 -0.184 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant, previous C-section, total number of premature births, tota l number of prenatal visits, and weight gain during pregnancy. Table 78 Multiple regression model of predictors assessed as present or absent during pregnancy and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Present/Absent Predictors (N = 427)* F-test 23.684; p < 0.001; R2 = 0.347 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -5.426 2.887 -10.199 Partner Social Support 0.497 0.319 0.774 Mother Social Support 0.394 0.362 0.428 Total Abuse 0.616 0.259 1.463 Pregnancy Wantedness 0.240 0.259 0.223 *Adjusted for body mass index, education le vel attained, gestational age of infant, total number of premature births, total number of prenatal visits, and weight gain during pregnancy.

PAGE 194

158 Tables 79 through 82 present results for predictors collected at the initial and final interviews. Neither model (excl uding or including physical work strain) of predictors from the initial interview c ontains significant associations between any predictor and birth weight. For t he final interview, no predictors are significantly associated with birth weight. However, when physical work strain is included in the model, physi cal or verbal abuse dur ing the second and third trimesters is significantly associated with birth weight. Abuse is associated with high birth weight compared with parti cipants who reported no abuse. The regression equation is stated: birth weight ( ) = -5.541 + 0.290(partner) + 0.167(mother) + -0.642(single) + 0.992(evermarried) + 1.148(abuse) + -0.184(depression) + -0.205(wanted) + -0.338(strain) + -0.209(drug) + -0.238(bmi) + 0.483(education) + 0.964(gestation) + 0.503(C-sect ion) + -1.216(premature) + 0.392(prenatal) + 0.174(weight) Finally, the difference in depression scores between the initial and final interviews is not significantly associated with birth weight (Table 83).

PAGE 195

159Table 79 Multiple regression model of predictors assessed from the initial interview during the first trimester and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Full Model of Initial Interview Predictors (N = 497)* Excluding Physical Work Strain Scale F-test 33.428; p < 0.001; R2 = 0.478 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -4.384 1.832 -10.489 Partner Social Support Scale 0.174 0.054 0.561 Mother Social Support Scale 0.256 0.056 1.159 Marital Status Single/Never Married-0.533 0.286 -0.993 Single/Ever Married0.630 0.465 0.853 Abuse t1 0.663 0.282 1.562 Depression Scale t1 -0.246 0.022 -0.276 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant, interview site, total number of premature births, total number of prenatal visits, and weight gain. Table 80 Multiple regression model of predictors assessed from the initial interview during the first trimester and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Full Model of Initial Interview Predictors (N = 230)* Including Physical Work Strain Scale F-test 16.591; p < 0.001; R2 = 0.488 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -4.339 2.615 -7.199 Partner Social Support Scale 0.149 0.092 0.242 Mother Social Support Scale 0.352 0.096 1.290 Marital Status Single/Never Married-0.279 0.421 -0.185 Single/Ever Married0.947 0.715 1.253 Abuse t1 0.523 0.458 0.598 Depression Scale t1 0.083 0.034 0.208 Physical Work Strain Scale -0.166 0.129 -0.212 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant, interview site, total number of premature births, total number of prenatal visits, and weight gain.

PAGE 196

160Table 81 Multiple regression model of the predictors assessed from the final interview during the third trimester and the birth weight of infants born to pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 427)* Excluding Physical Work Strain Scale F-test 17.503; p < 0.001; R2 = 0.352 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -5.429 2.919 -10.096 Partner Social Support Scale 0.274 0.058 1.297 Mother Social Support Scale 0.236 0.059 0.939 Marital Status Single/Never Married-0.612 0.288 -1.300 Single/Ever Married0.366 0.511 0.262 Abuse t2 0.516 0.316 0.841 Depression Scale t2 0.127 0.023 0.709 Pregnancy Wantedness 0.406 0.270 0.610 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant at birth, total number of premature births, to tal number of prenatal visits, and weight gain. Table 82 Multiple regression model of predictors assessed from the final interview during the third trimester and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Full Model of Final Interview Predictors (N = 139)* Including Physical Work Strain Scale F-test 7.483; p < 0.001; R2 = 0.429 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -5.451 5.221 -5.691 Partner Social Support Scale 0.290 0.098 0.857 Mother Social Support Scale 0.167 0.112 0.250 Marital Status Single/Never Married-0.642 0.550 -0.750 Single/Ever Married0.992 1.223 0.805 Abuse t2 1.148 0.668 1.974** Depression Scale t2 -0.184 0.043 -0.792 Pregnancy Wantednes s -0.205 0.505 -0.083 Physical Work Strain Scale -0.338 0.129 -0.889 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant at birth, previous C-section, total number of premature births, total number of prenatal visits, and weight gain. **p < 0.05.

PAGE 197

161Table 83 Multiple regression model of predictor difference scores between the initial and final interviews and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Full Model of Predictors Difference Scores (N = 433) F-test 0.067; p > 0.05; R2 = 0.000 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept 3.272 0.153 69.827 Difference in Depression Scale Scores 0.087 0.030 0.258 When abuse during the second and th ird trimesters is modeled with ethnicity, ethnicity is the only predictor si gnificantly associated with birth weight. White women in the sample have the hi ghest birth weight babies compared with Black women in the sample (Table 84). The regression equation is stated: birth weight ( ) = -2.927 + 0.597(abuse) + 1.0 48(ethnic) + -0.683(drug) + 0.222(bmi) + 0.430(education) + 0.964(gestation) + 0.175(weight) Table 84 Final multiple regression predictor model and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 139)* Including Physical Work Strain Scale F-test 22.901; p < 0.001; R2 = 0.337 Independent Variables Coefficient Estimate Standard Error t Statistic Intercept -2.927 3.243 -8.568 Abuse t2 0.597 0.315 1.135 Ethnicity 1.048 0.295 3.727** *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant at birth, and weight gain. **p < 0.001.

PAGE 198

162 The same analytic methodology used for predictors of high birth weight is also used for predictors of C-section. When all predictors are included in a full model, there are no predictors significantly associated with C-se ction (Table 85). When assessing the presence or absenc e of predictors throughout the pregnancy on C-section birth, hi story of physical or verbal abuse is significantly associated with C-sections (Table 86). Pa rticipants with a history of physical or verbal abuse are over two times more lik ely to have a C-section birth compared to participants who do not have a histor y of abuse (OR = 0.49; 95% CI 0.270.90).

PAGE 199

163Table 85 All predictors in one logistic regression model and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200149 Full Model of All Predictors (N = 203)* Independent Variables** Odds Ratio 95% Confidence Interval Partner Social Support Yes1.000 No13.108 0.778 220.959 Partner Social Support Scale 1.589 0.967 2.612 Mother Social Support Yes1.000 No4.107 0.244 69.190 Mother Social Support Scale 1.363 0.850 2.187 Marital Status t1 Married1.000 Single/Never Married0.889 0.188 4.215 Single/Ever Married4.142 0.251 68.358 Marital Status t2 Married1.000 Single/Never Married0.346 0.081 1.471 Single/Ever Married0.904 0.041 19.788 Total Abuse No1.000 Yes1.453 0.111 18.952 Abuse t1 No1.000 Yes2.184 0.377 12.663 Abuse t2 No1.000 Yes0.094 0.008 1.087 Depression Scale t1 1.302 0.931 1.821 Depression Scale t2 0.891 0.647 1.225 Difference in Depression Score 1.288 0.901 1.842 Physical Work Strain t1 0.969 0.672 1.399 Pregnancy Wantedness Yes1.000 No2.487 0.713 8.680 *Adjusted for age, alcohol and drug us e during pregnancy, body mass index, educat ion level attained, gestational age of infant, previous C-section, total number of liv e births, and total number of prenatal visits. **Physical work strain from the final interview is e xcluded due to the reduction of sample size and adequate power for analysis (N = 100). 49 As with previous tables, the large upper end of the 95% Co nfidence Intervals for specific predictors such as partner support, mother’s support, marital status, and physical or ve rbal abuse are due to the lack of variability between the dichotomous outcome of C-section and the categorical predict ors. The reduction in sample size also affected the variability.

PAGE 200

164Table 86 Logistic regression model of predictors assessed as present or absent during pregnancy and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Present/Absent Predictors (N = 425)* Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support 0.976 0.464 2.052 Mother Social Support 1.308 0.584 2.932 Total Abuse No1.000 Yes 0.498 0.274 0.903 Pregnancy Wantedness Yes1.000 No1.117 0.611 2.041 *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, interview site, previous C-section, total number of live births and total number of prenatal visits. None of the predictors are signific antly associated from the initial interview. Even when physical work strain is included in the model, no significant associations are indicated (Tables 87 and 88). Table 87 Logistic regression model of predictors assessed from the initial interview during the first trimester and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Initial Interview Predictors (N = 495)* Excluding Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.006 0.894 1.132 Mother Social Support Scale 1.090 0.958 1.241 Marital Status Married1.000 Single/Never Married0.568 0.307 1.049 Single/Ever Married1.167 0.428 3.180 Abuse t1 No1.000 Yes0.884 0.483 1.616 Depression Scale t1 1.035 0.989 1.084 *Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, previous C-section, total number of live births, and total number of prenatal visits.

PAGE 201

165Table 88 Logistic regression model of predictors assessed from the initial interview during the first trimester and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Initial Interview Predictors (N = 230)* Including Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.138 0.925 1.400 Mother Social Support Scale 1.084 0.869 1.353 Marital Status Married1.000 Single/Never Married0.509 0.204 1.273 Single/Ever Married2.534 0.553 11.616 Abuse t1 No1.000 Yes1.483 0.541 4.064 Depression Scale t1 1.068 0.995 1.147 Physical Work Strain Scale 0.947 0.705 1.272 *Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, previous C-section, total number of live births, and total number of prenatal visits. Tables 89 and 90 present results for t he final interview. Excluding the physical work strain scale, marital status is associated with C-section birth. Participants who single are over two ti mes less likely to have a C-section at delivery compared with partici pants who are currently married or living with a partner (OR = 0.37; 95% CI 0.18-0.72). When physical wo rk strain is included in the model, depression is a ssociated with C-section birth. For every one point increase in the depression scale, partici pants are 14% more likely to have a Csection birth compared with participants who have decreasing depression scores (OR = 1.14; 95% CI 1.031.27). The difference in depression scores is not associated with C-section births (OR = 0.98; 95% CI 0.93-1.03).

PAGE 202

166 Table 89 Logistic regression model of predictors assessed from the final interview during the third trimester and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 424)* Excluding Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 0.952 0.828 1.096 Mother Social Support Scale 1.013 0.880 1.167 Marital Status Married1.000 Single/Never Married 0.365 0.184 0.722 Single/Ever Married0.906 0.285 2.883 Abuse t2 No1.000 Yes0.755 0.370 1.541 Depression Scale t2 1.052 0.999 1.108 Pregnancy Wantedness Yes1.000 No1.003 0.526 1.910 *Adjusted for age. alcohol and drug us e during pregnancy, body mass index, educat ion level attained, gestational age of infant at birth, previous C-section, total number of live births, and total number of prenatal visits.

PAGE 203

167Table 90 Logistic regression model of predictors assessed from the final interview during the third trimester and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200150 Full Model of Final Interview Predictors (N = 140)* Including Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support Scale 1.018 0.784 1.323 Mother Social Support Scale 0.997 0.768 1.296 Marital Status Married1.000 Single/Never Married1.311 0.281 6.127 Single/Ever Married1.339 0.030 59.141 Abuse t2 No1.000 Yes0.739 0.160 3.400 Depression Scale t2 1.144 1.030 1.270 Pregnancy Wantedness Yes1.000 No2.474 0.624 9.815 Physical Work Strain Scale 0.766 0.539 1.090 *Adjusted for age. alcohol and drug us e during pregnancy, body mass index, educat ion level attained, gestational age of infant at birth, previous C-section, total number of live births, and total number of prenatal visits. The final model is presented in Table 91. Marital status is the only factor associated with C-section. Single participants are over two times less likely to have a C-section birth compared with marri ed participants or those living with a current partner (OR = 0. 46; 95% CI 0.21-1.00). 50 The wide 95% Confidence Interval for ever married wom en is due to the lack of variability between the predictor and outcome of Caesarean section. The reduction in sample size fr om the inclusion of work strain affected the variability.

PAGE 204

168Table 91 Final logistic regression predictor model and Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Full Model of Final Interview Predictors (N = 347)* Including Physical Work Strain Scale Independent Variables Odds Ratio 95% Confidence Interval Ethnicity Black1.000 White1.481 0.715 3.068 History of Abuse No1.000 Yes0.587 0.289 1.192 Marital Status Married1.000 Single/Never Married 0.462 0.213 1.002 Single/Ever Married1.127 0.367 3.465 Depression Scale t2 1.035 0.980 1.093 *Adjusted for age, alcohol and drug us e during pregnancy, body mass index, educat ion level attained, gestational age of infant at birth, previous C-section, total number of live births, and total number of prenatal visits. Hypothesis 4 Ethnic differences in associati ons between psychosocial and physical factors and urine sugar levels are the foci of analysis for this section. Due to the exploratory nature of this analysis, inte raction is assessed for each predictor separately only using the bi nomial measure of urine s ugar level (e.g., presence or absence). Consist ent with the Rothman’s Modern Epidemiology alpha is set at 0.20 for significance [174]. Only marital status at the initial an d final interviews, the physical work strain scale from the final interview, and the partner soci al support scale significantly interact with ethnicity. Tabl e 92 includes the model for marital status during the initial interview and urine s ugar levels. For comparative purposes, both groups of White and Bl ack single, never marri ed women and White married

PAGE 205

169 women are evaluated with Black married wom en as the reference. Black single women are two times more likely to hav e sugar in their urine compared with married Black women while White single women are 39% more likely compared with Black married women. White marri ed women are over two and half times more likely to have sugar spill compared with Black married women. Qualitative interaction is present between marital status and ethnicity; that is the direction of interaction is in opposite directions for Black versus White women. Clearly, the lowest risk group is Black married wom en, followed by White and Black single women. White married women are at highes t risk for urine sugar spill. Table 93 presents results of the inte raction between marital status from the final interview and ethnicity on urine sugar levels. Re sults are similar to those assessing marital status from the initial interview. Comparing all groups, it appears that Black married women no longer receive the mo st protective effect or have the lowest odds of urine sugar spill. Instead, White single women are at lowest risk. The highest risk group is Black single wo men, a change from the initial interview where White married women were most at risk for urine sugar spill.

PAGE 206

170Table 92 Logistic regression model of the interaction between marital status from the initial interview during the first trimester and ethnicity with urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Model of Interaction Terms and Predictors (N = 394)+ Independent Variables Odds Ratio 95% Confidence Interval Marital Status t1 Married1.000 Single/Never Married2.013 0.778 5.209 Single/Ever Married1.563 0.434 5.629 Ethnicity Black1.000 White2.546 0.970 6.685 Ethnicity*Marital Status t1 Single/Never Married0.271++ 0.75 0.973 Single/Ever Married0.423 0.073 2.450 +Adjusted for age and weight gain during pregnancy. ++p-value < 0.05. Table 93 Logistic regression model of the interaction between marital status from the final interview during the third trimester and ethnicity with urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Model of Interaction Terms and Predictors (N = 394)+ Independent Variables Odds Ratio 95% Confidence Interval Marital Status t2 Married1.000 Single/Never Married1.912 0.811 4.506 Single/Ever Married0.990 0.219 4.474 Ethnicity Black1.000 White1.757 0.745 4.142 Ethnicity*Marital Status t2 Single/Never Married0.287++ 0.071 1.161 Single/Ever Married0.617 0.068 5.560 +Adjusted for age, alcohol and drug use during pregnancy, prev ious C-section, total number of premature births, total number of live births, total number of pr egnancies, and weight gain during pregnancy. ++p-value < 0.10. Table 94 presents results of the in teraction between ethnicity and partner social support. Black women with no so cial support are the reference group.

PAGE 207

171 White women with either s upport or no support are both at higher risk of urine sugar spill compared with Bl ack women who receive no support. Black women who have support are over three times less likely to have urine sugar spill compared with Black wom en who have no support. The interaction between ethnicity and partner social support is quantitative indicating that regardless of support level, White women are at higher risk of urine sugar spill compared with Black women. Partner support is protec tive among Black women as expected; however, it increases the risk among the White women in the sample. Table 94 Logistic regression model of the interaction between partner social support and ethnicity with urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Model of Interaction Terms and Predictors (N = 394)+ Independent Variables Odds Ratio 95% Confidence Interval Partner Social Support 0.314++ 0.114 0.865 Ethnicity Black1.000 White1.143 0.673 1.940 Ethnicity*Partner Social Support 3.519+++ 0.843 14.685 +Adjusted for body mass index, education level attained, and weight gain during pregnancy. ++p-value < 0.05. +++p-value < 0.10. Ethnicity and physical work strain dur ing the second and third trimesters of pregnancy interact with urine sugar levels. Table 95 presents results. Again, White women appear to be at a higher risk for urine sugar spill. White women with a score of zero on the physical work strain scale are three and half times more likely to have urine sugar spill co mpared with Black women who score zero on the work strain scale. White wom en who score one point or higher on the scale are still over three times more lik ely to have urine s ugar spill compared with

PAGE 208

172 Black women. Black women who have a score of one point or higher on the scale are 24% more likely to have sugar in their urine compared with Black women who do not experience work strain. Figure 12 pictorially presents results of this qualitative interaction. As with partner support, increasing the work strain score decreases the risk of sugar sp ill among White women (OR = 2.782 for a score of 2; OR = 2.467 fo r a score of 3; etc.), and increases risk among Black women substantially (OR = 1.540 for a score of 2; OR = 1.912 for a score of 3; etc.). Table 95 Logistic regression model of the interaction between physical work strain during the second and third trimesters and ethnicity with urine sugar levels of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200151 Model of Interaction Terms and Predictors (N = 394)+ Independent Variables Odds Ratio 95% Confidence Interval Physical Work Strain t2 1.242 0.943 1.635 Ethnicity Black1.000 White3.535 0.664 18.819 Ethnicity*Physical Work Strain t2 0.715++0.458 1.114 +Adjusted for weight gain during pregnancy. ++p-value < 0.20. 51 The large range of the 95% Confidence Interval for ethnici ty and urine sugar levels is due to the lack of variability between the two dichotomous variables. The majori ty of urine sugar spills are among White women.

PAGE 209

173Figure 12 Logistic regression model of the inte raction between ethnicity and physical work strain in the second and third trimesters on urine sugar levels of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 (All possible scores; Black/0 score as reference), with triangles representing White women and squares representing Black women Hypothesis 5 That the associations between ps ychosocial and physical factors and pregnancy complications diffe r among ethnic groups is the final hypothesis under analysis. Interaction between predictors and ethnicity on birth weight and high birth weight are examined fi rst, followed by interactions with C-section births. Three predictors interact with ethnicity on high birth weig ht: history of physical or verbal abuse, physical or verbal abuse during the second and third trimesters, and social support of the participant’s mother. Table 96 presents the results of inte raction between ethnicity and history of abuse. The odds of White women hav ing a high birth weight baby are high 0 0.5 1 1.5 2 2.5 3 3.5 4 Strain = 0123456 Physical Work StrainOdds Ratios White Black

PAGE 210

174 regardless of a history of abuse (OR = 1.60 for non-abused women; OR = 4.08 for abused women) compared with Black non-abused women. Table 97 shows the results for physical or verbal abuse during the second and third trimesters of pregnancy with ethnicity on high birth weight. Similar to history of abuse, abuse during the second and third trimesters am ong White women greatly increases the risk of high birth weight (OR = 5.31) compared with Black non-abused women. Even White non-abused women have higher odds of high birth weight as an adverse birth outcome co mpared with Black non-abused women (OR = 1.58). Table 96 Logistic regression model of the interaction between history of physical or verbal abuse and ethnicity with high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990200152 Model of Interaction Terms and Predictors (N = 346)+ Independent Variables Odds Ratio 95% Confidence Interval History of abuse No1.000 Yes1.740 0.314 9.641 Ethnicity Black1.000 White 7.096++ 1.384 36.385 Ethnicity*History of abuse 0.226+++ 0.029 1.789 +Adjusted for education level attained, gestational age of the infant at birth, total number of prenatal visits, and weight gain during the pregnancy. ++p-value < 0.05. +++p-value < 0.20. 52 The large range in the 95% Confidence Interval for ethnici ty and high birth weight is due to the lack of variability between the two dichotomous variables. The majority of high birth weight infants are born to White women in the sample.

PAGE 211

175Table 97 Logistic regression model of the interaction between physical or verbal abuse during the second and third trimesters and ethnicity with high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200153 Model of Interaction Terms and Predictors (N = 346)+ Independent Variables Odds Ratio 95% Confidence Interval Abuse t2 No1.000 Yes0.404 0.045 3.654 Ethnicity Black1.000 White1.582 0.479 5.228 Ethnicity*Abuse t2 8.298++0.688 100.141 +Adjusted for alcohol and drug use during pregnancy, body mass index, education level attained, gestational age of the infant at birth, total number of prenatal visits, and weight gain during the pregnancy. ++p-value < 0.10. Social support of the participant’s mother interacts with ethnicity in association with high birth weight (Tabl e 98). The odds of having a high birth weight baby are increased for participants w ho receive social support from their mothers regardless of ethnicity (OR = 1. 167 for Blacks; OR = 0.418 for Whites; score of 1 compared with 0) compared with Black women who have no support from their mothers. To exemplify the interaction, Figure 13 shows the change in odds ratio per score on the social support scale. The lines converge demonstrating quantitative intera ction around a score of 3 (score of 2 versus 0: OR = 1.361 for Blacks and OR = 0.810 for Whites; score of 3 versus 0: OR = 1.587 for Blacks and OR = 1.570 for Whites; et c.) Essentially, no or close to no social support from a participant’s mother decreases risk of high birth weight; this is more pronounced among the White women than Black women with the 53 The large range in the 95% Confidence Interval between the in teraction term and high birth weight is due to the lack of variability among the dichotomous predictors and outcome. The majority of abused White women birthed high-weight babies in this sample. Further explanation is given in the following chapter: Structural Equation Modeling.

PAGE 212

176 protective effect changing to risk at a score of three. The odds increase for Black participants as well, except at a much slower rate. Table 98 Logistic regression model of the interaction between the mother’s social support scale and ethnicity with high birth weight infa nts of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Model of Interaction Terms and Predictors (N = 397)+ Independent Variables Odds Ratio 95% Confidence Interval Mother Social Support Scale 1.167 0.752 1.811 Ethnicity Black1.000 White0.215 0.006 7.621 Ethnicity*Mother Social Support Scale 1.662++0.840 3.288 +Adjusted for education level attained, gestational age of t he infant at birth, and total number of prenatal visits. ++p-value < 0.20. Figure 13 Logistic regression model of the interaction between ethnicity and the mother’s social support scale on high birth weight of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 (All possible scores; Black/0 score as reference), with triangles representing White women and squares representing Black women 0 2 4 6 8 10 12 Score of 0123456 Mother's Social SupportOdds Ratios White Black

PAGE 213

177 In contrast, when birth weight is analyzed as a continuous variable, interaction occurs between a different se t of predictors and ethnicity. Marital status, both at the initial and final inte rview, interacts with ethnicity on birth weight. Also, social support of the partner inte racts with ethnicity on birth weight. Examining the interaction between marita l status from the initial interview and ethnicity, White married women are the parti cipants that are most likely to have higher birth weight infants (Table 99). T he multiple regression equation follows: birth weight ( ) = -4.534 + 0.943(single) + 1.055(evermarried) + 0.654(ethnic*single) + 1. 021(ethnic*evermarried) + 1.286(ethnic*married) + -0. 500(druguse) + 0.218(bmi) + 0.278(education) + 0.828(gestati on) + -0.674(premature) + 0.292(prenatal) + 0.182(weight) Results are similar for marital status fr om the final interview (Table 100), and the regression equation is stated: birth weight ( ) = -5.510 + 0.90(single) + 0.92(evermarried) + 0.71(ethnic*single) + 0.83(ethnic*evermarried) + 1.34(ethnic*married) + 0.218( bmi) +0.322(education) + 0.965(gestation) + -0.525(prem ature) + 0.293(prental) + 0.170(weight)

PAGE 214

178Table 99 Multiple regression model of the interaction between marital status from the initial interview during the first trimester and ethnicity on the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Model of Interaction Terms and Predictors (N = 393)* F-test 31.349; p < 0.001; R2 = 0.482 Independent Variables Coefficient Estimate Standard Error t statistic Intercept -4.534 1.835 -11.206 Marital Status t1 Single/Never Married 0.943 0.451 1.973** Single/Ever Married1.055 0.652 1.710 Ethnicity 0a Ethnicity*Single/Never Married 0.654 0.474 0.903 Ethnicity*Single/Ever Married 1.021 0.817 1.275 Ethnicity*Married/Living with Partner 1.286 0.473 3.498*** *Adjusted for alcohol and drug use duri ng pregnancy, body mass index, education level attained, gestational age of infant at birth, total number of premature births, to tal number of prenatal visits, and weight gain. **p < 0.05. ***p < 0.01. a Set to zero because it is redundant. Table 100 Multiple regression model of the interaction between marital status from the final interview during the third trimester and ethnicity on the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Model of Interaction Terms and Predictors (N = 346)* F-test 16.826; p < 0.001; R2 = 0.335 Independent Variables Coefficient Estimate Standard Error t statistic Intercept -5.510 3.356 -9.048 Marital Status t2 Single/Never Married0.897 0.426 1.888 Single/Ever Married0.920 0.677 1.250 Ethnicity 0a Ethnicity*Single/Never Married 0.712 0.588 0.862 Ethnicity*Single/Ever Married 0.825 0.941 0.724 Ethnicity*Married/Living with Partner 1.338 0.437 4.099** *Adjusted for body mass index, education le vel attained, gestational age of infant at birth, total number of premature births, total number of prenatal visits, and weight gain. **p < 0.001. a Set to zero because it is redundant. Table 101 presents the association between partner social support, and ethnicity on birth weight. As is shown, for White women, an increase in partner

PAGE 215

179 support is associated with high er birth weight infants. In addition, including the interaction term, almost 50% of the varianc e of birth weight is explained by this model. The multiple r egression equation follows: birth weight ( ) = -4.377 + -0.81(part ner) + 0.30(ethnic) + 0.97(ethnic*partner) + 0.396(educat ion) + 0.842(gestation) + 0.308(prenatal) + 0.147(weight) Table 101 Multiple regression model of the interaction between the partner social support scale and ethnicity on the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Model of Interactions Terms and Predictors (N = 394)* F-test 51.081; p < 0.001; R2 = 0.481 Independent Variables Coefficient Estimate Standard Error t statistic Intercept -4.377 1.754 -10.923 Partner Social Support Scale -0.807 0.438 -1.488 Ethnicity 0.297 0.670 0.132 Ethnicity*Partner Social Support Scale 0.966 0.727 1.283** *Adjusted for education level attained, gestational age of infant at birth, total number of prenatal visits, and weight gain. **p < 0.20. Interaction between the predictors and ethnicity on C-section include marital status from the initial interview and, as with high birth weight, the social support of the participant’s mother. Contra ry to results from the urine sugar level analysis, interaction between ethnicity and ma rital status occurs specifically with single/ever married women, or women who have been in a legally documented relationship with a partner (Table 102). Black ever married women are 75% less likely to have a C-section birth compar ed with Black women currently married or living with a partner. White married (OR = 1.931) or single/ever married women

PAGE 216

180 (OR = 5.068) are both more likely to have a C-section birth compared with Black married women. In terms of a mother ’s social support, Table 103 presents results of the interaction analysis. Agai n, as participants’ social support from their mothers increases, they are more likely to have C-section birth. Specifically, the odds ratio for Black wo men with support increases from 27% to over four times more likely to have a C-section birth compar ed with Black women who have no social support from t heir mothers. White non-supported and supported women have a high risk of C-se ction birth compared with Black nonsupported women (range 6.627-6.166). For White women, as support increases, the odds of C-section decrease, though mini mally. The opposite effect occurs for Black women in the study. Table 102 Logistic regression model of the interaction between marital status from the initial interview during the first trimester and ethnicity with Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-200154 Model of Interaction Terms and Predictors (N = 396)+ Independent Variables Odds Ratio 95% Confidence Interval Marital Status t1 Married1.000 Single/Never Married0.950 0.326 2.767 Single/Ever Married0.572 0.093 3.518 Ethnicity Black1.000 White1.931 0.660 5.650 Ethnicity*Marital Status t1 Single/Never Married0.699 0.171 2.856 Single/Ever Married4.590++ 0.533 39.548 +Adjusted for age, alcohol and drug use during pregnancy, body mass index, education level attained, previous C-section, total number of live births, and total number of prenatal visits. ++p-value < 0.20. 54 The wide 95% Confidence Interval for ever married women is due to the lack of variability between the predictor and the outcome of Caesarean section. The reduction in sample size also affected the variability.

PAGE 217

181Table 103 Logistic regression model of the interaction between the mother’s social support scale and ethnicity with Caesarean section of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Model of Interaction Terms and Predictors (N = 396)+ Independent Variables Odds Ratio 95% Confidence Interval Mother Social Support Scale 1.267 0.977 1.643 Ethnicity Black1.000 White 6.627++ 1.502 29.230 Ethnicity*Mother Social Support Scale 0.780+++0.576 1.057 +Adjusted for education level attained, previous C-section, and total number of prenatal visits. ++p-value < 0.05. +++p-value < 0.20.

PAGE 218

182 CHAPTER 5 STRUCTURAL EQUATION MODELING 5.0 Introduction The structural equation analysis m odeling chapter consists of the description and composition of structural equation models directly constructed from significant findings from the results chapter (Chapter 4). The first section of the chapter briefly outlines the methodol ogy used in selecting and structuring each model, the second section is com posed of structural equation models for overall findings from the re sults chapter (e.g., exclud ing interaction terms), and the final section is an separate examinat ion of both ethnic groups, again, using significant findings from the results chapter. Figures ar e inserted at the beginning of each model testing procedure to describe pictorially the analytic procedure used to assess model fit. 5.1 Structural Equati on Modeling Methodology The structural equation modeling methodol ogy consists of first identifying the specific type of struct ural equation model to be c onstructed. The equation is based on review of the liter ature and assumptions requir ed in order to perform the analysis. After a discussion of the model type selected, a review of the assumptions for that specific model is lis ted with a brief descrip tion of the impact of each assumption on this analysis. Ne xt, the methodology specific to this

PAGE 219

183 dissertation is addressed including the al pha-level used in analysis, the strategy for variable removal, and the approac h for analyzing interaction terms. The original analytic strategy for the structural equation modeling section included the creation of latent constr ucts to assess a causal association55 between the predictor vari ables and the outcomes56. For example, the theoretical model contains component s such as the psychological and physiologic pathways. Each pathway is composed of specific predictors: the psychological pathway contains the m easurements of depressi on, verbal abuse, and pregnancy wantedness. These thr ee predictors compose the latent construct of the “psychological” pathway. The originally proposed confirmatory and latent path analysis consists of creati on of the latent cons truct via the three predictors, confirming the associations of the thr ee predictors and the latent construct, and then establishing a path between the latent construct and the outcomes of urine sugar levels, high birth weight, and C-section. Upon review of the current literatur e and comparison with measurements used in the data set, however, it was det ermined that confirmatory factor analysis required that the predictor variable measures be psychometrically sound and shown to be strongly associated with eac h outcome measure in the literature [173]. Based on the violation of these ma jor assumptions (i.e., each scale was tailored to fit the population being studied and therefore, no longer presented the 55 In the context of path analysis, a causal association is defined as statistically caus al based on covariation between predictor variables, t statistic values, and model fit. The paths are defined as causal if f ound statistically significant. Causality does NOT refer to the requi rements needed for causal inference. 56 See the methodology chapter for a further description of the original strategy for analysis.

PAGE 220

184 same reliability as in the literature; none of the predictors in this study had been associated in the literatur e with the outcomes of urine sugar levels, high birth weight, and C-section), another type of structural equati on modeling, path analysis of manifest variables, was re viewed. Path analysis of measured or manifest variables was determined to be the most appropriate structural equation modeling methodology. All path analyses are performed in Mplus version 2.14 [177]. Path analysis, specifically of measured variables, is a statistical method that tests whether the t heoretical model proposed accurately reflects the associations inherent in the data. Measured variables (e.g., manifest, antecedent, or exogenous variables) are tested for causal association with outcome variables (e.g., consequent variab les). Each manifest variable is correlated with all other manifest vari ables to determine their impact on each other, but not causal associations. All manifest variables are located on the left side of each model presented in this c hapter, and single-headed, straight arrows represent a uni-directional causal pathw ay to each consequent variable located on the right side of the model (see Figure 16 for an example). All correlations are presented in tables following the pat h analysis figures (see Table 107 for an example). Assumptions required to conduct a pat h analysis of manifest variables include normal distribution of the manifest and outco me variables, linearity, absence of multicollinearit y, absence of measurement error, sufficient sample

PAGE 221

185 size, overidentification of the model, inclusion of non-trivial causes, and unidirectional causality. All manifest and outcome variables should appear to be symmetric in order to be included in t he analysis. Any variable with a non-normal distribution must be transfo rmed prior to analysis. As discussed in the Results Chapter, birth weight appeared non-normally distributed. As a result, it was transformed prior to analysis. That same transformed variable is used in the path analysis. Relationships between manife st and outcome variables are assumed to be linear. This assumption applies to nominal and categorical ordinal data as well as continuous data. Interaction is not supported by path analysis modeling. Therefore, path analysis of Hypotheses 4 and 5 include analyzing the interaction term (e.g., ethnicity) in two separate models rather than as one variable in a comprehensive model. Multicollinear rela tionships are not supported by the path analysis model and separate models s hould be tested for each set of multicollinear variables. The next secti on describes the strategy for analysis of multicollinear variables in this data set. Independence of observations and a sample size of at least 200 are reco mmended to test a path analysis model. Each model must be either just-identified, or ideally, over-identified to test for goodness-of-fit. Hatcher defines a just-i dentified model as a model that tests “exactly as many linearly independent equat ions as unknowns” [173]. That is, each “variable in the model is interrela ted with every other variable, either through a causal path or a covariance” [1 ; page 160]. However, a just-identified model provides a perfect fit due to its test ing of all possible relationships in the

PAGE 222

186 model. Ideally, Hatcher recommends analysis of an over-identified model, which is a model that has more linear equati ons than unknowns. Only models that are either justor over-ident ified are included in this analysis. To assess the relationships between the path coefficients, non-trivial causes must be included in the analysis. To identify and include all non-trivial causes is not possible given the original purpose of each grant (e.g., focus on low birth weight infants). In order to isolate the psychosocial a nd physical factors and their independent effect on the outcomes of urine sugar le vels, high birth weight, and C-section; none of the confounding fact ors are included in the path analysis. The assumption of non-trivial c ause inclusion is partially violated by exclusion of these factors from analysis. In order for the model to be self-contained, no measured antecedent variables may be omitted from analysis. Only significantly associated predictors are in cluded in the path analysis; therefore, the models may not be completely self-contained. A final restriction of path analysis of manifest variables is uni-directional caus ality. All models mu st be recursive, or causal in one direction. All of the m odels analyzed in this chapter (e.g., Figure 16) are recursive. These major assu mptions are addressed in turn throughout the chapter and violations noted and discussed. All models in this chapter are eval uated using an alpha le vel of 0.10, are assessed for goodness-of-fit, separated for evaluation of multic ollinearity, and divided into separate model s for interaction analyses. Each set of models is

PAGE 223

187 composed of predictors significantly associated with the outcomes only57. In this study, due to the limited sample size (N = 506; decreases depending on missing values), alpha is set at 0.10 when eval uating the t statisti c of each manifest variable58. Increasing alpha to 0.10 also incr eases the power of the analysis. Goodness-of-fit is evaluated by checking the residuals (all should be less than 2.00), the p -value of the chi square goodness-of-fit index (>0.05), the value of the comparative fit index (CFI mu st be between 0.90-1.00), the R2 value for each model, and the significance level of each t statistic in the model (see Table 105 for an example). When the majority of the criteria are met, the model is determined to have a good fit. Rather than randomly modifying models t hat have a poor fit, two strategies are used. First, due to the mu lticollinearity of certain pr edictors (e.g., history of abuse and abuse during the second half of the pregnancy), separate models are created and compared. The model with th e better fit is chosen based on the above criteria and on the biological and cultur al plausibility of the impact of the manifest variables on the outcome59. If neither model pr esents a good fit, then a second strategy is employed; the least t heoretically significant manifest variable is removed from the model60. If the model still lacks a good fit, it is concluded that the theoretical model is incorrect and no further model testing is conducted. The reason for such a strategy is to av oid data-driven modifications. Data-driven 57 See Chapter 4, the Results Chapter for all significant associations. 58 Alpha is also larger than usual due to the expl oratory nature of this portion of the analysis. 59 This point if further explained in t he next section with specific models. 60 Significance does not refer to statistical significance. It refers to the impact of each variable on the outcome in terms of the theoretical model. Each removal is justified within the text.

PAGE 224

188 modifications occur when the investigat or relies on the data to determine the appropriate path model in the absence of a theory-based path model. The final set of path models includes dividing the sample into separate groups to address the last two hypotheses (e.g., Hypotheses 4 and 5). Since ethnicity is interactive and path analysis of manifest variables does not support analysis of interaction, separate models for Black and White women are constructed and analyzed. 5.2 Structural Equation Modeling for Overall Findings Overall inferential findings are su mmarized from results addressed in Hypotheses 1, 2, and 3. Table 104 describes the significant associations of each hypothesis. In order to evaluate the originally proposed model (Figure 14), Hypothesis 2 is the first path analysi s model constructed, followed by Hypotheses 3 and 1.

PAGE 225

189Table 104 Statistically significant associations between predictors and outcomes from hypotheses 1-3 in the results chapter for structural equation modeling of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Predictors and Outcomes by Hypothesis Hypotheses/OutcomesPredictors Hypothesis 1 High Birth Weight Higher Urine Sugar Levels (1+ or higher) Birth Weight Urine Sugar Level Hypothesis 2 Urine Sugar Level Mother Social Support Scale Depression Scale from the initial interview (t1)61* Depression Scale from the final interview (t2)* Marital Status from the initial interview (t1) History of Abuse+ Abuse from the final interview (t2)+ Physical Work Strain from the final interview (t2) Hypothesis 3 High Birth Weight Mother Social Support Scale Ethnicity Birth Weight Abuse from the final interview Ethnicity Caesarean Section Depression from the final interview (t2) Marital Status from the initial interview (t1) History of abuse Ethnicity Multicollinear predictors. + Multicollinear predictors. 61 As in previous chapters, t1 refers to data collected from the initial interview only, and t2 refers to data collected from the final interview only.

PAGE 226

190Figure 14 Proposed causal pathway model of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 5.2.1 Evaluation of Hypothesis 2 Hypothesis 2 tests the associatio n between specific psychosocial and physical stressors and urine sugar levels. Due to the associati on of two sets of multicollinear manifest variables, four models are compared and the model with the best fit chosen. Figure 15 describes the modeling procedures. In addition, physical work strain measur ed from the final interview only includes a subset of the sample, and is, therefore, added to t he model separately due to the reduction in sample size. Table 105 presents goodnessof-fit indices for all four models. Model 1 consists of an analysis of a mother’s social support, mari tal status of the participant, depression from the initial interview, and history of abuse; while Model 2 consists of social support, ma rital status, depressi on from the initial interview, but abuse from the final intervie w instead of history of abuse. Model 3, Ethnicity Depression Marital Status and Autonomy Social Support Physical Work Stress Physical and Verbal Abuse Urine Sugar Reading High Birth Weight Cesarean Section H1 H2 H3 H4 and H5 Pregnancy Wantedness

PAGE 227

191 again, includes a mother’s social suppor t and marital status, except depression as measured at the final interview is in cluded, and history of abuse rather than abuse from the final interview. Finally, Model 4 resembles Model 3 with the exception of history of abus e. Reviewing the statistics presented in Table 105, all models present a good fit and explain approximately the same amount of variance in urine sugar levels. Based on fu rther review of the t statistic of each manifest variable, the correlation and re sidual matrices, and the theoretical meaning of each combination of variabl es; it is determined that Model 4 represents the model wit h the best fit. Figure 15 Structural equation modeling procedure for Hypothesis 2 assessment of predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 2: Urine Sugar Levels• Choose the model with the best fit (2 sets of multicollinearvariables) among four models Depression from the initial interview Depression from the final interview History of abuse Abuse from the final interview A mother’s social support and marital status from the initial interview are included in all models • Use the same methodology to select the model with the best fit including physical work strain measured at the final interview (the same 2 sets of multicollinearvariables are analyzed) among the four models

PAGE 228

192Table 105 Goodness-of-fit indices for hypot hesis 2 assessing associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 2; N = 433 Excluding Physical Work Strain Model Chi-square Value Degrees of Freedom p -value Comparative Fit Index R2 1 5.622 4 0.23 1.00 0.03 2 5.312 4 0.26 1.00 0.03 3 5.189 4 0.27 1.00 0.03 4 4.849 4 0.30 1.00 0.03 Table 106 summarizes statistics specific to Model 4. In terms of model fit, only one of the manifest variables is statis tically significantly associated with the outcome, mother’s social support. All of the re siduals approximate zero, and when reviewing the correlation matrix, all va riables are slightly correlated except depression and abuse which are moderately correlated (Table 107; r = -0.48). The correlation indicates that as depressi on increases, verbal or physical abuse during pregnancy decreases. The model is just-identified according to Hatcher’s rule: the number of data points must ex ceed or equal the number of parameters to be estimated. The selected model contains ten data points and ten parameters to be estimated rendering it ju st-identified (i.e., the number of data points equals the number of parameters to be estimated). As a result, the comparative fit index equals 1.00 as the model is a perfect fit.

PAGE 229

193Table 106 Model 4 statistics of associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels (N = 433) Manifest Variable Estimate Standard Error t Statistic Mother Social Support Scale -0.052 0.031 -1.695* Marital Status 0.028 0.072 0.387 Depression t2** 0.014 0.013 1.135 Abuse t2 0.124 0.169 0.732 *p < 0.10. **t2 refers to data collected from the final interview only. Figure 16 illustrates the path analysis re sults. The solid arrow represents the manifest variable, mother’s social supp ort that is signific antly associated with the consequent variable of urine sugar le vels. The dotted line arrows represent non-significant manifest variables. As s hown, the slope of a mother’s social support and urine sugar level is negative indi cating an inverse relationship. That is, as a mother’s support increases, uri ne sugar is likely to decrease. This conclusion is consistent wit h the results in Chapter 4. However, the impact, as described by Hatcher, is minimal due to the minimal slope (-0 .052) and the small R2 value (0.03). Results of the path model indicate that there are other unaccounted predictors of urine sugar leve l. Most likely, many of these predictors are biologic and may include so me of the confounders in this study, and are, as a result, excluded from t he path analysis. This conclusion is consistent with previous analyses and agai n, supports the conclusion that the model is not highly predictive or causal.

PAGE 230

194Figure 16 Comprehensive modeling for predictors of urine sugar levels excluding physical work strain of pregnant women attending the Co unty Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 107 Correlation matrix for Model 4 assessing associations between predictors and urine sugar levels of pregnant women attendi ng the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Mother Social Support Marital Status Depression t2 Abuse t2 Mother Social Support 1.000 Marital Status 0.026 1.000 Depression t2 -0.010 0.080 1.000 Abuse t2 0.171 -0.003 -0.480 1.000 The following set of models is a re-tes ting of the four models previously presented with the addition of physical work strain. Adding physical work strain to the model violates one of the requi rements for conducting a path analysis, minimal sample size. When work strain is added to the model, the sample size is Mother Social Support Scale Marital Status Depression t2 Abuse t2 Urine Sugar Levels -0.052 0.028 0.014 0.124

PAGE 231

195 reduced to 144. Table 108 is a summary of all four models with the physical work strain scale added. The models are presented in the same order as in the previous analysis (Model 1: depression from the initial interview and history of abuse; Model 2: depression from the init ial interview and abuse from the final interview; Model 3: depression from the final interview and history of abuse; Model 4: depression from the final interv iew and abuse from the final interview). Although Model 3 presents a borderline chi-square goodnessof-fit score and explains 1% less variance than Models 1 and 2, when reviewing the t statistics, residuals, and correlations matrices, it pr esents a much bette r fit than the other three models. Table 108 Goodness-of-fit indices for hypot hesis 2 assessing associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 2; N = 144 Including Physical Work Strain Model Chi-square Value Degrees of Freedom p -value Comparative Fit Index R2 1 10.559 5 0.06 1.00 0.16 2 9.452 5 0.09 1.00 0.16 3 11.030 5 0.05 1.00 0.15 4 10.011 5 0.07 1.00 0.15 Table 109 presents statistics specific to Model 3. Based on t statistic results, the social support of a partici pant’s mother, depression during the second and third trimesters of pregnancy, and physi cal work strain during the latter half of the pregnancy are significantly causa lly associated with hi gher urine sugar levels. As a mother’s social support in creases, urine sugar levels are likely to decrease; this is consistent with the previous path analysis. Higher depression

PAGE 232

196 scores are associated with sugar spill in the urine, and increased perceived physical work strain is causally associ ated with urine sugar spill. All residuals approximate zero indicating a good model fit. When reviewing the correlation matrix in Table 110, both history of abuse and depression from the final interview (r = -0.44), and physical work strain from the final interview and depression (r = 0.32) are moderately correla ted. All other correlations are minimal. As previously noted, when depr ession increases, physical or verbal abuse during pregnancy is likely to decrease. However, as physical work strain during the latter half of the pregnancy increases, depression is likely to increase. Table 109 Model 3 statistics of associations between predictors and urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels (N = 144) Manifest Variable Estimate Standard Error t Statistic Mother Social Support Scale -0.091 0.054 -1.673* Marital Status 0.157 0.141 1.114 Depression Scale t2 0.045 0.025 1.841* History of abuse 0.052 0.279 0.187 Physical Work Strain Scale t2 0.122 0.065 1.888* *p < 0.10. Figure 17 displays the associations pict orially. Again, solid lines represent statistically significant causal associ ations, while dotted lines represent nonstatistically significant causal associat ions. The slopes indicate that both a mother’s social support and depression dur ing the second and third trimesters of pregnancy minimally influence urine sugar le vels; however, physical work strain, among this subset of the sample has a la rger impact on urine sugar spill (slope =

PAGE 233

197 0.122). Hatcher indicates that in order fo r the path coefficient to be theoretically meaningful, it must be > 0.32. None of t he coefficients in this model are greater than 0.32 indicating that the model is minimally explanatory. Although more variance is explained in this model than t hat excluding physical work strain, the amount explained is still minimal. Again, a larger amount of variance may be explained by the identified confounders. Based on the R2 and slopes, physical work strain during the second half of pregnancy explains the greatest amount of variance in urine sugar levels. Figure 17 Comprehensive modeling for predictors of urine sugar levels including physical work strain of pregnant women attending the Co unty Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Mother Social Support Scale Marital Status Depression t2 History of Abuse Urine Sugar Levels -0.091 0.157 0.045 0.122 Physical Work Strain Scale 0.052

PAGE 234

198Table 110 Correlation matrix for Model 3 assessing associations between predictors and urine sugar levels of pregnant women attendi ng the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Mother Social Support Depression t2 Marital Status Abuse t2 Work Strain t2 Mother Social Support 1.000 Depression t2 -0.102 1.000 Marital Status -0.093 0.159 1.000 History of Abuse 0.052 -0.438 -0.196 1.000 Physical Work Strain Scale t2 -0.074 0.323 -0.231 0.105 1.000 5.2.2 Evaluation of Hypotheses 3 and 1 Hypotheses 3 and 1 are evaluated and combined with the analysis of Hypothesis 2 to construct a comprehensiv e model that mirrors the theoretical model presented in Figure 1. Hypotheses 3 and 1 were initially tested alone, however, due to the size of the model (Hypothesis 1: one manifest and one consequent variable; Hypothesis 3: 2 manifest and one consequent variable) analysis is combined with the Hypothesis 2 model. Two models are presented, the first excluding the physical work st rain scale measure, and the second including the measure. Figure 18 present s the analysis strategy for the models. In terms of fit for the firs t analysis, the chi-square goodn ess-of-fit statistic is nonsignificant ( 2 = 3.607; p -value = 0.4612), the comparat ive fit index is 1.000 and the model is over-identified (15 data poi nts and 14 parameters to be estimated), and the R2 value is 0.04 for the portion of the model connecting the manifest variables to the mediator (intervening out come measure) of urine sugar level,

PAGE 235

199 and 0.29 for the manifest and mediator va riables connected to the consequent measure of high birth weight. Figure 18 Structural equation modeling procedure for Hypotheses 1 and 3 assessment for predictors of urine sugar levels and high birt h weight infants of pregnant women attending the County Health Department Prenatal Clin ic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 111 presents the model stat istics and Figure 19 presents results pictorially. In the causal pathway from the manifest variables to urine sugar levels, only the social support of the partici pant’s mother is significantly causally associated with urine sugar levels. As a mother’s support increases, the likelihood of urine sugar spill decreases. In the causal pathway from the manifest variables to high birth weight, a mother’s social support is significantly associated with high birth weight. As a mother’s support increases, the likelihood of a high birth weight infant increases. Finally, the mediating factor (urine sugar level) is significantly causally associated with high bi rth weight. Sugar spill in the urine is Hypotheses 1 and 3: High Birth Weight Only• Create the model based on results from Hypothesis 2 A mother’s social support Depression from the final interview Marital from the initial interview Abuse from the final interview A mother’s social support Ethnicity • Incorporate significant predictors of high birth weight • Use the same methodology to create a model including physical work strain measured at the final interview; replace abuse from the final interview with history of abuse for the urine sugar level pathway

PAGE 236

200 causally associated with high birth weight infants. Using Hatcher’s recommendations, only the pathway from uri ne sugar levels to high birth weight is influential with a path coefficient of 0.33 (> 0.32). When examining the correlation matrices in Tables 112 and 113, marital status and ethnicity are moderately correlated (r = 0.344), as well as depression and abuse during the latter half of the pregnancy (r = -0.396). All other variables are minimally correlated. The direction of the correla tions indicate that White women in the sample are more likely to be married, and that increasing depression decreases the likelihood of ver bal or physical abuse. Table 111 Model statistics of associations for predictors of urine sugar levels and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels and High Birth Weight (N = 347) Manifest Variable Estimate Standard Error t Statistic Pathway to Urine Sugar Level Mother Social Support Scale -0.068 0.036 -1.873* Marital Status -0.004 0.100 -0.043 Depression Scale t2 0.013 0.014 0.970 Abuse t2 0.110 0.187 0.589 Pathway to High Birth Weight Mother Social Support Scale 0.248 0.099 2.500* Ethnicity -0.418 0.367 -1.138 Urine Sugar Level to High Birth Weight 0.334 0.114 2.938* *p < 0.10.

PAGE 237

201Figure 19 Comprehensive modeling for predictors of urine sugar levels and high birth weight excluding physical work strain of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 112 Correlation matrix for predictors of urine sugar levels of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Mother Social Support Marital Status Depression t2 Abuse t2 Urine Sugar Levels 1.000 Mother Social Support -0.003 1.000 Marital Status 0.004 0.020 1.000 Depression t2 -0.103 0.062 0.155 1.000 Abuse t2 0.055 0.186 0.031 -0.396 1.000 Mother Social Support Scale Marital Status Depression t2 Abuse t2 Urine Sugar Levels -0.068 -0.004 0.110 -0.418 Ethnicity High Birth Weight 0.013 0.248 0.334

PAGE 238

202Table 113 Correlation matrix for predictors of high birth weight infants of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Mother Social Support Ethnicity Urine Sugar Levels 1.000 Mother Social Support 0.066 1.000 Ethnicity -0.002 0.186 1.000 Tables 114 through 115 present the same model with physical work strain during the second and third trimesters included. Figure 20 shows the model pictorially. The chisquare value for the model is non-significant ( 2 = 2.213; p value = 0.6963), the comparative fit index (CFI) is 1.000 for the over-identified model (21 data points and 19 paramet ers to be estimated), and the R2 values are 0.20 for the pathway from the manifest va riables to urine sugar levels and 0.39 for the pathway from the mani fest and mediator variables to high birth weight. All indices indicate a good fit. However, the sample size violates a major assumption of the path analysis (minimum of 200). When reviewing the pathway between the manifest variabl es and urine sugar leve ls, depression during the latter half of the pregnancy is the only signif icantly causally associated predictor. As depression increases, urine sugar leve ls are likely to increase. For the second pathway, ethnicity and urine sugar levels are significantly causally associated with high birth weight. Whit e women are likely to have high birth weight babies, and increases in urine s ugar are causally associated with high birth weight infants. The only path coeffi cient with theoretical meaning is the path

PAGE 239

203 between urine sugar levels and high birth we ight (0.32). The correlation matrices indicate no moderately or highly corre lated variables (Tables 115 and 116). Table 114 Model statistics of associations for predictors of urine sugar levels and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels and High Birth Weight (N = 109) Manifest Variable Estimate Standard Error t Statistic Pathway to Urine Sugar Level Mother Social Support Scale -0.045 0.074 -0.612 Marital Status 0.156 0.172 0.904 Depression Scale t2 0.053 0.028 1.920* History of abuse 0.402 0.344 1.168 Physical Work Strain t2 0.132 0.093 1.425 Pathway to High Birth Weight Mother Social Support Scale 0.294 0.222 1.326 Ethnicity -0.931 0.541 -1.720* Urine Sugar Level to High Birth Weight 0.319 0.180 1.778* *p < 0.10.

PAGE 240

204Figure 20 Comprehensive modeling for predictors of urine sugar levels and high birth weight infants including physical work strain of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 115 Correlation matrix for predictors of urine sugar levels of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Mother Social Support Marital Status Depression t2 History of Abuse Physical Work Strain t2 Urine Sugar Levels 1.000 Mother Social Support 0.063 1.000 Marital Status 0.005 -0.091 1.000 Depression t2 0.018 0.028 0.159 1.000 History of Abuse -0.027 0.265 -0.139 -0.290 1.000 Physical Work Strain t2 0.067 0.187 -0.277 0.303 0.298 1.000 Mother Social Support Scale Marital Status Depression t2 History of Abuse Urine Sugar Levels -0.045 0.156 0.132 -0.931 Ethnicity High Birth Weight 0.053 0.294 0.319 Physical Work Strain t2 0.402

PAGE 241

205Table 116 Correlation matrix for predictors of high birth weight infants of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Mother Social Support Ethnicity Urine Sugar Levels 1.000 Mother Social Support -0.011 1.000 Ethnicity 0.105 0.046 1.000 The next set of analyses resembles t he previous two models except birth weight is used as a continuous measure rather than as a ca tegorical measure (Figure 21). In addition, based on Table 10 4, the manifest variables for birth weight are ethnicity an d abuse during the second and third trimesters of pregnancy instead of mother’s social supp ort as in the analysis of high birth weight. Tables 117 through 119 describe the model excluding physical work strain during the latter half of the pr egnancy, and tables 120 through 122 include physical work strain. Figure 22 addr esses the first model, while Figure 23 describes the second model with work stress.

PAGE 242

206Figure 21 Structural equation modeling procedure for Hypotheses 1 and 3 assessment for predictors of urine sugar levels and the birt h weight of infants born to pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 117 presents results of the path model excluding physical work strain. The chi-square goodness-of-f it test is non-significant ( 2 = 5.698, p -value = 0.22), however, the CFI is less than 0.90 (CFI = 0.86; model is over-identified) and the R2 is only 0.02 for the pathway to urine sugar levels and 0.07 for the pathway to birth weight. After reviewin g all of the measures of fit, though the model appears to marginally fit the data, it is not as expl anatory as previous models where birth weight was dichotomized None of the manifest variables in the pathway to urine sugar are significantly causally associated with sugar in the urine, while ethnicity is the only signifi cantly associated antecedent variable in the pathway to birth weight. White wom en are likely to have higher birth weight babies. Urine sugar remains significantly causally associated with birth weight. Hypotheses 1 and 3: Birth Weight Only•Create the model based on results from Hypothesis 2 A mother’s social support Depression from the final interview Marital from the initial interview Abuse from the final interview Abuse from the final interview Ethnicity•Incorporate significant predictors of birth weight•Use the same methodology to create a model including physical work strain measured at the final interview; replace abuse from the final interview with history of abuse for the urine sugar level pathway

PAGE 243

207 Women with sugar in their urine have higher birth weight babies. Both path coefficients are theoretically meaningf ul with each greater than 0.32. Correlations are moderate for one pair of variables in Tables 118 and 119 (depression and abuse -0.42). Again, as depression increases, verbal or physical abuse during the second and third trimesters decreases. Table 117 Model statistics of associations for predictors of urine sugar levels and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels and Birth Weight (N = 347) Manifest Variable Estimate Standard Error t Statistic Pathway to Urine Sugar Level Mother Social Support Scale -0.051 0.036 -1.416 Marital Status -0.012 0.099 -0.122 Depression Scale t2 0.014 0.013 1.037 Abuse t2 0.043 0.187 0.232 Pathway to Birth Weight Abuse t2 0.624 0.385 1.623 Ethnicity -0.954 0.393 -2.430* Urine Sugar Level to Birth Weight 0.551 0.208 2.647* *p < 0.10.

PAGE 244

208Figure 22 Comprehensive modeling for predictors of urine sugar levels and the birth weight of infants excluding physical work strain of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 118 Correlation matrix for predictors of urine sugar levels of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Mother Social Support Marital Status Depression t2 Abuse t2 Urine Sugar Levels 1.000 Mother Social Support 0.030 1.000 Marital Status -0.018 0.025 1.000 Depression t2 -0.084 0.071 0.149 1.000 Abuse t2 0.073 0.182 0.004 -0.421 1.000 Mother Social Support Scale Marital Status Depression t2 Abuse t2 Urine Sugar Levels -0.051 -0.012 0.043 -0.954 Ethnicity Birth Weight 0.014 0.551 0.624

PAGE 245

209Table 119 Correlation matrix for predictors of the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Ethnicity Abuse t2 Urine Sugar Levels 1.000 Ethnicity -0.043 1.000 Abuse t2 -0.099 0.192 1.000 Table 120 summarizes results from the same analysis including physical work strain in the second and third trimes ters. As with the previous model, the chi-square goodness-of-fit te st is non-significant ( 2 = 8.051; p -value = 0.23), the CFI is low (CFI = 0.67), but the R2 values are higher ( R2 = 0.16 for urine sugar level path; R2 = 0.14 for birth weight path). A gain, the model indicates a poorer fit than the previous models, but this ma y be do to the smaller sample size (N = 109). Depression during the latter half of the pregnancy is significantly causally associated with urine sugar levels. Ess entially, as depression scores increase, the likelihood of sugar in t he urine increases. Ethnicity is significantly associated with birth weight; however, the path from uri ne sugar to birth weight is no longer significant. Ethnicity is the only th eoretically meaningf ul pathway with a coefficient of -1.63. When reviewing the correlation matrices, none of the variables are moderately or highl y correlated (Tables 121 and 122).

PAGE 246

210Table 120 Model statistics of associations for predictors of urine sugar levels and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels and Birth Weight (N = 109) Manifest Variable Estimate Standard Error t Statistic Pathway to Urine Sugar Level Mother Social Support Scale -0.010 0.078 -0.126 Marital Status 0.151 0.170 0.887 Depression Scale t2 0.048 0.028 1.711* History of Abuse 0.340 0.438 0.776 Physical Work Strain t2 0.124 0.093 1.331 Pathway to Birth Weight Abuse t2 1.772 1.234 1.436 Ethnicity -1.630 0.791 -2.060* Urine Sugar Level to Birth Weight 0.401 0.347 1.157 *p < 0.10. Figure 23 Comprehensive modeling for predictors of urine sugar levels and the birth weight of infants including physical work strain of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Mother Social Support Scale Marital Status Depression t2 History of Abuse Urine Sugar Levels -0.010 0.151 0.124 -1.630 Ethnicity Birth Weight 0.048 1.772 0.401 Physical Work Strain t2 0.340 Abuse t2

PAGE 247

211Table 121 Correlation matrix for predictors of urine sugar levels of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Mother Social Support Marital Status Depression t2 History of Abuse Physical Work Strain t2 Urine Sugar Level 1.000 Mother Social Support 0.088 1.000 Marital Status 0.002 -0.056 1.000 Depression t2 -0.063 -0.037 0.165 1.000 History of Abuse -0.101 -0.025 -0.174 -0.113 1.000 Physical Work Strain t2 -0.072 0.170 -0.234 0.293 0.211 1.000 Table 122 Correlation matrix for predictors of the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Ethnicity Abuse t2 Urine Sugar Levels 1.000 Ethnicity 0.054 1.000 Abuse t2 -0.093 0.235 1.000 The final path model for Hypotheses 1 through 3 is an analysis of manifest variables with C-section (Figure 24 pr esents the analysis strategy). Since Csection is not associated with either of the other outcome measures (e.g., urine sugar levels and birth weight) in the Resu lts Chapter, it is analyzed separately. The manifest variables included in this analysis are ethnicity, history of abuse, marital status, and depression during the la tter half of the pregnancy. Tables 123 through 124 show results, and Figure 25 pr esents results pictorially. The chisquare goodness-of-fit test is non-significant ( 2 = 8.300; p -value = 0.08), the CFI

PAGE 248

212 is 1.000 (the model is just-identifie d indicating a perfect fit), and the R2 value is 0.07. Ethnicity is the only theoretically significantly causally associated predictor of C-section (-0.53). White women are likely to have C-sections in this sample. Two pairs of manifest variables are m oderately correlated, ethnicity and marital status (r = 0.63) and hist ory of abuse and depression during the second and third trimesters of pregnancy (r = -0.48; Table 124). White women in the sample are the most likely to be married, and as with previous models, as depression increases, abuse is likely to decrease. Figure 24 Structural equation modeling procedure for Hypothesis 3 assessment of predictors on Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 3: Caesarean Section•Create the model Ethnicity Marital status from the initial interview History of abuse Depression from the final interview

PAGE 249

213Table 123 Model statistics of associations for predictors of Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on C-section (N = 347) Manifest Variable Estimate Standard Error t Statistic Ethnicity -0.533 0.204 -2.608* History of Abuse 0.184 0.185 0.995 Marital Status -0.067 0.110 -0.616 Depression t2 0.005 0.013 0.361 *p < 0.10. Figure 25 Comprehensive modeling for predictors of Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Ethnicity History of Abuse Marital Status Depression t2 C-section -0.533 0.184 -0.067 0.005

PAGE 250

214Table 124 Correlation matrix for predictors of Caesarean section births to pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Ethnicity History of Abuse Marital Status Depression t2 Ethnicity 1.000 History of Abuse 0.088 1.000 Marital Status 0.625 0.073 1.000 Depression t2 0.060 -0.476 0.124 1.000 5.3 Structural Equation Modeling for Interaction Terms Hypotheses 4 and 5 reassess Hypotheses 2 and 3 with the added interaction of ethnicity. Since path anal ysis does not support interaction within its models, a separate analysis is perform ed for each ethnic group to assess the former hypotheses. A major weakness of this strategy is the reduction in sample size. For each ethnic group, the sample r anges from 152 to 227. As a result, all models must be considered exploratory and all causal paths specific to this sample only62. Hypothesis 4 is considered first, and is the assessment of ethnicity on predictors of ur ine sugar spill. Hypothesis 5 follows and includes an analysis of ethnicity on the predictors of high birth wei ght and Caesarean section. As in the previous section, multicolli near terms are modeled separately and the model with the best fit is selected63. Table 125 outlines statistically significant predictors with ethnicity fr om the results chapter. 62 As was defined at the beginning of the chapter, ‘causal’ does not imply epidemiologic causality; it is terminology specific to path analysis of manifest variables indi cating a statistically causal relationship. 63 All analyses are briefly described to reduce repetition. The in itial analyses contain extensive detail. Refer to the section on Evaluation of Hypothesis 2 for a more detailed explanation of tables and figures.

PAGE 251

215Table 125 Statistically significant associations between predictors and outcomes from hypotheses 4 and 5 in the results chapter for structural equation modeling of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Predictors and Outcomes by Hypothesis Hypotheses/OutcomesPredictors Hypothesis 4 Urine Sugar Level Marital Status from the initial interview (t1)64* Marital Status from the final interview (t2)* Partner Social Support Scale Physical Work Strain from the final interview (t2) Hypothesis 5 High Birth Weight History of Abuse+ Abuse from the final interview (t2)+ Mother’s Social Support Scale Birth Weight Marital Status fr om the initial interview (t1)* Marital Status from the final interview (t2)* Partner Social Support Scale Caesarean Section Marital Status from the initial interview (t1) Mother’s Social Support Scale Multicollinear predictors. + Multicollinear predictors. 5.3.1 Evaluation of Hypothesis 4 The manifest variables included in t he path analysis of urine sugar levels are marital status from the in itial interview, marital status from the final interview, the partner social support scale, and ph ysical work strain during the second and third trimesters of pregnancy. Measures of marital status are multicollinear; therefore, two model s will be analyzed and the model with the best fit selected. Figure 26 presents the strategy used fo r analysis of the model and Table 126 presents results for both ethnic groups and both models of mari tal status. Model 1 includes partner social s upport and marital status from the initial interview. Model 2 includes partner social support and marital status from the final 64 As in previous chapters, t1 refers to data collected from the initial interview only, and t2 refers to data collected from the final interview only.

PAGE 252

216 interview. Neither model includes physica l work strain due to the sample size reduction. Figure 26 Structural equation modeling procedure for Hypothesis 4 assessment of the interaction between ethnicity and predictors on urine sugar levels of pregnant women attending the County Health Department Pr enatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 126 Goodness-of-fit indices for hypothesis 4 assessing the interaction between ethnicity and predictors on urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 4 Excluding Physical Work Strain Model Ethnicity Chisquare Value Degrees of Freedom p value Comparative Fit Index R2 1 Black 4.263 2 0.12 1.000 0.04 1 White 1.914 2 0.38 1.000 0.02 2 Black 4.172 2 0.12 1.000 0.04 2 White 2.476 2 0.29 1.000 0.03 Hypothesis 4: Urine Sugar Levels•Choose the model with the best fit (1 set of multicollinearvariables) among two models Marital status from the initial interview Marital status from the final interview A partner’s social support is included in all models•Physical work strain is excluded from all interaction analyses due to the reduction in sample size•Each model is divided by ethnicity for a total of two models per evaluation (e.g., four models must be compared above, two for Black women and two for White women

PAGE 253

217 Upon review of the goodness-of-fit st atistics, correlation matrices, and the t statistics, Model 1 present s the best fit. Model stat istics are presented in Table 127 and Figure 27. The only statistically si gnificant causal pathway is among the Black participant group. Single, neve r-married Black women are likely to have urine sugar spill. However, the pathwa y is not theoretically meaningful with a coefficient less than 0.32. When review ing the correlation matrices, among the Black sample, no variables are moder ately correlated. Among the White participant sample, partner support is moder ately correlated with marital status (r = -0.415; Table 128). That is, married women are less likely to perceive their partners as supportive. When physical work strain during the second and third trimesters is added to the model, the sample size is reduc ed to 78 for the Black participant sample and 31 for the White sample. Due to the reduction in sample size, no further path analyses are assessed including work strain in the model. Table 127 Model 1 statistics of the interaction between ethnicity and predictors on urine sugar levels of pregnant women attending the Co unty Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels by Ethnicity Manifest Variable Estimate Standard Error t Statistic Black (N = 227) Partner Social Support Scale 0.057 0.047 1.219 Marital Status t1* -0.216 0.129 -1.666** White (N = 170) Partner Social Support Scale -0.018 0.054 -0.325 Marital Status t1 0.181 0.135 1.344 *t1 refers to data collected from the initial interview only. **p < 0.10.

PAGE 254

218Figure 27 Comprehensive modeling for the interaction between ethnicity and predictors of urine sugar levels of pregnant women attendi ng the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 128 Correlation matrix for Model 1 assessing the interaction between ethnicity and predictors on urine sugar levels of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Partner Social Support Marital Status Black Partner Social Support Scale 1.000 Marital Status -0.211 1.000 White Partner Social Support Scale 1.000 Marital Status -0.415 1.000 5.3.2 Evaluation of Hypothesis 5 When combining analyses from the Hy pothesis 4 results with Hypothesis 5, two models are again reviewed as history of abuse and abuse during the second and third trimesters of pregnancy ar e multicollinear. As with the analysis Black Partner Social Support Scale Marital Status Urine Sugar Levels 0.057 -0.216 Partner Social Support Scale Marital Status Urine Sugar LevelsWhite -0.018 0.181

PAGE 255

219 of Hypothesis 3, manifest variables of uri ne sugar levels are added to the path of manifest variables and high birth weight (F igure 28). Urine sugar levels are also assessed in terms of high bi rth weight. Table 129 shows model results. Model 1 includes history of abuse, while Model 2 includes abuse during the second and third trimesters of pregnancy. Figure 28 Structural equation modeling procedure for Hypothesis 5 assessment of the interaction between ethnicity and predictors of urine sugar levels and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 5: High Birth Weight Only•Create the model based on results from Hypothesis 4 Marital from the initial interview A partner’s social support A mother’s social support is included in all models•Incorporate significant predictors of high birth weight; and choose the model with the best fit (1 set of multicollinearvariables) among four models due to the subgroupingby ethnicity History of abuse Abuse from the final interview

PAGE 256

220Table 129 Goodness-of-fit indices for hypotheses 4 and 5 assessing the interaction between ethnicity and predictors on urine sugar levels and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 5 Model Ethnicity Chi-square Value Degrees of Freedom p value Comparative Fit Index R2 Urine Sugar Levels R2 High Birth Weight 1 Black 4.646 4 0.33 0.122 0.04 0.04 1 White 3.418 4 0.49 1.000 0.04 0.57 2 Black 4.915 4 0.30 0.332 0.05 0.05 2 White 2.866 4 0.58 1.000 0.04 0.65 Model 2, or the manifest variables including abuse durin g the latter half of the pregnancy appears to have the bes t fit based on the goodness-of-fit statistics, the correlation matrices, and the t statistics. Table 130 and Figure 29 present the findings from Model 2. Among the Black subset, only marital status is significantly causally associated with urine sugar levels. Single, never-married Black women are likely to have urine sugar sp ill. However, this association is not theoretically meaningful with a coeffici ent of less than 0.32. There are no causally associated paths leading to high birth weight. In contrast, among the White subset, although there are no causal ly significant paths leading to urine sugar spill, a mother’s social support, abuse during the latter half of pregnancy, and urine sugar spill are all causally associ ated with high birth weight infants. White women who have their mother’s social support, who are abused during the second and third trimesters of pregnancy, and who have sugar in their urine are most likely to have a high birth weight in fant. These factors account for 65% of the variance of high birth weight in th is sample among White women, and all are

PAGE 257

221 theoretically meaningful. Reviewing t he correlation matrix, among the Black subset, mother’s social support and abus e during the latter half of the pregnancy are moderately correlated (r = 0.64); and among the White sample, none of the variables are moderately correlated (T ables 131 and 132). Among Black women in the sample, a mother’s support increas es with the likelihood of physical or verbal abuse during the la tter half of the pregnancy. Table 130 Model 2 statistics of the interaction between ethnicity and predictors on urine sugar levels and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels and High Birth Weight by Ethnicity Manifest Variable Estimate Standard Error t Statistic Black (N = 195) Urine Sugar Levels Partner Social Support Scale 0.061 0.051 1.197 Marital Status t1 -0.256 0.141 -1.820* High Birth Weight Mother’s Social Support Scale 0.056 0.169 0.333 Abuse t2 -0.366 0.618 -0.592 Urine Sugar on High Birth Weight 0.091 0.111 0.819 White (N = 152) Urine Sugar Levels Partner Social Support Scale -0.028 0.059 -0.469 Marital Status t1 0.233 0.149 1.567 High Birth Weight Mother’s Social Support Scale 0.430 0.236 1.821* Abuse t2 1.386 0.416 3.332* Urine Sugar on High Birth Weight 0.558 0.159 3.500* *p < 0.10.

PAGE 258

222Figure 29 Comprehensive modeling for the interaction between ethnicity and predictors of urine sugar levels and high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Table 131 Correlation matrix for the interaction between ethnicity and predictors of urine sugar levels of pregnant women attending the Co unty Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Partner Social Support Marital Status Black Urine Sugar Levels 1.000 Partner Social Support Scale 0.231 1.000 Marital Status -0.097 -0.226 1.000 White Urine Sugar Levels 1.000 Partner Social Support Scale -0.035 1.000 Marital Status 0.012 -0.364 1.000 Partner Social Support Scale Marital Status Urine Sugar LevelsBlack 0.061 -0.256 Mother’s Social Support Scale Abuse t2 High Birth Weight 0.056 -0.366 0.091 Partner Social Support Scale Marital StatusWhite-0.028 0.233 Mother’s Social Support Scale Abuse t2 Urine Sugar Levels High Birth Weight 0.430 1.386 0.558

PAGE 259

223Table 132 Correlation matrix for the interaction between ethnicity and predictors of high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Mother’s Social Support Scale Abuse t2 Black Urine Sugar Levels 1.000 Mother’s Social Support Scale -0.101 1.000 Abuse t2 -0.040 0.642 1.000 White Urine Sugar Levels 1.000 Mother’s Social Support Scale 0.029 1.000 Abuse t2 -0.018 -0.084 1.000 Due to the high amount of varian ce (65%) explained by the model for White women combined with the findings from the Results Chapter, the relationship between a mother’s social support and abuse during the second and third trimesters of pregnancy requires fu rther exploration. Referring to the previous chapter, a mother’s social suppor t is predictive of a high birth weight baby. This specific finding is counter to the association of social support and adverse outcomes in the literature. Upon review, it was determined that all models including a significant associati on between a mother’s social support and high birth weight also contained abuse as a predictor (although abuse is not significant in any model). The curr ent path analysis highlights a moderate correlation between the two variables, and a statistically significant causal association between the two predictors and high birth weight. An adjusted logistic regression model is shown in T able 133 testing the interaction between

PAGE 260

224 social support and abuse excluding ethnicity. The interaction is significant, and Figure 30 displays a graph of results. As displayed in the figure, wom en with the highest amount of social support from their mothers and who are ph ysically or verbally abused during the latter half of pregnancy are over ten times more likely to have high birth weight babies compared with women who lack social support and are not abused toward the end of their pregnancies. The figure indicates a positive trend between increasing support among abused wo men and the odds of having high birth weight infants. Table 133 Logistic regression model of the interaction between physical or verbal abuse during the second and third trimesters and the mother’s social support scale with high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Interaction Model (N = 428)+ Independent Variables Odds Ratios 95% Confidence Intervals Mother’s Social Support Scale 1.069 0.779 1.467 Abuse t2 0.031 0.000 2.112 Mother’s Social Support Scale*Abuse t2 2.471++ 1.108 5.509 +Adjusted for alcohol and drug use during pregnancy, body mass index, education level attained, gestational age of the infant, and weight gain during the pregnancy. ++p-value < 0.05.

PAGE 261

225Figure 30 Logistic regression model of the interaction between the mother’s social support scale and physical or verbal abuse during the second and third trimesters on high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 (All possible scores; nonabused/0 score as reference), with triangles representing non-abused women and squares representing abused women T o furt her ex plo re the int era ction between a mother’s social s upport and abuse, models including the interaction of ethnicity were analyzed. The interaction was tested between ethnicity and each predictor separately, the interaction between the predictors with ethnicity included in the model, and a three-way interaction among ethnicity, mother’s social support, and abuse. T he three-way interaction was found to be the most significant and explanatory model. Table 134 displays results in conjunction with Figure 31. The ethnic group most strongly affect ed by the interaction is White, abused women (OR 89.39 for a score of 6 on the social support scale). The Interaction Between Mother's Social Support and Abuse on High Birth Weight0 2 4 6 8 10 12 score of 0123456 Social SupportOdds Ratios No abuse Abuse

PAGE 262

226 graph indicates a trend among this group with the odds of high birth weight increasing exponentially with t he increase in a mother’s so cial support. All three other groups are closely layered with Wh ite, non-abused women exhibiting the highest risk of high birth weight in fants compared with both Black abused and non-abused women. These findings suppor t the path analysis model presented in Figure 29. To ensure that the White, abused par ticipant group was not driven by outlying cases, frequencies were reviewed. Approximately 33% of the sample reported a social support score of 6 (i.e ., the highest score on the scale). Among those reporting the highest score, 38% we re White. For White women with a score of 6, 22% reported physical or ve rbal abuse during the second and third trimesters of pregnancy. Among that group, 10% had hi gh birth weight babies. For Black women in the sample, 19% reported abuse, and none of those women had high birth weight infants. As a result, the subset in the analysis does not consist of a cluster of outlying cases65. 65 See the Discussion Chapter for an explanation of results.

PAGE 263

227Table 134 Logistic regression model of the three-way interaction between physical or verbal abuse during the second and third trim esters, ethnicity, and the mother’s social support scale with high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Interaction Model (N = 346)+ Independent Variables Odds Ratios 95% Confidence Interval Mother’s Social Support Scale 1.304 0.868 1.957 Abuse t2 0.332 0.033 3.386 Ethnicity 1.622 0.489 5.381 Mother’s Social Support Scale*Abuset2*Ethnicity 1.799++1.104 2.930 +Adjusted for alcohol and drug use during pregnancy, body mass index, education level attained, gestational age of the infant at birth, total number of prenatal visits, and weight gain during pregnancy. ++p-value < 0.05. Figure 31 Logistic regression model of the three-way interaction between the mother’s social support scale, ethnicity, and physical or verbal abuse during the second and third trimesters on high birth weight infants of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 (All possible scores; Black/non-abused/0 score as reference), with the straight line representing Black, non-abused women; squares representing Black, abused women; circles representing White, non-abused women; and triangles representing White, abused women Interaction of Ethnicity, Mother's Social Support, and Abuse on High Birth Weight0 10 20 30 40 50 60 70 80 90 100 score of 0123456 Social SupportOdds Ratio Black No abuse Black Abuse White No abuse White Abuse

PAGE 264

228 When including birth weight as a continuous measure instead of a dichotomous variable, the manifest variabl es predicting the path to birth weight change. Partner social support and marita l status from the initial and final interviews are in the pathway to birth we ight. Due to the multicollinearity of marital status, two models are again pr esented in Table 135 (Figure 32 presents the analysis strategy). Model 1 includes a ll manifest variables and marital status from the initial interview, while Model 2 includes mari tal status from the final interview. Upon review of all goodness-of-f it statistics, correla tion matrices, and t statistics, Model 1 presents a better fit. Figure 32 Structural equation modeling procedure for Hypothesis 5 assessment of the interaction between ethnicity and predictors of urine sugar levels and the birth weight of infants born to pregnant women attending the Co unty Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 5: Birth Weight Only•Create the model based on results from Hypothesis 4 Marital from the initial interview A partner’s social support A partner’s social support is included in all models•Incorporate significant predictors of high birth weight; and choose the model with the best fit (1 set of multicollinearvariables) among four models due to the subgroupingby ethnicity Marital status from the initial interview Marital status from the final interview

PAGE 265

229Table 135 Goodness-of-fit indices for hypotheses 4 and 5 assessing the interaction between ethnicity and predictors on urine sugar levels and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 5 Model Ethnicity Chi-square Value Degrees of Freedom* p value Comparative Fit Index R2 Urine Sugar Levels R2 High Birth Weight 1 Black 8.880 5 0.11 1.000 0.04 0.03 1 White 8.620 5 0.13 1.000 0.02 0.07 2 Black 1.387 2 0.50 1.000 0.01 0.01 2 White 0.684 2 0.71 1.000 0.01 0.09 *Degrees of freedom vary due to Model 1 being just-identified and Model 2 being over-identified. Table 136 presents findings from M odel 1 (Figure 33). Among the Black subset, marital status as in the previous model is causally associated with urine sugar spill. Black single, never-marri ed women are likely to have urine sugar spill. When examining the birth weig ht paths, urine sugar spill is causally associated with higher birth weight inf ants. The only theoretically meaningful path is between urine sugar levels and bi rth weight. Among White women, the only statistically significant causal path is between urine sugar spill and birth weight. Reviewing the co rrelation matrices (Tables 137 and 138), no variables are moderately correlated in the Black participant sample; however, among the White sample, partner social support and ma rital status in terms of urine sugar are moderately correlated (r = -0.42). That is, White women who are married are less likely to perceive they have their partner’s support.

PAGE 266

230Table 136 Model 2 statistics of the interaction between ethnicity and predictors on urine sugar levels and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on Urine Sugar Levels and Birth Weight by Ethnicity Manifest Variable Estimate Standard Error t Statistic Black (N = 227) Urine Sugar Levels Partner Social Support Scale 0.057 0.047 1.218 Marital Status t1 -0.216 0.129 -1.667* Birth Weight Partner Social Support Scale -0.123 0.099 -1.248 Marital Status t1 -0.054 0.335 -0.162 Urine Sugar on Birth Weight 0.575 0.309 1.860* White (N = 170) Urine Sugar Levels Partner Social Support Scale -0.017 0.054 -0.325 Marital Status t1 0.182 0.135 1.345 Birth Weight Partner Social Support Scale 0.118 0.156 0.754 Marital Status t1 -0.086 0.371 -0.233 Urine Sugar on Birth Weight 0.955 0.382 2.496* *p < 0.10.

PAGE 267

231Figure 33 Comprehensive modeling for the interaction between ethnicity and predictors of urine sugar levels and the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 19902001 Table 137 Correlation matrix for the interaction between ethnicity and predictors of urine sugar levels of pregnant women attending the Co unty Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Partner Social Support Marital Status Black Urine Sugar Levels 1.000 Partner Social Support Scale -0.006 1.000 Marital Status -0.004 -0.211 1.000 White Urine Sugar Levels 1.000 Partner Social Support Scale 0.000 1.000 Marital Status -0.023 -0.415 1.000 Partner Social Support Scale Marital Status Urine Sugar LevelsBlack 0.057 -0.216 Birth Weight -0.123 -0.054 0.575 Partner Social Support Scale Marital StatusWhite-0.017 0.182 Urine Sugar Levels Birth Weight 0.118 -0.086 0.955

PAGE 268

232Table 138 Correlation matrix for the interaction between ethnicity and predictors of the birth weight of infants born to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Urine Sugar Levels Partner Social Support Marital Status Black Urine Sugar Levels 1.000 Partner Social Support Scale -0.161 1.000 Marital Status 0.170 -0.207 1.000 White Urine Sugar Levels 1.000 Partner Social Support Scale 0.057 1.000 Marital Status -0.167 -0.321 1.000 The final path analysis presented in this chapter is an assessment of manifest variables and C-section (Figure 34). Only marital status and a mother’s social support are significantly associ ated with C-section and interactive with ethnicity. Table 139 presents model result s. Among the Black subset, the chisquare goodness-of-fit test is non-significant ( 2 = 3.436; p -value = 0.18), the CFI is 1.000 (model is ju st-identified), and the R2 equals 0.05. For the White sub sample, the chi-square goodness-of-f it test is non-significant ( 2 = 0.390; p -value = 0.39), the CFI equals 1.000 (model is just-identified), and the R2 is 0.004 for the model. Upon review of the goodness-of-fit indices, the correlation matrix, and the t statistics, the models are determined to fit; however, t he fit is poor. Although an explanation is provided, the models should not be considered causally representative of the path from predictor s to C-section (e.g., approximately zero variance is explained by ei ther model). Among the Black participant sample, social support of a participant’s mother is statistically causally associated with C-

PAGE 269

233 section (Figure 35). The association is not theoretically m eaningful (> 0.32). None of the manifest variables are signi ficantly associated with C-section among White women in the sample. None of the manifest variables are moderately or highly correlated in either ethnic group (Table 140). Figure 34 Structural equation modeling procedure for Hypothesis 5 assessment of the interaction between ethnicity and predictors on Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Hypothesis 5: Caesarean Section•Create the model Marital status from the initial interview A mother’s social support

PAGE 270

234Table 139 Model 2 statistics of the interaction between ethnicity and predictors on Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Manifest Variables on C-section by Ethnicity Manifest Variable Estimate Standard Error t Statistic Black (N = 227) Mother’s Social Support Scale 0.109 0.061 1.795* Marital Status t1 -0.060 0.130 -0.463 White (N = 170) Mother’s Social Support Scale 0.002 0.004 0.042 Marital Status t1 0.078 0.125 0.623 *p < 0.10. Figure 35 Comprehensive modeling for the interaction between ethnicity and predictors of Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Mother Social Support Scale Marital Status C-sectionBlack 0.109 -0.060 Mother Social Support Scale Marital StatusWhite0.078 C-section 0.002

PAGE 271

235Table 140 Correlation matrix for the interaction between ethnicity and predictors Caesarean section births to pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Correlation Matrix Variable Name Mother Social Support Marital Status Black Mother Social Support Scale 1.000 Marital Status 0.013 1.000 White Mother Social Support Scale 1.000 Marital Status 0.037 1.000

PAGE 272

236 CHAPTER 6 DISCUSSION 6.0 Introduction The following chapter summarizes t he major findings from both the Results and Structural E quation Modeling chapt ers, applies major findings in the context of the proposed theoretical framew ork, expands on the limitations of the study discussed in the Methodology chapter evaluates the consistency of major findings with the current epidemiologic lit erature as reviewed in the Literature Review Chapter, addresses the public heal th importance of the major findings, and suggests future directions for further study. The organization of this chapter is based on the topics listed above and eac h section is delineated by a heading that outlines each topical point. 6.1 Major Findings Two sets of major findings are addre ssed; inferential findings from the Results Chapter and path analysis models from the Structural Equation Modeling Chapter. Inferential results are divided into those addressing urine sugar levels and high birth weight, and those in relation to C-section. The path model results follow a similar methodology separating ur ine sugar levels and high birth weight from C-section.

PAGE 273

237 6.1.1 Inferential Results The primary finding of this dissertat ion is the association between high urine sugar spill and high birth weight. High amounts of spill in the urine are positively associated with the birth of high-weight infants compared with no detectable spill (OR = 3.25; 95% CI 1.30-8.10). In addi tion, when birth weight is treated as a continuous m easure, low and high amounts of spill are associated with increased birth we ight among infants. Referring to Hypothesis 2, the impact of social support of a participant’s mother is protective of urine sugar sp ill, whereas depression during the latter half of the pregnancy is predictive of urine sugar spill. Both measures remain significant regardless of the inclusion or exclusion of the multicollinear measures of history of abuse and abuse during t he second and third trimesters of pregnancy (Mother’s social support OR = 0.90, 95% CI 0.811.00; Depression OR = 1.04, 95% CI 1.00-1.09). When examining high birth weight (cat egorically), again, a mother’s social support is significantly associated with t he birth of a high-weight infant. In contrast to the association with urine sugar levels, the im pact of a mother’s social support is not protective of having high bi rth weight infants, but increases the odds (OR = 1.56; 95% CI 1. 13-2.17), especially a mother’s emotional support (OR = 2.03; 95% CI 1.14-6.63) Ethnicity is also si gnificantly associated with having high birth weight infants; White women in the sample are over two and a half times more likely to gi ve birth to a high-weight infant compared with Black

PAGE 274

238 women (OR = 2.74; 95% CI 1. 14-6.63). As a continuous measure, birth weight retains its association with ethnicity, wit h White women more likely to have high birth weight babies compared with Black wo men. In terms of C-section, only marital status, specifically single never-married women, is significantly associated66. The interaction of ethnicity and t he predictors of urine sugar spill are addressed in Hypothesis 4. Marital st atus (i.e., single never-married women), partner social support, and physical wo rk strain during the second and third trimesters of pregnancy significantly intera ct with ethnicity on urine sugar spill. Essentially, the highest risk group for ur ine sugar spill is White, married women compared with all other groups Again, White women ar e at an overall higher risk for urine sugar spill compared wit h Black women; however, White women with partner support are at the highest ri sk of urine sugar spill compared with all other groups. Physical work strain among women in the sample increases the risk of urine sugar spill. Working for pay at all during pregnancy affects both ethnic groups. The relationship becomes complex when examining the trends in physical work strain score and its affe ct on each ethnic group. White women present an increased risk for urine sugar spil l at the lower end of the strain scale although the upper end of the scale only decreases by a one and a half-fold change in the odds ratio (e.g., OR = 3.5 to 1.8), while Black women present an 66 For a more detailed description of reasons why predictors were not associated with C-section, see the Limitations section.

PAGE 275

239 increased risk as the scale increases (e .g., OR = 1.0 to 3.7). Although the interaction presents a different tr end depending on ethnicity, the larger implication is that while the odds fo r White women decrease, the odds still present a substantial risk of urine sugar spill if that woman works for pay outside the home at all during pregnancy. The odds for Black women increase as expected with the scale. The final hypothesis is an examinati on of interaction between ethnicity and the predictors on the outcome s of high birth weight an d C-section. Ethnicity interacts with a history of physical or verbal abuse, physical or verbal abuse during the latter half of the pregnancy, and a mother’s social support. Regardless of type of abus e (before or during pregnanc y), White women have a pronounced increase in risk of having a hi gh birth weight infant compared with Black women (OR ranges from 1.6 to 5. 3). White women who are physically or verbally abused are at the highest risk fo r a high birth weight infant compared with all other groups (OR hi story of abuse = 4.1; OR abuse during latter half of pregnancy = 5.3). In terms of a mother ’s social support, White women with support are at the highest risk of having a high birth weight baby (OR = 11.5) compared with all other groups67. When birth weight is examined as a continuous measure, marital status and the social support of a partner interact with ethnicity. White, ma rried women are at highest ri sk for having a high birth weight baby compared with all ot her groups, and among White women, 67 See the path analysis section for a further discussion of this phenomenon.

PAGE 276

240 increased partner support is associated wit h high birth weight infants. For Caesarean section, only marital status and the support of a participant’s mother interact with ethnicity. White, single ever-married women are at the highest risk of a C-section birth compared with all other groups and White women regardless of support are most likely to have a Csection, with increasing support among White women as the highes t risk group (OR = 6.6). 6.1.2 Structural Equa tion Modeling Results The path analysis models are a compliment to the inferential results and also a tool to explore further the relationships between the predictors and outcomes that may not be consistent with current literature. Path models are composed of all significant findings fr om the Results Chapter and include submodels (i.e., interaction analyses) of predi ctors that reduce sample size below the recommended sample size for modeling68. In terms of urine sugar levels, the full model only includes one significant path; a mother’s social support is causal ly associated with a decrease in urine sugar levels69. The impact of this path on expl aining the variance of the model is minimal (0.03), and therefore, not explanat ory in reference to the preventative factors associated with urine sugar spill70. The model for the group of women in the sample who worked for pay outside the home during their pregnancy presents more readily interpretable paths. A mother’s social support, depression 68 C-section results are not causally significan t, and are, therefore, excluded from discussion. 69 As defined in Chapter 5, causal here refers to stat istical causality in reference to path analysis modeling. 70 As explained in the Structural Equation Modeling Chapter any path coefficient less than 0.32 is not theoretically meaningful.

PAGE 277

241 during the last two trimesters of pregnancy, and work strain during the latter half of the pregnancy explain 15% of the vari ance of urine sugar spill. Again, a mother’s social support is protective, while depression and work strain present an increase in risk. Again as with the prev ious urine sugar model, none of the path coefficients reach a theoretica lly meaningful am ount (> 0.32). When examining the causal associations with urine sugar levels leading to high birth weight, the impact of a mother’s social supp ort is protective of high urine sugar levels, but a risk for high birt h weight. In addition, high urine sugar levels are predictive of having high birt h weight babies. The pathway to higher urine sugar levels is on ly minimally explained (R2 = 0.04); however, the pathway leading to high birth weight is more thoroughly explained (R2 = 0.29). In that pathway, high urine sugar le vels theoretically impact the birth of high-weight babies (r = 0.33). Results fr om the logistic regression model in the interaction section of the Structural Equation Modeli ng Chapter aid in further interpreting these results. When examin ing the interaction of a mo ther’s social support and physical or verbal abuse during the latter half of t he pregnancy, the odds of a high birth weight baby more than double for women receiving the highest amount of support compared with abused women rece iving less support (O R = 10.70). It appears that although social support is gener ally protective, when increased in a situation of abuse, it may exacerbate the likelihood of a woman birthing a highweight infant. Again, ex amination of the group of women who work for pay during the latter half of pregnancy, depression during the second and third

PAGE 278

242 trimesters is causally associated with high urine sugar spill, and high urine sugar spill is causally associated with high birt h weight babies. As with the former path analysis, a higher percentage variance is ex plained in the smaller sample of women compared with the entire sample ( R2 = 0.20 for the path between predictors and urine sugar levels; R2 = 0.39 for the path to hi gh birth weight). In conjunction with high urine sugar levels, ethni city is also causally associated with high birth weight infants, with White wom en more likely to have high birth weight babies (ethnicity r = -0.93; urine sugar r = 0.32). Birth weight as a continuous measure yields similar results. No predictors are significantly causally associated with high urine sugar levels, but both ethnicity and high urine sugar levels are causally associated with increasing birth weight (ethnicity r = -0.95; urine sugar r = 0.55). A gain, the variance explained by the path is minimal (0.07). In the sample of working women, however, urine sugar levels are no longer predictive of bi rth weight and ther efore the model is not as useful in explaining predictor s of high birth weight infants. To summarize, the urine sugar level path models alone predict few to no causal associations. Of the models predi cting high urine sugar levels and high birth weight, the overall model is the mo st interpretable with the causal pathway leading to high birth weight as the most explanatory path. The secondary model, though focusing on a smaller set of women who work during pregnancy, explains the highest magnitude of variance, and is highly predictive of both high urine sugar levels and high birth weight.

PAGE 279

243 The findings assessing the interaction effect of ethnicity on the pathways are the most explanatory of all the path models. Although analysis of urine sugar levels alone explains only a mi nimal amount of variance, analysis of predictors and high urine sugar levels leading to high birth weight infants is key to understanding the causal relationships between predictors and outcomes among White women in the sample71. Essentially, none of th e predictor or manifest measures are significantly associated with high urine sugar levels in any theoretically meaningful way. A mother’s social support, physical or verbal abuse, and high urine sugar levels are hi ghly causally associated with high birth weight infants. These th ree factors explain 65% of the variance of high birth weight infants. In suppor t, logistic regression results in that same path analysis section indicate that ethnicity, a mother ’s social support, and physical or verbal abuse during the latter half of the pregnancy all interact to significantly impact the birth of high-weight babies. Specifical ly, among White women who state they are physically or verbally abused during the second and third trimesters of pregnancy, birthing a high-weight infant is over eleven times more likely compared with White women who are not abused but re ceive the highest amount of social support, and over eighteen time s more likely compared with Black nonabused women who have that same level of a mother’s support. Essentially, ethnicity, a mother’s social support, and physical or verbal abuse during the latter 71 The path model for Black women yielded minimal informati on regarding predictors of higher urine sugar levels and not predictive causal pathways to high birth weight infants.

PAGE 280

244 half of the pregnancy are all effect meas ure modifiers, or interactive in exacerbating the risk of having a high birth weight infant. In conclusion, the path analysis models highlight the causal association between higher urine sugar levels and high birth weight. This association is present throughout each path model, and moder ately to highly predictive of high birth weight. Of specific interest are the results of the group of women who work for pay outside the home. Among wo men who work during pregnancy, depression during the second and third tr imesters and physical work strain are causally associated with having high birth weight infants. The most predictive models are those in which ethnicity divi des the sample. Among White women, a mother’s social support, high urine sugar levels, and abuse are highly causally associated with the birth of a high-weight infant. 6.2 Application of the Theoretical Framework Krieger’s ecosocial framework coupled with Berkman and Glass’ model is directly applicable to the major findings of this research. The model consists of measurements at the micro and macro levels, pathways of exposure to disease, and assessments of outcome measures. To contextualize the model, Krieger’s ecosocial framework is composed of i ndividual embodiment of exposure and disease; the pathways of embodiment incl uding the biologic, social and material; the cumulative interplay between the pathway s; and the accountability of results. The following discussion begins with results at the micro or individual level in

PAGE 281

245 conjunction with a brief discussion on em bodiment, followed by a description of significant pathways to the major outcomes of the study. In this research, exposure and diseas e at the individual level are primarily mediated by social support. To be specif ic, the support of a participant’s mother is key in protecting against high urine s ugar spill during pregnancy. However, as summarized in the Results Chapter, that same support appears to increase the risk of having a high birth weight baby. T hese results seem counterintuitive until analysis of the buffering effect of social support is reviewed by moving to the macro level of measurement, or the in teraction of ethnicity among this low income population. For t he two ethnic groups of Bla ck and White in this study, cultural practices provide insight into this complicated causal pathway. Among Black women in the study, social support acts as predicted; as social support increases, the likelihood of adverse event s decrease. Results of White women’s responses in the sample do not follow t he same pattern. Receiving a mother’s social support, while protective of high ur ine sugar spill increases the likelihood of a high-weight birth. Possible reasons for the difference in response between the two groups may be attributed to varying cultural practices among these two sub-populations [178]. Social support by a member of the “nuclear” family may be defined disparately among the two groups. For in stance, Black women in the sample may receive support from multiple sources; the extended family plays a large role in general among this group of women as opposed to the influence of specific

PAGE 282

246 nuclear family members such as the mo ther [31, 179-181]. These women are also more likely to be single, never-marr ied with multiple partners [30]. This partner base may also provide a signifi cant amount of support throughout the pregnancy and childbirth process [71]. As shown in the St ructural Equation Modeling Chapter, support fr om a partner and marital status are inversely correlated (r = -0.23), although the correlation is weak in this sample. That is, as support increases, participants are more likely to be single, never-married women. Among White women in the samp le, however, the social support based is more structured with a core group of people such as parents and current partner providing the bulk of support. As a result, White women may rely more on their mothers and thus the social in fluence and support cent ers on them as primary providers. The three pathways to embodiment are influenced by participant’s physiologic response to that social support; it is protective of urine sugar spill in both groups, but increases risk of high birt h weight infants am ong White women. In this research, the specific biologi c response to psychosocial and physical factors of interest is glucose intoleranc e and the by-product of sugar spill in the urine. The ultimate outcome is the ov er-processing of sugar by the fetus and resulting weight gain. A possible soci al response to the biologic change during pregnancy is to either increase support as necessary, or seek support through other individuals or institutions. The materialistic response in conjunction with the social response is to increase instru mental resources, again, putting a strain

PAGE 283

247 on the mothers of White women to prov ide both emotional and instrumental security during pregnancy. The psychological and physiologic pathways to disease which encompass the micro and macro levels, embodimen t, and the pathways to embodiment, are mediated by social support. Overall, abuse both verbal and physical plays an interactive role with social support in both pathways to disease. As discussed previously, abuse interacts with a mother’s social support to increase the risk of having a high birth weight baby. The odds are greatly increased among White, abused women (i.e., interaction is more than multiplicative), but minimally impact the odds among the other three groups (e .g., White/non-abused; Black/abused; Black/non-abused). The ‘cumulative inte rplay’ between levels of embodiment (the micro and macro) and the pathway to disease is through the theory that social support is intended to buffer adverse events. In this study, abuse is directly causally associated with havi ng high birth weight infants among White women. The social support of the mother is an attempt to buffer the effects of that abuse. The reason increased social support appears to increase the risk of birthing a high-weight baby is that as a st rategy, it is ineffective in buffering abuse. The social support provided is an intervention attempt by a core family member, but inconsequential in the caus al pathway from abuse to high birth weight. Further reasoning for the interaction of social support and abuse includes cultural differences in the social definiti ons of ‘social support’ and ‘abuse.’ As

PAGE 284

248 noted previously, social support is defined as emotional or instrumental, and may be provided by core or extended family members. Emotional and instrumental support may not be anticipated from t he same sources, or one may take precedence over the other. In contrast, the concept of abuse may differ among Whites and Blacks in the study. While physical abuse may be visually documented, verbal abuse is not as strai ghtforward an observation. White and Black women in the study may perceive t hese types of abuse differently, such as forms of physical abuse as deserved or verb al abuse as a form of neglect rather than uniformly defining both types. Therefor e, questions could be misinterpreted and responses differentially misclassified. Interplay is also exemplified amon g the group of working women in the sample. As shown in the path analysis, working women have combined factors of depression and the physical strain of wo rking that increase the risk of higher urine sugar spill and in turn, increase the ri sk of having high birth weight infants. This pathway is more direct and biologically plausible than that of social support, with depression during pregnancy logically a ssociated with the stress of working for pay while anticipating a child. Such stress causes a chain reaction within the body, leading to a decrease in insulin production, a lack of glucose processing, and a response from the fetus. In turn the fetus over-produces insulin, overprocesses glucose, and gains weight. To enhance the aspect of cultural in fluence at the macro level on the exposure-disease relationship, the impac t of these findings at the population

PAGE 285

249 level is fundamental to this discussion. Although these findings are specific to a group of low income, urban women in Alab ama, it is possible to discuss these pathways in the context of a larger popul ation. Due to a lack of resources, instrumental or material support is a major concern among poorer populations. Transportation to and from clinics, money for goods and services, and help during emergencies are all types of s upport that must be addressed through alternative methods by this population. In support, the working poor are especially vulnerable to resulting psyc hological stress from the lack of such resources. As a whole, the lifestyle of the urban poor is indicative of these results, specifically groups that utiliz e a limited social support network which results in problematic health outcomes lik e high urine sugar levels and high birth weight infants. Kr ieger’s work supports this interpretation based on her research of discrimination [163]. She posits that health disparities are based on social inequalities expressed through acts of ra cism and racial stereotyping leading to a lack of access to resources such as adequate health care, insurance, and treatment. This discrimination result s in both psychological and physiologic responses increasing susceptibility to disease or other adverse health events [182]. In the context of th is dissertation, those event s are increased urine sugar spill and resulting high birth weight infants. Figure 36 provides a summary of t heoretical results. Based on all findings, the macro level discussion of soci al structural conditions that affect disease is focused on ethnic differences within the sample, and micro level

PAGE 286

250 results addressing the psychosocial mec hanisms in the causal pathway to disease consist of social support, spec ifically the support of the mother during pregnancy. The psychological pathway is primarily composed of verbal abuse with depression as a factor fo r the working participants. The physiologic path is, again, comprised of physical abuse with physical work strain a focus for the working participants. Both paths lead to hi gh urine sugar levels and ultimately, to high birth weight babies. Figure 36 Revised Theoretical Model and Framework Based on Synthesis of Results The final component of the ecosocial fr amework is dissemination of results to the public. Responsibility for the di ssemination of these findings lies in the hands of health providers, practitioners, and researchers. In epidemiology, reporting these results to t he professional and lay communi ty is the responsibility Macro Level Micro Level Pathways Outcomes Social Structural ConditionsCulture -Interaction of ethnicityPsychosocial MechanismsSocial Support -Perceived Support (Mother’s)Psychological-Depression -Verbal AbusePhysiologic-Work Stress -Physical AbuseHigh Urine Sugar Levels High Birth Weight BabiesAccountability and Agency EmbodimentPathway To Embodiment Cumulative Interplay (all levels)

PAGE 287

251 of each investigator, and although bias is unavoidable in both research and interpretation of findings, co mmunication of results is necessary to modify public health practice72. The following section addressi ng study limitations describes the biases involved in this research, and the implications of that bias in interpretation. 6.3 Study Limitations The major limitations of this dissert ation are selection bias, response bias, generalizability of results, limited interpretation due to uncollected data, and impact of other co-morbid events. As addressed in the Methodology Chapter, selection bias impacts the dat a through the eligibility of participants, interpretation of missing data, combination of both data sets, and interpretation of results. Response bias, although impacting the re sults to a lesser degree, affects the interpretation and theoretical discussion of major findings. Data not included in the original grant protocol such as dis ease diagnosis data, multiple measures of screening tests, or other factors detailing recorded medical procedures affected the methodology of the current analysis and interpretation of results. Collection of specific data at defined time points dur ing pregnancy limited the interpretability of results. In addition, a specific co-morbidity that may have impacted final results is the effect of low-weight births among women in the sample (<= 2500 grams or 2.5 kilograms). 72 Further explanation is given in the P ublic Health Implications section.

PAGE 288

252 The selection of participants severely limited the number of women eligible for study. Due to the interview structure (e .g., two interviews during the course of a pregnancy), women further along in t heir pregnancies were ineligible to participate in the study. Women who neglect to receive adequate prenatal care or care early in their pregnancy are at the highest risk for adverse outcomes and logically are of most interest to researc hers of birth complications. However, due to the interviewing strategy of recruiti ng and interviewing during prenatal visits, these women were not included in the samp le. In effe ct, the sample consists primarily of the healthiest women who s eek prenatal care early during pregnancy and who are the most likely to receiv e adequate monitoring and screening tests for diseases of pregnancy or adverse outcomes. As reviewed in the Results Chapter the types of missing data are in part due to the sampling strategy and the struct ure of the interview schedule. Again, some data are missing due to the types of participants sampled. Any women who changed clinics during their pregnancy became lost to follow-up, and if the change was not local, were less likely to have a follow-up interview. Due to the nature of the study design, women who mi scarried after the initial interview or gave premature birth (< 32 weeks) prior to the second interview, contributed to missing measurements at the final intervie w. Findings are only applicable to women who have almost full-term pregnanc ies (give birth >= 32 weeks gestation) and receive consistent care at the same clinic. Further, since all the women in the sample received care at the local Co unty Health Departm ent prenatal clinic,

PAGE 289

253 results are also only applicable to lo w-income, Medicaid insured women who receive adequate prenatal care and carry t heir pregnancies approximately to fullterm. A major issue both at the onset of this study and throughout the analysis is the combination of two separate data se ts utilizing similar methodologies. The Methodology and Results Chapters address this issue by county to county comparisons, state to county compar isons, and controlling for confounding by site-specific characteristics. Both study sites are comparable in terms of demographic characteristics; however, bot h of the initial gr ant-funded studies had different foci. The Birth Weight Study measured the impact of psychosocial and physical stressors on low birth wei ght babies among chil dbearing-age women (20-34 years of age). The Healthy Start Evaluation measured the effectiveness of an intervention among high-risk teenage gi rls (14-18 years of age). The major disparity between these two samples is that they measure the effects of stressors on two very different risk groups In order to address this concern, a site-specific variable was created and treat ed as a potential confounder during analysis. This variable would remove the effects of ri sk-specific behaviors, health status, and age as possible factors presenting major differences between the two samples. Resulting analyses indicated that site did not confound the exposure-disease associations, and further, that age as part of the site variable or as measured alone in each sample, did not confound results.

PAGE 290

254 Due to the limitations presented in te rms of selection of participants and analysis of collected data, ma jor findings must be conservatively interpreted. Findings are only generalizable to urban, low income women in the South. In terms of the cultural practices discussed in earlier sections, interpretation of support mechanisms in reference to expos ure and disease is applicable to this defined population only. Responses of participants may be r egionally or locally specific. How participants defined words and phrases used in the interviews is explicit to the population under study. As reflected in the discussion addressing ‘abuse,’ interpretations of questions may vary across ethnic groups rendering all interpretation applicable to identified, pre-defined groups. Application of the theoretical framework is al so population-specific. If other ethnic groups, for example Hispanics or Native Americans we re included in the sample, overall findings may have been different from those conclusions reached studying Blacks and Whites only. Major findings, therefore, are gener alizable to these ethnic groups only, and are onl y regionally applicable. Specific data such as history or current diagnosis of gestational diabetes, pre-eclampsia, and eclampsia; glucos e tolerance test results and other laboratory tests; and multiple screening re sults including urine sugar levels and ketone levels were uncollected but woul d have enhanced current study findings. Disease diagnoses past and current coul d have been controlled in analysis as confounding factors. However, this wa s not possible given the data collected.

PAGE 291

255 As with disease diagnosis, control of t he confounding effects of laboratory tests may have strengthened significant associati ons. Finally, including the full range of urine sugar level tests would have increased the robustne ss of the analysis, allowed analysis over t he entire pregnancy instead of one time point for each participant73, and, again, strengthened cu rrent results. Multiple factors are included in a physician’s decision to perform a Caesarean section during childbirth. T he facilities readily available at the hospital, the instruments in the labor and delivery room, insuranc e, fetal distress, contraction intensity, and c linical style are a few exam ples of peripheral factors involved in the decision to perform a C-se ction surgery. None of these data were collected under the grant pr otocols, and are therefor e missing in the current analysis. The influence of such factors is unknown, but may have been highly influential and as a result, decreased the influence of the factors in the current study on Caesarean section. Exclusion of these data did not have a large impact on significant associations, but would have supplemented the results found in this dissertation, and may account for the lack of statistical significance in terms of the identified predi ctors and C-section. As discussed in the Literature Review Chapter, there is a defined set of risk factors that contribute to having a lo w birth weight infant. Although a specific set of biologic factors were controlled in this analysis (i.e., confounding factors), the affect of including low-weight births with the normal weight births may have 73 Recall that only the highest urine sugar reading was used in analysis regardless of when it occurred during pregnancy. The average gestational age of the infant at highest urine sugar reading was 29 weeks.

PAGE 292

256 influenced results. It is anticipated t hat any impact would have biased the odds ratios towards the null value as including these cases would lessen the influence of the predictors on high-weig ht births. To address this issue, all final models were re-analyzed excluding all low and very lo w-weight births from the sample (N = 56). Appendix F lists the results of a ll analyses. As hypothesized, including low birth weight infants in the analysis had little to no impact on final models. Odds ratios in all analyses changed less than three hundredt hs of a decimal place (0.03), and all remained statistically significant. 6.4 Study Strengths Of major importance is a hi ghlighting of strengths spec ific to this study in terms of design, data collection, and analys is. The selection of a prospective cohort is ideal for studying pregnancy-relat ed outcomes. Due to the duration of a pregnancy, a large number of participants ma y be recruited in a comparatively shorter amount of time than with other disease outcomes of a non-infectious nature such as cancer, heart disease, or mental disorders. In terms of data collection, many biases traditionally associated with the cohort design were controlled based on a rigorous methodology. Attrition bias was limited due to the detailed follow-up of all par ticipants, even those that le ft the state immediately prior to giving birth. The likelihood of interviewer bias decreased with multiple levels of interviewer training and observati on before individual interviewers began conducting sessions alone. By training in terviewers to establish rapport with participants throughout each interview, reporting bias was minimized.

PAGE 293

257 Methodologies in data analysis included meas uring both urine sugar levels and birth weight in multiple ways (e.g., or dinal versus binary; continuous versus categorical), multiple a ssessments of potentially conf ounding factors, and using multiple modeling techniques to verify findings. Potential confounders were generally assessed and then assessed in te rms of their impact on each predictoroutcome association. Finally, structural equation modeling was used as a tool to supplement and further define the associatio ns found in the inferential analysis. 6.5 Consistency with Current Literature All major research associating any type of glucose intolerance and high birth weight are focused on diagnosis and treatment of gestat ional or Type II diabetes. None of the epidemiologic st udies reviewed endeavored to associate urine sugar screening with high birth weight infants. A reason for such exclusion may result from recommendations from the American Diabetes Association and the World Health Organization. The ADA recommends that urine glucose monitoring, a preliminary screening test not be used to monitor or diagnose gestational diabetes [183] as does the WHO [93]. Since high birth weight infants are primarily identif ied with women who are diagnos ed with gestational diabetes, studies focus on glucose intolerance and diagnosis, then diagnosis and high birth weight. Studies do indicate an associ ation between diagnosis and high birth weight, and with the preliminar y test and diagnosis. It is logical, therefore, that the screening test which is a pre-curs or to the diagnostic test would be associated with diagnosis and result ing high birth weight infants.

PAGE 294

258 This dissertation indicates a biologic ally plausible link between elevated urine sugar levels of the mother and high-we ight births. In term s of Hill’s criteria, although consistency with current research is weak, if the original premise of the causal pathway is examined, increased urine sugar excluding glucose tolerance testing and diabetes diagnosis is causally associated with having a high birth weight baby. The causal pathway does not include, however, as in other research, subsequent Caesarean section [ 184]. The strength of the association between urine sugar levels and high birth weight is high wit h an odds ratio of greater than 2.00, the maximum assu med difference in risk between high and normal weight infants at the beginning of this research. Also temporal sequencing is supported by both the measurem ent of urine sugar levels prior to childbirth and by the literal requirement that each pa rticipant be pregnant during the interviewing portion of the study and must have had a viable, live birth in order to be included in the outcome phas e of the study. A nalogy with another similar population is again, supported if the popul ation of reference is low income pregnant women. In this specific ur ban, southern population of Black and White women only, analogy to similar studies is weak due to the lack of research specifically of this population and the a ssociation between urine sugar levels and high birth weight. As with low-weight births, social s upport impacted study results; however, the opposite effect was observed among the physically abused white women in the study. Previously addressed in this chapter, the associ ation between social

PAGE 295

259 support and high birth weight as with studies of low birth weight was expected to protect against problematic outcomes. In stead, it exacerbat ed the outcome. Therefore, the original premise that there was a U-shaped distribution between birth weight and morbidity may not be entir ely correct for this predictor [12]. However, in support of this assertion, and based on previous research associating depression and physical work stra in with low-weight births, this study also found an association between thes e socio-cultural predictors and highweight births. In addition, as with Nor beck and Anderson’s study, social support affected Black and White women differently [70]. These findings corroborate the concept of social support as a buffer to adverse outcomes, interacting with other factors to exacerbate an exis ting association. 6.6 Public Health Implications The public health impact of these findings includes three recommendations: urine sugar screening, a non-invasive required test, may be used to identify higher-risk Black and White women; the sociocultural measure of social support is both protective and in teractively predictive of problematic pregnancy outcomes; and intervention during pregnancy may reduce the proportion of highweight bi rths among this population. The urine sugar spill test is given at each prenatal visit. As described in the Methodology chapter, the urine sugar test is administered during the visit using a dipstick in a urine specimen to determine the amount of sugar expelled. The test identifies categories of spill in cluding none, trace am ounts, 1+, 2+, 3+,

PAGE 296

260 and 4+ or higher. Results from this disse rtation indicate that Black or White women who have a spill of 1+ or higher during their pregnancy are at a higher risk of giving birth to a high-weight infant compared with women who have no detectable sugar spill duri ng their pregnancy (OR 3.30; 95% CI 1.35-8.08). Among these normally low risk women, ur ine sugar monitoring at each prenatal visit is predictive of high birth weight with 84.4% efficiency. While the positive predictive value of this screening test is moderate (32.3% ), the negative predictive value (87.6%) and specificity (95.2%) are high. Clinically, these findings support ruling out thos e women who are at little to no risk of having high birth weight infants (e.g., women with no/trace urine sugar), and, in turn, identifying a new group of “high risk” wo men who should be cl osely monitored for the duration of their pregnancy. Sociocultural measures are rarely applied in medical practice due to the lack of epidemiologic evidence to support implementation, and the practicality of administering and interpreting such measures [185]. However, as shown in the results of this dissertation, these nonbiologic measures do impact the risk of disease and are modifiable if identifi ed early enough in the pregnancy process. Social support is both protective and a buffe r to the major outcome s in this study. Assessing the support status of women who ent er the prenatal clinic provides an opportunity for practitioners to reduce expos ure to harmful stressors more likely to affect birth outcomes compared with an intervention based solely on biologic measurements.

PAGE 297

261 A major goal of public health practice is prevention. Ideally, practitioners aim prevention efforts to be effective prio r to the onset of disease. While the former screening discussion is a type of se condary prevention, identifying women among this population who receive little to no social support, or who are being verbally or physically abused is a form of primary prevention. This identification is primary for low income, White wo men; however, increased social support protects against urine sugar spill for both groups in the study Identifying women who lack support or who are abused for intervention is a way to utilize both physiologic and socio-cultural methodologies in primary preventi on of both urine sugar spill and resulting high birth weight in fants. Practitioners in public health need to modify procedures in patient intake to evaluate both biologic and nonbiologic factors, and utilize low-risk scr eening tools to identify and implement policy directed at women who are at risk for these problematic birth outcomes. 6.7 Further Research Among the many different paths futu re research may take, of primary interest is exploring the association between low-risk, non-invasive screening tools and their predictability of birth co mplications. Secondary research should focus on other psychosocial and physical fa ctors in predicting birth complications and other biologic measures that may predict such outcomes. Another prospective cohort shoul d be implemented with a focus on collecting all screening test data at each prenatal visit, all laboratory results including glucose-tolerance test resu lts, diagnosis notes, and birth outcome

PAGE 298

262 measures. Other screening test data such as ketone levels, blood pressure, and weight at each visit would be helpful in supplementing the current results and also in identifying other non-invasive te sts that may predict the outcome of high birth weight. Laboratory data including ur inalysis and blood test results provide continuous measures with which to compar e in-clinic screening tests. Also, specific diagnostic tests like the glucosetolerance test provides a continuous measure of glucose levels to supplement the categorical urine sugar levels. Finally, diagnosis of pregnancy-related cond itions such as gestational diabetes or hypertension could be examined in the causal pathway between urine sugar spill and high birth weight infants to strengthen the causal association between urine sugar and birth complications. In addition, other stress indicators such as hormone levels should be measured throughout the pr egnancy. A modified design taking into account women who delay prenatal care would capture a larger group of women, and measuring the stressors more than t wice during pregnancy would provide data for trend analysis. The mechanisms leading to birth complications that were found significant in this study should be expanded in further research. For instance, social support should be assesse d in multiple ways, not just using a single scale. Also, abuse should be more clearly defined and specified as to whether it is perceived or observed within the clinic. Such modifications in the study design would yield more detaile d and possibly causal results than found currently. To strengthen cu rrent findings, biologic plausibility, consistency with

PAGE 299

263 current literature in terms of diseas e diagnosis, elaboration on a dose-response relationship, and replication of these resu lts would all provide further evidence for the association between stressors and ur ine sugar levels and resulting highweight infants.

PAGE 300

264 CHAPTER 7 CONCLUSION 7.0 Introduction In the literature review of this di ssertation, being born high birth weight was suggested as having long term effects on health status such as development of childhood and adult onset obesity and other morbid conditions later in life. Additional support was provided by the Centers for Disease Control and Prevention which highlighted t he mortality or ‘actual causes of death’ associated with poor diet and poor physical activity as second in the nation to smoking in the year 2000 [186]. To address this iss ue, the CDC has raised obesity to epidemic status, and implemented a 2015 national health objecti ve of lowering obesity among healthy Americans to 15% [187]. Identifying early markers for obesity such as high birth weight and creati ng preventative programs centered on reducing the prevalence of these births may aid in achieving this goal. The maternal precursors to high-weight bi rth such as socio-cultural factors and biologic measures such as urine s ugar levels are key to understanding the pathway leading to high birth weight and of fer additional options for intervention at the most early stages of life. This dissertation explained such fact ors associated with high birth weight controlling for biologic factors such as body mass index, history of high birth

PAGE 301

265 weight babies, previous C-section, age, and alcohol and drug use during pregnancy. The major findings indicate that urine sugar screening is a tool that may be used to identify a group of high ri sk women independent of ethnicity or age. In fact, if a high sugar spill is re corded at any point in the pregnancy (1+ or higher) it is predictive of a high weight bi rth. Such a screening tool is invaluable to practitioners who may use it to closely monitor a woman throughout her pregnancy and intervene by diet modifi cation, weight monitoring, or recommendations for stress reduction. Secondary conclusions indicate that ethnicity affects how women in this study coped with identified psychosocial and physical factors. White women in the study were more likely to have high ur ine sugar spill, and to have high-weight babies compared with Black women in the study. Further, among this ethnic group, a subset of White women w ho are abused during pregnancy, and who seek social support from another primary family mem ber, specifically their mothers, are highly likely to have a hi gh birth weight baby compared with both other non-abused White women and Bl ack women in the study. Such a finding is important in the c ontext of public health as practitioners attempt to prevent adverse outcomes. Therefore, identifyi ng these women at an early point during pregnancy may not only impact the weight of the infant, but may impact the health and safety of the woman if abuse is identified and intervention attempted in conjunction with the mother of the pregnant woman.

PAGE 302

266 Future research may further define the psychosocial and physical mechanisms that interact to cause glucose intole rance and weight gain of the infant. Tertiary findings include the associat ion of specific psycho-social or protective factors that impact both urine sugar spill and high birth weight. Overall, the social support of t he pregnant woman’s mother is needed in protecting against adverse outcomes. Al so, physical and verbal abuse during pregnancy, depression during pregnancy, and ph ysical work strain all negatively impact the pregnancy process, and contribu te to high urine sugar levels and birthing high weight babies. Again, understanding these factor s and identifying them early during pregnan cy is a preventive strategy that public health practitioners can utilize to facilitate change among women in this population. The results of this dissertation cont ribute to the epidemiologic literature through the use of newly developed st atistical modeling techniques, and a prospective study design leading to the outcome of high birth weight. While inferential statistics enable a researcher to infer causality through the strength of the association, causal modeling comp liments those findings through the creation and testing of a plausible causal pathw ay connecting exposures to disease. Structural equation modeling c annot replace the use of in ferential statistics, but adds to interpretation of those findings. Previous research of high birth weight has been primarily retrospective in nature through the use of historical cohorts and case-controls study designs. Direct observation of both psycho-social and physical factors prior to high weight births adds to the understanding of

PAGE 303

267 exposures in predicting the magnitude of birth weight and their impact on change in birth weight. Studying both the biologic and societ al stressors, and the response to those stressors will not only add to the sci entific understanding of the interaction of these mechanisms, but will enable m edical and public health practitioners to affect change in the health of women and their children, and possibly the health of those children during adol escence and later adulthood.

PAGE 304

268 REFERENCES 1. Vital Statistics of the U.S., 1999, Natality. 1999 National Center for Health Statistics. 1, http://www.cdc.gov/nchs/ data/statab/t991x26.pdf 2. NHS Maternity Statistics, England: 2001-02 ed. S.D. Department of Health. 2003, London: Department of Health. 3. Low Birth Weight (Less than 2,500 gr ams), by sex, Canada, provinces and territories, annual. 2002 CANSIM II. May 27, 2003, http://cansim2.statcan.ca 4. Cesarean Fact Sheet. 1995-1998 childbirth.org. May 05, 2003, http://www.childbirth.org/section 5. Menacker, F. and S. Curtin, Trends in Cesarean Birth and Vaginal Birth after Previous Cesarean, 1991-99. National Vital Statistics Reports, 2001. 49 (13): p. 1-16. 6. Health System Performance. 2002 Canadian Inst itute for Health Information. May 25, 2003, http://www.statcan.ca /english/freepub/82-221XIE/01002/tables.htm#health 7. Nortzon, F., et al., International collaborative e ffort (ICE) on birth weight, plurality, perinatal, and in fant mortality. Differenc es in obstetrical delivery practice: Norway, Scotl and, and the United States. Acta Obstet Gynecol Scand, 1991. 70 (6): p. 451-60. 8. Alabama County Selection Map 2002 U.S. Census Bureau. http://quickfacts.census.gov/qfd/maps/alabama_map.html 9. State and County Quick Facts: Alabama. 2003 U.S. Census Bureau. September 3, 2003, http://quickfacts.census. gov/qfd/states/01000.html

PAGE 305

269 10. State and County Quick Facts: Tuscaloosa County. 2003 U.S. Census Bureau. Septem ber 3, 2003, http://quickfacts.census.gov/qfd/states/01/01125.html 11. State and County Quick Facts: Mobile County. 2003 U.S. Census Bureau. September 3, 2003, http://quickfacts.census. gov/qfd/states/01/01097.html 12. Colhoun, H.M. and N. Chaturvedi, A Life Course Approach to Diabetes in A Life Course Approach to Women's Health D. Kuh and R. Hardy, Editors. 2002, Oxford Press: London. p. 121-40. 13. Egeland, G., R. Skjaerven, and L. Irgens, Birth characteristics of women who develop gestational diab etes: population based study. Br Med J, 2000. 321 : p. 546-7. 14. McCance, D., et al., Birth weight and non-insulin dependent diabetes: thrifty genotype, thrifty phenotype, or surviving small baby genotype? Br Med J, 1994. 308 : p. 1652-8. 15. Dabelea, D., et al., Birth weight, type 2 diabetes, and insulin resistance in Pima Indian children and young adults. Diabetes Care, 1999. 22 (6): p. 944-50. 16. Hahn, R.A., The Role of Society and Culture in Sickness and Healing in Sickness and Healing: An Anthropological Perspective 1995, Yale University Press: Ne w Haven. p. 76-98. 17. Okafor, C.B., Maternal and child health project in Nigeria. J Natl Black Nurses Assoc, 2003. 14 (2): p. 51-8. 18. Ujah, I.A., et al., How safe is motherhood in Nigeria?: the trend of maternal morbidity in a te rtiary health institution. East Afr Med, 1999. 76 (8): p. 436-9. 19. Okolocha, C., et al., Socio-cultural factors in maternal morbidity and mortality: a study of a semi-ur ban community in southern Nigeria. 52, 1998. 5 : p. 293-7.

PAGE 306

270 20. Essen, B., et al., Are some perinatal deaths in immigrant groups linked to suboptimal perinatal care services? BJOG, 2002. 109 (6): p. 677-82. 21. Diaz, S., et al., Acceptability of emergency cont raception in Brazil, Chile, and Mexico. 2 Facilitating factors versus obstacles. Cad Saude Publica, 2003. 19 (6): p. 1729-37. 22. Lei, H., S. Wen, and M. Walker, Determinants of caesarean delivery among women hospitalized for childbirth in a remote population in China. Journal of Obstetric and Gynaecology of Canada, 2003. 25 (11): p. 937-43. 23. Shrestha, S., Socio-cultural factors infl uencing adolescent pregnancy in rural Nepal. Int J Adolesc Med Health, 2002. 14 (2): p. 101-9. 24. Staugard, F., Traditional and modern maternity care in Botswana cooperation or confrontation. Botswa Natl Health Bull, 1985. 1 (1): p. 105-13. 25. Ghubash, R. and M. Abou-Saleh, Postpartum psychiatric illness in Arab culture: prevalence and psychosocial correlates. BR J Psychiatry, 1997. 171 : p. 65-8. 26. Rankin-Esquer, L., et al., Autonomy and relatedness in marital functioning. J Marital Fam Ther, 1997. 23 (2): p. 175-90. 27. Waring, E. and J. Reddon, The measurement of intimacy in marriage: the Waring Intimacy Questionnaire. J Clin Psychol, 1983. 39 (1): p. 53-7. 28. Helsing, K.J., M. Szklo, and G. Comstock, Mortality after bereavement. American Journal of Epidemiology, 1981. 114 : p. 41-52. 29. Helsing, K.J., M. Szklo, and G. Comstock, Factors associated with mortality after widowhood. American Journal of Public Health, 1981. 71 : p. 802-9. 30. Stack, C.B., All our kin 1974, New York: Harper and Row.

PAGE 307

271 31. Boone, M., Capital Crime: Black Infant Mortality in America 1989, Beverly Hills: Sage. 32. Berhane, Y., et al., Women's health in a rural se tting in societal transition in Ethiopia. Social Science and Medicine, 2001. 53 (11): p. 1525-39. 33. Luo, Z., R. Wilkin s, and M. Kramer, Disparities in pregnancy outcomes according to marital and cohabitation status. Obstet Gynecol, 2004. 103 (6): p. 1300-7. 34. Kabir, A.A., et al., Unnecessary cesarean delivery in Louisiana: An analysis of birth certificate data. Am J Obstet Gynecol, 2004. 190 (1): p. 10-9. 35. Miller, W., The intendedness and wantedn ess of the first child in The First Child and Family Formation W. Miller and L. Ne wman, Editors. 1978, Carolina Population Center : Chapel Hill. p. 209-43. 36. Mhyrman, A., P. Rant akallio, and M. Isohanni, Unwantedness of a pregnancy and schizophrenia in the child. BR J Psychiatry, 1996. 169 : p. 637-40. 37. Myhrman, A., P. Ols en, and P. Rantakallio, Does the wantedness of a pregnancy predict a child's educational attainment? Fam Plan Perspect, 1995. 27 (3): p. 116-9. 38. Bustan, M. and A. Coker, Maternal attitude towa rd pregnancy and the risk of neonatal death. Am J Public Health, 1994. 84 (3): p. 411-4. 39. Oropesa, R., et al., Prenatal care among Puerto Ricans on the United States mainland. Social Science and Medicine, 2000. 51 (12): p. 1723-39. 40. Pagnini, D. and N. Reichman, Psychosocial factors and the timing of prenatal care among women in Ne w Jersey's HealthStart program. Fam Plan Perspect, 2000. 32 (2): p. 56-64.

PAGE 308

272 41. Hickey, C., et al., Low prenatal weight gain am ong low-income women: what the risk factors? Birth, 1997. 2 : p. 102-8. 42. Morris, N., J. Ud ry, and C. Chase, Reduction of low birth weight rates by the prevention of unwanted pregnancies. Fam Plan Resume, 1977. 1 (1): p. 252-3. 43. Sable, M., et al., Pregnancy wantedness and adverse pregnancy outcomes: Differences by ra ce and Medicaid status. Fam Plan Perspect, 1997. 29 (2): p. 76-81. 44. Monk, C., et al., Fetal heart rate reactivity di ffers by women's psychiatric status: an early marker for developmental risk? J Am Acad Child Adolesc Psychiatry, 2004. 43 (3): p. 283-90. 45. Anoop, S., et al., Maternal depression and low maternal intelligence as risk factors for malnutrition in chil dren: A community based case-control study from South India. Arch Dis Child, 2004. 89 (4): p. 325-9. 46. Rahman, A., et al., Mothers' mental health and infant growth: a casecontrol study from Rawalpindi, Pakistan. Child Care Health Dev, 2004. 30 (1): p. 21-7. 47. Jacobson, S., et al., Maternal recall of alcohol, cocaine, and marijuana use during pregnancy. Neurotoxicol Teratol, 1991. 13 (5): p. 535-40. 48. Zuckerman, B., et al., Maternal depressive symptoms during pregnancy, and newborn irritability. J Dev Behav Pediatr, 1990. 11 (4): p. 190-4. 49. Paarlberg, K., et al., Psychosocial predictors of low birthweight: a prospective study. Br J Obstet Gynaecol, 1999. 106 (8): p. 834-41. 50. Copper, R., et al., The preterm prediction stud y: Maternal stress is associated with spontaneous preterm birth at less than thirty-five weeks' gestation. National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. Am J Obstet Gynecol, 1996. 175 (5): p. 1286-92.

PAGE 309

273 51. Hedegaard, M., et al., The relationship between psychological distress during pregnancy and birth weight for gestational age. Acta Obstet Gynecol Scand, 1996. 75 (1): p. 32-9. 52. Brooke, O., et al., Effects on birth weight of smoking, alcohol, caffeine, socioeconomic factors, and psychosocial stress. BMJ, 1989. 298 : p. 795801. 53. Wu, J., et al., Mood disturbance in pregnancy and the mode of delivery. Am J Obstet Gynecol, 2002. 187 (4): p. 864-7. 54. Peorksen, A. and D. Petitti, Employment and low bi rth weight in black women. Social Science and Medicine, 1991. 33 (11): p. 1281-6. 55. Nurminen, T., et al., Physical work load, fetal development and course of pregnancy. Scand J Work Environ Health, 1989. 15 (6): p. 404-14. 56. Hansteen, I., H. Kjuus, and S. Fandrem, Spontaneous abortions of known karyotype related to occupational and environmental factors: A casereferent study. Int J Occup Environ Health, 1996. 2 (3): p. 195-203. 57. Klonoff-Cohen, H., J. Cross, and C. Pieper, Job stress and preeclampsia. Epidemiology, 1996. 7 (3): p. 245-9. 58. Henriksen, T., M. Hedegaard, and N. Secher, The relation between psychosocial job strain, and preter m delivery and low birthweight for gestational age. Int J Epidemiol, 1994. 23 (4): p. 764-74. 59. Bohn, D., Lifetime and current abus e, pregnancy risks, and outcomes among Native American women. J Health Care Poor Underserved, 2002. 13 (2): p. 184-98. 60. Renker, P., Physical abuse, social support, self-care, and pregnancy outcomes of older adolescents. J Obstet Gynecol Neonatal Nurs, 1999. 28 (4): p. 377-88.

PAGE 310

274 61. Curry, M., B. Doyle, and J. Gilhooley, Abuse among pregnant adolescents: Differences by developmental age. Am J Matern Child Nurs, 1998. 23 (3): p. 144-50. 62. Huth-Bocks, A., A. Levendosky, and G. Bogat, The effects of domestic violence during pregnancy on maternal and infant health. Violence Vict, 2002. 17 (2): p. 169-85. 63. Neggers, Y., et al., Effects of domestic viol ence on preterm birth and low birth weight. Acta Obstet Gynecol Scand, 2004. 83 (5): p. 455-60. 64. Valladares, E., et al., Physical partner abuse during pregnancy: A risk factor for low birt h weight in Nicaragua. Obstet Gynecol, 2002. 100 (4): p. 700-5. 65. Murphy, C., et al., Abuse: A risk factor for low birth weight? A systematic review and meta-analysis. CMAJ, 2001. 164 (11): p. 1567-72. 66. Kearney, M., et al., Health behaviors as mediators for the effect of partner abuse on infant birth weight. Nurs Res, 2004. 53 (1): p. 36-45. 67. Altarac, M. and D. Strobino, Abuse during pregnancy and stress because of abuse during pregnan cy and birth weight. J Am Med Womens Assoc, 2002. 57 (4): p. 208-14. 68. Rachana, C., et al., Prevalence and complications of physical violence during pregnancy. Eur J Obstet Gynecol Reprod Biol, 2002. 103 (1): p. 269. 69. Berenson, A., et al., Perinatal morbidity associated with violence experienced by pregnant women. Am J Obstet Gynecol, 1994. 170 (6): p. 1766-9. 70. Norbeck, J. and N. Anderson, Psychosocial predictors of pregnancy outcomes in low-income bla ck, Hispanic, and white women. Nurs Res, 1989. 38 (4): p. 204-9.

PAGE 311

275 71. Lespinasse, A.A., et al., Maternal support in the delivery room and birthweight among African-American women. Journal of National Medical Association, 2004. 96 (2): p. 187-95. 72. Norbeck, J., J. De Joseph, and R. Smith, A randomized trial of an empirically-derived social support inte rvention to prevent low birthweight among African American women. Social Science and Medicine, 1997. 45 (5): p. 797-9. 73. Da Costa, D., et al., Psychosocial predictors of labor/delivery complications and infant birth weigh t: A prospective multivariate study. J Psychosom Obstet Gynaecol, 2000. 21 (3): p. 137-48. 74. Feldman, P., et al., Maternal social support predi cts birth weight and fetal growth in human pregnancy. Psychosom Med, 2000. 62 (5): p. 715-25. 75. Jesse, D., D. Wallace, and W. Seaver, A holistic approach to riskscreening in pregnancy with Appalachian women. J Holist Nurs, 2002. 20 (2): p. 133-51. 76. Dejin-Karlsson, E., et al., Association of a lack of psychosocial resources and the risk of giving birth to small for gestational age infants: A stress hypothesis. BJOG, 2000. 107 (1): p. 89-100. 77. Pryor, J., et al., Stress and lack of social support as risk factors for smallfor-gestational-age birth. Acta Paediatr, 2003. 92 (1): p. 62-4. 78. Sheehan, T., Stress and low birth weight: A structural mo deling approach using real life stressors. Social Science and Medicine, 1998. 47 (10): p. 1503-12. 79. Hahn, R.A., Divisions of Labor: Obstetrici an, Woman, and Society in Williams Obstetrics, 1903-1989 in Sickness and Healing: An Anthropological Perspective 1995, Yale University Press: New Haven. p. 209-33.

PAGE 312

276 80. Pavittranon, S., et al., Laboratory case identification of arsenic in Ronpibul village, Thailand (2000-2002). J Environ Sci Health Part A Toxic Hazard Subst Environ Eng, 2003. 38 (1): p. 213-21. 81. Orem, J., et al., Renal glycosuria treated as di abetes mellitus: case report. East Afr Med J, 1997. 74 (11): p. 740-2. 82. Scholl, T.O., et al., Maternal glucose concentrati on influences fetal growth, gestation, and pregnancy complications. Am J Epidemiol, 2001. 154 (6): p. 514-20. 83. Saravanan, G., L. Pari, and S. Venkateswaran, Effect of cogent db, a herbal drug, on plasma insulin and hepatic enzymes of glucose metabolism in experimental diabetes. Diabetes Obes Metab, 2002. 4 (6): p. 394-8. 84. Kumar, C., et al., Modulatory effect of a butyric acid-a product of dietary fiber fermentation in experimen tally induced diabetic rats. J Nutr Biochem, 2002. 13 (9): p. 522. 85. Than, S., et al., Bone marrow transplantation as a strategy for treatment of non-insulin-dependent diabetes me llitus in KK-Ay mice. J Exp Med, 1992. 176 (4): p. 1233-8. 86. Koh, K., et al., Knowledge and concerns of newly diagnosed NIDDM patients in Singapore. Diabetes Res Clin Pract, 1991. 12 (1): p. 11-7. 87. Metz, L., et al., Gastric tolerance of high-dose pulse oral prednisone in multiple sclerosis. Neurology, 1999. 53 (9): p. 2093-6. 88. Kohzuki, M., et al., Antihypertensive and renal-protective effects of losartan in streptozotocin diabetic rats. J Hypertens, 1995. 13 (1): p. 97103. 89. Gestational Diabetes Mellitus. Diabetes Care, 2004. 27 (Suppl 1): p. S88S90.

PAGE 313

277 90. Medical tests and exams: Glucose urine. 2004 American Accreditation HealthCare Commission. June 30, 2004, 2004, http://health.allrefer.com/heal th/glucose-urine-info.html 91. Abrams, R. and D. Coustan, Gestational diabetes update. Clinical Diabetes, 1990. 8 (4): p. 19-25; 30. 92. Coustan, D., Diabetes in America 2nd ed. Vol. 95. 1995, Bethesda: National Institutes of Health, Nationa l Institute of Diabetes and Digestive Kidney Disease. 93. Reinauer, H., Philip D. Home, Ariyur S. Kanagasabapathy, and Claus-Chr. Heuck, World Health Organization: Laboratory Diagnosis and Monitoring of Diabetes Mellitus 2002 2003, Geneva: World Health Organization Documents Productions Services. 94. Combs, C., E. Gunder son, and J. Kitzmiller, Relationship of fetal macrosomia to maternal postprandial glucose control during pregnancy. Diabetes Care, 1992. 15 : p. 1251-7. 95. Jovanovic-Peterson, L., C. Peterson, and G. Reed, Maternal postprandial glucose levels and infant birth wei ght: the Diabetes in Early Pregnancy Study. Am J Obstet Gynecol, 1991. 164 : p. 103-11. 96. Tamas, G. and Z. Kerenyi, Gestational diabetes: current aspects on pathogenesis and treatment. Exp Endocrinol Diabetes, 2001. 109 (Suppl 2): p. S400-11. 97. Langer, O., L. Brustm an, and A. Anyaegbunam, The significance of one abnormal glucose tolerance test value on adverse outcomes in pregnancy. Am J Obstet Gynecol, 1987. 157 : p. 758-63. 98. Spellacy, W.N., Diabetes Mellitus and Pregnancy in Danforth's Obstetrics and Gynecology J.R. Scott, et al., Editor s. 1999, Lippincott Williams and Wilkins: Philadelphia.

PAGE 314

278 99. Siccardi, D.C., Obstetrics and Gynecology: Gestational Diabetes. 2004 MedStudents. June 30, 2004, 2004, http://www.medstudents.c om.br/ginob/ginob4.htm 100. Star, W.L., Gestational Diabetes Mellitus in Ambulatory Obstetrics: Protocols for Nurse Practitioners/Nurse-Midwives W.L. Star, et al., Editors. 1990, School of Nursing, Univ ersity of California, San Francisco: San Francisco. 101. Tharpe, N., Clinical Practice Guidelines for Midwives 2001-2002 Edition 2001, East Boothbay: Midwife Publications. 102. O'Sullivan, J., et al., Screening criteria for high-risk gestational diabetes patients. Am J Obstet Gynecol, 1973. 116 : p. 895. 103. Silverman, B., T. Rizzo, and N. Cho, Long-term prospective evaluation of offspring of diabetic mothers. Diabetes, 1991. 40 : p. 121-5. 104. Xiong, X., et al., Gestational diabetes mellitus : prevalence, risk factors, maternal and infant outcomes. International J Gynecology Obstet, 2001. 75 (3): p. 221-8. 105. Merchant, K., J. Villar, and E. Kestler, Maternal height and newborn size relative to risk of intrapartum c aesarean delivery and perinatal distress. Br J Obstet Gynec, 2001. 108 : p. 689-96. 106. McMahon, M.J., C.V. Ananth, and R.M. Liston, Gestational diabetes mellitus. Risk factors, obstetric complications and infant outcomes. J Reprod Med, 1998. 43 (4): p. 372-8. 107. Taber's Cyclopedic Medical Dictionary 18th ed, ed. E.J. Egan. 1997, Philadelphia: F.A. Davis Company. 108. The Merck Manual of Diagnosis and Therapy 16th ed, ed. R. Berkow. 1992, Rathway: Merck Re search Laboratories.

PAGE 315

279 109. He, Q., et al., Risk factors of obesity in pre school children in China: a population-based case--control study. Int J Obes Relat Metab Disord, 2000. 24 (11): p. 1528-36. 110. Gallaher, M.M., et al., Obesity among Mescalero preschool children. Association with maternal obesity and birth weight. Am J Dis Child, 1991. 145 (11): p. 1262-5. 111. Ekbom, A., Growing evidence that several human cancers may originate in utero. Semin Cancer Biol, 1998. 8 (4): p. 237-44. 112. Robison, L.L., et al., Birth weight as a risk factor for childhood acute lymphoblastic leukemia. Pediatr Hematol Oncol, 1987. 4 (1): p. 63-72. 113. Kaatsch, P., et al., German case control study on childhood leukaemia-basic considerations, methodolo gy and summary of the results. Klin Padiatr, 1998. 210 (4): p. 185-91. 114. Gillman, M.W., et al., Maternal gestational diabetes, birth weight, and adolescent obesity. Pediatrics, 2003. 111 (3): p. e221-6. 115. Mikulandra, F., et al., The effect of high birth weig ht (400 g or more) on the weight and height of adult men and women. Coll Antropol, 2000. 24 (1): p. 133-6. 116. Curhan, G.C., et al., Birth weight and adult hypertension, diabetes mellitus, and obesity in US men. Circulation, 1996. 94 (12): p. 3246-50. 117. Curhan, G.C., et al., Birth weight and adult hypertension and obesity in women. Circulation, 1996. 94 (6): p. 1310-5. 118. Okcu, M.F., et al., Birth weight, ethnicity, and occurrence of cancer in children: a population-based, incident ca se-control study in the State of Texas, USA. Cancer Causes Control, 2002. 13 (7): p. 595-602.

PAGE 316

280 119. Ou, S.X., et al., Birth characteristics, mate rnal reproductive history, hormone use during pregnancy, and risk of childhood acute lymphocytic leukemia by immunophenotype (United States). Cancer Causes Control, 2002. 13 (1): p. 15-25. 120. Murray, L., et al., Association of early life factors and acute lymphoblastic leukaemia in childhood: historical cohort study. Br J Cancer, 2002. 86 (3): p. 356-61. 121. Yeazel, M.W., et al., High birth weight and risk of specific childhood cancers: a report from t he Children's Cancer Group. J Pediatr, 1997. 131 (5): p. 671-7. 122. Cnattingius, S., et al., Prenatal and neonatal risk factors for childhood lymphatic leukemia. J Natl Cancer Inst, 1995. 87 (12): p. 908-14. 123. Schuz, J., et al., High-birth weight and other ri sk factors for Wilms tumour: results of a population-based case-control study. Eur J Pediatr, 2001. 160 (6): p. 333-8. 124. Innes, K., T. Byers, and M. Schymura, Birth characteristics and subsequent risk for breast canc er in very young women. Am J Epidemiol, 2000. 152 (12): p. 1121-8. 125. Sandhu, M.S., et al., Self-reported birth weight and subsequent risk of colorectal cancer. Cancer Epidemiol Biomarkers Prev, 2002. 11 (9): p. 935-8. 126. Wei, J.N., et al., Low birth weight and high birth weight infants are both at an increased risk to have type 2 diabetes among schoolchildren in Taiwan. Diabetes Care, 2003. 26 (2): p. 343-8. 127. Savona-Ventura, C. and M. Chircop, Birth weight influence on the subsequent development of ges tational diabetes mellitus. Acta Diabetol, 2003. 40 (2): p. 101-4.

PAGE 317

281 128. Webster, L.A., et al., Prevalence and determinants of caesarean section in Jamaica. J Biosoc Sci, 1992. 24 (4): p. 515-25. 129. Wikstrom, I., et al., Traumatic injury in large-for-date infants. Acta Obstet Gynecol Scand, 1988. 67 (3): p. 259-64. 130. Maayan-Metzger, A., R. Mazkereth, and J. Kuint, Fever in healthy asymptomatic newborns during the first days of life. Arch Dis Child Fetal Neonatal Ed, 2003. 88 (4): p. F312-4. 131. Sutton, L., et al., Do very sick neonates born at term have antenatal risks? Infants ventilated primarily for lung disease. Acta Obstet Gynecol Scand, 2001. 80 (10): p. 917-25. 132. Mogren, I., et al., Reproductive factors have low impact on the risk of different primary brain tumours in offspring. Neuroepidemiology, 2003. 22 (4): p. 249-54. 133. Hestbaek, L., et al., Is low back pain in youth associated with weight at birth? A cohort study of 8000 Danish adolescents. Dan Med Bull, 2003. 50 (2): p. 181-5. 134. Gunnarsdottir, I., et al., Size at birth and coronary artery disease in a population with high birth weight. Am J Clin Nutr, 2002. 76 (6): p. 1290-4. 135. Chali, D., F. Enquselassie, and M. Gesese, A case-control study on determinants of rickets. Ethiop Med J, 1998. 36 (4): p. 227-34. 136. McFarland, L.V., et al., Erb/Duchenne's palsy: a consequence of fetal macrosomia and method of delivery. Obstet Gynecol, 1986. 68 (6): p. 7848. 137. Gunnell, D., et al., Patterns of fetal and childhood growth and the development of psychosis in young males: a cohort study. Am J Epidemiol, 2003. 158 (4): p. 291-300.

PAGE 318

282 138. Bergmann, R.L., et al., Secular trends in neonatal macrosomia in Berlin: influences of potential determinants. Paediatric and Perinatal Epidemiology, 2003. 17 : p. 244-9. 139. Naeye, R., Maternal body weight and pregnancy outcome. Am J Clin Nutr, 1990. 52 : p. 273-9. 140. Johnson, J., J. Longm ate, and B. Frentzen, Excessive maternal weight and pregnancy outcome. Am J Obstet Gynecol, 1992. 167 : p. 352-72. 141. Dorner, G. and A. Plagemann, Perinatal hyperinsulinism as possible predisposing factor for diabet es mellitus, obesity and enhanced cardiovascular risk later in life. Hormone and Metabolism Research, 1993. 26 : p. 213-21. 142. Bo, S., et al., Obesity or diabetes: what is wo rse for the mother and for the baby? Diabetes Metab, 2003. 29 (2): p. 175-8. 143. Lu, G.C., et al., The effect of the increasing prevalence of maternal obesity on perinatal morbidity. Obstet Gynecol, 2001. 185 (4): p. 845-9. 144. Keily, J.L., S. Yu, and D.L. Rowley, Low birth weight and intrauterine growth retardation: From data to ac tion CDC's public health surveillance for women, infants, and children 1994, Centers for Disease Control and Prevention: Atl anta. p. 185-202. 145. Dole, N., et al., Maternal stress and preterm birth. Am J Epidemiol, 2003. 157 (1): p. 14-24. 146. Haram, K., J. Pi rhonen, and P. Bergsjo, Suspected big baby: a difficult clinical problem in obstetrics. Acta Obstet Gynecol Scand, 2002. 81 : p. 185-94. 147. Persson, B. and U. Hanson, Neonatal morbidities in gestational diabetes mellitus. Diabetes Care, 1998. 21 (Suppl 2): p. B79-84.

PAGE 319

283 148. Maymon, E., et al., Peripartum complications in grand multiparous women: para 6-9 versus para > = 10. Eur J Obstet Gynecol Reprod Biol, 1998. 81 (1): p. 21-5. 149. Jain, L., C. Ferre, and D. Vidyasagar, Cesarean delivery of the breech very-low-birth-weight infan t: does it make a difference? J Matern Fetal Med, 1998. 7 (1): p. 28-31. 150. Ostlund, I., et al., Maternal and fetal outcome s if gestational impaired glucose tolerance is not treated. Diabetes Care, 2003. 26 (7): p. 2107-11. 151. Jardim, O., et al., Delivery in diabetic pregnancy. Ann 1st Super Sanita, 1997. 33 (3): p. 329-32. 152. Aucott, S., et al., Rigorous management of in sulin-dependent diabetes mellitus during pregnancy. Acta Diabetol, 1994. 31 (3): p. 126-9. 153. Lawoyin, T., A prospective study on some factors which influence the delivery of large babies. J Trop Med Hyg, 1993. 96 (6): p. 352-6. 154. Teberg, A., et al., Nursery survival for inf ants of birth weight 500-1500 grams during 1982-1995. J Perinatol, 2001. 21 (2): p. 97-106. 155. Kjos, S., K. Berkowitz, and B. Kung, Prospective delivery of reliably dates term infants of diabetic mothers without determination of fetal lung mortality: comparison to historical cohort. J Matern Fetal Neonatal Med, 2002. 12 (6): p. 433-7. 156. Albrechtsen, S., et al., Factors influencing deliv ery method in breech presentation. Acta Obstet Gynecol Scand, 1998. 77 (4): p. 416-21. 157. Tong, L., et al., Physician cesarean delivery rates and risk-adjusted perinatal outcomes. Obstet Gynecol, 2003. 101 (6): p. 1204-12.

PAGE 320

284 158. Poma, P., Effects of obstetrician characteristics on cesarean delivery rates: A community hospital experience. Am J Obstet Gynecol, 1999. 180 : p. 1364-72. 159. Guillemette, J. and W. Fraser, Differences between obstetricians in cesarean section rates and the management of labor. Br J Obstet Gynec, 1992. 99 : p. 105-8. 160. Lagrew, D. and J. Adashek, Lowering cesarean section rate in a private hospital: Comparing of individual physicians' rates, risk factors, and outcomes. Am J Obstet Gynecol, 1998. 178 : p. 1207-14. 161. Krieger, N., Theories for social epidemiolo gy in the 21st century: an ecosocial perspective. International Journal of Epidemiology, 2001. 30 : p. 668-77. 162. Berkman, L.F. and T. Glass, Social integration, so cial networks, social support, and health in Social Epidemiology L.F. Berkman and I. Kawachi, Editors. 2000, Oxford University Press: New York. p. 137-73. 163. Berkman, L.F., et al., From Social Integration to Health: Durkheim in the New Millennium. Social Science and Medicine, 2000. 51 : p. 843-57. 164. Selected Maternal and Child H ealth Statistics: Alabama 1999 2001, Montgomery: Alabama Department of P ublic Health, Center for Health Statistics. 165. Dressler, W.W., Camila D. Borges Mauro C. Balieiro, and Jose Ernesto Dos Santos, Measuring Cultural Consonanc e: Examples with Special Reference to Measurement Theory in Anthropology in Field Methodology in Anthropology 2004. 166. Oliver, M.R., Kaaron Yeghiazarian, Min Zheng, Mori Jamshidian, Igor Koyfman, nQuery Advisor 1995. 167. Hennekens, C.H. and J.E. Buring, Epidemiology in Medicine ed. S.L. Mayrent. 1987, Boston: Little, Brown, and Company. 383.

PAGE 321

285 168. Tjarden, P. and N. Thoennes, Prevalence, Incidence, and Consequences of Violence Against Women: Findings from the National Violence Against Women Survey 1998, National Institute of Justice, Centers for Disease Control and Prev ention. p. 16. 169. Surveillance and Research: Examples of Translations of Pregnancy Risk Assessment Monitoring System data. 2002 National Center for Chronic Disease and Health Promotion. May 8, 2003, http://www.cdc.gov/nccdphp/drh/surveil.htm 170. Cohabitation, Marriage, Divorce, and Re-marriage in the United States. 2002 Department of Health and Human Services. April 11, 2003, http://www.cdc.gov/nchs/dat a/series/sr_23/sr23_022.pdf 171. Depression Research at the National Institute of Mental Health. 1999 National Institute of Ment al Health. June 3, 2003, http://www.nimh.nih.gov 172. Gordis, L., Epidemiology 2nd ed. 2000, Philadelphia: Sanders Company. 173. Hatcher, L., A Step-by-Step Approach to Usi ng the SAS System for Factor Analysis and Structur al Equation Modeling 3rd ed. 1994, Cary: SAS Institute, Inc. 174. Rothman, K.J. and S. Greenland, Modern Epidemiology 2nd ed. 1998, Philadelphia: Lippincott, Williams, and Wilkins. 175. Neter, J., et al., Applied Linear Statistical Models ed. 4th. 1996, Boston: WCB McGraw-Hill. 176. Stokes, M.E., C.S. Davis, and G.G. Koch, Categorical Data Analysis Using the SAS System 5th ed. 1999, Cary: SAS Institute, Inc. 177. Muthen, L.K. and B.O. Muthen, Mplus Statistical Analysis with Latent Variables 1998, National Institute on Al cohol Abuse and Alcoholism, National Institutes of Health: Los Angeles.

PAGE 322

286 178. Lincoln, K.D., L.M. Chatters, and R. Taylor, Psychological distress among black and white Americans: differential effects of social support, negative interaction and personal control. Journal of Health and Social Behavior, 2003. 44 (3): p. 390-407. 179. Mullings, L., Households headed by women: The politics of race, class, and gender in Conceiving the new world order: The global politics of reproduction F.D. Ginsberg and R. Rapp, Ed itors. 1995, University of California Press: Berkeley. p. 122-39. 180. Krishnakumar, A. and M.M. Black, Family processes within threegeneration households and adole scent mothers' satisfaction with father involvement. Journal of Family Psychology, 2003. 17 (4): p. 488-98. 181. Aronowitz, T. and D. Morrison-Beedy, Resilience to Risk-Taking Behaviors in Impoverished African American Girl s: The Role of Mother-Daughter Connectedness. Research in Nursing and Health, 2004. 27 (1): p. 29-39. 182. Stuber, J., et al., The Association between multiple domains of discrimination and self-assessed health: a multilevel analysis of Latinos and blacks in four low-income New York City neighborhoods. Health Service Research, 2003. 38 (6): p. 1735-59. 183. Association, A.D., Gestational Diabetes Mellitus. Diabetes Care, 2004. 27 : p. S88-S90. 184. Li, T., et al., Physician Cesarean section deliv ery rates and risk-adjusted perinatal outcomes. Obstet Gynecol, 2003. 101 (6): p. 1204-12. 185. Peterson, A. and D. Lupton, The New Public Health: Health and self in the age of risk 1996, London: Sage Publications. 186. CDC Fact Sheet: Actual Causes of Death in the United States, 2000. 2004 Centers for Disease Control and Pr evention. June 17, 2004, 2004, http://www.cdc.gov/nccdphp/fa ctsheets/death_causes2000.htm

PAGE 323

287 187. Overweight and Obesity. 2004 Centers for Disease Control and Prevention. June 17, 2004, 2004, http://www.cdc.gov/nccdphp/dnpa/obesity/

PAGE 324

288 APPENDICES

PAGE 325

289 APPENDIX A: DESCRIPTION OF PREDICTOR VARIABLES Psychosocial Predictors – Instruments from the Low Birth Weight Study (Tuscaloosa sample) are presented first fo llowed by instruments from the Healthy Start Evaluation Grant (Mobile sample). Social support (partner and mother) – Tuscaloosa sample, initial interview I’m going to ask you some questions about the people you know, and how much they can help you when you need it. Fo r each question asked, please name all the people you can think of who you know for certain you can count on. Emotional 1. Please name all of the people you are close to in your life who make you feel liked or loved. 2. Please name all of the people you are close to who make you feel important. 3. Please name all of the people you k now for certain you could go to for comfort if you were upset about something. Instrumental/Material 4. Who would help you if you needed a ride to the doctor or to work? 5. Who would help you if you needed to borrow some money? 6. Who would help you if you were sick for a long time and couldn’t get out of bed? Name Age Sex Relat.How long known? 1. Love 2. Impt 3. Upst 4. Ride 5. $ 6. Sick 1. 2. etc. Social support (partner and mother) – Tuscaloosa sample, final interview I’m going to ask you some questions about the people you know, and how much they can help you when you need it. Last time, when I asked you to “name all of the people you are close to in your life who you know for certain you can count on”, you named [ REPEAT ALL NAMES] A. …is there anyone you would like to add to this list? B. …is there anyone you would like to take off this list?

PAGE 326

290 APPENDIX A (CONTINUED) (IF NAMES ADDED, ask each question for each new name:) a. Does ________ (new name[s]) make you feel liked or loved? b. Does ________ make you feel important? c. Could you go to _________ for comfort if you were upset about something? d. Would you ask _________ for a ri de to the doctor or to work? e. Would you go _________ if you needed to borrow some money? f. Would ________ help you if you were sick for a long time and couldn’t get out of bed? ADD: Name Age Sex Relat.How long known? 1. Love 2. Impt 3. Upst 4. Ride 5. $ 6. Sick 1. 2. etc. DELETE: Name Age Sex Relat.How long known? 1. Love 2. Impt 3. Upst 4. Ride 5. $ 6. Sick 1. 2. etc. Social support (partner and mother) – Mobile sample, initial interview Now, I’d like to ask you some questions about people you care about and who care about you. Ask questions 1-6 completely, in cluding questions for sex, age, relation, and years known. 1. Please name all the people you think would listen to you talk about your feelings. (fill in name, age, relationship to informant, amount of time known, and check box labeled hear) 2. Please name all the people who you feel very close to. (check box labeled love) 3. Please name all the people who think that you are okay just the way you are. (check box labeled okay) 4. Please name all the people who woul d help you out if you were sick and could not get out of bed. (check box labeled sick) 5. Please name all the people who y ou think would give you a ride somewhere if you needed it. (check box labeled ride)

PAGE 327

291 APPENDIX A (CONTINUED) 6. Please name all the people who y ou think would loan you something you need including money. (check box labeled give) Name Sex Age RelationKnown Hear Love Okay Sick Ride Give 1. 2. etc. For analysis purposes, a 6-pont scale was constructed for each the mother and partner. If the mother or partner were not listed as s upporters, they were given a score of zero. The scale was treated as a continuous variable in analysis. Depression scale (CESD) – Tuscaloo sa sample, init ial interview I am going to read a list of ways you may have felt or behaved. Please tell me how often you have felt this way during the past week a. less than 1 day (never or almost never) b. 1 to 2 days c. 3-4 days d. 5 to 7 days (most of the time) 1. ____ You were bothered by things that usually don’t bother you. 2. ____ You had trouble keeping your mind on what you were doing. 3. ____ You felt depressed. 4. ____ You felt hopeful about the future. 5. ____ You thought your life was a failure. 6. ____ You felt lonely. 7. ____ You enjoyed life. 8. ____ You had crying spells. 9. ____ You felt sad. 10. ____ You felt that people do not like you. Depression scale (CESD) – Tuscaloosa sample, final interview I am going to read a list of ways you may have felt or behaved. Please tell me how often you have felt this way during the past week a. less than 1 day (never or almost never) b. 1 to 2 days c. 3-4 days d. 5 to 7 days (most of the time) 1. ____ You were bothered by things that usually don’t bother you. 2. ____ You had trouble keeping your mind on what you were doing.

PAGE 328

292 APPENDIX A (CONTINUED) 3. ____ You felt depressed. 4. ____ You felt hopeful about the future. 5. ____ You thought your life was a failure. 6. ____ You felt lonely. 7. ____ You enjoyed life. 8. ____ You had crying spells. 9. ____ You felt sad. 10.____ You felt that people do not like you. Depression scale (CESD) – Mobile sample, initial interview I’m going to read you a list of ways you may have felt or behaved. Please circle how often you have felt this way during the last 7 days (past week) a. Less than 1 day (never or almost never) b. 1 to 2 days c. 3 to 4 days d. 5 to 7 days (most of the time) 1. ____ You were bothered by things that usually don’t bother you. 2. ____ You had trouble keeping your mind on what you were doing. 3. ____ You felt depressed. 4. ____ You felt hope when you thought about the future. 5. ____ You thought you life was a failure. 6. ____ You felt lonely. 7. ____ You enjoyed life. 8. ____ You had crying spells. 9. ____ You felt sad. 10. ____ You felt that people do not like you. Depression scale (CESD) – Mobile sample, fina l interview I’m going to read you a list of ways you may have felt or behaved. Please circle how often you have felt this way during the last 7 days (past week) a. Less than 1 day (never or almost never) b. 1 to 2 days c. 3 to 4 days d. 5 to 7 days (most of the time) 1. ____ You were bothered by things that usually don’t bother you. 2. ____ You had trouble keeping your mind on what you were doing. 3. ____ You felt depressed. 4. ____ You felt hope when y ou thought about the future.

PAGE 329

293 APPENDIX A (CONTINUED) 5. ____ You thought you life was a failure. 6. ____ You felt lonely. 7. ____ You enjoyed life. 8. ____ You had crying spells. 9. ____ You felt sad. 10.____ You felt that people do not like you. For analysis purposes, the scales were treated as continuous variables. Marital status – Tuscaloosa sample, initia l interview Now I’d like to know a little about your personal life. Which word best describes your marital status? (READ OPTIONS TO PATIENT) ____ single ____ unmarried but living (s taying) with partner ____ common law ____ married ____ separated ____ divorced ____ widowed Marital status – Tuscaloosa sample, final interview Now I’d like to know a little bit about your personal life. The fi rst time we talked you told me… [REVIEW WHAT SHE SAID LAST TIME eg. “you were married”, “you had cut off with your boyfriend”, “you were not seeing anyone”] Has there been any change in your personal life since then?........YES NO (IF YES) How has it changed? _______ __________________ ____________ When did this happen? _______ __________________ ___________ Marital status – Mobile sample, initial interview Please tell me if you are: Dating Engaged Married Some other kind of relationship _____________ (Widowed, Divorced, etc.) Marital status – Mobile sample, final interview Now I’m going to ask you a few questions about a relationship you might have right now. The first ti me we talked you told me you were ‘____________’ [refer to x1 s.s. for status]

PAGE 330

294 APPENDIX A (CONTINUED) Have things changed since then? YES [If Yes, continue] NO [If No, go to section XX] Are you seeing anyone right now? YES NO Please tell me if you are: Dating Engaged Married Some other kind of relationship ________________ (Widowed, Divorce, etc.) For analysis purposes, the variable was co ded as single, single ever-married, and married/living with partner. Autonomy – Tuscaloosa sample, final interview PERSONAL BELIEFS A. GENDER ROLE ORIENTATION TOWARD WORK This next section has to do with how you feel things should be between men and women. What you think does not have to match how things really are for you. I will read each sentence. Tell me if you agree or do not agree with each. 1. Men should spend the same amount of time as women in caring for children and the home AGREE DISAGREE 2. Men and women should be equal, but the husband should have the final say on all the big decisions AGREE DISAGREE 3. To be good at either one, a wo man must choose either marriage or a career, but not both AGREE DISAGREE 4. A woman should work outside the home only if her income is needed by the family AGREE DISAGREE 5. These days, men and women are tr eated the same at work when it comes to their pay and moving up AGREE DISAGREE 6. Women should be paid the same as men for doing the same jobs AGREE DISAGREE 7. It is the natural duty of the woman to provide the love and caring for the family AGREE DISAGREE

PAGE 331

295 APPENDIX A (CONTINUED) B. AUTHORITARIANISM These next few questions are like the last ones. They are about how you think people should act. After I read each sent ence, tell me if you agree or do not agree with each. 1. A trusted person in aut hority tells you to do someth ing. You should do it even if you do not see the r eason for it AGREE DISAGREE 2. You should treat experts with resp ect even if you do not think much of them AGREE DISAGREE 3. A person should get a “second opinion” when not sure about a doctor’s advice AGREE DISAGREE (IF HAS WORKED AT ALL DURING PREGNANCY:) 4. A person should speak up against t he boss when the boss acts unfairly AGREE DISAGREE For analysis purposes, the scale was treated as a continuous variable. Pregnancy Wantedness – Tuscal oosa sample, final interview These last questions are about birth control and you getting pregnant. 1. When you found out you were pregnant di d you really feel like you wanted to have a baby? YES NO 2. Do you feel like you w ant the baby now? YES NO Pregnancy Wantedness – Mobile sample, initial interview Now I’d like you to circle or fill-in your answers to the next few questions. 1. When you first found out you were pregnant, did you really feel like you wanted to have a baby? YES NO 2. Do you feel like you want this baby now? YES NO Pregnancy Wantedness – Mobile sample, final interview Now I’d like you to circle or fill-in your answers to the next few questions. 1. When you first found out you were pregnant, did you really feel like you wanted to have a baby? YES NO

PAGE 332

296 APPENDIX A (CONTINUED) 2. Do you feel like you want this baby now? YES NO Verbal abuse – Tuscaloosa sample, initial interview These last few questions are about how you are treated by other people. Is there anyone who often says things to you that hurt you? YES NO (IF YES) Please tell me who. __________________________ _____________ How often do they say hurtfu l things? _________________________ Verbal abuse – Tuscaloosa sample, final interview These next few questions are about how you are treated by other people. Is there anyone who often says things to you that hurt you? YES NO Please tell me who. ________________________ ________________ How often do they say hurtful things? DAILY WEELLY MONTHLY Verbal abuse – Mobile sample, initial interview Now, I’d like you to circle whether you Agree or Disagree with the next few questions I’m going to ask you. We’ll st art on page XX with section XX, question number X. 1. Is there anyone who often sa ys things to you that hurt you? YES NO A. Please circle who. If there’s more than one person, please circle all that apply. a. No one says hurtful things to me. b. Ex-Boyfriend c. Boyfriend d. Mother e. Father f. Sister g. Brother h. Other Relative i. Friend j. Enemy k. Stranger

PAGE 333

297 APPENDIX A (CONTINUED) B. How often do they say hurtful things? Please circle and connect your answer to the person who says hurtful things. Let’s look at the box below this question for an example of what to do when there is more than one person in your answer. a. No one says hurtful things to me. b. Daily c. Weekly d. Monthly e. Not often Verbal abuse – Mobile sample, final interview Now I’d like to ask you some personal questions. I’d like for you to keep answering the booklet in front of you. 1. Since your first prenatal visit, is t here anyone who often says things to you that hurt you? YES NO A. Please circle who. If there’s more than one person, please circle all that apply. a. No one says hurtful things to me. b. Ex-Boyfriend c. Boyfriend d. Mother e. Father f. Sister g. Brother h. Other Relative i. Friend j. Enemy k. Stranger B. How often do they say hurtful things? Please circle and connect your answer to the person who says hurtful things. Let’s look at the box below this question for an example of what to do when there is more than one person in your answer. a. No one says hurtful things to me. b. Daily c. Weekly d. Monthly e. No often

PAGE 334

298 APPENDIX A (CONTINUED) For the purposes of analysis, verbal and physical abuse were combined to form one variable for each of the initial and fi nal interviews. Whether or not abuse occurred was measured, intensity of abus e was not analyzed. The variable was dichotomous, and was characterized as eith er having a history of abuse (initial interview did not distinguish bet ween pre-pregnancy a nd pregnancy abuse), and abuse during the second and thir d trimesters of pregnancy. Physical Predictors are presented below in the same order as the psychosocial predictors. Physical abuse – Tuscaloosa sample, initia l interview 1. Within the past year, have you been hit, slapped, kicked, or hurt by someone? YES NO 2. Within the past year, have you been hit, slapped, kicked, or hurt by someone? YES NO 3. Since you’ve been pregnant, have you been hit, slapped, kicked, or hurt by someone? YES NO (IF YES to either 2 or 3) 4. Could you please tell me who hurt you? __ ________________________ 5. Where on your body did they hurt you? (use body map) ______________ Physical abuse – Tuscaloosa sample, final interview 1. Since our first interview, have you been hit, slapped, kicked, or hurt by someone? YES NO (IF YES) 2. Please tell me who hur t you. ____________ _______________________ 3. Where on your body did they hurt you? (use body map) ______________ Physical abuse – Mobile sample, initial interview 1. Have you ever been hit, slapped, kicked, or hurt by someone? YES NO

PAGE 335

299 APPENDIX A (CONTINUED) A. Please circle who. If there’s more than one person, please circle all that apply. a. I’ve never been hit, slapped, kicked, or hurt by someone. b. Ex-Boyfriend c. Boyfriend d. Mother e. Father f. Sister g. Brother h. Other Relative i. Friend j. Enemy k. Stranger B. Please circle the last time they hurt you. If th ere is more than one person, please connect who it is to each answer you circle. a. I’ve never been hit, slapped, kicked, or hurt by someone. b. Today c. In the past 7 days (week) d. In the past month (30 days) e. In the past year f. More than one year ago 3. How many times have you been hit, slapped, kicked, or hurt by someone? a. Never b. Only 1 time c. More than once 4. If you have been hurt, please circle w here they hurt you on the body map below. If there is more than one per son please write down their name next to the circled body part. [Body map is located on informant’s interview guide] 5. Please write down how old you were t he first time you we re hurt. If you’ve never been hurt, leave this question blank.

PAGE 336

300 APPENDIX A (CONTINUED) Physical abuse – Mobile sample, final interview 1. Since your first prenatal visit, hav e you been hit, slapped, kicked, or hurt by someone? YES NO A. Please circle who. If there’s more than one person, please circle all that apply. a. I’ve never been hit, slapped, kicked, or hurt by someone. b. Ex-Boyfriend c. Boyfriend d. Mother e. Father f. Sister g. Brother h. Other Relative i. Friend j. Enemy k. Stranger B. Please circle the last time they hurt you. If there is more than one person, please connect who it is to the last time they hurt you. a. I’ve never been hit, slapped, kicked, or hurt by someone. b. Today c. In the past 7 days (week) d. In the past month (30 days) 2. How many times have you been hit, slapped, kicked, or hurt since the last time we talked? a. Never b. Only 1 time c. More than once 3. If you have been hit or hurt, please ci rcle where on the body map below. If there is more than one person, pleas e write down their name next to the circled body part. [Body map is located on informant’s interview guide] For the purposes of analysis, verbal and physical abuse were combined to form one variable for each of the initial and fi nal interviews. Whether or not abuse occurred was measured, intensity of abuse was not analyzed. The variable was

PAGE 337

301 APPENDIX A (CONTINUED) dichotomous, and was characterized as eith er having a history of abuse (initial interview did not distinguish bet ween pre-pregnancy a nd pregnancy abuse), and abuse during the second and thir d trimesters of pregnancy. Physical work strain scale – Tu scaloosa sample, in itial interview The following questions are about work you do that you get paid for. 1. Is your work physically difficult? YES NO 2. At your job, are you always on the move? YES NO 3. Does the work you do on the job cause you to worry a lot? YES NO 4. Do you get enough breaks during work hours? YES NO 5. Can you take a break whenever you need one? YES NO 6. At work, can you make a 10 minute personal phone call whenever you wish? YES NO 7. Can you receive a personal visitor for 10 minutes? YES NO Physical work strain scale – Tu scaloosa sample, final interview The following questions are still about work you do that you get paid for. Please remember to answer each question by thinki ng about your job since the last time I talked to you. 1. Is your work physically difficult? YES NO 2. At your job, are you always on the move? YES NO 3. Does the work you do on the job cause you to worry a lot? YES NO

PAGE 338

302 APPENDIX A (CONTINUED) 4. Do you get enough breaks during work hours? YES NO 5. Can you take a break whenever you need one? YES NO 6. At work, can you make a 10 minute personal phone call whenever you wish? YES NO 7. Can you receive a personal visitor for 10 minutes? YES NO Physical work strain scale – M obile sample, fi nal interview Now, I’m going to ask you some questions about jobs you might have had. Let me know if you need me to repeat a question. 1. (Is/was) your work hard physically? YES NO 2. At your job, (are/wer e) you always on the move? YES NO 3. (Does/did) the work you do on the job cause you to worry a lot? YES NO 4. (Do/did) you get enough breaks during work hours? YES NO 5. (Can/would) you take a br eak whenever you want/ed to? YES NO 6. At work, (can/could) you make a 10 minute phone call whenever you wish? YES NO 7. (Can/could) you receive a personal visitor for 10 minutes? YES NO For the purposes of analysis, the scale was treated as a continuous variable.

PAGE 339

303 APPENDIX B: DESCRIPTION OF PO TENTIALLY CONFOUNDING FACTORS Potentially confounding factors were assessed during the interview by questioning the participant, collected thr ough review of medical charts, or calculated during the analysis phase of the dissertation. The assessment of each factor is described prior to its definition. Age – calculated during analysis Age = exact age (month/day/year) at the time of the initial interv iew – date of birth Body Mass Index – calculated during analysis BMI = (pre-pregnant weight/height in inches*height in inches)*703 Education level attained – calculated during analysis 1 = Less than a high school education, but at least a middle school education (<= 8th grade) 2 = Less than a high school education, but more than a middle school education (9th – 12th grade without graduation or GED) 3 = High school education (12th grade graduation or GED) 4 = Post-high school education (some associ ates, certificate, or college courses) Pre-pregnant weight – medical chart review Recorded from the first prenatal visit; either verbally reported by participant if the gestational age of the infant was greater than eight weeks, or if less than eight weeks, the weight of the parti cipant at that first visit Interview site – calculated during analysis 1 = Tuscaloosa county 2 = Mobile county Total number of pregnancies – inte rview and medical chart review How many times have you been pregnant in all? This includes this pregnancy and any miscarriages and abortions you may have had. _________________ Taken from medical chart and treated as a continuous number. Total number of live bi rths – interview and medical chart review How many children have you ( had/given birth to)? _______________ What is the date of birth of each child you had? _________________ Taken from medical chart and treated as a continuous number.

PAGE 340

304 APPENDIX B (CONTINUED) Previous Caesarean section – interv iew and medical chart review Of the children you have had, how many were born by C-section? __________ Taken from medical chart and treated as a continuous number. Total number of abortions or miscarri ages – interview and medical chart review How many pregnancies have you had which ended in miscarriage? _________ (IF YES) When did (this/these) occur? _____________________ Ho w far along were y ou? ________________________ How many pregnancies have you had whic h ended in (induced) abortion? ____ (IF YES) When did (this/these) occur? _____________________ Taken from medical chart and treated as a continuous number. Total number of premature births – interview and medical chart review Of the children you have had, how many were born prematur ely? ___________ Taken from medical chart and treated as a continuous number. Gestational age – medical chart review Gestational age of the infant at birth as recorded on the delivery form in each participant’s chart Weight gain during pregnancy – calculated during analysis Weight gain = pre-pregnant weight – weight at the final prenatal visit Total number of prenatal visi ts – medical chart review All attended visits were added together to create a continuous measure Alcohol and drug abuse during pregnancy – interview and calculated during analysis Tuscaloosa sample, initial interview How often did you drink […] during the past year? How much do you drink […] at a time? Alcohol Most Days 3-4/ week 1-2/ week 1-2/ month 1-2/ 6 mo. 1-2/ year None at all Amount Beer or malt liquor Wine or wine coolers Hard liq./mixed drink

PAGE 341

305 APPENDIX B (CONTINUED) Since you’ve been pregnant, how often did you drink […]? Since you’ve been pregnant, how much do you drink […] at a time? Alcohol Most Days 3-4/ week 1-2/ week 1-2/ month 1-2/ 6 mo. 1-2/ year None at all Amount Beer or malt liquor Wine or wine coolers Hard liq./mixed drink When was the last time you used any str eet drugs (such as marijuana, cocaine, etc.) or any drug which requires the use of a needl e? _____________________ [If used, for each type drug ask the following questions] How often did you use […] during the past year? Did you use any other sort of drugs (speed, PCP, heroin)? Drugs Most Days 3-4/ week 1-2/ week 1-2/ month 1-2/ 6 mo. 1-2/ year None at all Amount Marijuana Cocaine or Crack Other _________ Since you’ve been pregnant, how often do you use […]? Since you’ve been pregnant, do you use any other sort of drugs (speed, PCP, heroin)? Drugs Most Days 3-4/ week 1-2/ week 1-2/ month 1-2/ 6 mo. 1-2/ year None at all Amount Marijuana Cocaine or Crack Other _________ Tuscaloosa sample, final interview In the past few months, how often do you drink […]? How much […] do you drink at a time? Alcohol Most Days 3-4/ week 1-2/ week 1-2/ month 1-2/ 6 mo. None at all Amount Beer or malt liquor Wine or wine coolers Hard liq./mixed drink

PAGE 342

306 APPENDIX B (CONTINUED) In the past few months, how often do you use […]? In the past few months, have you used any other sort of drugs (speed, PCP, heroin)? Drugs Most Days 3-4/ week 1-2/ week 1-2/ month 1-2/ 6 mo. None at all Amount Marijuana Cocaine or Crack Other _________ Mobile sample, initial interview Have you ever drank alcohol? YES NO In the past year? YES NO In the past 3 months? YES NO In the past 7 days (week)? YES NO Please circle every type of alcohol that you drink: a. Hard liquor, like rum, tequila, or vodka b. Beer or malt liquor c. Wine or wine coolers d. I don’t drink alcohol When you drink you: a. Get drunk b. Feel tipsy c. Don’t feel any different d. I don’t drink alcohol Have you ever used marijuana? YES NO In the past year? YES NO In the past 3 months? YES NO In the past 7 days (week)? YES NO

PAGE 343

307 APPENDIX B (CONTINUED) When you smoked last time, did you: a. Take one puff b. Take 2 or more puffs c. I’ve never smoked marijuana Have you ever sniffed or inhaled glue, gasoline, or paint to get high? YES NO Could you circle or write down what you did? a. Glue b. Gasoline c. Paint d. Other __________________ e. I’ve never inhaled glue, gas oline, or paint to get high Could you circle how many times you did last time? a. Inhaled 1 time b. Inhaled more than 1 time c. I’ve never inhaled glue, gas oline, or paint to get high Have you ever used cocaine or crack? YES NO In the past year? YES NO In the past 3 months? YES NO Mobile sample, final interview Have you drank alcohol since your first prenatal visit? YES NO In the past month (30 days)? a. No b. 1 time c. More than 1 time Please circle every type of alcohol that you drink: a. Hard liquor, like rum, tequila, or vodka b. Beer or malt liquor c. Wine or wine coolers d. I haven’t drank alcohol

PAGE 344

308 APPENDIX B (CONTINUED) When you drink you: a. Get drunk b. Feel tipsy c. Don’t feel any different d. I haven’t drank alcohol Have you used marijuana since your first prenatal visit? YES NO In the past month (30 days)? a. No b. 1 time c. More than 1 time When you smoked last time, did you: a. Take 1 puff b. Take 2 or more puffs c. I haven’t smoked marijuana Have you sniffed or inhaled glue, gasoline, or paint to get high since your first prenatal visit? YES NO Could you circle or write down what you did? a. Glue b. Gasoline c. Paint d. Other ________________________ e. I haven’t inhaled glue, gasoline, or paint to get high Could you circle how many ti mes you did it last time? a. Inhaled 1 time b. Inhaled more than 1 time c. I haven’t inhaled any glue, gasoline, or paint Have you used cocaine or crack si nce your first prenatal visit? YES NO

PAGE 345

309 APPENDIX B (CONTINUED) Calculations for analysis From the initial interview, the week of gestation was taken into account, and if the participant stated they did not drink or do drugs during their pregnancy, but they were 12 weeks pregnant and had consumed al cohol or marijuana within the past 3 months, their answers were changed to ‘yes.’ Responses from the final and initial in terviews were combined to create an alcohol and drug use variable that accounted for use throughout pregnancy.

PAGE 346

310 APPENDIX C: DESCRIPTION OF OUTCOME MEASURES Outcome measures were either assessed by reviewing the medical chart of the participant or the delivery sheet at the hospital of delivery. Urine sugar levels Ordinal measure 0 = No detectable level of sugar in the urine 1 = Low levels of sugar in the urine (trace) 2 = High levels of sugar in the urine (1+ or higher) Binary measure 0 = No/Low levels of sugar in the urine 1 = High levels of sugar in the urine (1+ or higher) or 0 = No detectable levels of sugar in the urine 1 = Any level of sugar in the urine (trace or higher) Birth weight Continuous measure Reported in grams on the c hart, but converted to kil ograms for ease of analysis after transformation. Binary measure 0 = All other birth weights (< 4000 grams/4 kilograms) 1 = High birth weight ( >= 4000 grams/4 kilograms) or in supplemental analyses 0 = Normal birth weights (2500 – 3999 grams/2.5 – 3.999 kilograms) 1 = High birth weight ( >= 4000 grams/4 kilograms) Caesarean section 0 = Vaginal birth 1 = Caesarean section

PAGE 347

311 APPENDIX D: MODEL FIT RESULTS FOR ANALYSES Table 141 Model fit statistics for each logistic regression model table from Chapter 4, the Results Chapter including the Chi-square goodness of fit statistic and statistical significance Table Number Chi-square Goodness of Fit Statistic Degrees of Freedom pvalue Table 60 15.012 8 0.059 Table 61 43.242 29 0.043 Table 62 38.075 13 0.000 Table 63 32.956 14 0.003 Table 64 32.205 15 0.006 Table 65 47.223 17 0.000 Table 66 20.491 18 0.306 Table 67 26.920 12 0.008 Table 68 27.511 12 0.007 Table 69 27.135 12 0.007 Table 70 27.793 12 0.006 Table 71 44.339 26 0.014 Table 72 46.123 12 0.000 Table 73 67.147 16 0.000 Table 74 35.108 16 0.004 Table 75 53.068 16 0.000 Table 76 39.095 18 0.003 Table 85 61.637 26 0.000 Table 86 66.184 13 0.000 Table 87 79.772 15 0.000 Table 88 44.371 16 0.000 Table 89 76.949 17 0.000 Table 90 25.785 18 0.105 Table 91 71.023 15 0.000 Table 92 20.986 7 0.004 Table 93 32.062 12 0.001 Table 94 28.163 8 0.000 Table 95 13.974 4 0.007 Table 96 35.529 9 0.000 Table 97 38.734 11 0.000 Table 98 49.889 8 0.000 Table 102 75.637 14 0.000 Table 103 51.080 8 0.000

PAGE 348

312 APPENDIX E: RESIDUAL PLOTS OF MULTIPLE REGRESSION ANALYSES The following residual plot represents the distribution of the data for the analysis on page 132 of the equation: birth weight ( ) = -4.478 + 0.956(ethnici ty) + 0.493(education) + 0.855(gestation) + 0.944(prenatal) The first plot is a scatterplot followed by another scatterplot with low birth weight infants removed from analysis. In the firs t scatterplot, there are clear outliers that appear to affect the distribution of the re siduals (Figure 37). The reason for this effect is known as a truncation problem wit h the data. That is stillbirths were not included in the analysis, ther efore, the lowest birth we ights all clustered at the lower end of the distribution. In Figure 38, it is clear that once removed, the distribution of the residual s appears more random. It is noted, however, that among the higher birth weights, the variabili ty is greater. The high birth weight infants only comprise 7% of the sample, as a result, this variability does not greatly affect the regre ssion line. The equation for Figure 38 is listed below: birth weight ( ) = -20.526 + -0.975(ethni c) + 0.277(education) + 0.792(gestation) + 0.009(prenatal)

PAGE 349

313 APPENDIX E (CONTINUED) Figure 37 The residual scatterplot of birth weight and ethnicity controlling for confounding factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dependent Variable: Birth Weightresidual plotsModel: Intercept + Ethnic + Education + Gestation + Prenatal Predicted 20 10 0 -10Std. Residual3 2 1 0 -1 -2 -3 Figure 38 The residual scatterplot of birth weight and ethnicity excluding low birth weight infants and controlling for confounding factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dependent Variable: Birth Weightwithout low birth weight infantsModel: Intercept + Ethnic + Education + Gestation + Prenatal Predicted 16 14 12 10 8 6Std. Residual3 2 1 0 -1 -2 -3

PAGE 350

314 APPENDIX E (CONTINUED) The next plot represents the equat ion on page 137, for the formula: birth weight ( ) = -4.450 + 0.785(sugar sp ill) + 0.415(education) + 0.855(gestation) + 0.979(prenatal) Again, the first plot is a scatterplot of residuals using the entire data set and urine sugar as the predictor vari able (Figure 39). The same problem exists as with the first set of scatterplots. When the low bi rth weight infants are removed, the trend is resolved (Figure 40). The second pl ot is a represents that analysis and the equation for the plot is listed: birth weight ( ) = -20.746 + 0.689(sugar) + 0.133(education) + 0.764(gestation) + 0.009(prenat) Figure 39 The residual scatterplot of birth weight and urine sugar levels controlling for confounding factors of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dependent Variable: Birth Weightresidual plotsModel: Intercept + Sugar + Education + Gestation + Prenatal Predicted 20 10 0 -10Std. Residual4 3 2 1 0 -1 -2 -3

PAGE 351

315 APPENDIX E (CONTINUED) Figure 40 The residual scatterplot of birth weight and urine sugar levels excluding low birth weight infants and controlling for confounding factors of pregnant women attending the County Health Department Prenatal Clin ic in Tuscaloosa and Mobile Counties, AL 1990-2001 Dependent Variable: Birth Weightwithout low birth weight infantsModel: Intercept + Sugar + Education + Gestation + Prenatal Predicted 16 14 12 10 8 6 4Std. Residual4 3 2 1 0 -1 -2 -3

PAGE 352

316 APPENDIX F: RE-ANALYSIS EXCLUDING LOW BIRTH WEIGHT INFANTS As highlighted in the Discussion C hapter, inclusion of another adverse outcome (e.g., low birth weight infants) as part of the reference group in the analysis of this dissertation may have im pacted results. Therefore, all major findings were re-analyzed excluding low birt h weight infants. Approximately 10% of women in the sample gave birth to a low-weight infan t. When those women are removed from analysis, t he sample size is reduce to 457, and the percentage of high-weight births changes fr om 6% to 7% (N = 31). Removing low birth weight infants fr om the sample has minimal effect on the odds ratios in each model. Initial anal ysis of presence or absence of urine sugar levels indicates the same lack of a ssociation (OR including LBW infants = 0.49; OR excluding LBW infants = 0. 49). When examini ng the association between ordinal levels of urine sugar and high birth weight, the associations remain the same with a change of 0.03 in the odds ratios (Low sugar levels OR including LBW infants = 0.89; Low sugar levels OR excluding LBW infants = 0.88; High sugar levels OR including LBW infants = 3.25; High sugar levels OR excluding LBW infants = 3.28). When the high spill group is analyzed alone with high birth weight infants, the odds ratios are different by 0.04 (OR including LBW infants = 3.30; OR excluding LBW infant s = 3.34). The lack of association between Caesarean section and high birth weight remains ( 2 including LBW infants = 4.298; 2 excluding LBW infants = 2.271; p > 0.05 for both).

PAGE 353

317 APPENDIX F (CONTINUED) In assessment of Hypothesis 3, the tw o significant predictors of high birth weight infants are a mot her’s social support and ethnicity. Table 142 presents the changes in the odds ratios by variable between the original models and the re-analyzed models excluding lo w birth weight infants ( 10%) from the sample. As shown, the change in the odds ratios is less than 0.03, and both associations remained significant. Table 142 Hypothesis 3 comparison of original odds ratios and re-analyzed odds ratios excluding low-weight births of pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Odds Ratio Comparisons (N = 456) Predictor Variable Odds Ratio Including LBW Infants Odds Ratio Excluding LBW Infants Mother’s Social Support Scale 1.561 1.558 Ethnicity Black1.000 White2.809 2.780 Mother’s Emotional Social Support 2.027 2.015 Ethnicity Black1.000 White2.743 2.724 Table 143 presents similar comparat ive findings of results within the dissertation and those same models excludi ng the 10% of low-we ight births. The table number refers to the table in the te xt of the Results and Structural Equation Modeling Chapters. Hypothesis 5, the analysis of ethnicity as an interaction term with each predictor on high birth weight infants, is addressed under the column heading ‘Table Number.’ Again, the change in the interaction terms is minimal at less than 0.02. For the supplemental analysis in conjunction with the path

PAGE 354

318 APPENDIX F (CONTINUED) models, the change in the odds ratio intera ctions is also less than 2%. Excluding low-weight births has minimal to no effect on these dissertation findings.

PAGE 355

319 APPENDIX F (CONTINUED) Table 143 Hypothesis 5 and supplemental analyses of the original interaction of ethnicity on predictors and high-weight births and analyses excluding low-weight births among pregnant women attending the County Health Department Prenatal Clinic in Tuscaloosa and Mobile Counties, AL 1990-2001 Odds Ratio Comparisons (New Sample Sizes Included with Each Table) Table Number Predictor Variable Odds Ratio Including LBW Infants Odds Ratio Excluding LBW Infants Table 96 (N = 329) Physical or Verbal Abuse No Abuse1.000 History of Abuse1.740 1.687 Ethnicity Black1.000 White7.096 6.859 Ethnicity*History of Abuse0.226 0.234 Table 97 (N = 329) Physical or Verbal Abuse No Abuse1.000 Abuse t20.404 0.407 Ethnicity Black1.000 White1.582 1.569 Ethnicity*Abuse t2 8.298 8.274 Table 98 (N = 364) Mother’s Social Support Scale 1.167 1.168 Ethnicity Black1.000 White0.215 0.216 Ethnicity*Mother’s Social Support Scale 1.662 1.657 Table 133 (N = 401) Mother’s Social Support Scale 1.069 1.071 Physical or Verbal Abuse No Abuse1.000 Abuse t20.031 0.032 Mother’s Social Support Scale*Abuse t2 2.471 2.452 Table 134 (N = 329) Mother’s Social Support Scale 1.304 1.302 Physical or Verbal Abuse No Abuse1.000 Abuse t20.332 0.333 Ethnicity Black1.000 White1.622 1.603 Mother’s Social Support Scale*Abuse t2*Ethnicity 1.799 1.799

PAGE 356

ABOUT THE AUTHOR Charlan Day Kroelinger received her Ba chelor’s Degree in Anthropology from Auburn University in 1995. She then pursued a Master’s Degree at The University of Alabama in Anthropology. Wh ile at The University of Alabama, she gained experience in research by directi ng a National Institutes of Health Grant on psychosocial and physical stressors and low birth weight infants. She successfully received her degree in 1997. She then worked as a Project Director for a Healthy Start Evaluation Grant from 1997 to 1999. While in the Ph.D. program at the University of South Florida, Ms. Kroelinger worked on research grants with faculty and taught graduate level courses in her field of study, Epidemiol ogy and Biostatistics. Her research focuses on women’s reproductive health. She has been offered a Post-Doctoral Fellow position in the Department of Ep idemiology and Biostatistics at the University of South Flori da that will enable her to c ontinue pursuing has research interests.


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 2200433Ka 4500
controlfield tag 006 m d s
007 cr mn
008 040811s2004 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000413
035
(OCoLC)56546874
9
AJS2461
b SE
040
FHM
c FHM
090
RA425 (ONLINE)
1 100
Kroelinger, Charlan Day.
0 245
Biomedical and psychosocial determinants of problematic birth outcomes
h [electronic resource] /
by Charlan Day Kroelinger.
260
[Tampa, Fla.] :
University of South Florida,
2004.
502
Thesis (Ph.D.)--University of South Florida, 2004.
504
Includes bibliographical references.
500
Includes vita.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 356 pages.
520
ABSTRACT: The primary objective of this study was to evaluate the associations between psychosocial stressors, urine sugar levels, and subsequent birth outcomes, specifically high birth weight babies and Caesarean section births. In a prospective cohort study, 506 Black and White women of childbearing age were followed for the duration of one pregnancy in Tuscaloosa and Mobile counties in Alabama from 1990 to 2001. Participants were interviewed twice throughout pregnancy, during the first and third trimesters, respectively, and birth outcome data were collected via medical chart reviews. Six percent (6.1%) of the women in the sample had a high birth weight baby, and 18.4% received a C-section during childbirth. Adjusted logistic regression results indicate that urine sugar levels are predictive of high-weight births, with women who have higher urine sugar levels were more than three times likely to birth a high weight baby compared with women who have no detectable urine sugar spill (OR 3.25; 95% CI 1.30, 8.10). In addition, the interaction of familial social support throughout pregnancy, physical or verbal abuse during the second and third trimesters, and ethnicity is significantly associated with increased risk of having a high birth weight baby. For C-section, single participants are over two times less likely to receive a C-section during childbirth compared with currently married participants (OR 0.46; 95% CI 0.21-1.00). Examining structural equation modeling results; pathways leading from urine sugar levels, physical or verbal abuse during the latter half of the pregnancy, and a mother's social support among White participants are indicative of high weight births (R = 0.65). White abused women who receive their mother's social support are more likely to have a high birth weight baby compared with both White and Black women who are not abused and receive the same amount of social support. Recommendations to public health practitioners include primary prevention through promotion of familial support during pregnancy, secondary prevention through urine sugar screening at every prenatal visit, and direct intervention by identifying and inquiring about instances of suspected abuse during pregnancy.
590
Adviser: Mason, Thomas J.
653
high birth weight.
psychosocial stressors.
pregnancy screening tests.
pregnancy stressors.
reproductive epidemiology.
690
Dissertations, Academic
z USF
x Public Health
Doctoral.
773
t USF Electronic Theses and Dissertations.
949
FTS
SFERS
ETD
RA425 (ONLINE)
nkt 8/27/04
4 856
u http://digital.lib.usf.edu/?e14.413