Citation
Neighborhood Environment and Hypertensive Disorders of Pregnancy

Material Information

Title:
Neighborhood Environment and Hypertensive Disorders of Pregnancy
Creator:
Hu, Hui
Publisher:
University of Florida
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Epidemiology
Committee Chair:
COTTLER,LINDA B
Committee Co-Chair:
PRINS,CINDY A
Committee Members:
HENDERSON,BARRON H
LU,XIAOMIN

Subjects

Subjects / Keywords:
air
disorders
disparity
environment
hypertension
hypertensive
neighborhood
pollution
pregnancy
racial
socioeconomic

Notes

General Note:
Hypertensive disorders of pregnancy (HDP) is a common pregnancy complication which has been linked to adverse health outcomes in both mothers and infants. In the US, African Americans have a disproportionately high incidence of HDP, and an increasing trend in racial disparities related to it has been observed in recent years. Although a few risk factors have been identified, the biological mechanisms underlying HDP are largely unknown. More importantly, these known risk factors cannot explain the increasingly large racial disparities. To address these knowledge gaps, a population-based retrospective study was done using the Florida Vital Statistics Birth Record dataset to further examine the relationship between neighborhood environment and HDP and how neighborhood environment contributes to racial disparities in HDP. Firstly, the association between pregnancy, ozone (O3) and HDP was examined. Increased odds of HDP were observed with higher O3 exposure during the first two trimesters after adjusting for confounders. In the distributed lag models, elevated odds of HDP were observed with increased exposure to O3 during the 1st to 24th weeks of pregnancy, with higher odds observed during earlier pregnancy. Secondly, the association between neighborhood socioeconomic status (SES) and HDP was assessed. Compared with women living in neighborhoods with higher SES, those living in neighborhoods with lower SES had significantly higher odds of HDP. In addition, living in neighborhoods with a lower percentage of residents at the same address as the previous year, a lower proportion of females working in professional occupations, lower median household values, and higher percentage of non-Hispanic Blacks were associated with increased odds of HDP. Finally, mediation analyses were done to examine how racial disparities in HDP are mediated by neighborhood environmental factors. Compared with women of other race/ethnicities, African American women had significantly higher odds of HDP; 35% of the racial disparities in HDP could be explained by neighborhood environmental factors including O3 exposure during the first two trimesters, neighborhood SES, urbanity, and racial residential segregation.

Record Information

Source Institution:
UFRGP
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:
8/31/2018

Downloads

This item has the following downloads:


Full Text

PAGE 1

NEIGHBORHOOD ENVIRONMENT AND HYPERTENSIVE DISORDERS OF PREGNANCY By HUI HU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2016

PAGE 2

2016 Hui Hu

PAGE 3

To my wife, Yun Shen, and parents, Jianlin Hu and Huiping Shi

PAGE 4

4 ACKNOWLEDGMENTS This dissertation was a product of many years of training and earnestly hard work. It was also a product which would not have been possible without the help of so many people in so many ways. I am truly indebted to ea ch and every one of my dissertation committee members for their tremendous mentorship and support. I want to especially thank Dr. Xiaohui Xu for his tremendous guidance, support, and mentorship over the years. I am highly grateful for his willingness to s erve as my research mentor when I started the program without a strong background in epidemiology. He has taught me to become a strong and productive epidemiologist. I will always be grateful for the opportunities, encouragements, and trusts he has given m e over the years. I am deeply grateful to Dr. Linda B. Cottler, Chair of my dissertation committee for always being positive, encouraging, and supportive. I would like to express my gratitude to her for offering me the opportunities to work with her over the years, and for her generosity to serve as the Chair of my dissertation committee. Without her encouragement and valuable feedback I would not have been able to complete this journey. Dr. Cottler, thank you! I am hugely indebted to Dr. Maria R. Khan for her guidance and support over the years. Dr. Khan has served as my academic advisor for my first two years of the program, and her patience, felicity, and acceptance have been invaluable in my graduate career. Dr. Khan, thank you! I would also like to thank my dissertation committee members: Dr. Cindy A. Prins, Dr. Barron H. Henderson, and Dr. Xiaomin Lu. They have given me guidance,

PAGE 5

5 encouragement, thoughtf ul criticisms, and time despite their busy schedule. Furthermore, I would like to express my gratitude to Dr. Volker Mai and Dr. Xinguang Chen, two of the many who have been instrumental to my graduate training. Dr. Mai has served as my academic advisor ov er the past two years. Dr. Chen has collaborated with me on many research projects. Their guidance, vision, and support have helped me to become a stronger epidemiologist. I also thank the Florida Department of Health for offering m e the access to the Vita l Statistics Birth R ecords data. I am highly indebted and thoroughly grateful for the opportunity to work with the faculty, staff, and fellow students in the Department of Epidemiology at the University of Florida. I specially want to thank Dr. Sandie U. H a who has been a great research teammate over the years. Lastly, and perhaps most importantly, many thanks and appreciation should go to my parents, Jianlin Hu and Huiping Shi, who worked hard to give me the opportunity to seek my dreams I sincerely thank them for their unconditional love and sacrifices. Their trust and support have played the most important part in my life. In addition, I would like to extend my deepest gratitude to my wife, Yun Shen, for her support, understanding, and patience.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 BACKGROUND ................................ ................................ ................................ ...... 14 Public H ealth Significance of Hypertensive Disorders of Pregnancy ...................... 14 Theoretical Framework ................................ ................................ ........................... 15 Risk Factors of Hypertensive Disorders Pregnancy ................................ ................ 16 Air Pollution and Hypertensive Disorders of Pregnancy ................................ ......... 17 Neighborhood Socioeconomic Status and Hypertensive D isorders of Pregnancy .. 19 Racial Disparities in Hypertensive Disorders of Pregnancy ................................ .... 22 Knowledge Gap ................................ ................................ ................................ ...... 23 Research Objectives ................................ ................................ ............................... 24 2 OZONE AND HYPERTENSIVE DISORDERS OF PREGNANCY IN FLORIDA: IDENTIFYING CRITICAL WINDOWS OF EXPOSURE ................................ .......... 28 Introduction ................................ ................................ ................................ ............. 28 Materials and Methods ................................ ................................ ............................ 30 Study Sample ................................ ................................ ................................ ... 30 Outcome Assessme nt ................................ ................................ ...................... 31 Ozone Exposure Assessment ................................ ................................ .......... 32 Covariates ................................ ................................ ................................ ........ 32 Statistical Analyses ................................ ................................ .......................... 33 Sensitivity Analyses ................................ ................................ .......................... 34 Results ................................ ................................ ................................ .................... 34 Discussion ................................ ................................ ................................ .............. 35 Conclusion ................................ ................................ ................................ .............. 38 3 NEIGHBORHOOD SOC IOECONOMIC STATUS AND HYPERTENSIVE DISORDERS OF PREGNANCY IN FLORIDA ................................ ........................ 44 Introduction ................................ ................................ ................................ ............. 44 Materials and Methods ................................ ................................ ............................ 46 Study Sample ................................ ................................ ................................ ... 46

PAGE 7

7 Outcome Assessment ................................ ................................ ...................... 47 Neighborhood Socioeconomic Status Assessment ................................ .......... 47 Covariates ................................ ................................ ................................ ........ 48 Statistical Analyses ................................ ................................ .......................... 49 Sensitiv ity Analyses ................................ ................................ .......................... 49 Results ................................ ................................ ................................ .................... 5 0 Discussion ................................ ................................ ................................ .............. 51 Conclusion ................................ ................................ ................................ .............. 53 4 RACIAL DISPARITIES IN HYPERTENSIVE DISORDERS OF PREGNANCY MEDIATED BY NEIGHBORHOOD ENVIRONMENTAL FACTORS ....................... 59 Introduction ................................ ................................ ................................ ............. 59 Materials and Methods ................................ ................................ ............................ 60 Study Sample ................................ ................................ ................................ ... 60 Outcome Assessme nt ................................ ................................ ...................... 61 Urbanity Assessment ................................ ................................ ....................... 61 Neighborhood Socioeconomic Status Assessment ................................ .......... 62 Ozone Exposure Assessment ................................ ................................ .......... 62 Racial Residential Segregation Assessment ................................ .................... 63 Covariates ................................ ................................ ................................ ........ 64 Statistical Analyses ................................ ................................ .......................... 64 Results ................................ ................................ ................................ .................... 66 Discussion ................................ ................................ ................................ .............. 67 Conclusion ................................ ................................ ................................ .............. 69 5 CONCLUSIONS ................................ ................................ ................................ ..... 74 Summary of Research Objectives ................................ ................................ .......... 74 Accomplishments of this Dissertation ................................ ................................ ..... 75 Determination of Association between O 3 Exposure during the Three Pre defined Exposure Windows of Pregnancy and HDP ................................ ..... 75 Identification of Critical Windows of O 3 Exposure during Pregnancy for HDP .. 75 Determination of Association between Neighborhood SES and HDP .............. 76 Identification of Individual Neighborhood Socioeconomic Characteristics Predictive of HDP ................................ ................................ .......................... 76 Identification of Racial Disparities in HDP between Af rican Americans and Non African Americans ................................ ................................ ................. 77 Determination of the Proportions of Racial Disparities in HDP Contributed by Neighborhood Environmental Factors Including O 3 Exposure, Neighborhood SES, urban ity, and Racial Residential Segregation ............... 77 Limitations ................................ ................................ ................................ ............... 78 Future Directions ................................ ................................ ................................ .... 80 Use the Socio Ecological Model for Prevention ................................ ................ 80 Use mHealth to Reduce Bias and Facilitate Translational Research ............... 81 Use Innovative Data Sources ................................ ................................ ........... 83

PAGE 8

8 LIST OF REFERENCES ................................ ................................ ............................... 84 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 97

PAGE 9

9 LIST OF TABLES Table page 2 1 Maternal characteristics by hypertensive disorders of pregnancy (HDP) status among w omen with conception date during 2005 2007 in Florida, USA. ................................ ................................ ................................ ................... 39 2 2 O 3 exposure by hypertensive disorders of pregnancy (HDP) status among women with conception date during 2005 2007 in Florida, USA ....................... 40 2 3 Association between ozone (O 3 ) and hypertensive disorders of pregnancy (HDP) by pregnancy period of exposure among women with conception date during 2005 2007 in Florida, USA. ................................ ................................ ..... 41 2 4 Sensitivity analyses: fixed effects model vs. mixed effects model. ..................... 42 3 1 Neighborhood Deprivation Index (NDI) and maternal characteristics by hypertensive disorders of pregnancy (HDP) status among women with conception date during 2005 2007 in Florida, USA. ................................ ........... 54 3 2 ORs for risk of hypertensive disorders of pregnancy (HDP) by NDI among women with conception date during 2005 2007 in Flor ida, USA. ....................... 55 3 3 ORs for risk of hypertensive disorders of pregnancy (HDP) by individual neighborhood SES characteristics among w omen with conception date during 2005 2007 in Florida, USA. ................................ ................................ ..... 56 3 4 Sensitivity analyses: fixed effects model vs. mixed effects mo del. ..................... 57 4 1 Maternal characteristics by hypertensive disorders of pregnancy (HDP) status among women with conception date during 2005 2007 in Florida, USA. ................................ ................................ ................................ ................... 70 4 2 ORs of hypertensive disorders of pregnancy (HDP) mediated by neighborhood environmental factors among women with conception date during 2005 2007 in Florida, USA. ................................ ................................ ..... 71

PAGE 10

10 LIST OF FIGURES Figure page 1 1 A socio ecological model of factors influencing HDP. ................................ ......... 27 2 1 Adjusted ORs for hypertensive diso rders of pregnancy (HDP) with each 5 ppb increase in weekly exposure to ozone (O 3 ) among women with conception dates during 2005 2007 in Florida, USA.. ................................ ........ 43 3 1 Neighborhood Deprivation Index (NDI) by census tracts in Florida, USA. .......... 58 4 1 Directed acyclic graph .. ................................ ................................ ...................... 72 4 2 Overview of the mediation analyses using the inverse odds ratio weighting (IORW) method. ................................ ................................ ................................ 73

PAGE 11

11 LIST OF ABBREVIATIONS ACS American Community Survey API Application Programming Interface AUC Area Under the Curve BMI Body Mass Index CDC Centers for Disease Control and Prevention CI Confidence Interval DAG Directed Acyclic Graph EPA Environmental Protection Agency FDOH Florida Department of Health GDM Gestational Diabetes Mellitus HBM Hierarchical Bayesian Model HDP Hypertensive Disorders of Pregnancy ICC Intraclass Correlation Coefficient IORW Inverse Odds Ratio Weighting LASSO Least Absolute Shrinkage and Selection Operator mHealth Mobile Health NDI Neighborhood Deprivation Index O 3 Ozone OR Odds Ratio ppb Parts per billion SEM Socio Ecological Model SES Socioeconomic Status US United States

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy NEIGHBORHOOD ENVIRONMENT AND HYPERTENSIVE DISORDERS OF PREGNANCY By Hui Hu August 2016 Chair: Linda B. Cottler Major: Epidemiology Hypertensive disorders of pregnancy (HDP) is a common pregnancy complication which has been linked to adverse health outcomes in both mothers and infants. In the US, African Americans have a disproportionately high incidence of HDP, and an increasing trend in racial disparities related to it has been observed in recent years. Although a few risk factors have been identified, the biological mechanisms underlying HDP are largely unknown. More importantly, these known risk factors cannot explain the increasing ly large racial disparities To address these knowledge gaps, a population based retrospective study was done using the Florida Vital Statistics Birth Record dataset to further examine the relationship between neighborhood environment and HDP and how neigh borhood environment contributes to racial disparities in HDP. Firstly, the association between pregnancy ozone ( O 3 ) and HDP was examined. I ncreased odds of HDP were observed with higher O 3 exposure during the first two trimesters after adjusting for confounders In the distributed lag models, elevated odds of HDP were observed with increased exposure to O 3 during the 1 st to 24 th weeks of pregnancy with h igher odds observed during earlier pregnancy.

PAGE 13

13 Secondly, the association betwe en neighborhood socioeconomic status ( SES ) and HDP was assessed. Compared with women living in neighborhoods with higher SES those living in neighborhood s with low er SES had significantly higher odds of HDP. In addition, living in neighborhoods with a low er percentage of residents at the same address as the previous year, a lower proportion of females working in professional occupations, lower median household values, and higher percentage of non Hispanic Blacks were associated with increased odds of HDP. Finally, m ediation analyses were done to examine how racial disparities in HDP are mediated by neighborhood environmental factors. Compared with women of other race/ethnicit ies African American women had significantly higher odds of HDP ; 35 % of the racial disparities in HDP c ould be explained by neighborhood environmental factors including O 3 exposure during the first two trimesters neighborhood SES, urbanity, and racial residential segregation.

PAGE 14

14 CHAPTER 1 BACKGROUND Public Health Significance of Hypertensive Disorders of Pregnancy H ypertensive disorders of pregnancy (HDP) are the most common medical problem s encountered during pregnancy, seen among up to 10% of all pregnancies (Duley 2009) The National High Blood Pressure Education Program Working Group on High Blood Pressure in Pregnancy classified HDP into four c ategories : chronic hypertension, preeclampsia eclampsia, preeclampsia superimposed on chronic hypertension, and gestational hypertension (National High Blood Pressure Education Program 2000) These medical problems are characterized by high blood pressure, usually after 20 weeks o f gestation because blood volume change during pregnancy lead s to higher stress on the cardiovascular system (Yoder et al. 2009) Chronic hypertension is high blood pressure occurr ing either before pregnancy or diagnosed within the first 20 weeks of gestation, defined as systolic blood pressure higher than 140 mm Hg or dias tolic blood pressure higher than 90 mm Hg on 2 occasions more than 24 hours apart (Mammaro et al. 2009) Gest ational hypertension, also known as pregnancy induced hypertension, is the development of hypertension after 20 weeks of pregnancy. Preeclampsia is characterized by the new onset of hypertension and proteinuria after 20 weeks of gestation, and eclampsia is characterized by the onset of convulsions (Mammaro et al. 2009) HDP is an important risk factor for increased neonatal and maternal morbidity and mortality (Duley 2009; Lo et al. 2013) Among mothers, HDP is associated with pitting edema, endothelial abnormalities, liver and renal dysfunction, and increased risk of cardiovascular disease, stroke and T ype II diabetes later in life (Bauer and Cleary

PAGE 15

15 2009; Bellamy et al. 2007; Duley 2009; Wang et al. 2012) Among infants, maternal HDP is associated with highe r risks of small for gestational age, preterm delivery, low birthweight, and hospitalization for a wide range of neonatal diseases (Allen et al. 2004; Wu et al. 2009) In the United States, p reeclampsia alone contributes to about 25% of all medi cally indicated preterm deliveries (Ananth and Vint zileos 2006; Goldenberg et al. 2008) Theoretical Framework The Socio Ecological Model (SEM) by Bronfenbrenner (2009) can be applied to the investigation of the risk of HDP. The SEM suggests that health outcomes can be influenced by numerous individual and environmental factors. Using SEM as the theoretical framework, risk factors of HDP can be categorized into two groups as shown in Figure 1 1 : individual risk factors and environmental risk factors. Risk factors at the individual level include genetics and innate characteristics that are not part of the exogenous environment, while risk factors that are external to an indivi dual are regarded as environmental risk factors. There are four subcategories underlying the environmental risk factors, including the biological environment (i.e. living organisms), the physical environment (i.e. physical attributes of the environment), t he social environment (i.e. socioeconomic context), and the chemical environment (i.e. chemical substances). Compared with risk factors at the individual level, environmental risk factors are more amendable since interventions at both the individual leve l and the environmental level can be designed to improve health outcomes. In this dissertation, we assessed the association between neighborhood environment and HDP after controlling for individual

PAGE 16

16 risk factors. Specifically, we focused on the social and c hemical environment due to a lack of evidence linking risk factors at physical or biological environment al level to HDP. Risk Factors of Hypertensive Disorders Pregnancy Despite serious consequences, the biological mechanisms underlying HDP are largely un known. Known risk factors for HDP include maternal factors such as new paternity, obesity, age 35 years or more, adolescent pregnancy, insulin resistance, pre pregnancy hypertension or diabetes mellitus, pregnancy related factors such as multiple gestation placental abnormalities, weight gain, and gestational diabetes mellitus (GDM), family history of pre eclampsia as well as African American race (Wolf et al. 2004) To reduce the continuously increasing incidence of HDP since the 1990s in the US (Wallis et al. 2008) more modifiable risk factors for HDP need to be identified. The associations between neighborhood environment and adverse birth outcomes have been widely reported. Previous studies have consistently observed the association between air pollution exposure during pregnancy and adverse birth outcomes such as preterm delivery, low birthweight, small for gestational age, and birth defects Since it is biologically plausible t hat the underlying pathways linking air pollution to these adverse birth outcomes may be similar to the potential mechanisms between air pollution and pregnancy complications, recent studies started to focus on pregnancy complications and found pregnancy a ir pollution exposure to be associated with higher risk of HDP and GDM (Hu et al. 2014; Hu et al. 2015; Malmqvist et al. 2013) In addition to the physical environment, other environmental factors have also been linked to adverse outcomes in pregnancy. In the US, women liv ing in neighborhoods with low socioeconomic status have higher risk of preterm birth (Holzman et al. 2009; Messer et

PAGE 17

17 al. 2008; O'Campo et al. 2008) low birthweight (Nkansah Amankra et al. 2010; Schempf et al. 2009) small for gestational age (Elo et al. 2009; Farley et al. 2006) and neural tube defects (Grewal et al. 2009; Wasserman et al. 1998) Consistent findings were reported in other areas of the world (Agyemang et al. 2009; Gray et al. 2008; Gudmundsson et al. 1997; Janghorbani et al. 2006; Vrijheid et al. 2000) However, only a few studies have examined the effects of neighborhood social environment on HDP (Agyemang et al. 2009; Clausen et al. 2006; Gudmundsson et al. 19 97; Messer et al. 2012; Vinikoor Imler et al. 2012) and inconsistent findings were reported Furthermore, racial disparities in HDP exist in the US (Tanaka et al. 2007) with racial residential segregation regarded as the fundamental cause of racial disparities in health (Williams and Collins 2001) While the association between racial residential segregation and adverse birth outcomes ha s been extensively reported (Anthopolos et al. 2014; Mason et al. 2009; Messer et al. 2010; Walton 2009) little is known about its effects on HDP. Air Pollution and H ypertensive Disorders of Pregnancy The adverse health effects of ambient air pollution on mortality and morbidity have been extensively studied in both adults and children across different geographic areas of the world (Barnett et al. 2006; Bell et al. 2004; Dominici et al. 2003; Gold et al. 1999; Hung et al. 2012; Jerrett et al. 2005; Lai et al. 2013; Middleton et al. 2008; Pope et al. 1991; Pope and Kanner 1993; Pope 1999; Raaschou Nielsen et al. 2013; Samoli et al. 2007; Wietlisbach et al. 1996) Considerable consistency across studies has been observed for many health endpoints including total mortality, cardiopulmonary mortality and morbidity. More importantly, ambient air pollution has also been linked to hypertension (Coogan et al. 2012; Dong et al. 2013; Foraster et al. 2014) Compared with men, studies around the world have suggested that women may be more

PAGE 18

18 susceptible to the negative effects of air pollution (Bell et al. 2014; Park and Wang 2014) Recent studies have also consistently linked air pollution exposure to adverse birth outcomes including intrauterine growth restriction, low birth weight, premature delivery and birth defects (Ha et al. 2014; Lai et al. 2013; Smarr et al. 2013; Vinikoor Imler et al. 2 014; Vrijheid et al. 2011) Given the potential shared biological mechanisms underlying these birth outcomes and pregnancy complications, there are recent concerns that air pollution may also play a role in the development of adverse pregnancy complications such as GDM and HDP (Hu et al. 2014; Hu et al. 2015; Malmqvist et al. 2013) The biological mechanism underlying air pollution and HDP is largely unknown. However, several potential pathways have been suggested by previous studies. Evidence from animal and toxicological experiments showed that physiologic responses to environmental factors are associated w ith hypertension. Inhalation of air pollutants such as PM 2.5 has been found to increase oxidative stress, lipid peroxidation and inflammation level among pregnant women as well as the general population (Ghio et al. 2012; Lee et al. 2011; Nagiah et al. 2015; Slama et al. 2008) These inflammatory responses have been known to increase the risk of hypertension (Rodrigo et al. 2011) and they may also lead to end othelial dysfunction, autonomic imbalance, and altered blood rheology (Bind et al. 2012; Brook et al. 2004; Huang et al. 2012; Nodari et al. 2006) all of which can increase the risk of hypertension. Recent meta analyses of existing studies on air pollution and HDP suggested an overall p ositive association (Hu et al. 2014; Pedersen et al. 2014) However, results are still inconsistent among individual studies. Due to sign ificant heterogeneities among

PAGE 19

19 existing studies on the effect of air pollution on pregnancy complications, more studies need to be conducted to better investigate this relationship. Several important limitations existed in previous studies. First, exposure assessment in about half of previous studies relied on sparsely located stationary air monitors. The poor spatial resolution of this method is unable to adequately capture the spatial contrasts in air pollution levels and participants living far away from these monitors are always excluded from the analyses, which increases the exposure misclassification and may also introduce selection bias. Some other studies used traffic density and distance from major roads as a proxy for exposure to pollution, which ma y introduce substantial exposure misclassifications (van den Hooven et al. 2009) Secondly, susceptible exposure windows have not been established for HD P Identifying critical exposure windows will not only improve our understanding s of the underlying biological mechanisms between air pollution and HDP, but also help policy makers to design and implement more targeted and efficient preventive strategies. Neighborhood Socioeconomic Status and Hypertensive Disorders of Pregnancy Socioeconomic status has been recognized as an important determinant of health for many decades (Marmot 2005) Previous studies showed that living in a disadvantaged neighborhood has been associated with health behaviors such as perinatal substance use and gambling (Finch et al. 1999; Welte et al. 2004) health intermediates such as partner violence a nd pediatric injury (Cunradi et al. 2000; Shenassa et al. 2004) and health outcomes including hypertension and cardiovascular disease (Cozier et al. 2007; Cubbin et al. 2000; Cubbin et al. 2006; Morenoff et al. 2007; Mujahid et al. 2008; Roux et al. 2001) cancer incidence (Yost et al. 2001) and excess mortality (Doubeni et al. 2012; Jaffe et al. 2005) Perinatal health research increasingly

PAGE 20

20 reported the association between living in deprived neighborhood environments and adverse birth outcomes such as preterm birth, low birthweight, small fo r gestational age, and birth defects (Agyemang et al. 2009; Elo et al. 2009; Farley et al. 2006; Gray et al. 2008; Grewal et al. 2009; Gudmundsson et al. 1997; Holzman et al. 2009; Janghorbani et al. 2006; Messer et al. 2008; Nkansah Amankra et al. 2010; O'Campo et al. 2008; Schempf et al. 2009; Vrijheid et al. 2000; Wasserman et al. 1998) A recent meta analysis on seven studies including 2,579,032 pregnancies found that compared with women living in neighborhoods with the least deprived quintile, those living in the most deprived neighborhood quintile have significantly higher incidence of preterm birth, small for gestational age, and stillbirth (Vos et al. 2014) While the associations between neighborhood SES and adverse birth outcomes as well as hypertension in the general population have been well established, only a f ew studies have investigated the relationship between neighborhood deprivation and HDP (Agyemang et al. 2009; Clausen et al. 2006; Gudmundsson et al. 1997; Messer et al. 2012; Vinikoor Imler et al. 2012) and inconsistent results were found. Vinikoor Imler et al. observed a positive association between neighborhood deprivation and HDP in the US (Vinikoor Imler et al. 2012) and another study in Norway found that low income areas had higher rates of preeclampsia (Clausen et al. 2006) However, a Sweden study observed higher rates of preeclampsia in areas with higher incomes (Gudmundsson et al. 1997) and another study in Netherlands reported no association between HDP and neighborhood median income as well as unemployment (Agyemang et al. 2009) More studies are needed to further investig ate the association between neighborhood SES and HDP.

PAGE 21

21 These inconsistent findings may be caused by the social differences among different countries studied across the world as well as the different categories of neighborhood characteristics examined in the se studies. Neighborhood SES is multifaceted, and multiple categories need to be considered to assess it comprehensively. Rajaratnam et al. identified 12 categories of neighborhood characteristics that have been routinely examined in perinatal health studi es (Rajaratnam et al. 2006) such as income, employment, family structure, po pulation composition, housing, mobility, education, occupation, and social resources. Several composite indexes have been developed to capture the multifaceted characteristics of neighborhood SES, including the Townsend Deprivation Index (Townsend et al. 1988) the Jarman score (Jarman 1983) the Index of Multiple Deprivation (Noble et al. 2004) the Carstairs Morris score (Carstairs and Morris 1990) and the standardized Neighborhood Deprivation Index (NDI) (Messer et al. 2006) The standardized ND I developed by Messer et al. is widely used in the US, especially in perinatal health studies (Messer et al. 2006) This index uses the US Census data and covers five domains including income/poverty, education, empl oyment, housing, and occupation. Although the use of data reduction techniques to generate these composite indexes provides a comprehensive assessment of neighborhood SES, the results from these studies are hard to be interpreted to guide interventions bec ause no individual neighborhood characteristics predictive of the outcome can be identified. Studies using variable selection methods are needed to better identify the individual neighborhood SES characteristics associated with HDP.

PAGE 22

22 Racial Disparities in Hypertensive Disorders of Pregnancy The burden of HDP falls disproportionately on African Americans (Tanaka et al. 2007) A 10 year longitudinal study in New York State reported that 8.5% of African American women had HDP during pregnancy, compared with only 5.5% and 6. 2% in White and Hispanic women, respectively (Tanaka et al. 2007) More importantly, the study found an increasing trend of racial disparities in HDP that cannot be explained by known risk factors such as new paternity, multiple gestation, and extreme reproductive age. The State of Florida also witnessed increasingly lar ge racial disparities in HDP. Data from the Florida Vital Statistics Birth Records show that the racial disparities in HDP have been two times larger in 2012 compared with 2004. In 2004, 5.57% and 5.00% African American and White women had HDP during pregn ancy, respectively, while in 2012, 7.21% and 5.54% African American and White women had HDP, respectively. Given the increasingly large racial disparities in HDP, it is urgent and necessary to investigate the underlying factors associated with this trend w hich could help public health agencies develop more targeted and efficient health policies and interventions to reduce both racial disparities and risks of HDP among Floridians Racial residential segregation has been regarded as the fundamental cause of racial disparities in health (Williams and Collins 2001) In the US, it usually refers to the residential separation of African American neighborhoo ds from those of other racial/ethnicity groups. Although overt discrimination in housing markets were made illegal by the Civil Rights Act in 1968, racial residential segregation persisted in forms such as racial steering and lending discrimination (Mendez et al. 2013) Previous studies suggested that the average residential context of African American communities is worse than the worst resi dential context for Whites (Sampson and Wilson 1995;

PAGE 23

23 Williams and Collins 2001) Various studies have linked racial residential segregation to adverse birth outcomes in the US (Anthopolos et al. 2011; Anthopolos et al. 2014; Bell et al. 2006; Grady 2006; Grady and Ramrez 2008; Osypuk and Acevedo Garcia 2008) In a parallel literature, racial residential segregation has been associated with hypertension in the general population (Kershaw et al. 2011) However, no study has examined the association between racial residential segregation and HDP. More importantly, it is necessary to assess the underlying risk factors related to the enlarging racial disparities in HDP and the potential association between racial residential segregation and HDP in Florida. Factors mediating the pathways should be associated with both HDP risks and distributed unevenly in different racial/ethnicity groups. Neighborhood socioeconomic status and air pollution are two candidates that match the criteria (Kearney and Kiros 2009; Pollock and Vittas 1995; Stretesky and Hogan 1998) However, few stud ies have directly addressed how neighborhood environment contribut es to racial disparities in HDP. Mediation analysis is increasingly used recently in health disparity studies given its advantages to separate and quantify the direct and indirect effects (Bennett et al. 2012; Chatterjee et al. 2011; Mugavero et al. 2009) Identifying the mediators will not only improve our understanding of the underlying causal mechanisms, but also help to design and implement better invention stra tegies. Knowledge Gap As discussed in the previous sections, a lthough several risk factors have bee n identified (Wolf et al. 2004) the biological mechanism underlying HDP is still largely unknown, and none of these known risk factors can explain increasingly l arge racial disparities. Studies designed to identify modifiable risk factors for HDP and to

PAGE 24

24 investigate how they contribute to racial disparities in HDP are needed to guide public health interventions and improve pregnancy outcomes. Emerging evidence has linked neighborhood environmental factors such as air pollution exposure and nei ghborhood SES to adverse birth outcomes (Agyemang et al. 2009; Elo et al. 2009; Farley et al. 2006; Gray et al. 2008; Grewal et al. 2009; Gudmundsson et al. 1997; Han sen et al. 2009; Holzman et al. 2009; Janghorbani et al. 2006; Messer et al. 2008; Nkansah et al. 2005; Vrijheid et al. 2000; Wasserman et al. 1998) It is plausible that the underlying pathways between neighborhood environment and adverse birth outcomes may be similar to the relationship between neighborhood environment and HDP. Our preliminary studies found an association between air pollution exposure and HDP (Hu et al. 2014; Xu et al. 2014) and other studies have linked neighborhood SES to HDP (Agyemang et al. 2009; Clausen et al. 2006; Gudmundsson et al. 1997; Messer et al. 2012; Viniko or Imler et al. 2012) However, inconsistent results were observed and no firm conclusion ca n be made so far. Furthermore, racial disparities in HDP have not been well studied and factors contributing to the disparities are largely unknown. Research Objectives To address the s e knowledge gaps, a population based retrospective cohort study was done to investigate the association between neighborhood environment and HDP using the Florida Vital Statistic Birth Record dataset which includes all pregnant women residing in Florida with a conception date between January 1, 2005 and December 31, 2 007. The main aims of this dissertation were : 1. To investigate t he association between O 3 exposure during pregnancy and hypertensive disorder s of pregnancy.

PAGE 25

25 1) Hypothesis 1.1 : E xposure to O 3 during the first two trimesters of p regnancy is associated with hypertensive disorder s of pregnancy. 2) Hypothesis 1.2 : Early pregnancy is the most critical window of exposure 2. To examine the association between neighborhood socioeconomic status and hypertensive disorder s of pregnancy. 1) Hypothesis 2.1 : Neighborhood socioeconomic status assessed by the Standardized Neighborhood Deprivation Index is associated with hypertensive disorder s of pregnancy. 2) Hypothesis 2.2 : Individual neighborhood socioeconomic characteristics such as income and poverty are pre dictive of hypertensive disorder s of pregnancy. 3. To examine the racial disparities in hypertensive disorder s of pregnancy and how neighborhood environmental factors mediate the pathways. 1) Hypothesis 3.1: African American w omen have higher risks of hyperten sive disorder s of pregnancy compared with non African Americans 2) Hypothesis 3.2 : The association between race and hypertensive disorder s of pregnancy is mediated by neighborhood environmental factor s including O 3 exposure during pregnancy, neighborhood socioeconomic status, urbanity, and racial residential segregation In other words, this dissertation aimed to answer the following research questions: 1. What are the effect estimates for the association between O 3 exposure during the three pre defined wind ows of pregnancy (i.e. the first trimester, the second trimester, and the first and second trimesters) and HDP? 2. What are the most critical exposure windows of O 3 during pregnancy for HDP? 3. What are the overall effect estimates for the association between neighborhood SES and HDP? 4. Which individual neighborhood socioeconomic status is predictive of HDP? 5. What are racial disparities in HDP between African Americans and non African Americans?

PAGE 26

26 6. What proportions of the racial disparities in HDP are contributed by neighborhood environmental factors including O 3 exposure during pregnancy neighborhood socioeconomic status, urbanity, and racial residential segregation? This dissertation is the first to utilize large population based data to assess the susceptible exposure windows to O 3 during pregnancy for HDP. Additionally, it is the first study to assess how individual neighborhood SES characteristics are associated with HDP. More importantly, it investigates how neighborhood environmental factors may contribute to the increasingly large racial disparities in HDP. Results from this study provide important guidance for public health agencies to design and implement more efficient and targeted intervention strategies to reduce the HDP incidence and to mitigate raci al disparities in HDP

PAGE 27

27 Figure 1 1 A s ocio ecological model of factors influencing HDP.

PAGE 28

28 CHAPTER 2 OZONE AND HYPERTENSIVE DISORDERS OF PREGNANCY IN FLORIDA: IDENTIFYING CRITICAL WINDOWS OF EXPOSURE Introduction H ypertensive disorders of pregnancy (HDP) are among the most common medical problems encountered during pregnancy, affecting up to 10% of all pregnancies (Duley 2009; Miller and Carpenter 2015) HDP is classified into four categories, including chronic hypertension, preeclampsia eclampsia, preeclampsia superimposed on chronic hypertension, and gestational hypertension (National High Blood Pressure Education Program 2000) HDP is characterized by high blood pressure, usu ally after 20 weeks of gestation because blood volume change during pregnancy may lead to higher stress on the cardiovascular system (Yoder et al. 2009) It is considered a risk factor for increasing both neonatal and maternal morbidity and mortality (Allen et al. 2004; Bauer and Cleary 2009; Bellamy et al. 2007; Duley 2009; Lo et al. 2013; Wang et al. 2012; Wu et al. 2009) In the United States, p reeclampsia alone contributes to about 25% of all medically indicated preterm deliveries (Ananth and Vintzileos 2006; Goldenberg et al. 2008; Romero et al. 2014) Despite serious consequences, the biological mechanisms underlying HDP remain to be determined. Kno wn risk factors for HDP include maternal factors such as new maternity, obesity, age 35 years or more, adolescent pregnancy, pre pregnancy hypertension or diabetes mellitus, pregnancy related factors such as multiple gestation, placental abnormalities, wei ght gain, gestational diabetes mellitus (GDM), family history of pre eclampsia as well as African American race (Wolf et al. 2004) To reduce the continuously increasing rate of incidence of HDP (Wallis et al. 2008) a better understanding of modifiable risk factors for HDP is needed to guide the designs of interventions.

PAGE 29

29 Ambient air pollution has also been linked to hypertension in the general population (Coogan et al. 2012; Dong et al. 2013; Foraster et al. 2014) Compared with men, women may be more susceptible to the negat ive effects of air pollution due to the hormones and pulmonary structural and morphologic differences (Bell et al. 2014; Harms 2006; Park and Wang 2014) Rec ent studies have also consistently found associations between air pollution exposure and adverse birth outcomes, such as intrauterine growth restriction, low birth weight, premature delivery, or birth defects (Ha et al. 2014; Lai et al. 2013; Smarr et al. 2013; Vinikoor Imler et al. 2014; Vrijheid et al. 2011) Given the toxicological effects of air pollution such as increasing oxidative stress, lipid peroxi dation and inflammation (Ghio et al. 2012; Lee et al. 2011; Nagiah et al. 2015; Slama et al. 2008) there are concerns that air pollution may also play a role in the developm ent of HDP (Hu et al. 2014; Malmqvist et al. 2013) Ozon e (O 3 ) is the air pollutant of the greatest concern to the state of Florida (Florida Department of Environmental Protection 2012) and recent meta analyses of existing studies on ozone and HDP suggest an overall positive association (Hu et al. 2014; Pedersen et al. 2014) However, results are still inconsistent among individual studies. Several important limitations exist from previous studies. First, exposure assessment in many studies relied on sparsely located stationary air monit ors (Mobasher et al. 2013; Olsson et al. 2013; van den Hooven et al. 2011; Vinikoor Imler et al. 2012; Xu et al. 2013; Zhai et al. 2012) The poor spatial resolution of this method constrained the capability to adequately capture the spatial contrasts in O 3 levels, leading to increased exposure misclassification bias. In addition, participants living far away from these monitors are always excluded from the analyses, which increases

PAGE 30

30 selection bias. Other studies used traffic densi ty and distance from major roads as a proxy for exposure to pollution, which may introduce substantial exposure misclassification bias as well (van den Hooven et al. 2009) Secondly, critical exposure windows to O 3 have not been established for HDP. Identification of critical exposure windows is needed to improve the understanding of the underlying biological mechanisms between O 3 and HDP and to help inform the designs and implementations of targeted and effective preventive strategies. Environmental Public Health Tracking Network (U.S. EPA 2014) to assess ambient O 3 levels, and linked it to the Florida Vital Statistics Birth Record dataset to investigate the association between HDP and O 3 among all eligible women residing in Florida with conception dates between January 1, 2005 and December 31, 2 007. More importantly, we assessed critical pregnancy windows for O 3 exposure using a distributed lag model to reduce the influences of autocorrelation and collinearity in weekly O 3 exposure. Materials and Methods Study Sample Birth record data were obtai ned from the Bureau of Vital Statistics, Office of Health Statistics and Assessment, Florida Department of Health ( http://www.floridahealth.gov/certificates/certificates/ Jacksonville, Florida). The data included all re gistered live births in Florida between January 1, 2005 and December 31, 2008 (n=917,788). Women with residential addresses outside Florida (n=4,632) were s initially geocoded by Florida Department of Health using ArcGIS v10.1, and 864,247 records (94.6%) were successfully geocoded. We further geocoded the addresses that failed to be geocoded

PAGE 31

31 by DOH using the Google Maps API (Application Programming Interfac package in R and a total of 913,048 records (99.9%) were successfully geocoded. Women whose residential address could not be geocoded were excluded (n=108). To avoid fixed cohort bias ( Strand et al. 2011 ) women were included based on their conception date instead of delivery date. Conception date was calculated using delivery date and gestational age which is mainly determined by ultrasound. When ultrasound data was not available, clinical examination or last menstrual period was used to estimate gestational age. Among the 913,048 women who delivered during 2005 2008, a total of 691,011 women had the conception date between Januar y 1, 2005 and December 31, 2007. In addition, women were excluded if they had non singleton deliveries (n=21,609) or pre pregnancy hypertension (n=10,590). Women whose births had a birthweight <500 g or >5000 g (n=621), or with a gestational age <26 weeks (n=2,662) were also excluded. A total of 655,529 women were included in the analyses. Outcome Assessment During the collections of Vital Statistics Birth Record data, the medical history of each woman was checked, and diagnoses of pre pregnancy hypertension, gestational hypertension or preeclampsia, and eclampsia were abstracted for data analyses. Gestational hypertension was determined as the development of hypertension after 20 weeks of pregnancy, and preeclampsia was defined as the new onset of hypertension and proteinuria after 20 weeks of gestation. Eclampsia was determined by the onset of convulsions. Si milar to previous environmental studies on HDP ( Hu et al. 2014 ) the restricted definition of HDP was used in this study, which included gestational hypertension, preeclampsia, and/or eclampsia.

PAGE 32

32 Ozone Exposure Assessment O 3 Health Tracking Network ( U.S. EPA 2014 ) which were assessed using the hierarchical Bayesian space time statistical model (HBM) during 2001 2008 with a daily temporal resolution and a spatial resolution of 12km12km across the continental areas in the US ( McMillan et al. 2010 ) The HBM approach combines the Air Quality System monitoring data with the Community Multiscale Air Quality modeled data, which includes emission, meteorology, and chemical modeling components, to predict air quality data for a specific time period and spatial scale ( McMillan et al. 2010 ) corres ponding grid of the HBM data. Exposures were calculated as daily concentrations averaged over each of the first two trimesters (trimester 1: 1 13 weeks and trimester 2: 14 26 weeks) and over both the first and second trimesters (1 26 weeks) determined by g estational age and delivery date of each woman. In addition, weekly average levels of O 3 exposure during the first and second trimesters were calculated to assess the critical windows of exposure during pregnancy. Covariates Information on maternal charact eristics such as age, race/ethnicity, education, marital status, pregnancy smoking status, pre pregnancy body mass index (BMI), season and year of conception were obtained directly from the births records. Maternal age at delivery was categorized into six groups, with 5 year increments for women aged 20 40 years old as well as two additional groups for < Race/ethnicity was categorized as non Hispanic White, non Hispanic Black, Hispanic, and others. In addition, dichotomous variables were used to indicate marital status and

PAGE 33

33 pregnancy smoking status. Maternal education was divided into three categories: high school. Pre pregnancy BMI was categorized into four groups: underweight (<18.5), normal (18.5 24.9), overweight (25 Season [warm (June November) or cool (Decembe r May)] and year (2005, 2006, or 20 07) of conception were also treated as categorical variables. Statistical Analyse s Distribution of categorical covariates and continuous exposures between women with HDP and those without HDP were examined. Firstly, we c onducted logistic regression models to investigate the association between O 3 exposure during the three predefined gestational windows (i.e. trimester 1, trimester 2, and trimesters 1&2) and odds of HDP. The level of O 3 exposure was analyzed as both contin uous and categorical (i.e. quartiles) variable s Both an unadjusted model and an adjusted model controlling for maternal age, race/ethnicity, education, marital status, pregnancy smoking, pre pregnancy BMI, season and year of conception were used. Odds rat ios (ORs) and 95% confidence intervals (CIs) were obtained for O 3 during each specific pregnancy window Consistent with other studies ORs and 95% CIs were reported for each 5 ppb increase in continuous O 3 exposure. Secondly, we used exposure window s. The associations between HDP and each of the 26 weekly O 3 exposures was examined controlling for other weekly O 3 exposures. Similar to the study by Darrow et al. ( 2011 ) we assumed an underlying cubic structu re, and we constrained the 26 weekly specific effect estimates to follow the shape of a natural cubic spline with a knot at lag 13 to reduce the influences of collinearity on the estimates. All statistical analyses were conducted using R 3.2.2.

PAGE 34

34 Sensitivity Analyses Sensitivity analyses were conducted to test the robustness of our results. To account for the correlations at census tract level, we further fitted mixed effects models with random intercept for each census tract. Then we made comparisons between the resul ts from the sensitivity analyses and our original results to check whether the potential correlations at census tract level have influenced the observed effects. Results Among the 655,529 women included in the analyses, 31,362 (4.8%) women had HDP. A tota l of 613,032 women including 29,286 HDP cases had complete data for all covariates. Table 2 1 shows the maternal characteristics by HDP status. Women with HDP were less likely than those without HDP to be between 25 to 34 years old, and more likely to belo ng to non Hispanic Black racial/ethnic categories. Compared with women without HDP, HDP cases were less likely to be married and to have smoked during pregnancy. In addition, HDP cases had higher pre pregnancy BMI than women without HDP. Table 2 2 shows t he distribution of exposures to O 3 during the first two trimesters of pregnancy. Women with HDP had a higher exposure to O 3 compared with those without HDP during the three pre defined exposure windows (p<0.001 for each comparison). Table 2 3 shows unadju sted and adjusted ORs from logistic regression models. Positive associations between HDP and O 3 exposure were consistently observed in all models and exposure windows. In the adjusted model, increased odds of HDP for each 5 ppb increase in O 3 exposure were observed during all the three exposure windows (OR Trimester1 =1.04, 95% CI: 1.03, 1.06 ; OR Trimester2 =1.03, 95% CI: 1.02 1.04 ;

PAGE 35

35 OR Trimester1&2 =1.07, 95% CI: 1.05, 1.08 ). Consistent results w ere observed in the analyses when O 3 exposure was categorized. Comp ared with women exposed to the lowest quartile of O 3 women with O 3 exposure in the other quartiles had higher odds of HDP. Figure 2 1 displays adjusted ORs of HDP for each 5 ppb increase in weekly exposure to O 3 during the first two trimesters of pregnan cy (26 weeks). Consistent with the previous logistic regression models, the results from the distributed lag models also showed positive associations between HDP and O 3 exposure across the first 24 weeks of gestation, with higher odds of HDP observed in ea rly pregnancy. Table 2 4 shows the comparisons between the original results from the fixed effects model and the results of the sensitivity analyses using the mixed effects model. Consistent results were observed in the mixed effects model with intraclass correlation coefficient s (ICC s ) ranging from 0.059 to 0.06 3 Discussion In this retrospective cohort study, we found a consistent pattern of elevated odds of HDP with increased exposure to O 3 during the first two trimesters. Such associations persisted with adjustment for confounders including maternal age, race/ethnicity, education, marital status, pregnancy smoking, pre pregnancy BMI, season and year of conception. Furthermore, the distribut ed lag models showed that early pregnancy was the most critical window for O 3 exposure during pregnancy. The results of this study add to the emerging evidence linking O 3 to HDP. The observed association between HDP and exposure to O 3 in early pregnancy i s consistent with other studies. Our previous meta analyses on O 3 and HDP found that exposure to O 3 during the first trimester is associated with increased odds of HDP

PAGE 36

36 (OR=1.05, 95% CI: 1.02, 1.06 for each 5 ppb increase in O 3 ), while no significant associ ation was observed for O 3 exposure during the second trimester ( Hu et al. 2014 ) In this study, we observed a slightly larger effect size of O 3 on HDP during the first trimester as well as a significant association during the second trimester. This is expected since previous studies assessed O 3 exposure using data from stationary monitors, which may lead to non differential exposure misclassification and bias the effect towards null. The biological mechanisms underlying O 3 and HDP have been suggested to be multifaceted. Evidence from animal and toxicological experiments showed that physiologic responses to environmental factors are associated with hypertension ( Nagiah et al. 2015 ) Inhalation of air pollutants has been found to increase oxidative stress, lipid peroxidation and inflammation level among pregnant women ( Ghio et al. 2012 ; Lee et al. 2011 ; Nagiah et al. 2015 ; Slama et al. 2008 ) These inflammatory responses have been known to increase the risk of hypertension ( Rodrigo et al. 2011 ) and they may also lead to endothelial dysfunction, autonomic imbalance, and altered blood rheology ( Bind et al. 2012 ; Brook et al. 2004 ; Huang et al. 2012 ; Nodari et al. 2006 ) all of which can increase the risk of hypertension. Blood pressure levels of women without HDP usually fall during the first trimester, and then gradually increase to the pre pregnancy level after reaching the lowest point in mid pregnancy ( Ayala et al. 1997 ; Hermida et al. 2001 ) However, a different blood pressure pattern during pregnancy has been observed among women with HDP ( Hermida et al. 2001 ) Instead of the fall during first trimester, HDP cases have stable blood pressure levels during the first half of pregnancy and then continuously increases until their deliveries. The

PAGE 37

37 differences in the blood pressur e patterns suggest that HDP usually develops at an early stage of pregnancy, which are consistent with the elevated effect size observed in this study. Vasoconstriction has been regarded as one of the mechanisms underlying HDP, and previous studies found t hat vasoconstriction in HDP develops early in pregnancy ( O'Brien 1990 ; Orpana et al. 1996 ) Given O 3 vasoconstriction observed in the general population ( Brook et al. 2002 ) it is plausible that vasoconstriction may be the mechanism underlying the o bserved higher odds in early pregnancy period. The findings from our study have important implications for health interventions in pregnancy period, and therefore knowin g the critical exposure windows is important for the design of an efficient and targeted intervention in the future. Future studies are needed to examine the potential use of emerging technologies to reduce risks of HDP, such as the wearable environment tr ackers and indoor air purifiers ( Chen et al. 2015 ) Our study has several s trengths. First, the sample size is large, and a high geocoded rate was achieved in this study. Second, the HBM air pollution data used in the analysis covered the whole study area and they had a daily temporal resolution and a 12km12km spatial resolution allowing us to include all participants regardless of the distance to air monitors. Third, different from previous studies, we used distributed lag models to assess critical exposure windows in addition to investigating the association in the predefined exposure windows. Several limitations need to be noted. First, the diagnosis dates of HDP were not available for analysis. Future studies with more detailed data on pregnancy outcomes

PAGE 38

38 and diagnoses time are needed to better understand the effects of O 3 on different adverse pregnancy and birth outcomes by using more accurate exposure windows and comparing competing risks on outcomes. Second, instead of investigating gestational hypertension, preeclampsia, and eclampsia separately, we assessed them together as HDP because of the potential differences in disease coding and diagnosis. The variation in HDP coding has been suggested to be an issue since increased gestational hypertension and reduced mild preeclampsia and eclampsia are compensated more ( Savitz et al. 2015 ) In addition, it is possible that HDP may be underdiagnosed in the vital statistics records data. However, this misclassifica tion is likely to be non differential (Hu et al. 2015) Furthermore, we assessed O 3 addresses at delivery, while information on residential history, daily mobility, and behavior patterns were not available, which may introduce misclassifications of exposure. Future studies with improved exposure assessments are needed to better quantify the effects of O 3 on HDP. Conclusion Using Florida birth vital statistics records, we found that exposure to O 3 during pregnancy was associated with increased odds of HDP, with stronger association noted during early pregnancy periods.

PAGE 39

39 Table 2 1. Maternal characteristics by hypertensive disorders of pregnancy (HDP) status among women with conception date during 2 005 2007 in Florida, USA. Maternal Characteristics HDP (n=31,362) n (%) No HDP (n=624,167) n (%) Total (n=655,529) n (%) Maternal age (years) <20 4,089(13.0) 67,388(10.8) 71,477(10.9) 20 24 8,325(26.5) 164,444(26.4) 172,769(26.4) 25 29 7,973(25.4) 170,781(27.4) 178,754(27.3) 30 34 6,244(19.9) 134,271(21.5) 140,515(21.4) 35 39 3,678(11.7) 71,101(11.4) 74,779(11.4) 1,053(3.4) 16,156(2.6) 17,209(2.6) Missing 0(0.0) 26(0.0) 26(0.0) Race/ethnicity Non Hispanic White 14,634(46.7) 279,996(44.9) 294,630(45.0) Non Hispanic Black 7,426(23.7) 108,974(17.5) 116,400(17.8) Hispanic 1,924(6.1) 4,9681(8.0) 51,605(7.9) Others 7,378(23.5) 185,502(29.7) 192,880(29.4) Missing 0(0.0) 14(0.0) 14(0.0) Maternal education High school 14,853(47.4) 298,307(47.8) 313,160(47.8) Missing 149(0.5) 3,343(0.5) 3,492(0.5) Marital status Married 16,159(51.5) 343,198(55.0) 359,357(54.8) Not married 15,199(48.5) 280,873(45.0) 296,072(45.2) Missing 4(0.0) 96(0.0) 100(0.0) Smoking during pregnancy No 28,916(92.2) 570,831(91.5) 599,747(91.5) Yes 2,419(7.7) 52,887(8.5) 55,306(8.4) Missing 27(0.1) 449(0.1) 476(0.1) Pre pregnancy BMI Underweight (<18.5) 776(2.5) 31,191(5.0) 31,967(4.9) Normal (18.5 24.9) 10,589(33.8) 305,043(48.9) 315,632(48.2) Overweight (25.0 29.9) 7,942(25.3) 141,831(22.7) 149,773(22.9) 10,093(32.2) 108,143(17.3) 118,236(18.0) Missing 1,962(6.3) 37,959(6.1) 3,9921(6.1) Season of conception Warm 15,327(48.9) 306,634(49.1) 321,961(49.1) Cool 16,035(51.1) 317,533(50.9) 333,568(50.9) Year of conception 2005 10,077(32.1) 205,628(32.9) 215,705(32.9) 2006 10,611(33.8) 212,402(34.0) 223,013(34.0) 2007 10,674(34.0) 206,137(33.0) 216,811(33.1)

PAGE 40

40 Table 2 2 O 3 exposure by hypertensive disorders of pregnancy (HDP) status among women with conception date during 2005 2007 in Florida, USA (n=31,362 with HDP, n=624,167 without HDP, and total n= 655,529). Exposure Trimester 1 Trimester 2 Trimesters 1 and 2 Statistics HDP No HDP Total HDP No HDP Total HDP No HDP Total O 3 (ppb) Mean SD 39.07 6.64 38.63 6.76 38.65 6.76 39.02 6.66 38.57 6.69 38.59 6.69 39.06 5.20 38.61 5.33 38.63 5.33 Median 38.39 37.89 37.91 38.29 37.81 37.84 38.84 38.39 38.42 IQR 9.37 9.41 9.41 9.52 9.37 9.38 7.52 7.82 7.80

PAGE 41

41 Table 2 3 Association between ozone (O 3 ) and hypertensive disorders of pregnancy (HDP) by pregnancy period of exposure among women with conception date during 2005 2007 in Florida, USA. Unadjusted Model Adjusted Model a O 3 Exposure n (HDP/Total) OR (95% CI) n (HDP/Total) b OR (95% CI) Trimester 1 Continuous (each 5 ppb increase) 31,362/655,529 1.05(1.04, 1.06 ) 29,286/613,032 1.04(1.03, 1.06 ) Quartile 1 (21.2 33.7 ppb) 6,841/163,885 Reference 6,182/149,297 Reference Quartile 2 (33.8 37.9 ppb) 8,000/163,883 1.18(1.14, 1.22) 7,528/154,680 1.13(1.09, 1.17) Quartile 3 (38.0 43.1 ppb) 8,174/163,879 1.21(1.17, 1.25) 7,706/153,717 1.18(1.14, 1.23) Quartile 4 (43.2 57.4 ppb) 8,347/163,882 1.23(1.19, 1.27) 7,870/155,338 1.19(1.15, 1.25) Trimester 2 Continuous (each 5 ppb increase) 31,362/655,529 1.05(1.04, 1.06 ) 29,286/613,032 1.03(1.02, 1.04 ) Quartile 1 (21.2 33.6 ppb) 6,996/163,891 Reference 6,288/148,506 Reference Quartile 2 (33.7 37.8 ppb) 7,973/163,874 1.15(1.11, 1.19) 7,520/154,514 1.09(1.06, 1.13) Quartile 3 (37.9 43.0 ppb) 7,915/163,883 1.14(1.10, 1.18) 7,422/154,363 1.08(1.04, 1.12) Quartile 4 (43.1 57.4 ppb) 8,478/163,881 1.22(1.18, 1.26) 8,056/155,649 1.14(1.10, 1.18) Trimesters 1&2 Continuous (each 5 ppb increase) 31,362/655,529 1.08(1.07, 1.09 ) 29,286/613,032 1.07(1.05, 1.08 ) Quartile 1 (23.2 34.8 ppb) 6,748/163,883 Reference 6,036/148,205 Reference Quartile 2 (34.9 38.4 ppb) 7,859/163,882 1.17(1.13, 1.21) 7,369/153,480 1.13(1.09, 1.15) Quartile 3 (38.5 42.6 ppb) 8,333/163,883 1.25(1.21, 1.29) 7,916/155,640 1.19(1.15, 1.17) Quartile 4 (42.7 53.1 ppb) 8,422/163,881 1.26(1.22, 1.30) 7,965/155,707 1.20(1.23, 1.25) a Adjusted for maternal age, race, education, marital status, pregnancy smoking status, pre pregnancy BMI, season of conception, and year of conception. b Women with complete data for all covariates.

PAGE 42

42 Table 2 4 Sensitivity analyses: fixed effects model vs. mixed effects model n (HDP/Total) a Fixed effects Model b OR (95% CI) Mixed effects Model c OR (95% CI) O 3 Exposure Trimester 1 Continuous (each 5 ppb increase) 29,286/613,032 1.04(1.03, 1.06 ) 1.04(1.03, 1.05) Quartile 1 (21.2 33.7 ppb) 6,182/149,297 Reference Quartile 2 (33.8 37.9 ppb) 7,528/154,680 1.13(1.09, 1.17) 1.11(1.07, 1.15) Quartile 3 (38.0 43.1 ppb) 7,706/153,717 1.18(1.14, 1.23) 1.17(1.12, 1.21) Quartile 4 (43.2 57.4 ppb) 7,870/155,338 1.19(1.15, 1.25) 1.17(1.12, 1.22) Trimester 2 Continuous (each 5 ppb increase) 29,286/613,032 1.03(1.02, 1.04 ) 1.03(1.02, 1.04) Quartile 1 (21.2 33.6 ppb) 6,288/148,506 Reference Quartile 2 (33.7 37.8 ppb) 7,520/154,514 1.09(1.06, 1.13) 1.08(1.04, 1.12) Quartile 3 (37.9 43.0 ppb) 7,422/154,363 1.08(1.04, 1.12) 1.06(1.02, 1.10) Quartile 4 (43.1 57.4 ppb) 8,056/155,649 1.14(1.10, 1.18) 1.12(1.08, 1.16) Trimesters 1&2 Continuous (each 5 ppb increase) 29,286/613,032 1.07(1.05, 1.08 ) 1.06(1.05, 1.08) Quartile 1 (23.2 34.8 ppb) 6,036/148,205 Reference Quartile 2 (34.9 38.4 ppb) 7,369/153,480 1.13(1.09, 1.15) 1.12(1.08, 1.16) Quartile 3 (38.5 42.6 ppb) 7,916/155,640 1.19(1.15, 1.17) 1.17(1.13, 1.22) Quartile 4 (42.7 53.1 ppb) 7,965/155,707 1.20(1.23, 1.25) 1.18(1.13, 1.23) a Women with complete data for all covariates. b Adjusted for maternal age, race, education, marital status, pregnancy smoking status, pre pregnancy BMI, season of conception, and year of conception. c Mixed effects model with rando m intercept for each census tract and adjusted for maternal age, race, education, marital status, pregnancy smoking status, pre pregnancy BMI, season of conception, and year of conception.

PAGE 43

43 Figure 2 1 Adjusted ORs for hypertensive disorders of pregnancy (HDP) with each 5 ppb increase in weekly exposure to ozone (O 3 ) among women with conception dates during 2005 2007 in Florida, USA. The circle reflects the central estimate; the horizontal line represents the 95% CI.

PAGE 44

44 CHAPTER 3 NEIGHBORHOOD SOCIOECONOMIC STATUS AND HYPERTENSIVE DISORDERS OF PREGNANCY IN FLORIDA Introduction H ypertensive disorders of pregnancy (HDP) are common pregnancy complications, which occur among up to 10% of all pregnant women ( Duley 2009 ) HDP is characterized by high blood pressure, usually after 20 weeks of gestation when the change in blood volume may result in stress on the cardiovascular system ( Yoder et al. 2009 ) HDP is associated with an increase in neonatal and maternal morbidity and mortality ( Allen et al. 2004 ; Bauer and Cleary 2009 ; Bellamy et al. 2007 ; Duley 2009 ; Lo et al. 2013 ; Wang et al. 2012 ; Wu et al. 2009 ) Since the 1990s, there has been an increase in the incidence of HDP in the US ( Wallis et al. 2008 ) Given the adverse effects of HDP, there is a need to understand modifiable risk factors for HDP. Socioeconomic status (SES) is an important determinant of health ( Marmot 2005 ) Low neighborhood SES has been linked to an increased risk of hype rtension and cardiovascular diseases in the general population ( Cozier et al. 2007 ; Cubbin et al. 2000 ; Cubbin et al. 2006 ; Morenoff et al. 2007 ; Mujahid et al. 2008 ; Roux et al. 2001 ) On the other hand, perinatal epidemiologists have also found an association between living in deprived neighborhood environments and adverse birth outcomes such a s preterm birth, low birthweight, small for gestational age, and birth defects ( Agyemang et al. 2009 ; Elo et al. 2009 ; Farley et al. 2006 ; Gray et al. 2008 ; Grewal et al. 2009 ; Gudmundsson et al. 1997 ; Holzman et al. 2009 ; Janghorbani et al. 2006 ; Messer et al. 2008 ; Nkansah Amankra et al. 2010 ; O'Campo et al. 2008 ; Schempf et al. 2009 ; Vrijheid et al. 2000 ; Wasserman et al. 1998 ) While the association between the neighborhood SES and adverse birth outcomes (as well as hypertension in the general

PAGE 45

45 population) have been well established, only a few studies have investigated the relationship be tween neighborhood deprivation and HDP ( Agyemang et al. 20 09 ; Clausen et al. 2006 ; Gudmundsson et al. 1997 ; Messer et al. 2012 ; Vinikoor Imler et al. 2012 ) Due to the social differences among different countries studied across the world as well as the different categories of neighborhood characteristics examined in these studies, the results on the association between neighborhood deprivation and HDP h ave been inconsistent. Neighborhood SES is multifaceted, and multiple categories need to be considered to assess it comprehensively. Several composite indi ces have been developed to capture the multifaceted characteristics of neighborhood SES, including t he Townsend Deprivation Index ( Townsend et al. 1988 ) the Jarman score ( Jarman 1983 ) the Index of Multiple Deprivation ( Noble et al. 2004 ) the Carstairs Morris score ( Carstairs and Morris 1990 ) and the standardized Neighborhood Deprivation Index (NDI) ( Messer et al. 2006 ) Although the use of data reduction techniques to generate these composite indexes provides a comprehensive assessment of the neighborhood SES, the results from these studies are sometimes difficult to interpret because no indiv idual neighborhood characteristics reliably predictive of the outcome can be identified. To address these limitations, we obtained data of census tract level neighborhood SES characteristics from the 2006 2010 American Community Survey and linked them to t he Florida Vital Statistics Birth Record dataset to examine the association between the neighborhood SES and HDP among all eligible women residing in Florida and with a conception date between January 1, 2005 and December 31, 2007. In addition to assessing neighborhood SES by using a composite index (i.e.

PAGE 46

46 NDI), we further investigated the association between individual neighborhood SES characteristic and HDP using the LASSO model. Materials and Methods Study Sample Birth record data were obtained from the Bureau of Vital Statistics, Office of Health Statistics and Assessment, Florida Department of Health (DOH, http://www.floridahealth.gov/cert ificates/ certificates/ Jacksonville, Florida). All registered live births in Florida, USA between January 1, 2005 and December 31, 2008 (n=917,788) were included in the data. Women with residential addresses outside Florida (n=4,632) were excluded. Mothe geocoded by Florida Department of Health using ArcGIS v10.1, and 864,247 records (94.6%) were successfully geocoded. The remaining addresses were geocoded using the Google Maps application programming inter face (API). A total of 913,048 records (99.9%) were successfully geocoded. The census tracts corresponding to the geocoded Women whose residential address could not be geocoded we re excluded (n=108). To avoid fixed cohort bias ( Strand et al. 2011 ) women were included based on their conception date instead of delivery date. Among the 913,048 women who delivered during 2005 2008, a total of 691,011 women had the conception dates between January 1, 2005 and December 31, 2007. In addition, women were excluded if they had non singleton deliveries (n=21,609), pre pregnancy h ypertension (n=10,590), births with birthweights <500 g or >5000 g (n=621), or with a gestational age <26 weeks (n=2,662).

PAGE 47

47 Outcome Assessment During the collection of Vital Statistics Birth Record data, the medical history of each woman w as checked by th e staff, and diagnoses of the mutually exclusive categories of HDP ( pre pregnancy hypertension, gestational hypertension or preeclampsia, and eclampsia ) were abstracted for data analyses Gestational hypertension was determined as the development of hypert ension after 20 weeks of pregnancy; preeclampsia was defined as the new onset of hypertension and proteinuria after 20 weeks of gestation. Eclampsia was determined by the onset of convulsions. Similar to previous environmental studies on HDP, the restricte d definition of HDP was used in this study, which only includes gestational hypertension, preeclampsia, and/or eclampsia ( Hu et al. 2014 ) Neighborhood Socioeconomic Status Assessment Information on the neighborhood SES characteristics were obtained from the 2006 2010 American Community Survey (ACS) using the ACS 5 year API through the tract level characteristics covering seven different domains of n eighborhood SES were obtained: poverty domain (i.e. percent of households in poverty, percent of households earning <$30,000 per year, and percent of households with no vehicle), occupation domain (i.e. percent of males in management, percent of males in p rofessional occupations, percent of females in management, and percent of females in professional occupations), housing domain (i.e. percent of rented housing, percent of vacant housing, percent of renter or owner costs in excess of 50% of income, and medi an household value), employment domain (i.e. percent unemployed, and percent of males no longer in work force), education domain (i.e. percent earning less than a high school education), racial composition domain (i.e.

PAGE 48

48 percent of non Hispanic blacks), and residential stability domain (i.e. percent in same residence in the last year, and percent residents 65 years and above). Principal component analysis was conducted to combine these indicator variables at the census tract level into the standardized Neighb orhood Deprivation Index (NDI) developed by Messer et al. ( 2006 ) using the first factor loadings. The NDI was categorized into quartiles in census tract level. Women with missing NDI were excluded in the analyses ( n=3,581). A total of 651,948 women were included in the analyses. Covariates Maternal age, race/ethnicity, education, marital status, pregnancy smoking status, pre pregnancy body mass index (BMI), season and year of conception were obtained directly from t he births records. We categorized maternal age at delivery into six groups, with 5 year increments for women aged 20 40 years old as well as two additional categories: non Hispanic Whit e, non Hispanic Black, Hispanic, and others. In addition, dichotomous variables were used to indicate marital status and pregnancy smoking status. Maternal education was categorized as: high school. Pre pregnan cy BMI was categorized to four groups: underweight (<18.5), normal (18.5 24.9), overweight (25 spatially linked to the urbanity data and used to determine whether they resided in urban or rural areas. Season [warm (June November) or cool (December May)] and year (2005, 2006, or 20 07) of conception were also treated as categorical variables.

PAGE 49

49 Statistical Analyse s Distribution of categorical covariates and exposures between women with HDP and those without HDP were examined. Two sets of analyses were conducted to examine the association between neighborhood SES and HDP. Firstly, we used logistic regression models to determine the association b etween NDI and HDP. Both an unadjusted model and an adjusted model controlling for maternal age, race/ethnicity, education, marital status, pregnancy smoking, pre pregnancy BMI, urbanity, season and year of conception were conducted. Odds ratios (ORs) and 95% confidence interval were reported. Secondly, we determined individual neighborhood SES characteristics that were predictive of HDP using a two stage procedure. Neighborhood SES characteristics that were found to be predictive of HDP were screened first by using the logistic regression model with an penalty (LASSO). The LASSO can improve the prediction accuracy and interpretability of the statistical model by performing both variable selections and regularizations ( Tibshirani 1996 ) The tuning parameter was determined by a 10 fold cross validation based on the c statistics (AUC). Variables with nonzero coefficient s were then refitted using the unpenalized logistic regress ion model to avoid bias and approximation errors. All statistical analyses were conducted using R 3.2.2. Sensitivity Analyses To account for the multilevel structure of the data and potential correlations at census tract level, we performed sensitivity ana lyses using mixed effects models with random intercept for each census tract. We compared the original results to the sensitivity analyses to check whether the potential correlations at census tract level have influenced the observed effects.

PAGE 50

50 Results Amon g the 651,948 women included in the analyses 31,126 (4.8%) women had HDP, and 609,738 had complete data for all covariates (n = 29,062 with HDP). Table 3 1 shows the distribution of NDI and maternal characteristics by HDP status. Women with HDP were more likely than those without HDP to live in a neighborhood with higher NDI, to be with higher pre pregnancy BMI, and to belong to non Hispanic Black racial/ethnic categories. In addition, women with HDP were less likely than women without HDP to be within the age between 20 to 34 years, to be married, to smoke during pregnancy, and to live in urban areas. Figure 3 1 shows the spatial distributions of NDI by census tracts in the study area. Table 3 2 shows unadjusted and adjusted ORs from the first set analyse s on NDI. In the unadjusted model, compared with women living in neighborhoods with NDI in the lowest quartile, those living in neighborhoods with a higher NDI had increased odds of HDP (OR Quartile2 : 1.23, 95% CI: 1.18, 1.28; OR Quartile3 : 1.26, 95% CI: 1.2 1, 1.30; OR Quartile4 : 1.35, 95% CI: 1.31, 1.40). Attenuated but consistent ORs were observed in the adjusted model (OR Quartile2 : 1.15, 95% CI: 1.10, 1.20; OR Quartile3 : 1.14, 95% CI: 1.10, 1.19; OR Quartile4 : 1.19, 95% CI: 1.14, 1.24). Table 3 3 shows the r esults from the second set of analyses on individual neighborhood SES characteristics. In the first variable screening stage, three neighborhood characteristics had zero coefficient s from the LASSO, and they were excluded from the unpenalized Logistic mode l. The results from the unpenalized Logistic model showed that compared with women living in neighborhoods with a low proportion of females in professional occupations, those living in neighborhoods with a high proportion of females in professional occupat ions had decreased odds of HDP

PAGE 51

51 (OR: 0.98, 95% CI: 0.96, 0.99 for per 10 percent increase). In addition, women living in neighborhoods with higher residential stability (OR: 0.96, 95% CI: 0.94, 0.98 for per 10 percent increase) and higher median household v alue had decreased odds of HDP (OR: 0.95, 95%CI: 0.94, 0.96 for each 50,000 dollar increase). On the other hand, women living in neighborhoods with higher percentages of non Hispanic Blacks had increased odds of HDP (OR: 1.02, 95% CI: 1.01, 1.03 for each 1 0 percent increase). No statistically significant association was found between HDP and other individual ne ighborhood SES characteristics. Table 3 4 shows the comparisons between the results from the original analyses and the sensitivity analyses. Consiste nt results were observed in the sensitivity analyses using the mixed effects model. The intraclass correlation coefficient (ICC) is 0.059 indicating a very small effect of clusters Discussion In this retrospective cohort study, we found that women living in neighborhoods with lower SES had higher odds of HDP. This association is independent of individual level SES and other potential confounders such as pre pregnancy BMI and pregnancy smoking status. Further analyses on the individual neighbo rhood SES characteristics found that the percentages of females in professional occupations and non Hispanic blacks, median household value, and residential stability of neighborhoods are predictive of HDP. These findings add to the evidence linking neighb orhood SES to HDP. The results from our analyses are consistent with previous studies. The neighborhood SES is an important health determinant, and living in a disadvantaged neighborhood has been associated with behaviors such as perinatal substance use a nd

PAGE 52

52 gambling ( Finch et al. 1999 ; Welte et al. 2004 ) health intermediates such as partner violence and pediatric injury ( Cunradi et al. 2000 ; Shenassa et al. 2004 ) and health outcomes including hypertension and cardiovascular disease ( Cozier et al. 2007 ; Cubbin et al. 2000 ; Cubbin et al. 2006 ; Morenoff et al. 2007 ; Mujahid et al. 2008 ; Roux et al. 2001 ) cancer incidence ( Yost et al. 2001 ) and excess mortality ( Doubeni et al. 2012 ; Jaffe et al. 2005 ) A recent meta analysis of seven studies including 2,579,032 pregnancies found that compared with women living in neighborhoods with the least deprived quint ile, those living in the most deprived neighborhood quintile h ave significantly higher rates of preterm birth, small for gestational age, and stillbirth ( Vos et al. 2014 ) Similar to the association identified here, Vinikoor Imler et al. ( 2012 ) reported a positive association between neighborhood deprivation and HDP in the US. Another study in Norway also found that women living low income areas had higher rates of preeclampsia than women living in high income areas ( Clausen et al. 2006 ) Our study has several strengths. First, this study included a large number of women across Florida, and a number of confounders of HDP were adjusted in the analysis. Second, in addition to assessing neighborhood SES using a composite index (i.e. NDI), we f urther assessed individual neighborhood SES characteristics that are predictive of HDP, and four individual characteristics were identified. These findings provide insights into the important characteristics of SES that may inform targeted interventions in the future for vulnerable population subgroups, such as pregnant women living in a low SES neighborhood with a residential instability. Besides these strengths, several limitations need to be noted. Firstly, we assessed neighborhood SES characteristics u sing the 2006 2010 ACS 5 year data.

PAGE 53

53 Secondly, information on residential history or daily mobility was not available. These limitations may lead to non differential misclassification bias. In addition, although a number of confounders have been included in this study, we were not able to account for other potential confounders such as diet ( Dubowitz et al. 2008 ) Future studies are needed to confirm our findings with more detailed informatio n to address these limitations. Conclusion Using Florida birth vital statistics records, we found that women living in neighborhoods with a lower SES index had higher odds of H DP than those living in high SES neighborhoods. Several individual neighborhood SES characteristics such as a high residential instability, a low proportion of females working in professional occupations, low median household values, and a high proportion of non Hispanic Black residents were found to be predictive of HDP. Future studies are warranted to confirm the findings.

PAGE 54

54 Table 3 1. Neighborhood Deprivation Index (NDI) and maternal characteristics by hypertensive disorders of pregnancy (HDP) status among women with conception date during 2005 2007 in Florida, USA. Maternal Characteristics HDP (n=31,126) n (%) No HDP (n=620,822) n (%) Total (n=651,948) n (%) Neighborhood Deprivation Index Quartile 1 [ 7.22, 1.66] 4,418 (14.2) 109,000 (17.6) 113,418 (17.4) Quartile 2 ( 1.66, 0.18] 6,985 (22.4) 140,350 (22.6) 147,335 (22.6) Quartile 3 ( 0.18, 1.39] 8,460 (27.2) 166,280 (26.8) 174,740 (26.8) Quartile 4 (1.39, 11.10] 11,263 (36.2) 205,192 (33.1) 216,455 (33.2) Maternal age (years) <20 4,063 (13.1) 67,001 (10.8) 71,064 (10.9) 20 24 8,228 (26.4) 163,179 (26.3) 171,407 (26.3) 25 29 7,915 (25.4) 169,747 (27.3) 177,662 (27.3) 30 34 6,209 (19.9) 133,823 (21.6) 140,032 (21.5) 35 39 3,661 (11.8) 709,300 (11.4) 74,591 (11.4) 1,050 (3.4) 16,116 (2.6) 17,166 (2.6) Missing 0 (0.0) 26 (0.0) 26 (0.0) Race/ethnicity Non Hispanic White 14,545 (46.7) 278,768 (44.9) 293,313 (45.0) Non Hispanic Black 7,363 (23.7) 108,196 (17.4) 115,559 (17.7) Hispanic 1,906 (6.1) 49,371 (8.0) 51,277 (7.9) Others 7,312 (23.5) 184,473 (29.7) 191,785 (29.4) Missing 0 (0.0) 14 (0.0) 14 (0.0) Maternal education High school 14,756 (47.4) 296,915 (47.8) 311,671 (47.8) Missing 148 (0.5) 3,332 (0.5) 3,480 (0.5) Marital status Married 16,040 (51.5) 341,332 (55.0) 357,372 (54.8) Not married 15,082 (48.5) 279,395 (45.0) 294,477 (45.2) Missing 4 (0.0) 95 (0.0) 99 (0.0) Smoking during pregnancy No 28,692 (92.2) 567,672 (91.4) 596,364 (91.5) Yes 2,407 (7.7) 52,703 (8.5) 55,110 (8.5) Missing 27 (0.1) 447 (0.1) 474 (0.1) Pre pregnancy BMI Underweight (<18.5) 772 (2.5) 31,010 (5.0) 31,782 (4.9) Normal (18.5 24.9) 10,511 (33.8) 303,492 (48.9) 314,003 (48.2) Overweight (25.0 29.9) 7,875 (25.3) 141,065 (22.7) 148,940 (22.8) 10,017 (32.2) 107,565 (17.3) 117,582 (18.0) Missing 1,951 (6.3) 37,690 (6.1) 39,641 (6.1) Urbanity Urban 28,377 (91.2) 573,257 (92.3) 601,634 (92.3) Rural 2,749 (8.8) 47,565 (7.7) 50,314 (7.7) Season of conception Warm 15,212 (48.9) 304,984 (49.1) 320,196 (49.1) Cool 15,914 (51.1) 315,838 (50.9) 331,752 (50.9) Year of conception 2005 9,980 (32.1) 204,496 (32.9) 214,476 (32.9) 2006 10,540 (33.9) 211,208 (34.0) 221,748 (34.0) 2007 10,606 (34.1) 205,118 (33.0) 215,724 (33.1)

PAGE 55

55 Table 3 2 ORs for risk of hypertensive disorders of pregnancy (HDP) by NDI among women with conception date during 2005 2007 in Florida, USA. Unadjusted Model Adjusted Model a NDI n(HDP/Total) b OR (95% CI) n(HDP/Total) b OR (95% CI) Quartile 1 [ 7.22, 1.66 ] 4,418/113,418 Reference 4,125/105,981 Reference Quartile 2 ( 1.66, 0.18 ] 6,985/147,335 1.23(1.18, 1.28) 6,599/139,426 1.15(1.10, 1.20) Quartile 3 ( 0.18, 1.39 ] 8,460/174,740 1.26(1.21, 1.30) 7,924/163,949 1.14(1.10, 1.19) Quartile 4 (1.39, 11.1 0 ] 11,263/216,455 1.35(1.31, 1.40) 10,414/200,382 1.19(1.14, 1.24) a Adjusted for maternal age, race, education, marital status, pregnancy smoking status, pre pregnancy BMI, urbanity, season of conception, and year of conception. b Women with complete data for all covariates.

PAGE 56

56 Table 3 3 ORs for risk of hypertensive disorders of pregnancy (HDP) by individual neighborhood SES characteristics among women with conception date during 2005 2007 in Florida, USA. LASSO Model Unpenalized Logistic Model Neighborhood SES Characteristics Coefficient a OR (95% CI) a Poverty Percent of households in poverty 0.99(0.96, 1.01) Percent of households earning <$30,000 per year 0.99(0.97, 1.01) Percent of households with no vehicle 0.98(0.95, 1.01) Occupation Percent of males in management 1.01(0.99, 1.03) Percent of males in professional occupations 0 Percent of females in management 1.01(0.99, 1.04) Percent of females in professional occupations 0.98(0.96, 0.99) Housing Percent of rented housing 0.99(0.98, 1.01) Percent of vacant housing 1.00(0.98, 1.01) Percent of renter or owner costs in excess of 50% of income 1.00(0.99, 1.00) Median household value (each 50,000 dollars increase) 0.95(0.94, 0.96) Employment Percent unemployed 0.98(0.93, 1.03) Percent of males no longer in work force 0 Education Percent earning less than a high school education 1.00(0.98, 1.02) Racial composition Percent of non Hispanic blacks 1.02(1.01, 1.03) Residential stability Percent in same residence in the last year 0.96(0.94, 0.98) Percent residents 65 years and above 0 a Adjusted for maternal age, race, education, marital status, pregnancy smoking status, pre pregnancy BMI, urbanity, season of conception, and year of conception. Coefficient, OR, and 95% CI for each 10 percent increase in neighborhood SES characteristics or each 50,000 dollar increase in median household value.

PAGE 57

57 Table 3 4 Sensitivity analyse s: fixed effects model vs. mixed effects model n(HDP/Total) a Fixed effects Model b Mixed effects Model c NDI Quartile 1 [ 7.22, 1.66 ] 4,418/113,418 Reference Reference Quartile 2 ( 1.66, 0.18 ] 6,985/147,335 1.15(1.10, 1.20) 1.16(1.15, 1.21) Quartile 3 ( 0.18, 1.39 ] 8,460/174,740 1.14(1.10, 1.19) 1.15(1.10, 1.21) Quartile 4 (1.39, 11.10 ] 11,263/216,455 1.19(1.14, 1.24) 1.21(1.15, 1.27) a Women with complete data for all covariates. b Adjusted for maternal age, race, education, marital status, pregnancy smoking status, pre pregnancy BMI, urbanity, season of conception, and year of conception. c Mixed effects model with random intercept for each census tract and a djusted for maternal age, race, education, marital status, pregnancy smoking status, pre pregnancy BMI, urbanity, season of conception, and year of conception.

PAGE 58

58 Figure 3 1 Neighborhood Deprivation Index (NDI) by census tracts in Florida, USA.

PAGE 59

59 CHAPTER 4 RACIAL DISPARITIES IN HYPERTENSIVE DISORDERS OF PREGNANCY MEDIATED BY NEIGHBORHOOD ENVIRONMENTAL FACTORS Introduction H ypertensive disorders of pregnancy (HDP) are common pregnancy complications, characterized by high blood pressure, usually after 20 weeks of gestation when the change in blood volume lead s to higher stress on the cardiovascular system ( Yoder et al. 2009 ) HDP have been linked to increased neonatal and maternal morbidity and mortality ( Allen et al. 2004 ; Bauer and Cleary 2009 ; Bellamy et al. 2007 ; Duley 2009 ; Lo et al. 2013 ; Wang et al. 2012 ; Wu et al. 2009 ) The burden of HDP falls disproportionately on African Americans ( Tanaka et al. 2007 ) A 10 year longitudinal study in New York State reported that 8.5% of African American women had HDP during pregnancy, compared with 5.5% and 6.2% in White and Hispanic women, respectively ( Tanaka et al. 2007 ) More importantly, Tanaka et al. found an increasing trend of racial disparities in HDP that appear to be independent of the influences of other risk factors such as new paternity, multiple gestation, a nd extreme reproductive age. The State of Florida also had shown increasingly large racial disparities in HDP. Data from the Florida Vital Statistics Birth Records indicate that the racial disparities in HDP have been two times greater in 2012 compared wit h 2004. In 2004, 5.57% and 5.00% African American and White women had HDP during pregnancy, respectively, while in 2012, 7.21% and 5.54% African American and White women had HDP, respectively. Given an increase in racial disparities in HDP, it is importan t to identify the underlying factors associated with this trend in order to inform the development of targeted and useful health policies and interventions to reduce both racial disparities

PAGE 60

60 and risks of HDP in Florida. Neighborhood environmental factors su ch as racial residential segregation, urbanity, neighborhood socioeconomic status (SES), and air pollution exposure are potential mediators for racial disparities in HDP since they are both associated with HDP risks and distribute unevenly between African Americans and non African Americans ( Kearney and Kiros 2009 ; Pollock and Vittas 1995 ; Stretesky and Hogan 1998 ; Williams and Collins 2001 ) However, few study has directly addressed how these neighborhood environmental factors contribute to the racial disparities in HDP. In this stud y, we used the Florida Vital Statistic Birth Record dataset to assess the mediation effects of neighborhood environmental factors (i.e. racial residential segregation, urbanity, neighborhood SES, and ozone exposure) on racial disparities in HDP among all e ligible women residing in Florida with conception dates between January 1, 2005 and December 31, 2007. Materials and Methods Study Sample We obtained the Florida Vital Statistics Birth Record dataset from the Bureau of Vital Statistics, Office of Health S tatistics and Assessment, Florida Department of Health ( http://www.floridahealth.gov/certificates/certificates/ Jacksonville, Florida), which include all registered live births in Flo rida, USA between January 1, 2005 and December 31, 2008 (n=917,788). Records with maternal residential addresses located outside Florida (n=4,632) were excluded. Maternal residential address at delivery was initially geocoded by Florida Department of Healt h (FDOH) using ArcGIS v10.1, and 864,247 records (94.6%) were successfully geocoded. We further geocoded the addresses that failed to be geocoded by FDOH using the Google Maps application

PAGE 61

61 programming interface (API), and a total of 913,048 records (99.9%) were successfully geocoded. We excluded women whose residential address could not be geocoded (n=108). The census tracts corresponding to the geocoded maternal addresses were then determined using the 2010 TIGER/Line Shapefiles. To avoid fixed cohort bias ( Strand et al. 2011 ) we included women based on their conception date instead of delivery date (n=691,011). We further excluded women with n on singleton deliveries (n=21,609), pre pregnancy hypertension (n=10,590), birth s with birthweight <500 g or >5000 g (n=621), gestational age <26 weeks (n=2,662), or with missing race/ethnicity (n=14). A total of 655,515 women were included. Outcome Assessment The diagnoses of pre pregnancy hypertension, gestational hypertension or preeclampsia, and eclampsia were obtained during the collections of Vital Statistics Birth Record data by the staff. Gestational hypertension was determined as the developm ent of hypertension after 20 weeks of pregnancy, and preeclampsia was defined as the new onset of hypertension and proteinuria after 20 weeks of gestation. Eclampsia was determined by the onset of convulsions. Similar to previous environmental studies on H DP, we used the restricted definition of HDP which only includes gestational hypertension, preeclampsia, and/or eclampsia. Urbanity Assessment Urbanity was assessed using data from the 2010 Census. The US Census Bureau defines territories as urban areas if they encompass at least 2,500 people, and addresses were linked to the urbanity data and used to determine whether they resided in urban or rural areas.

PAGE 62

62 Neighborhood Socioeconomic Status Assessment Information on neighborhood SES characteristics was obtained from the 2006 2010 American Community Survey (ACS). A total of 17 census tract level characteristics covering seven different domains of neighborhood SES were obta ined: poverty (i.e. percent of households in poverty, percent of households earning <$30,000 per year, and percent of households with no vehicle), occupation (i.e. percent of males in management, percent of males in professional occupations, percent of f emales in management, and percent of females in professional occupations), housing (i.e. percent of rented housing, percent of vacant housing, percent of renter or owner costs in excess of 50% of income, and median household value), employment (i.e. perc ent unemployed, and percent of males no longer in work force), education (i.e. percent earning less than a high school education), racial composition (i.e. percent of African Americans), and residential stability (i.e. percent in same residence in the l ast year, and percent residents 65 years and above). We conducted the principal component analysis to combine information from these indicators at the census tract level into the standardized Neighborhood Deprivation Index (NDI) developed by Messer et al. ( 2006 ) using the first factor loadings. Ozone Exposure Assessment We obtained ambient ozone (O 3 Environmental Public Health Tracking Network ( U.S. EPA 2014 ) which were assessed using the hierar chical Bayesian space time statistical model (HBM) to combine the Air Quality System monitoring data with the Community Multiscale Air Quality modeled data. The data is available during 2001 2008 with a daily temporal resolution and a spatial resolution of 12km12km across the continental areas in the US ( McMillan et al.

PAGE 63

63 2010 ) The HBM data were spatiotemporally linked to each woman to calculate O 3 exposure during the first two trimesters (1 26 weeks) of t he pregnancy, which were determined by gestational age and delivery date of each woman. Racial Residential Segregation Assessment The spatial measure of neighborhood level racial isolation developed by Anthopolos et al. ( 2011 ) was used to assess rac ial residential segregation. The Census block group level population count data by race were obtained from the 2006 2010 American Community Survey. Let denote the entire region of Florida, denote the Census block group, and denote the area of block group Let indicate mutually exclusive race/ethnicity groups. was then defined as the total population count in block group and let denote the total population count of group in block group We then used to denote a first order adjacency matrix with if and share a boundary and otherwise. was set to a constant to indicate group. The racial isolation of group in block group can then be defined as: It can be interpreted as the average percentage of group in the local environment of as defined by the adjacency matrix. Bordering block groups located outside the State of Florida were also included to correct for the edge effects. To be consistent with previous studies ( Anthopolos et al. 2011 ; Anthopolos et al. 2014 ) the constant was set to 1.

PAGE 64

64 Covariates Maternal age, race/ethnicity, education, marital status, pregnancy smoking status, pre pregnancy body mass index (BMI), season and year of conception were obtained directly from the births records. We categorized maternal age at delivery into six groups, with 5 year increments for women aged 20 40 years old as well as two additional categories: African Americans and non African Americ ans. In addition, dichotomous variables were used to indicate marital status and pregnancy smoking status. Maternal education was categorized as: high school. Pre pregnancy BMI was categorized to four groups: u nderweight (<18.5), normal (18.5 24.9), overweight (25 Season [warm (June November) or cool (December May)] and year (2005, 2006, or 20 07) of conception were also treated as categorical variables. Statistical Analyse s Distributio n of race/ethnicity, mediators, and covariates between women with HDP and those without HDP were examined. To assess the potential correlations at census tract level, the intraclass correlation coefficient (ICC) was calculated using a mixed effects model with random intercept for each census tract and fixed slopes for exposure, mediators, and covariates We obtained an ICC of 0.06, indicating a very small effect of clusters. Therefore, fixed effects models were used in the anal yses. We first fitted an unadjusted logistic regression model to investigate racial differences in odds of HDP. Odds ratios (ORs) and 95% confidence interval were reported. An adjusted model controlling for maternal age, education, marital status, pregnanc y smoking, pre pregnancy BMI, season and year of conception were then used. The

PAGE 65

65 Inverse Odds Ratio Weighting (IORW) method developed by Nguyen et al. ( 2015 ) was used to assess the mediation effects of neighborhood environmental fac tors. Figure 4 1 shows the directed acyclic graph (DAG) depicting the relationship between the exposure (i.e. race/ethnicity), the outcome (i.e. HDP), the mediators (i.e. neighborhood environmental factors including ozone urbanity neighb orhood SES and racial residential segregation ), and the confounders (i.e. pre pregnancy maternal characteristics). The IORW can be used with any standard regression models and can be easily implemented in standard statistical software with weig hted regression. It avoids the difficulties of specifying a model for the joint conditional density of multiple mediators by using weights to indicate the relationship between exposure and multiple mediators. The IORW method was implemented to estimate the direct and indirect effects as follows (Figure 4 2): (1) model the exposure conditional on the mediators and the confounders, (2) compute an IORW weight for the exposed group by taking the inverse of the predicted odds from step 1 and set the IORW weight to 1 for the reference group, (3) fit a weighted model for the outcome conditional on the exposure and the covariates using the IORW weight obtained in step 2 to estimate the direct effect of exposure, and then (4) estimate the total effect of exposure by fitting a standard model for the outcome conditional on the exposure and the covariates The indirect effects of the exposure were then obtained by subtracting the direct effects from the total effects, and standard errors for the estimated direct and indi rect effects were derived by using 1,000 bootstrap replications. All statistical analyses were conducted using R 3.2.2 (R Foundation for Statistical Computing, Vienna, Austria).

PAGE 66

66 Results Among the 655,515 women included in the analyses, 31,362 (4.8%) women had HDP, and 609,738 had complete data for all mediators and covariates (n = 29,062 with HDP). Table 4 1 shows the distribution of maternal characteristics by HDP status. African American women had a higher prevalence of HDP (6.4%) compared with non Africa n American women (4.4%). In addition, HDP cases were more likely than those without HDP to live in a neighborhood with higher racial residential segregation score and NDI, to have higher ozone exposure during the first two trimesters of pregnancy to live in rural areas, and to be with higher pre pregnancy BMI Furthermore, women with HDP were less likely than those without HDP to be between 20 to 34 years of age to be married, and to smoke during pregnancy. Table 4 2 shows the unadjusted and adjusted ORs between race/ethnicity and HDP. In the unadjusted model, African American women had increased odds of HDP compared with non African American women (OR: 1.47, 95% CI: 1.43, 1.51). Consistent results were observed in the adjusted model after controlling for maternal age, education, marital status, pregnancy smoking status, pre pregnancy BMI, and season and year of conception (OR: 1.23, 95% CI: 1.19, 1.27). Table 4 4 1) well as the estimated proportion of total effects mediated by neighborhood environmental factors. For African American women, the estimates of the total OR of 1.23 (95% CI: 1.19, 1.27) was decomposed into a direct OR of 1.14 (95% CI: 1.09, 1.20) and an indirect OR of 1.07 (95% CI: 1.03, 1.12), with 35% (95% CI: 15%, 55%) of

PAGE 67

67 the total effects mediated through neighborhood environmental factors including ozone exposure, urb anity, neighborhood SES, and racial residential segregation. Discussion To our knowledge, this is the first population based study to examine the mediating effects of neighborhood environment on racial disparities in HDP. We found that African American wom en had significantly higher odds to have HDP compared with non African American women. More importantly, the results from mediation analyses show that 35% of racial disparities in HDP can be explained by the differential exposure to neighborhood environmen tal factors, including ozone exposure, urbanity, neighborhood SES, and racial residential segregation, between African American women and non African American women. Neighborhood environment is an important health determinant that has been recognized for a long time ( Malmstrm et al. 1999 ) However, limited stud ies have been done to assess how it may contribute to the increasingly large racial disparities in HDP. Our findings suggest that the improvement on neighborhood environmental factors such as ozone exposure, urbanity, neighborhood SES, and racial residential segreg ation can greatly reduce the racial disparities in HDP. Racial residential segregation has been regarded as the fundamental cause of racial disparities in health ( Williams and Collins 2001 ) In the US, it usually refers to the residential separation of African American neighborhoods from non African Americans. In spite of the Civil Rights Act in 1968 which made overt discrimination in housing markets illegal racial residential segregation persisted in forms such as racial steering and discriminatory lending ( Mendez et al. 2013 ) Althou gh no study has been done to assess the mediating effects of racial residential segregation on racial disparities in

PAGE 68

68 HDP, various studies have linked racial residential segregation to adverse birth outcomes in the US ( Anthopolos et al. 2011 ; Anthopolos et al. 2014 ; Bell et al. 2006 ; Grady 2006 ; Grady and Ramrez 2008 ; Osypuk and Acevedo Garcia 2008 ) In a parallel literature, racial resi dential segregation has been associated with hypertension in the general population ( Kershaw et al. 2011 ) On the other hand, neighborhood environmental factors such as ozone exposure and neighborhood SES have been linked to HDP and many adverse health comes and b ehaviors ( Cozier et al. 2007 ; Cubbin et al. 2000 ; Cubbin et al. 2006 ; Cunradi et al. 2000 ; Doubeni et al. 2012 ; Finch et al. 1999 ; Hu et al. 2014 ; Jaffe et al. 2005 ; Morenoff et al. 2007 ; Mujahid et al. 2008 ; Roux et al. 2001 ; Shenassa et al. 2004 ; Welte et al. 2004 ; Yost et al. 2001 ) Studies suggest ed that the average residential context of African American communities is worse than the worst residential context for Whites ( Sampson and Wilson 1995 ; Williams and Collins 2001 ) Th is evidence support s our findings that neighborhood environment is an important mediator for racial disparities in HDP. More efforts should be given on neighborhood level interventio ns such as the smart growth by t he American Planning Association ( Duany et al. 2010 ) the community revitalization initiatives ( Anderson et al. 2 003 ; Williams 2007 ) ( Alfieri 2013 ) Our study has several strengths. First, by using the Florida Vital Statistics Birth Record dataset, we are able to i nclude all eligible women who delivered baby in the study period. Second, we assessed how neighborhood environment may contribute to racial disparities in HDP by mediation analyses. Mediation analysis is increasingly used recently in health disparity studi es given its advantages to separate and quantify the

PAGE 69

69 direct and indirect effects ( Bennett et al. 2012 ; Chatterjee et al. 2011 ; Mugavero et al. 2009 ) Identifying the mediators will not only improve the understandings of the underlying causal mechanisms, but also help to design and implement better intervention strategies. Besides these strengths, several limitations need to be noted. Firstly, information on residential history or daily mobility was not available, which may have led to non differential misclassification bias. Secondly, only a limited number of n eighborhood environmental factors were examined in this study. In addition, although a number of confounders have been included in this study, we were not able to account for other potential confounders such as diet ( Dubowitz et al. 2008 ) In spite of these limitations, this study provides important evidence that an unneglectable proportion of racial disparities in HDP was mediated by neighborhood environment. Future interventional studies are warranted to improve the neighborhood environment for African Americans and to reduce the racial disparities in HDP. Conclusion Using Florida birth vital statistics records, we found that 35 % of the racial disparities in HDP can be explained by differential exposure to neighborhood environmental factors such as ozone, urbanity, neighborhood SES, and racial residential segregation. Interventional studies are needed to improve neighborhood environment and to reduce racial disparities in HDP.

PAGE 70

70 Table 4 1. Maternal characteristics by hypertensive disorders of pregnancy (HDP) status among women with conception date during 2005 2007 in Florida, USA. Maternal Characteristics HDP (n=31,362) n (%) /mean(sd) No HDP (n=624,153) n (%) /mean(sd) Total (n=655,515) n (%) /mean(sd) Race/ethnicity African American 7,426 (23.7) 108,974 (17.5) 116,400 (17.8) Non African American 23,936 (76.3) 515,179 (82.5) 539,115 (82.2) Racial Residential Segregation Score 0.21 (0.23) 0.19 (0.21) 0.19 (0.21) Neighborhood Deprivation Index 0.79 (2.36) 0.56 (2.37) 0.57 (2.37) Ozone exposure during first two trimesters 39.06 (5.20) 38.61 (5.33) 38.63 (5.33) Urbanity Urban 28,609 (91.2) 576,479 (92.4) 605,088 (92.3) Rural 2,753 (8.8) 47,674 (7.6) 50,427 (7.7) Maternal age (years) <20 4,089 (13.0) 67,386 (10.8) 71,475 (10.9) 20 24 8,325 (26.5) 164,441 (26.3) 172,766 (26.4) 25 29 7,973 (25.4) 170,778 (27.4) 178,751 (27.3) 30 34 6,244 (19.9) 134,265 (21.5) 140,509 (21.4) 35 39 3,678 (11.7) 71,101 (11.4) 74,779 (11.4) 1,053 (3.4) 16,156 (2.6) 17,209 (2.6) Missing 0 (0.0) 26 (0.0) 26 (0.0) Maternal education High school 14,853 (47.4) 298,300 (47.8) 313,153 (47.8) Missing 149 (0.5) 3,343 (0.5) 3,492 (0.5) Marital status Married 16,159 (51.5) 343,191 (55.0) 359,350 (54.8) Not married 15,199 (48.5) 280,866 (45.0) 296,065 (45.2) Missing 4 (0.0) 96 (0.0) 100 (0.0) Smoking during pregnancy No 28,916 (92.2) 570,818 (91.5) 599,734 (91.5) Yes 2,419 (7.7) 52,886 (8.5) 55,305 (8.4) Missing 27 (0.1) 449 (0.1) 476 (0.1) Pre pregnancy BMI Underweight (<18.5) 776 (2.5) 31,191 (5.0) 31,967 (4.9) Normal (18.5 24.9) 10,589 (33.8) 305,035 (48.9) 315,624 (48.1) Overweight (25.0 29.9) 7,942 (25.3) 141,826 (22.7) 149,768 (22.8) 10,093 (32.2) 108,143 (17.3) 118,236 (18.0) Missing 1,962 (6.3) 37,958 (6.1) 39,920 (6.1) Season of conception Warm 15,327 (48.9) 306,630 (49.1) 321,957 (49.1) Cool 16,035 (51.1) 317,523 (50.9) 333,558 (50.9) Year of conception 2005 10,077 (32.1) 205,621 (32.9) 215,698 (32.9) 2006 10,611 (33.8) 212,398 (34.0) 223,009 (34.0) 2007 10,674 (34.0) 206,134 (33.0) 216,808 (33.1)

PAGE 71

71 Table 4 2 ORs of hypertensive disorders of pregnancy (HDP) mediated by neighborhood environmental factors among women with conception date during 2005 2007 in Florida, USA. Race/ethnicity Non African American African American Unadjusted model n(HDP/Total) b 23,936/539,115 7,426/116,400 OR (95% CI) Reference 1.47(1.43, 1.51) Adjusted model n(HDP/Total) b 22,270/502,504 6,792/107,234 Total effect Reference 1.23(1.19, 1.27) Direct effect Reference 1.14(1.09, 1.20) Indirect effect Reference 1.07(1.03, 1.12) Proportion mediated 0.35(0.15, 0.55) a Adjusted for maternal age, education, marital status, pregnancy smoking status, pre pregnancy BMI, season of conception, and year of conception. b Women with complete data for all mediators and covariates.

PAGE 72

72 Figure 4 1 Directed acyclic graph. Neighborhood environmental factors including ozone, neighb orhood socioeconomic status (SES), and racial residential segregation (RRS) are hypothesized to mediate racial disparities in hypertensive disorder of pregnancy (HDP).

PAGE 73

73 Figure 4 2 Overview of the mediation analyses using the inverse odds ratio weighting (IORW) method.

PAGE 74

74 CHAPTER 5 CONCLUSIONS Summary of Research Objectives As outlined in Chapter 1, neighborhood environment may independently contribute to the development of HDP aside from other known risk factors. This dissertation aimed to address the knowledge gap in the literature by assessing the association between neigh borhood environment and HDP and examining how neighborhood environment may contribute to the racial disparities in HDP. To achieve these objectives, we firstly investigated the association between O 3 exposures during pregnancy and HDP and identified critic al exposure windows. Secondly, we examined the association between neighborhood SES and HDP using the standardized NDI. We also further identified individual neighborhood socioeconomic characteristics that are predictive of HDP using the Lasso model. Final ly, due to the large racial disparities in HDP between African Americans and non African Americans, we further assessed how O 3 exposure during pregnancy and neighborhood SES may contribute to racial disparities by mediation analyses using the Inverse Odds Ratio Weighting method. More specifically, we sought to answer the following research questions: 1. What are the effect estimates for the association between O 3 exposure during the three pre defined windows (i.e. the first trimester, the second trimester, and the first and second trimesters) of pregnancy and HDP? 2. What are the most critical exposure windows of O 3 during pregnancy for HDP? 3. What are the overall effect estimates for the association between neighborhood SES and HDP? 4. Which individual neighborhood socioeconomic status are predictive of HDP?

PAGE 75

75 5. What are racial disparities in HDP between African Americans and non African Americans? 6. What proportions of the racial disparities in HDP are contributed by neighborhood environmental factors including O 3 exposure during pregnancy neighborhood socioeconomic status, urbanity, and racial residential segregation? Accomplishments of this Dissertation This dissertation has accomplished the objectives outlined in Chapter 1 and successfully addressed the researc h questions above. The findings from this dissertation contribute to the limited literature on neighborhood environment and HDP. The following sections summarized and discussed the ac hievements of this dissertation Determination of Association between O 3 Exposure during the Three Pre defined Exposure Windows of Pregnancy and HDP In the state of Florida, O 3 is the air pollutant of the greatest concern, and recent studies suggested an overall positive association between O 3 exposure during pregnancy and HDP However, several limitations existed in previous studies such as the use of data from stationary air monitors which may introduce both information and selection bias. To address these limitations, in Chapter 2, we linked the Florida Vital Statistics Birt h Records data to the hierarchical Bayesian space time modelled O 3 data 3 during the three pre defined windows: the first trimester, the second trimester, and the first and second trimesters. We observed positive ass ociations between O 3 exposure and HDP during all these thre e pre defined exposure windows. Identification of Critical Windows of O 3 Exposure during Pregnancy for HDP Another significant limitation in previous studies on O 3 exposure and HDP is the lack of i dentifications of critical exposure windows. Therefore, in addition to the traditional logistic regression models performed to determine the overall association

PAGE 76

76 between O 3 exposure during pregnancy and HDP we further used the constrained distributed lag models to identify critical O 3 exposure windows for HDP in Chapter 2. We found that early pregnancy was the most critical window for O 3 exposure. Our findings contribu te to the literature since this is the first to identify critical O 3 exposure windows for HDP. Future studies can confirm these findings and further investigate the mechanisms underlying them such as the potential effects of O 3 on vasoconstriction in HDP. Determination of Association between Neighborhood SES and HDP Neighborhood SES is an imp ortant health determinant. While neighborhood SES has been found to be associated with both adverse birth outcomes and hypertension in the general population only a few studies examined the association between neighborhood SES and HDP. To answer the third research question, in Chapter 3, we generated the standardized Neighborhood Deprivation Index (NDI) based on seventeen census tract level SES characteristics, and we found that women living in neighborhoods with lower SES had higher odds of HDP after cont rolling for individual level SES and other confounders including pre pregnancy BMI and pregnancy smoking status Identification of Individual Neighborhood Socioeconomic Characteristics Predictive of HDP A big limitation of previous studies on neighborhood SES and HDP is the lack of evidence to identify individual neighborhood SES characteristics predictive of HDP. To address the fourth research question, we used the regularized logistic regression model with an penalty (LASSO) to determine predictive neighborhood SES characteristics, and we found that women living in neighborhoods with a high residential instability, a

PAGE 77

77 low proportion of females working in professional occupations, low median household values, and a high proportion of non His panic Blacks residents have higher odds of HDP. The identifications of these neighborhood socioeconomic characteristics that are predictive of HDP can be used to better inform community interventions in the future Identification of Racial Disparities in HDP between African Americans and Non African Americans In Chapter 4, we assessed the racial disparities in HDP. Among the 116,400 African American women included in the study, 7,426 (6.4%) had HDP. While only 23,936 (4.43%) out of the 539,115 non African American women developed HDP. Consistent results were found after adjusting for maternal age, education, marital status, pregnancy smoking status, pre pregnancy BMI, and season and year of conception African American women had a higher odds of HDP (OR: 1 .23, 95% CI: 1.19, 1.27) compared with non African American women. These findings are consistent with the racial disparities observed in other areas in US. Determination of the Proportions of Racial Disparities in HDP Contributed by Neighborhood Environmental Factors Including O 3 Exposure, Neighborhood SES, urbanity, and Racial Residential Segregation After the identification of racial disparities in HDP, we further assessed how neighborhood environmental factors examined in Chapter 2 and Chapter 3 as well as racial residential segregation may contribute to the higher odds of HDP observed among African American women. We used the inverse O dds Rati o Weighting method to perform the mediation analyses, and we found that 3 5 % of racial disparities in HD P can be explained by the differential exposure to neighborhood environmental factors including O 3 exposure during pregnancy neighborhood SES, urbanity, and racial residential s egregation

PAGE 78

78 Limitations As we discussed above, a major limitation of the cur rent study as well as existing literature is measurement errors in the outcome, exposure, and covariates. For example, most studies assessed air pollution exposure based on the residential addresses. Such method introduces substantial exposure misclassification bias. In this dissertation, although we used the HBM output to assess ozone exposure, substantial error s activ ity pattern. While studies using personal air monitors have been conducted, these studies usually have small sample sizes and take a long time and high expenses to be completed in the traditional epidemiological study settings (Jedrychowski et al. 2004) Such studies are hard to be scaled up, and more importantly, they may introduce selection bias that make the results hard to be generalizable. In addition, more detailed and innovative data sources are needed to assess neighborhood SES. In this disserta tion, we assessed neighborhood SES using the 5 year ACS data from 2006 2010. The ACS is the major data source for assessment of neighborhood SES in US currently. The ACS is a national survey using a series of monthly samples to produce estimates for differ ent size of areas and time periods. While the data provided a good estimate of neighborhood SES at census tract level in the 5 year period, it could not capture more detailed temporal changes in neighborhood SES (e.g. monthly or annual estimates). Furtherm ore, in this dissertation, we defined neighborhoods using census tracts. Although this is a routine practice in the field, better definitions of neighborhoods are required in future studies.

PAGE 79

79 Similarly more refined measurement of HDP is warranted in futu re studies using detailed clinical data accounting for timing of HDP onset and high quality blood pressure measurements. However, i n the traditional epidemiological study settings, there seems to be a tradeoff between measurement errors and selection bias es In studies with large sample sizes (e.g. utilizing vital statistics data or other large birth cohort), selection bias is minimal while substantial measurement errors may exist. On the other hand, in studies with small sample sizes (e.g. enrolling parti cipants from community or clinical settings), measurement errors can be minimized while substantial selection biases may exist. In a recent study on air pollution and HDP, Savitz et al. (2015) stated that we are quickly reaching the limits of benefiting from new studies of similar natures to those have done before. Innovative methods and data sources are needed to simultaneously reduce measurement errors and selection bias and to bring a new perspective to the topic. Another major limitation in the field is the lack of individual level interventions designed to reduce environmental risk factors of HDP. As the theoretical framework SEM s uggests, environmental risk factors are more amenable than individual risk factors because we can design interventions at both levels. However, the lack of individual level interventions on these environmental risk factors largely hinders the improvement o f racial disparities and the burden of HDP. The lack of interventions may be due to many reasons. Traditionally, epidemiological research starts with descriptive studies, and then move s on to etiological studies. Only after the confirmations of etiological effects of a risk factor, interventional studies can be finally achieved There is

PAGE 80

80 a long time lag between etiological studies and interventional studies, and i nnovative techniques are needed to combine etiological studies with interventional studies. Future Directions Use the Socio Ecological Model for Prevention Th e SEM considers the comple x relationship between individual and neighborhood environmental factors, allowing us to understand the range of factors that put pregnant women at risk for HDP. The overlapping half rings in Figure 1 1 illustrate how factors at one level influence factors at another level. Using the SEM as the theoretical framework, this dissertation identified several risk factors at the neighborhood environment level. Our findi ngs suggest that the improvement on these neighborhood environmental factors may greatly reduce the burden of HDP as well as the racial disparities in HDP. The SEM can also be used as a framework for prevention. In order to protect pregnant women from HDP it is necessary to act across multiple levels of the model at the same time. Environmental level. A number of environmental level interventions have been developed in the past decade to improve neighborhood environment, such as the smart growt h by t he Am erican Planning Association (Duany et al. 2010) the community revitalization initiatives (Anderson et al. 2003; Williams 2007) (Alfieri 2013) Future studies are warranted to evaluate the effects of these interventions as well as t o develop more preventive strategies using innovative technologies such as mHealth (mobile health) Individual level. As discussed above, there is a lack of individual level interventions to prevent HDP by reducing individual exposure to the neighborhood e nvironmental risk factors. The emergence of mHealth technologies provides a great

PAGE 81

81 opportunity for personalized, localized, and on demand health interventions (Riley et al. 2011) For example, the use of wearable environment trackers with indoor air purifier has t s exposure to air pollution (Ch en et al. 2015) More efforts are needed to design and integrate interventions at both individual factors identified in this dissertation and to prevent HDP. Use mHealth to Reduce Bias and Facilitate Translational Research In addition to prevention, mHealth technologies can also be used to s imultaneously reduce measurement error and selection bias as well as to integrate etiological studies with individual level interventions, which can greatly reduce the long time lag in the traditional translational research settings mHealth is defined as the research and practice of medicine and public health supported by mobile devices (Adibi 2015) These devices provide a wide range of functions to users including mobile communication, access to internet and multimedia contents, and millions of software applications. mHealth is a rapidly expanding area in healthcare that emerged only a few years ago (Blaya et al. 2010) and the recent developments of wearable devi ces and mobile sensors further increase s its potential to improve health research and practice. mHealth technologies offer many advantages to both health research and practice, including modernized automatic and semi automatic data collections (Walther et al. 2011) reduced cost s of participants recruitment and follow ups (Kumar et al. 2013) and reduced costs of healthcare (Gurol Urganci et al. 2012)

PAGE 82

82 By collecting data from mobile devices with wearable trackers such as the TZOA environment tracker and the Omron Project Zero w earable blood pressure monitor studies using mHealth technologies can obtain real time data from study participants remotely, which can be easily scaled up to enroll a large number of participants in the studies. The mHealth technologies can be combined w ith multistage probabilistic sampling methods to address both the measurement and selection bias simultaneously. Take the measurement of air pollution exposure as an example, real time exposure surface can be generated using exposure data streamed from in environment trackers. With a large number of participants carrying such trackers, the generated surface can provide both a great spatial coverage and an accurate estimate of the exposure. The real time exposure surface can also be used to guide environmental level as well as individual level interventions. For example, the real time exposure surface will enable policy makers to identify exposure sources more easily and lead to a more effective enforcement of environmental laws. On the other hand, at individual level, navigation systems can be designed to recommending routes for cyclists and pedestrians based on exposure levels in addition to the traditional time and distance metrics. S tudies fully utilizing the power of mHealth technologies can be used to simultaneously assess disease etiology and evaluate intervention efficacy. Recently, a European group developed a mHealth platform, mMamee, for assessing maternal environmental exposures (Karagiannaki et al. 2015) which is a potential prototype platform that can be used to integrate self reported data from parti cipants and real time data streams from monitors, wearable devices, and other data

PAGE 83

83 sources. The platform makes it possible to conduct large scale etiological research and individual level interventions in the near future Use Innovative Data Sources In co mbination with the mHealth technologies, innovative data sources such as satellite images are increasingly popular in the assessment of environmental exposures (Lillesand et al. 2014) It is anticipated that satellite images will cover nearly every place on Earth every day by 2017 (Draper 2016) Future studies are warranted to fully utilize such data to improve the assessment of environmental exposures.

PAGE 84

84 LIST OF REFERENCES Adibi S. 2015. Mobile health: A technology road map:Springer. Agyemang C, Vrijkotte T, Droomers M, Van der Wal M, Bonsel G, Stronks K. 2009. The effect of neighbourhood income and deprivation on pregnancy outcomes in amsterdam, the netherlands. Journal of epidemiology and community health 63:755 760. Alfieri AV. 2013. Paternalistic interventions in civil rights and poverty law: A case study of environmental justice. Michigan law review 112:2014 2013. Allen VM, Joseph K, Murphy K E, Magee LA, Ohlsson A. 2004. The effect of hypertensive disorders in pregnancy on small for gestational age and stillbirth: A population based study. BMC pregnancy and childbirth 4:17. Ananth CV, Vintzileos AM. 2006. Maternal fetal conditions necessitatin g a medical intervention resulting in preterm birth. American journal of obstetrics and gynecology 195:1557 1563. Anderson LM, Scrimshaw SC, Fullilove MT, Fielding JE, Services TFoCP. 2003. The health. American journal of preventive medicine 24:12 20. Anthopolos R, James SA, Gelfand AE, Miranda ML. 2011. A spatial measure of neighborhood level racial isolation applied to low birthweight, preterm birth, and birthweight in north carolina. Spatial a nd spatio temporal epidemiology 2:235 246. Anthopolos R, Kaufman JS, Messer LC, Miranda ML. 2014. Racial residential segregation and preterm birth: Built environment as a mediator. Epidemiology 25:397 405. Ayala DE, Hermida RC, Mojn A, Fernndez JR, Silva I, Ucieda R, et al. 1997. Blood pressure variability during gestation in healthy and complicated pregnancies. Hypertension 30:611 618. Barnett AG, Williams GM, Schwartz J, Best TL, Neller AH, Petroeschevsky AL, et al. 2006. The effects of air pollution on hospitalizations for cardiovascular disease in elderly people in australian and new zealand cities. Environmental health perspectives 114:1018 1023. Bauer ST, Cleary KL. 2009. Cardiopulmonary complications of pre eclampsia. Seminars in perinatology 33:158 165. Bell JF, Zimmerman FJ, Almgren GR, Mayer JD, Huebner CE. 2006. Birth outcomes among urban african american women: A multilevel analysis of the role of racial residential segregation. Social science & medicine 63:3030 3045.

PAGE 85

85 Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F. 2004. Ozone and short term mortality in 95 us urban communities, 1987 2000. Jama 292:2372 2378. Bell ML, Zanobetti A, Dominici F. 2014. Who is more affected by ozone pollution? A systematic review and meta analysis. American jour nal of epidemiology 180:15 28. Bellamy L, Casas JP, Hingorani AD, Williams DJ. 2007. Pre eclampsia and risk of cardiovascular disease and cancer in later life: Systematic review and meta analysis. Bmj 335:974. Bennett AC, Rankin KM, Rosenberg D. 2012. Does a medical home mediate racial disparities in unmet healthcare needs among children with special healthcare needs? Maternal and child health journal 16:330 338. Bind MA, Baccarelli A, Zanobetti A, Tarantini L, Suh H, Vokonas P, et al. 2012. Air pollution a nd markers of coagulation, inflammation, and endothelial function: Associations and epigene environment interactions in an elderly cohort. Epidemiology 23:332 340. Blaya JA, Fraser HS, Holt B. 2010. E health technologies show promise in developing countrie s. Health affairs 29:244 251. Bronfenbrenner U, Bronfenbrenner U. 2009. The ecology of human development: Experiments by nature and design:Harvard university press. Brook RD, Brook JR, Urch B, Vincent R, Rajagopalan S, Silverman F. 2002. Inhalation of fine particulate air pollution and ozone causes acute arterial vasoconstriction in healthy adults. Circulation 105:1534 1536. Brook RD, Franklin B, Cascio W, Hong Y, Howard G, Lipsett M, et al. 2004. Air pollution and cardiovascular disease: A statement for he althcare professionals from the expert panel on population and prevention science of the american heart association. Circulation 109:2655 2671. Carstairs V, Morris R. 1990. Deprivation and health in scotland. Health bulletin 48:162 175. Chatterjee R, Yeh H C, Shafi T, Anderson C, Pankow JS, Miller ER, et al. 2011. Serum potassium and the racial disparity in diabetes risk: The atherosclerosis risk in communities (aric) study. The American journal of clinical nutrition 93:1087 1091. Chen R, Zhao A, Chen H, Zh ao Z, Cai J, Wang C, et al. 2015. Cardiopulmonary benefits of reducing indoor particles of outdoor origin: A randomized, double blind crossover trial of air purifiers. Journal of the American College of Cardiology 65:2279 2287.

PAGE 86

86 Clausen T, yen N, Henriksen T. 2006. Pregnancy complications by overweight and residential area. A prospective study of an urban norwegian cohort. Acta obstetricia et gynecologica Scandinavica 85:526 533. Coogan PF, White LF, Jerrett M, Brook RD, Su JG, Seto E, et al. 2012. Air poll ution and incidence of hypertension and diabetes mellitus in black women living in los angeles. Circulation 125:767 772. Cozier YC, Palmer JR, Horton NJ, Fredman L, Wise LA, Rosenberg L. 2007. Relation between neighborhood median housing value and hyperten sion risk among black women in the united states. American journal of public health 97:718 724. Cubbin C, Hadden WC, Winkleby MA. 2000. Neighborhood context and cardiovascular disease risk factors: The contribution of material deprivation. Ethnicity & dise ase 11:687 700. Cubbin C, Sundquist K, Ahln H, Johansson S E, Winkleby MA, Sundquist J. 2006. Neighborhood deprivation and cardiovascular disease risk factors: Protective and harmful effects. Scandinavian journal of public health 34:228 237. Cunradi CB, C aetano R, Clark C, Schafer J. 2000. Neighborhood poverty as a predictor of intimate partner violence among white, black, and hispanic couples in the united states: A multilevel analysis. Annals of epidemiology 10:297 308. Darrow LA, Klein M, Strickland MJ, Mulholland JA, Tolbert PE. 2011. Ambient air pollution and birth weight in full term infants in atlanta, 1994 2004. Environmental health perspectives 119:731. Dominici F, McDermott A, Zeger SL, Samet JM. 2003. Airborne particulate matter and mortality: Ti mescale effects in four us cities. American journal of epidemiology 157:1055 1065. Dong GH, Qian ZM, Xaverius PK, Trevathan E, Maalouf S, Parker J, et al. 2013. Association between long term air pollution and increased blood pressure and hypertension in ch ina. Hypertension 61:578 584. Doubeni CA, Schootman M, Major JM, Torres Stone RA, Laiyemo AO, Park Y, et al. 2012. Health status, neighborhood socioeconomic context, and premature mortality in the united states: The national institutes of health aarp diet and health study. American journal of public health 102:680 688. Draper. 2016. Draper satellite image chronology. Available: https://www.kaggle.com/c/draper satellite image chronolo gy Duany A, Speck J, Lydon M. 2010. The smart growth manual:McGraw Hill New York. Dubowitz T, Heron M, Bird CE, Lurie N, Finch BK, Basurto Dvila R, et al. 2008. Neighborhood socioeconomic status and fruit and vegetable intake among whites,

PAGE 87

87 blacks, and mexican americans in the united states. The American journal of clinical nutrition 87:1883 1891. Duley L. 2009. The global impact of pre eclampsia and eclampsia. Seminars in perinatology 33:130 137. Elo IT, Culhane JF, Kohler IV, O'Campo P, Burke JG, Messe r LC, et al. 2009. Neighbourhood deprivation and small for gestational age te rm births in the united states. Paediatric and perinatal epidemiology 23:87 96. Farley TA, Mason K, Rice J, Habel JD, Scribner R, Cohen DA. 2006. The relationship between the neig hbourhood environment and adverse birth outcomes. Paediatric and perinatal epidemiology 20:188 200. Finch BK, Kolody B, Vega WA. 1999. Contextual effects of perinatal substance exposure among black and white women in california. Sociological Perspectives 4 2:141 156. matter air quality trends. Available: https://www.dep.state.f l.us/air/air_quality/new_ozone/florida_ozone_pm_trends.pdf [accessed Dec 1, 2015. Foraster M, Basagana X, Aguilera I, Rivera M, Agis D, Bouso L, et al. 2014. Association of long term exposure to traffic related air pollution with blood pressure and hyperte nsion in an adult population based cohort in spain (the regicor study). Environmental health perspectives 122:404 411. Ghio AJ, Carraway MS, Madden MC. 2012. Composition of air pollution particles and oxidative stress in cells, tissues, and living systems. Journal of toxicology and environmental health Part B, Critical reviews 15:1 21. Gold DR, Damokosh AI, Pope CA, 3rd, Dockery DW, McDonnell WF, Serrano P, et al. 1999. Particulate and ozone pollutant effects on the respiratory function of children in south west mexico city. Epidemiology 10:8 16. Goldenberg RL, Culhane JF, Iams JD, Romero R. 2008. Epidemiology and causes of preterm birth. Lancet 371:75 84. Grady SC. 2006. Racial disparities in low birthweight and the contribution of residential segregation: A multilevel analysis. Social science & medicine 63:3013 3029. Grady SC, Ramrez IJ. 2008. Mediating medical risk factors in the residential segregation and low birthweight relationship by race in new york city. Health & place 14:661 677.

PAGE 88

88 Gray R, Bonellie S Chalmers J, Greer I, Jarvis S, Williams C. 2008. Social inequalities in preterm birth in scotland 1980 2003: Findings from an area based measure of deprivation. BJOG : An International Journal of Obstetrics & Gynaecology 115:82 90. Grewal J, Carmichael SL Song J, Shaw GM. 2009. Neural tube defects: An analysis of neighbourhood and individual level socio economic characteristics. Paediatric and perinatal epidemiology 23:116 124. Gudmundsson S, Bjrgvinsdttir L, Molin J, Gunnarsson G, Marsal K. 1997. Socio economic status and perinatal outcome according to residence area in the city of malm. Acta obstetricia et gynecologica Scandinavica 76:318 323. Gurol Urganci I, de Jongh T, Vodopivec Jamsek V, Car J, Atun R. 2012. Mobile phone messaging for communicating results of medical investigations. The Cochrane Library. Ha S, Hu H, Roussos Ross D, Haidong K, Roth J, Xu X. 2014. The effects of air pollution on adverse birth outcomes. Environmental research 134C:198 204. Hansen CA, Barnett AG, Jalaludin BB, Morgan GG 2009. Ambient air pollution and birth defects in brisbane, australia. PloS one 4:e5408. Harms CA. 2006. Does gender affect pulmonary function and exercise capacity? Respiratory physiology & neurobiology 151:124 131. Hermida RC, Ayala DE, Iglesias M. 2001 Predictable blood pressure variability in healthy and complicated pregnancies. Hypertension 38:736 741. Holzman C, Eyster J, Kleyn M, Messer LC, Kaufman JS, Laraia BA, et al. 2009. Maternal weathering and risk of preterm delivery. American journal of pub lic health 99:1864 1871. Hu H, Ha S, Roth J, Kearney G, Talbott EO, Xu X. 2014. Ambient air pollution and hypertensive disorders of pregnancy: A systematic review and meta analysis. Atmospheric environment 97:336 345. Hu H, Ha S, Henderson BH, Warner TD, R oth J, Kan H, et al. 2015. Association of atmospheric particulate matter and ozone with gestational diabetes mellitus. Environmental health perspectives 123:853. Huang W, Zhu T, Pan X, Hu M, Lu SE, Lin Y, et al. 2012. Air pollution and autonomic and vascul ar dysfunction in patients with cardiovascular disease: Interactions of systemic inflammation, overweight, and gender. American journal of epidemiology 176:117 126. Hung LJ, Chan TF, Wu CH, Chiu HF, Yang CY. 2012. Traffic air pollution and risk of death fr om ovarian cancer in taiwan: Fine particulate matter (pm2.5) as a proxy marker. Journal of toxicology and environmental health Part A 75:174 182.

PAGE 89

89 Jaffe DH, Eisenbach Z, Neumark YD, Manor O. 2005. Does living in a religiously affiliated neighborhood lower m ortality? Annals of epidemiology 15:804 810. Janghorbani M, Stenhouse E, Millward A, Jones RB. 2006. Neighborhood deprivation and preterm birth in plymouth, uk. The Journal of Maternal Fetal & Neonatal Medicine 19:85 91. Jarman B. 1983. Identification of underprivileged areas. British medical journal (Clinical research ed) 287:130. Jedrychowski W, Bendkowska I, Flak E, Penar A, Jacek R, Kaim I, et al. 2004. Estimated risk for altered fetal growth resulting from exposure to fine particles during pregnancy: An epidemiologic prospective cohort study in poland. Environmental health perspectives:1398 1402. Jerrett M, Burnett RT, Ma R, Pope CA, 3rd, Krewski D, Newbold KB, et al. 2005. Spatial analysis of air pollution and mortality in los angeles. Epidemiology 16 :727 736. Karagiannaki K, Chonianakis S, Patelarou E, Panousopoulou A, Papadopouli M. Mmamee: A mhealth platform for monitoring and assessing maternal environmental exposure. In: Proceedings of the Computer Based Medical Systems (CBMS), 2015 IEEE 28th Inte rnational Symposium on, 2015, IEEE, 163 168. Kearney G, Kiros G E. 2009. A spatial evaluation of socio demographics surrounding national priorities list sites in florida using a distance based approach. International journal of health geographics 8:33. Ker shaw KN, Roux AVD, Burgard SA, Lisabeth LD, Mujahid MS, Schulz AJ. 2011. Metropolitan level racial residential segregation and black white disparities in hypertension. American journal of epidemiology:kwr116. Kumar S, Nilsen WJ, Abernethy A, Atienza A, Pat rick K, Pavel M, et al. 2013. Mobile health technology evaluation: The mhealth evidence workshop. American journal of preventive medicine 45:228 236. Lai HK, Tsang H, Wong CM. 2013. Meta analysis of adverse health effects due to air pollution in chinese po pulations. BMC public health 13:360. Lee PC, Talbott EO, Roberts JM, Catov JM, Sharma RK, Ritz B. 2011. Particulate air pollution exposure and c reactive protein during early pregnancy. Epidemiology 22:524 531. Lillesand T, Kiefer RW, Chipman J. 2014. Remo te sensing and image interpretation:John Wiley & Sons. Lo JO, Mission JF, Caughey AB. 2013. Hypertensive disease of pregnancy and maternal mortality. Current opinion in obstetrics & gynecology 25:124 132.

PAGE 90

90 Malmqvist E, Jakobsson K, Tinnerberg H, Rignell Hyd bom A, Rylander L. 2013. Gestational diabetes and preeclampsia in association with air pollution at levels below current air quality guidelines. Environmental health perspectives 121:488 493. Malmstrm M, Sundquist J, Johansson S E. 1999. Neighborhood envi ronment and self reported health status: A multilevel analysis. American journal of public health 89:1181 1186. Mammaro A, Carrara S, Cavaliere A, Ermito S, Dinatale A, Pappalardo EM, et al. 2009. Hypertensive disorders of pregnancy. Journal of Prenatal Me dicine 3:1 5. Marmot M. 2005. Social determinants of health inequalities. The Lancet 365:1099 1104. Mason SM, Messer LC, Laraia BA, Mendola P. 2009. Segregation and preterm birth: The effects of neighborhood racial composition in north carolina. Health & p lace 15:1 9. McMillan N, Holland D, Morara M, Feng J. 2010. Combining numerical model output and particulate data using bayesian space time modeling. Environmetrics 21:48 65. Mendez DD, Hogan VK, Culhane JF. 2013. Stress during pregnancy: The role of insti tutional racism. Stress Health 29:266 274. Messer LC, Laraia BA, Kaufman JS, Eyster J, Holzman C, Culhane J, et al. 2006. The development of a standardized neighborhood deprivation index. Journal of Urban Health 83:1041 1062. Messer LC, Vinikoor LC, Laraia BA, Kaufman JS, Eyster J, Holzman C, et al. 2008. Socioeconomic domains and associations with preterm birth. Social science & medicine 67:1247 1257. Messer LC, Oakes JM, Mason S. 2010. Effects of socioeconomic and racial residential segregation on preterm birth: A cautionary tale of structural confounding. American journal of epidemiology 171:664 673. Messer LC, Vinikoor Imler LC, Laraia BA. 2012. Conceptualizing neighborhood space: Consistency and variation of associations for neighborhood factors and pre gnancy health across multiple neighborhood units. Health & place 18:805 813. Middleton N, Yiallouros P, Kleanthous S, Kolokotroni O, Schwartz J, Dockery DW, et al. 2008. A 10 year time series analysis of respiratory and cardiovascular morbidity in nicosia, cyprus: The effect of short term changes in air pollution and dust storms. Environmental health : a global access science source 7:39. Miller MA, Carpenter M. 2015. Hypertensive disorders of pregnancy. In: Medical management of the pregnant patient:Spring er, 177 193.

PAGE 91

91 Mobasher Z, Salam MT, Goodwin TM, Lurmann F, Ingles SA, Wilson ML. 2013. Associations between ambient air pollution and hypertensive disorders of pregnancy. Environmental research 123:9 16. Morenoff JD, House JS, Hansen BB, Williams DR, Kaplan GA, Hunte HE. 2007. Understanding social disparities in hypertension prevalence, awareness, treatment, and control: The role of neighborhood context. Social science & medicine 65:1853 1866. Mugavero MJ, Lin H Y, Allison JJ, Giordano TP, Willig JH, Raper J L, et al. 2009. Racial disparities in hiv virologic failure: Do missed visits matter? Journal of acquired immune deficiency syndromes (1999) 50:100. Mujahid MS, Roux AVD, Morenoff JD, Raghunathan TE, Cooper RS, Ni H, et al. 2008. Neighborhood characteristi cs and hypertension. Epidemiology 19:590 598. Nagiah S, Phulukdaree A, Naidoo D, Ramcharan K, Naidoo RN, Moodley D, et al. 2015. Oxidative stress and air pollution exposure during pregnancy: A molecular assessment. Human & experimental toxicology 34:838 847. National High Blood Pressure Education Program. 2000. Report of the national high blood pressure education program working group on high blood pressure in pregnancy. American journal of obstetrics and gynecology 183:s1 s22. Nguyen QC, Osypuk TL Schmidt NM, Glymour MM, Tchetgen EJT. 2015. Practical guidance for conducting mediation analysis with multiple mediators using inverse odds ratio weighting. American journal of epidemiology 181:349 356. Nkansah Amankra S, Luchok KJ, Hussey JR, Watkins K, Liu X. 2010. Effects of maternal stress on low birth weight and preterm birth outcomes across neighborhoods of south carolina, 2000 2003. Maternal and child health journal 14:215 226. Noble M, Wright G, Dibben C, Smith G, McLennan D, Anttila C, et al. 200 4. Indices of deprivation 2004. Report to the Office of the Deputy Prime Minister London: Neighbourhood Renewal Unit. Nodari S, Corulli A, Manerba A, Metra M, Apostoli P, Dei Cas L. 2006. Endothelial damage due to air pollution. Heart international 2:115. O'Brien WF. 1990. Predicting preeclampsia. Obstetrics & Gynecology 75:445 452. O'Campo P, Burke JG, Culhane J, Elo IT, Eyster J, Holzman C, et al. 2008. Neighborhood deprivation and preterm birth among non hispanic black and white women in eight geographic areas in the united states. American journal of epidemiology 167:155 163. Olsson D, Mogren I, Forsberg B. 2013. Air pollution exposure in early pregnancy and adverse pregnancy outcomes: A register based cohort study. BMJ open 3.

PAGE 92

92 Orpana AK, Avela K, Ranta V, Viinikka L, Ylikorkala O. 1996. The calcium dependent nitric oxide production of human vascular endothelial cells in preeclampsia. American journal of obstetrics and gynecology 174:1056 1060. Osypuk TL, Acevedo Garcia D. 2008. Are racial disparities in preterm birth larger in hypersegregated areas? American journal of epidemiology 167:1295 1304. Park SK, Wang W. 2014. Ambient air pollution and type 2 diabetes: A systematic review of epidemiologic research. Current environmental health reports 1:275 286. Pedersen M, Stayner L, Slama R, Sorensen M, Figueras F, Nieuwenhuijsen MJ, et al. 2014. Ambient air pollution and pregnancy induced hypertensive disorders: A systematic review and meta analysis. Hypertension 64:494 500. Pollock PH, Vittas ME. 1995. Who bea rs the burdens of environmental pollution? Race, ethnicity, and environmental equity in florida. Social Science Quarterly 76:294 310. Pope CA, 3rd, Dockery DW, Spengler JD, Raizenne ME. 1991. Respiratory health and pm10 pollution. A daily time series analy sis. The American review of respiratory disease 144:668 674. Pope CA, 3rd, Kanner RE. 1993. Acute effects of pm10 pollution on pulmonary function of smokers with mild to moderate chronic obstructive pulmonary disease. The American review of respiratory dis ease 147:1336 1340. Pope CA, 3rd. 1999. Mortality and air pollution: Associations persist with continued advances in research methodology. Environmental health perspectives 107:613 614. Raaschou Nielsen O, Andersen ZJ, Beelen R, Samoli E, Stafoggia M, Wein mayr G, et al. 2013. Air pollution and lung cancer incidence in 17 european cohorts: Prospective analyses from the european study of cohorts for air pollution effects (escape). The Lancet Oncology 14:813 822. nal and child health and neighborhood context: The selection and construction of area level variables. Health & place 12:547 556. Riley WT, Rivera DE, Atienza AA, Nilsen W, Allison SM, Mermelstein R. 2011. Health behavior models in the age of mobile interv entions: Are our theories up to the task? Translational behavioral medicine 1:53 71. Ritz B, Yu F, Fruin S, Chapa G, Shaw GM, Harris JA. 2002. Ambient air pollution and risk of birth defects in southern california. American journal of epidemiology 155:17 2 5. Rodrigo R, Gonzalez J, Paoletto F. 2011. The role of oxidative stress in the pathophysiology of hypertension. Hypertension research : official journal of the Japanese Society of Hypertension 34:431 440.

PAGE 93

93 Romero R, Dey SK, Fisher SJ. 2014. Preterm labor: One syndrome, many causes. Science 345:760 765. Roux AVD, Merkin SS, Arnett D, Chambless L, Massing M, Nieto FJ, et al. 2001. Neighborhood of residence and incidence of coronary heart disease. New England Journal of Medicine 345:99 106. Samoli E, Touloumi G, Schwartz J, Anderson HR, Schindler C, Forsberg B, et al. 2007. Short term effects of carbon monoxide on mortality: An analysis within the aphea project. Environmental health perspectives 115:1578 1583. Sampson RJ, Wilson WJ. 1995. Toward a theory of rac e, crime, and urban inequality. Race, crime, and justice: A reader:177 190. Savitz DA, Elston B, Bobb JF, Clougherty JE, Dominici F, Ito K, et al. 2015. Ambient fine particulate matter, nitrogen dioxide, and hypertensive disorders of pregnancy in new york city. Epidemiology 26:748 757. Schempf A, Strobino D, O'Campo P. 2009. Neighborhood effects on birthweight: An exploration of psychosocial and behavioral pathways in baltimore, 1995 1996. Social science & medicine 68:100 110. Shenassa ED, Stubbendick A, Brown MJ. 2004. Social disparities in housing and related pediatric injury: A multilevel study. American journal of public health 94:633 639. Slama R, Darrow L, Parker J, Woodruff TJ, Strickland M, Nieuwenhuijsen M, et al. 2008. Meeting report: Atmospheric pollution and human reproduction. Environmental health perspectives 116:791 798. Smarr MM, Vadillo Ortega F, Castillo Castrejon M, O'Neill MS. 2013. The use of ultrasound measurements in environmental epidemiological studies of air pollution and fetal gro wth. Current opinion in pediatrics 25:240 246. outcomes: A review of the literature. Environmental health perspectives:375 382. bient air pollution and pregnancy outcomes: A review of the literature. Environmental health perspectives:375 382. Strand LB, Barnett AG, Tong S. 2011. Methodological challenges when estimating the effects of season and seasonal exposures on birth outcomes BMC medical research methodology 11:49. Stretesky P, Hogan MJ. 1998. Environmental justice: An analysis of superfund sites in florida. Soc Probs 45:268.

PAGE 94

94 Tanaka M, Jaamaa G, Kaiser M, Hills E. 2007. Racial disparity in hypertensive disorders of pregnancy in new york state: A 10 year longitudinal population based study. American journal of public health 97:163. Tibshirani R. 1996. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society Series B (Methodological):267 288. To wnsend P, Phillimore P, Beattie A. 1988. Health and deprivation: Inequality and the north:London. U.S. EPA. 2014. Air quality data for the cdc national environmental public health tracking network. Available: http://www.epa.gov/heasd/research/cdc.html [accessed 21 March 2014]. van den Hooven EH, Jaddoe VW, de Kluizenaar Y, Hofman A, Mackenbach JP, Steegers EA, et al. 2009. Residential traffic exposure and pregnancy related outcomes: A prospective bi rth cohort study. Environmental health : a global access science source 8:59. van den Hooven EH, de Kluizenaar Y, Pierik FH, Hofman A, van Ratingen SW, Zandveld PY, et al. 2011. Air pollution, blood pressure, and the risk of hypertensive complications duri ng pregnancy: The generation r study. Hypertension 57:406 412. Vinikoor Imler LC, Gray SC, Edwards SE, Miranda ML. 2012. The effects of exposure to particulate matter and neighbourhood deprivation on gestational hypertension. Paediatric and perinatal epide miology 26:91 100. Vinikoor Imler LC, Davis JA, Meyer RE, Messer LC, Luben TJ. 2014. Associations between prenatal exposure to air pollution, small for gestational age, and term low birthweight in a state wide birth cohort. Environmental research 132:132 1 39. Vinikoor Imler LC, Gray SC, Edwards SE, Miranda ML. 2012. The effects of exposure to particulate matter and neighbourhood deprivation on gestational hypertension. Paediatr ic and perinatal epidemiology 26:91 100. Vos AA, Posthumus AG, Bonsel GJ, Steeger neighborhoods and adverse perinatal outcome: A systematic review and meta analysis. Acta obstetricia et gynecologica Scandinavica 93:727 740. Vrijheid M, Dolk H, Stone D, Abramsky L, Alberman E, Scott J. 2000. Socioeconomic inequalities in risk of congenital anomaly. Archives of disease in childhood 82:349 352. Vrijheid M, Martinez D, Manzanares S, Dadvand P, Schembari A, Rankin J, et al. 2011. Ambient air pollution and risk of congenital anomalies: A systematic review and me ta analysis. Environmental health perspectives 119:598 606.

PAGE 95

95 Wallis AB, Saftlas AF, Hsia J, Atrash HK. 2008. Secular trends in the rates of preeclampsia, eclampsia, and gestational hypertension, united states, 1987 2004. American journal of hypertension 21: 521 526. Walther B, Hossin S, Townend J, Abernethy N, Parker D, Jeffries D. 2011. Comparison of electronic data capture (edc) with the standard data capture method for clinical trial data. PloS one 6:e25348. Walton E. 2009. Residential segregation and birt h weight among racial and ethnic minorities in the united states. Journal of health and social behavior 50:427 442. Wang IK, Tsai IJ, Chen PC, Liang CC, Chou CY, Chang CT, et al. 2012. Hypertensive disorders in pregnancy and subsequent diabetes mellitus: A retrospective cohort study. The American journal of medicine 125:251 257. Wasserman CR, Shaw GM, Selvin S, Gould JB, Syme SL. 1998. Socioeconomic status, neighborhood social conditions, and neural tube defects. American journal of public health 88:1674 16 80. Welte JW, Wieczorek WF, Barnes GM, Tidwell M C, Hoffman JH. 2004. The relationship of ecological and geographic factors to gambling behavior and pathology. Journal of Gambling Studies 20:405 423. Wietlisbach V, Pope CA, 3rd, Ackermann Liebrich U. 1996. Air pollution and daily mortality in three swiss urban areas. Sozial und Praventivmedizin 41:107 115. Williams DR, Collins C. 2001. Racial residential segregation: A fundamental cause of racial disparities in health. Public health reports 116:404. Willia ms N. 2007. The guide to community preventive services what works to promote health? Occupational medicine 57:75 75. Wolf M, Shah A, Jimenez Kimble R, Sauk J, Ecker JL, Thadhani R. 2004. Differential risk of hypertensive disorders of pregnancy among hispanic women. Journal of the American Society of Nephrology 15:1330 1338. Wu CS, Nohr EA, Bech BH, Vestergaard M, Catov JM, Olsen J. 2009. Health of children born to mothers who had preeclampsia: A population based cohort study. American journal of obste trics and gynecology 201:269 e261 269 e210. Xu X, Hu H, Ha S, Roth J. 2013. Ambient air pollution and hypertensive disorder of pregnancy. Journal of epidemiology and community health. Xu X, Hu H, Ha S, Roth J. 2014. Ambient air pollution and hypertensive d isorder of pregnancy. Journal of epidemiology and community health 68:13 20. Yoder SR, Thornburg LL, Bisognano JD. 2009. Hypertension in pregnancy and women of childbearing age. The American journal of medicine 122:890 895.

PAGE 96

96 Yost K, Perkins C, Cohen R, Morr is C, Wright W. 2001. Socioeconomic status and breast cancer incidence in california for different race/ethnic groups. Cancer Causes & Control 12:703 711. Zhai D, Guo Y, Smith G, Krewski D, Walker M, Wen SW. 2012. Maternal exposure to moderate ambient carb on monoxide is associated with decreased risk of preeclampsia. American journal of obstetrics and gynecology 207:57 e51 59.

PAGE 97

97 BIOGRAPHICAL SKETCH Hui Hu received his Doctor of Philosophy from the Department of Epidemiology at the University of Florida in August 2016 He also recei ved his Bachelor of Science in n ursing from Fudan University in 2012. Hui has been actively involve d in biomedical research since he was an undergraduate stud ent at Fudan University, where he was two studies focusing on liver cancer patients and nursing students. After the completion of his BS, Hui joined the PhD program in the Department of Epidemiolog y at the Un iversity of Florida, where he was involved in several NIH funded studies with different research groups. Coming from one of the most highly polluted countr ies in the world, China, research is primarily focused on evaluating health effects of hazardou s environmental exposure, especially on adverse pregnancy and birth outcomes. In addition to the research in environmental epidemiology, he has also worked with Dr. Linda B. Cottler and Dr. Maria R. Khan on substance use and HIV prevention.