57...! 5.... I. 1.1 I... It?! 2.. 31:”. I . a :3: 7:11”?! I 535.. I . .1... a . . S. ..:.....Vi....h....n.n.ah Mbuhfm: z..- .13."... ILWiltfials.Ih§$9lI1.Imutv.Dh-“Ihiv , .cw...m.a...§un~.n....ru. fizrfi. firm“. a. .5 .fi . :flh...rmu?¢!.“. .v. . .1353 Is: {nah grit“... ..II.Im.II....I GI... . Ilzkurffi E 3.1-2.7“? 1. Marina” 1.3.1.3., . t. a. 2....- I 3.....II 72.33.10 If: I ‘1 .l .32. I)! n.91.l.E:I-§I...I III"? .. I.) .t 4..)I.iS...¢II.!.II .§ $.13}: in. .Ith:II.£\I : IIII. iwflti 5... . an .IZdII. .339 z THESIS 200% My 571% 7 TH (’ LIBRARY Miclz ‘ litate UniveISii‘y' This is to certify that the thesis entitled PERCEIVED AND OBJECTIVE NEIGHBORHOOD CONDITIONS: RELATIONS WITH PRETERM DELIVERY presented by MARY JACQUELINE KLEYN has been accepted towards fulfillment of the requirements for the MS. degree in Epidemiology / I/I : Infra. I I: if." /" ;’ ' ,~" :' ;' IL ' - \g Git/(1k L‘ K 2/“ ZZ' ‘ r Major Professor’s Signature ‘3 ‘1' flzflifi/ g 2 0 [)f Date MSU is an affinnative—action, equal-opportunity employer .ouuo—._._ .I-n-I-I-I-IOIQI-I-u-gpI-.-I-I-_l_-_L-I-0-.--I-I-I-I-I-I.g. - - PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE PERCEIVED AND OBJECTIVE NEIGHBORHOOD CONDITIONS: RELATIONS WITH PRETERM DELIVERY By Mary Jacqueline Kleyn A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Epidemiology 2008 ABSTRACT PERCEIVED AND OBJECTIVE NEIGHBORHOOD CONDITIONS: RELATIONS WITH PRETERM DELIVERY By Mary Jacqueline Kleyn The Pregnancy Outcomes and Community Health Study was used to explore the relations between perceptions of neighborhood disorder, objective measures of neighborhood deprivation and preterm delivery (PTD) grouped by week (<35, 35-36 weeks) and clinical circumstance (spontaneous, medically indicated). Due to differing distributions of disorder and deprivation scores, race/ethnic-specific groups were used: Black/Hispanic (B/H) (743 term, 127 preterm) and White/Other (W/O) (1852 term, 200 preterm). In the B/H group, the top decile of neighborhood disorder score was associated with increased odds of medically indicated PTD (OR=2.8 95% CI: 1.2, 6.1 ), and the highest compared to the lowest tertile of neighborhood deprivation score was associated with increased risk of all PTD <35 weeks (OR=2.3 95% CI: 1.1, 4.9). Disorder and deprivation were not associated with PTD in the W/O group. These results suggest unique associations between perceived and objective neighborhood conditions and PTD in black and Hispanic women. To Andrew-Thanks for all of your love, support, and editing! iii ACKNOWLEDGMENTS I want to thank Dr. Claudia Holzman for all of her advice and encouragement during my graduate study. She has truly made my time at Michigan State University a wonderful experience. I also appreciate the support of the other members of my graduate committee. Dr. Sue Grady provided invaluable help with the geocoding process and provided me with insightful comments that greatly improved this manuscript. Dr. Hwan Chung gave me advice on statistical approaches. Dr. Janet Eyster served as an “unofficial" committee member and also helped with the geocoding process. I would also to thank the entire POUCH team. I have been blessed to work with such a wonderful group of kind and talented researchers for the past two years. Lastly, I would like to thank my family-especially my husband and my parents. Their love and support has allowed me to achieve all of my academic goals. TABLE OF CONTENTS List of Tables ............................................................................................ vi List of Figuresvu LITERATURE REVIEW ........................................................................... 1 Perceived and objective neighborhood conditions: relations with preterm delivery .............................................................................................. 17 AIMS AND HYPOTHESES ..................................................................... 17 MATERIALS AND METHODS ................................................................ 17 RESULTS ........................................................................................... 24 DISCUSSION ...................................................................................... 27 WORKS CITED ................................................................................... 46 LIST OF TABLES Table 1. Maternal characteristics of the POUCH study sample (N=2,922)a by race/ethnicity group ........................................................................ 37 Table 2. Preterm delivery characteristics of the POUCH study sample (N=2,922) by race/ethnicity group .......................................................... 38 Table 3. Associations between neighborhood disorder score and PTD subtypes by clinical circumstances, stratified on race/ethnicity group.............38 Table 4. Associations between neighborhood disorder score and PTD subtypes by week, stratified on race/ethnicity group ................................... 38 Table 5. Associations between neighborhood deprivation score tertiles and PTD subtypes by clinical circumstances, stratified on race/ethnicity group.....39 Table 6. Associations between neighborhood deprivation score tertiles and PTD subtypes by week, stratified on race/ethnicity group............................39 Table 7. Concordance of neighborhood disorder and deprivation score tertiles in Black/Hispanic group, n (cell percentage) ................................... 40 Table 8. Concordance of neighborhood disorder and deprivation score tertiles in White/Other group, n (cell percentage) ....................................... 40 Table 9. Associations between neighborhood disorder score and PTD subtypes by clinical circumstances, stratified on race/ethnicity group (same top decile cut point for both groups) .............................................. 40 Table 10. Associations between neighborhood disorder and PTD subtypes by week, stratified on race/ethnicity group (same top decile cut point for both groups) ............................................................. 41 Table 11. Associations between neighborhood deprivation score tertiles and PTD subtypes by clinical circumstances, stratified on race/ethnicity group (same tertile cut points for both groups) ........................................... 41 Table 12. Associations between neighborhood deprivation score tertiles and PTD subtypes by week, stratified on race/ethnicity group (same tertile cut points for both groups) ............................................................ 42 vi LIST OF FIGURES Figure 1. Modified Version of Section VII of the Korbin and Coulton Instrument ......................................................................................... 43 Figure 2. Comparison of Race/Ethnicity—specific Top Decile of Neighborhood Disorder Score ............................................................... 44 Figure 3. Comparison of Neighborhood Deprivation Score by Race/Ethnicity Group ............................................................................ 45 vii LITERATURE REVIEW Preterm birth, defined by the World Health Organization as the delivery of an infant between 20 and 37 weeks of gestation, is a leading cause of perinatal mortality in Europe and North America (Berkowitz & Papiemik, 1993). Preterm delivery (PTD) is a major problem in developing countries. Determining the exact rates of PTD has been difficult, but estimates have placed PTD rates as high as 25% in developing countries (Steer, 2005). Among industrialized countries, the United States has dramatically higher rates of PTD (Joseph, Huang, Liu, Ananth, Allen, Sauve et al., 2007). In 2004, more than 500,000 babies in the United States were born preterm. This represents 12.5% of all births in that year. PTD may result in pulmonary dysfunction or visual impairment in the neonate, as well as nearly one-half of all perinatal—related neurological disabilities (Martin, Hamilton, & Sutton, 2006; Wen, Smith, Yang, & Walker, 2004). In the United States, PTD is responsible for 75% of neonatal morbidity and 70% of neonatal mortality (Challis, Lye, Gibb, Whittle, Patel, & Alfaidy, 2001). In addition to cost in terms of human life, PTD is also very costly for the health-care system. The hospital costs of caring for preterm infants in the United States have been estimated to be $5.8 billion for the first year of life alone (Russell, Green, Steiner, Meikle, Howse, Poschman et al., 2007) PTD may be further divided by clinical circumstances. Approximately 40- 50% of all preterm births occur due to spontaneous preterm labor of unknown etiology, while premature rupture of membranes is implicated in 25-40% of all PTD (Tucker, Goldenberg, Davis, Copper, Winkler, & Hauth, 1991). Although these medical risk pathways may be distinct, spontaneous preterm labor and premature rupture of membranes have many common risk factors and may be grouped together to form a category consisting of all spontaneous PTD (Pickett, Abrams, & Selvin, 2000). The remaining 20-25% of all preterm deliveries are medically indicated to decrease adverse maternal or fetal outcomes. Maternal hypertension, preeclampsia, poor fetal growth, and fetal distress are all possible reasons for medically intervening in pregnancy before a woman has completed 37 weeks' gestation (Ananth & Vintzileos, 2006). Despite medical advances and efforts directed towards reducing the incidence of PTD, the PTD rate in the United States has risen from 10.6% in 1990 to 12.7% in 2005 (Goldenberg & Rouse, 1998; Hamilton, Martin, & Ventura, 2006). Multifetal gestations may exert a strong influence on the rate of preterm birth due to the elevated risk of PTD associated with multifetal gestations and the increasing frequency of multiple births. However, among singleton births only, the rate of PTD still increased from 9.7% in 1990 to 10.8% in 2004 (Martin, Hamilton, & Sutton, 2006). The most dramatic increase in PTD rate was seen in infants delivered at 34-36 weeks (Hamilton, Martin, & Ventura, 2006). These late-term preterm infants comprise over 70% of all preterm births (Martin, Hamilton, & Sutton, 2006). Although these infants are at lower risk than infants born <34 weeks, late preterm infants have more medical problems than infants born >37 weeks gestation (McIntire & Leveno, 2008; Wang, Dorer, Fleming, & Catlin, 2004). According to the National Center for Health Statistics, preterm birth is more common in younger and older mothers. For women younger than 20 years of age, the average PTD rate from 2002-2004 was 14.3%. During the same time period, the PTD rate was 16.3% for women aged 40 or greater, and the rate was lowest for women 20 to 29 years of age (11.7%) (National Center for Health Statistics). Preterm birth is also more common in certain racial and ethnic groups. For example, from 2002-2004, Black women experienced a PTD rate of 17.8%. In that same time period, White and Asian women had rates of PTD 11.3% and 10.4%, respectively (National Center for Health Statistics). Although there have been many mechanistic studies about the etiology of PTD, many questions remain-especially about the pathophysiological mechanisms of PTD. Several mechanisms that have been proposed include: activation of the maternal orfetal hypothalamic—pituitary—adrenal (HPA) axis; inflammation and/or infection; decidual hemorrhage; and, pathologic uterine distention (Lockwood & Kuczynski, 1999). Whether these mechanisms work independently orjointly is still an area of active investigation. Many risk factors for PTD have been previously identified through epidemiological studies, though some are more strongly and/or consistently associated with PTD. Multifetal pregnancy has been reliably identified as a risk factor for PTD (Strauss, Paek, GenzeI-Boroviczeny, Schulze, Janssen, & Hepp, 2002; Wen, Demissie, Yang, & Walker, 2004). Previous preterm delivery is another strong risk factor for PTD (Adams, Elam-Evans, Wilson, & Gilbertz, 2000; Adams, Samo, Harlass, Rawlings, & Read, 1995). Black race is also a strong risk factor for PTD, as Black women experience rates of PTD that are nearly double the rates of any other racial/ethnic group (Mercer, Goldenberg, Das, Moawad, Iams, Meis et al., 1996; Schieve & Handler, 1996). Infections have also been shown to be a risk factor for early PTD (Goldenberg, Iams, Mercer, Meis, Moawad, Copper et al., 1998). While these maternal (i.e., individual-level) risk factors are very important, they still do not fully explain the etiology of PTD and why the rates of PTD are so high in the United States, particularly among Black women. In recent years, neighborhood context (i.e., the neighborhood in which mothers live and infants are born into) has been hypothesized to affect the risk of PTD, even after controlling for the effects of maternal and infant risk factors. Neighborhood context, typically described by level of deprivation or group- Ievel variables, has been linked to PTD. For example, in the United Kingdom, Smith (2007) reported a dose-response relationship between decile of deprivation and incidence rate ratio for PTD. When compared to women living in the least deprived decile, women living in the most deprived decile had an incidence rate for PTD almost twice as high. This result was found in both extremely preterm births (22-28 weeks) and very preterm births (22-32 weeks) and remained consistent throughout the 10-year period of study (19942003). This study, however, did not have access to data on maternal age or parity, and therefore, could not control for these potentially important individual-level variables which may have slightly attenuated the association between deprivation and PTD. A similar study controlled for individual—level covariates and found a smaller, but still significant, increased risk (adjusted OR=1.16, 95% Cl 1.03, 1.32) of PTD for women living in the highest deprivation category compared to women living in the lowest (Smith, Shah, White, Pell, Crossley, & Dobbie, 2006). This study included all preterm births, and not only very preterm births as in the previous study, which may also partially explain the weaker association. While some studies have found significant associations between neighborhood deprivation and PTD (Janghorbani, Stenhouse, Millward, & Jones, 2006; Smith, Shah, White et al., 2006; Smith, Draper, Manktelow et al., 2007), others have not found these associations. Craig (2002) studied the relation between neighborhood deprivation and PTD in New Zealand. This study found that between 1980 and 1999, PTD rates among singletons, particularly infants >28 weeks of gestation, increased more dramatically among women living in less deprived areas than among women living in more deprived areas. One possible explanation for the increased PTD rates among more affluent women is an increase in the use of assisted reproductive technology. This technology may result in an increased number of multifetal gestations. As a result of the uneven increase in rates, the deprivation disparity in PTD that existed in New Zealand in 1980 disappeared in the 19908. These individual-level studies have failed to consistently demonstrate or fully explain the possible association between neighborhood conditions and PTD. A limitation in the previous studies is a failure to report on the clustering of the study population within neighborhoods. If clustering of women and infants is present, multilevel modeling should be used, so that both individual-level and neighborhood-level factors may be considered simultaneously. On the individual-level, PTD could result from various pathways if biological and behavioral factors are influenced by neighborhood-level variables. For example, the characteristics of a neighborhood, both physical and social, may induce or alleviate stress in the lives of its inhabitants (Farley, Mason, Rice, Habel, Scribner, & Cohen, 2006). Stress induced by feeling unsafe in one’s neighborhood may be reduced by the introduction of street lights to improve the physical environment or by the formation of neighborhood watch groups to improve the social environment. Neighborhoods may also affect other aspects of life such as nutrition or substance abuse through access to grocery stores or ease of obtaining illegal substances. Neighborhoods may affect prenatal care utilization, as women who live in neighborhoods with high deprivation are at increased risk for receiving delayed or no prenatal care (Cubbin, Marchi, Lin, Bell, Marshall, Miller et al., 2008). Since neighborhood characteristics affect risk factors for PTD, multi-Ievel analysis is often used to account for population-level, as well as individual-level, factors when studying what role neighborhoods play in PTD. Previously, multilevel modeling has been used to study the relation between neighborhood context and mortality or self—rated health (Bosma, van de Mheen, Borsboom, & Mackenbach, 2001; Jaffe, Eisenbach, Neumark, & Manor, 2005; Singh & Siahpush, 2002; van Jaarsveld, Miles, & Wardle, 2007). Multilevel modeling is now being used to explore the relation between neighborhoods, by type (rural vs. urban) or characteristics (high vs. low deprivation), and birth outcomes. The effects of neighborhoods on PTD may differ by the type of community. For example, women living in relatively large rural cities in central Pennsylvania had a decreased risk of PTD compared to women living in urban communities (aOR=0.73, 95% Cl 0.60, 0.89), although this protective rural effect was only seen in women living in more populated rural areas and not in women living in small rural towns or women living in isolated small rural areas (Hillemeier, Weisman, Chase, & Dyer, 2007). For both urban and rural areas, a significant trend of increasing PTD has been found when neighborhood median income decreases, although trends for other adverse birth outcomes are more consistent in the urban areas (Luo, Kierans, Wilkins, Liston, Mohamed, & Kramer, 2004). Race/ethnicity may also affect the degree to which neighborhood conditions influence the risklof PTD. Messer (2006) found that the largest proportion of White women lived in block groups in the lowest neighborhood deprivation quartile, and increasing neighborhood deprivation was not associated with PTD among White women. In contrast, the largest proportion of Black women lived in block groups in the highest neighborhood deprivation quartile, and neighborhood deprivation was associated with PTD among Black women. This study suggests that neighborhood deprivation may affect the risk of PTD more strongly for Black women than for White women, perhaps due to the larger number of Black women living in neighborhoods with high deprivation levels. However, a different study found a stronger association between neighborhood deprivation and PTD among non-Hispanic Whites than among non-Hispanic Blacks (O'Campo, Burke, Culhane, Elo, Eyster, Holzman et al., 2008). The discrepancy in that finding compared to the results of other studies was attributed to differences in the methods of assigning deprivation, study design, and sample sizes. Other studies have found that neighborhoods may affect Black women and White women in different ways. For Black women, living in a wealthier census tract (i.e., > $30,000/year median income) was associated with a reduced risk for PTD (Kaufman, Dole, Savitz, & Herring, 2003). However, this reduced risk based on wealth of census tract was not found among White women. While the majority of studies on neighborhood context characterize neighborhood deprivation using census data, one study sought to describe neighborhoods without the use of census data. The Pregnancy, Infection, and Nutrition (PIN) study used trained personnel to qualitatively assess the neighborhood conditions of low-income pregnant non-Hispanic White and non- Hispanic Black women (Laraia, Messer, Kaufman, Dole, Caughy, O'Campo et al., 2006). This study also found that more Black women compared to White women lived in neighborhoods with poorer conditions. Non-Hispanic White women lived in neighborhoods with less litter and graffiti and more sidewalks than neighborhoods of non-Hispanic Black women. In addition to the general neighborhood context, specific neighborhood conditions may not influence the risk of PTD for Black and White women in the same manner. For example, among Black women, a significant positive association between neighborhood poverty and very preterm birth was found (Reagan & Salsberry, 2005), but among White women in the same study, an increasing fraction of workers in professional occupations was associated with decreased risk of PTD. For White women, neighborhood poverty was not a risk factor for PTD. In another study, median household income was found to be significantly associated with PTD among Black women, but not among White women (Pickett, Ahern, Selvin, & Abrams, 2002). The lack of consistency in the findings regarding neighborhood deprivation and PTD may be due to several factors. First, the term “neighborhood" in these studies may indicate areas of very different sizes. For example, two of the studies defined neighborhoods as block groups, which are typically comprised of approximately 1,500 people (Laraia, Messer, Kaufman et al., 2006; Messer, Kaufman, Dole et al., 2006). Several other studies defined neighborhoods as census tracts or census area units (Ahern, Pickett, Selvin, & Abrams, 2003; Craig, Thompson, & Mitchell, 2002; Farley, Mason, Rice et al., 2006; Kaufman, Dole, Savitz et al., 2003; O'Campo, Burke, Culhane et al., 2008; Pickett, Ahern, Selvin et al., 2002; Reagan & Salsberry, 2005). Block groups form census tracts, and census tracts typically contain around 4,000 people. Neighborhoods may also be assigned using zip codes, which encompass areas of approximately 30,000 people (Smith, Shah, White et al., 2006; Smith, Draper, Manktelow et al., 2007). Other unique methods have also been used to divide the populations into area-based neighborhoods (Janghorbani, Stenhouse, Millward et al., 2006; Luo, Kierans, Wilkins et al., 2004; Luo, Wilkins, & Kramer, 2006). The neighborhood size has been found to affect the results obtained. Krieger (2003) compared socioeconomic measures at the census block group, census tract, and zip code level and found that measures at the zip code level detected attenuated effects than measures at the block group or census tract level. Subramanian (2006) conducted a similar study and found that census tract socioeconomic measures and block group socioeconomic measures were stronger predictors of birth weight than zip code measure. Furthermore, census tract measures were even stronger predictors of birth weight than block group socioeconomic measures. That study demonstrates that census tracts, rather than block groups, may be the better neighborhood unit when examining pregnancy outcomes. According to Subramanian (2006), this may be “due to the importance of pregnancy and maternal health-related services and their policy relevance for identifying low income areas or medically underserved areas, both of which tend to operate in relatively larger than smaller spaces.” The term “deprivation" may also differ in meaning. Of the studies examined, only two used the same definition of deprivation (Luo, Kierans, Wilkins et al., 2004; Luo, Wilkins, & Kramer, 2006). Deprivation may be assigned using the Townsend Material Deprivation Index, Carstair’s Socioeconomic Deprivation Index, median tract income, household size-adjusted average income per single person, or many other ways. Variables used to assign a deprivation status may be used alone or may be included in a composite index with other variables. Additionally, the method of modeling deprivation may affect the results. A study in the United Kingdom found an association between neighborhood deprivation, modeled continuously, and risk of PTD. Janghorbani (2006) used the Townsend score to calculate neighborhood deprivation and found a 7% increase in the relative risk of PTD for every unit increase in the Townsend score. When deprivation was categorized into tertiles, no significant association was found between deprivation and PTD. In addition to continuous and tertiles used by Janghorbani (2006), neighborhood deprivation has been modeled using quartiles, quintiles, 7 levels, and deciles (Craig, Thompson, & Mitchell, 2002; Luo, Wilkins, & Kramer, 2006; Messer, Kaufman, Dole et al., 2006; O'Campo, Burke, Culhane et al., 2008; Smith, Shah, White et al., 2006; Smith, Draper, Manktelow et al., 2007). In addition to using census data to objectively measure neighborhood context, residents’ perceptions of their own neighborhoods have been used to examine how neighborhoods may affect health (Aneshensel & Sucoff, 1996; Chandola, 2001; Feldman & Steptoe, 2004; Wen, Hawkley, & Cacioppo, 2006; Wilcox, Bopp, Oberrecht, Kammerrnann, & McElmurray, 2003; Wilson, Elliott, Law, Eyles, Jerrett, & Keller-Olaman, 2004). In the field of perinatal epidemiology, however, this method has been rarely used. Petrou (2007) found that women living in neighborhoods of medium or high self-perceived deprivation 9 months after giving birth were more likely to report fair or poor health status than women living in neighborhoods of low self-perceived deprivation. A case- control study of very low birth weight infants born to Black mothers found a significant association between maternal self-rated unfavorable residential environment and delivery of a very low birth weight infant (Collins, David, Symons, Handler, Wall, & Andes, 1998). This association remained significant after adjusting for negative maternal behaviors (smoking, alcohol use, and illicit drug use). To the best of my knowledge, the association between prospectively- assessed perceived neighborhood conditions and PTD has only been examined in one study. Dole (2003) studied maternal stressors, including perception of neighborhood safety, in relation to PTD. The study found no significant associations between perceived neighborhood safety and PTD grouped by clinical circumstances (spontaneous and medically indicated). Like objective neighborhood conditions, the use of subjective neighborhood perceptions encounters difficulties defining a “neighborhood." To the best of my knowledge, no study has provided a clear definition of “neighborhood” to the participants. Therefore, the responses may vary based on what each participant deems to be her “neighborhood.” Another difficulty in using neighborhood perceptions to characterize a neighborhood is that a different questionnaire is used in each study. One study asked participants to rate the safety of neighborhoods using 6 questions (Wilcox, Bopp, Oberrecht et al., 2003), while another study focused on perceived neighborhood strain using 10 questions (Feldman & Steptoe, 2004). Petrou (2007) used a composite measure from 6 questions, and Collins (1998) used a composite measure from 8 questions to obtain a neighborhood deprivation score. Further complicating the difficulties in comparing results across studies, neighborhood conditions may also be divided into physical or social components (Wilson, Elliott, Law et al., 2004). The correlation between objective and perceived measures of neighborhood conditions has been an issue of debate. Intuitively, the two measures should be highly correlated, as the objective neighborhood context should drive the perceived neighborhood context. Sampson and Raudenbush (2004) conducted a study in Chicago that illustrates the complexity of the relations between these two measures. The study used trained personnel to conduct systematic observations of neighborhoods to obtain objective measures of disorder and surveys designed to obtain perceptions of disorder from residents. The objective measure of disorder was significantly correlated with the perception of disorder, which supports the belief that the two measures should be highly similar. However, after controlling for objective disorder, concentration of poverty and race composition of the neighborhood were significantly positively related to perceptions of disorder. This study suggests that perceptions of neighborhood disorder are influenced by both objective neighborhood disorder and neighborhood racial context. Several mechanisms have been proposed to describe the link between neighborhood context and PTD. These mechanisms include premature aging, behavioral responses, and stress. According to Wen (2006), “the ‘accelerated aging hypothesis’...specifles that psychological and physiological responses to demanding environmental stimuli operate as key mechanisms linking social conditions with health." The neighborhood environment may also affect risk of PTD by influencing certain behaviors, such as use of illicit drugs or alcohol, which could potentially L contribute to early aging (Farley, Mason, Rice et al., 2006). Cigarette smoking is a behavior that may be modified by neighborhood environment through stress levels, accessibility to cigarettes, or social tolerance of pregnancy smoking (Pickett, Wakschlag, Rathouz, Leventhal, & Abrams, 2002). Thus, cigarette smoking could potentially mediate associations between neighborhood conditions and PTD. However, neighborhood socioeconomic characteristics have been found to affect the risk of PTD for both Black and White women, independent of cigarette smoking, indicating that neighborhood conditions influence risk of PTD through more than just differing levels of pregnancy smoking (Ahern, Pickett, Selvin et al., 2003). Neighborhood context may influence stress levels. Unfavorable perceptions of one’s neighborhood may be a chronic stressor (Collins, David, Symons et al., 1998), and individuals living in lower SES neighborhoods have reported heightened vigilance for future potential threats (Feldman & Steptoe, 2004). High levels of anxiety, perceived stress, depression, and neutral or negative attitudes toward pregnancy have been associated with shorter gestations (Copper, Goldenberg, Das, Elder, Swain, Norman et al., 1996; Dole, Savitz, Hertz-Picciotto, Siega-Riz, McMahon, & Buekens, 2003; Steer, Scholl, Hediger, & Fischer, 1992). However, these findings should be interpreted carefully, since the definitions of the exposure may not be consistent and the studies may not differentiate acute and chronic stress. Stressful life events, which may be influenced by neighborhood context, have been linked to increased risk of PTD in some studies (Berkowitz & Kasl, 1983; Dole, Savitz, Hertz-Picciotto et al., 2003; Newton, Webster, Binu, Maskrey, & Phillips, 1979). Other studies, however, have not found this association (Stein, Campbell, Day, McPherson, & Cooper, 1987) or found it only in White women (Berkowitz & Kasl, 1983). Wadhwa (2001) reviewed the role of stress in PTD and concluded that maternal psychosocial stress could influence PTD through several pathways. These pathways include neuroendocrine processes, such as activation of the hypothalamic—pituitary—adrenal axis, immune/inflammatory processes, such as susceptibility to infections, or vascular processes, such as pregnancy-induced hypertension or preeclampsia. The previous research on objectively-measured neighborhood conditions and PTD suggests significant associations, but invites further studies to continue exploring these associations. Only a few studies have divided PTD by weeks of gestation (Janghorbani, Stenhouse, Millward et al., 2006; Reagan & Salsberry, 2005; Smith, Draper, Manktelow et al., 2007). Also, the vast majority of studies grouped all preterm births together, regardless of clinical circumstances. Of the three studies that addressed clinical circumstances, two restricted their study populations to include only spontaneous preterm births (Pickett, Ahern, Selvin et al., 2002; Smith, Shah, White et al., 2006). Only one study included both spontaneous and medically indicated preterm deliveries as separate groups (Dole, Savitz, Hertz-Picciotto et al., 2003). Examining the associations between neighborhood conditions and PTD, grouped by week of gestation at birth and by clinical circumstances, may elucidate the mechanisms(s) by which neighborhood context affects risk of PTD. Also, the use of perceptions of neighborhood environment, in addition to objective measures, will be useful for several reasons. Many of the previous studies of neighborhood perception and health have been cross-sectional, so causal inference has been difficult to establish. A prospective study may alleviate some of this difficulty. Also, the correlation between objective and subjective measures of neighborhood conditions has been a source of debate, and this needs further exploration. Lastly, if perceptions truly do reflect more than just objective neighborhood conditions, the associations with PTD may differ between the two measures. The following study was designed to shed light on the complex roles that neighborhoods play as potential determinants of PTD. 16 Perceived and objective neighborhood conditions: relations with preterm delivery AIMS AND HYPOTHESES The overall aim of this thesis was to examine the associations between risk of preterm delivery subtypes, grouped by clinical circumstances (spontaneous, medically indicated) and by week (<35, 35-36 weeks), in relation to two measures of neighborhood conditions: 1) perceived neighborhood disorder ascertained at mid-pregnancy; and 2) an objective measure of neighborhood deprivation based on 2000 US census data. Additional analyses considered the concordance/discordance of the neighborhood disorder and neighborhood deprivation measures. I hypothesized that, of the two measures, deprivation would be more strongly associated with PTD. Census tract information is an objective method used to summarize neighborhood conditions. In contrast, a woman's opinion of her neighborhood may be altered differentially by neighborhood conditions, by specific personality traits, or by individual characteristics such as income (Wen, Hawkley, & Cacioppo, 2006). MATERIALS AND METHODS Study Population This study included participants from the Pregnancy Outcomes and Community Health (POUCH) Study, a prospective cohort study (Holzman, Bullen, Fisher, Paneth, & Reuss, 2001). The POUCH study enrolled women in the 15-27th week of pregnancy from five Michigan communities from 1998 to 2004. These communities included urban, suburban, and rural areas. Eligibility criteria included matemal age 215 years, screened for maternal serum alpha- fetoprotein between 15 and 22 weeks' gestation, singleton pregnancy with no known congenital or chromosomal abnormalities, no history of pre-pregnancy diabetes, and English-speaking. Ascertainment of both a neighborhood disorder and deprivation score was also necessary for inclusion in this particular study. Data Collection At 15-27 weeks' gestation, participants met with a study nurse in their community. At this meeting, biological samples such as urine, plasma, and serum were collected. The participants also gave a detailed interview and took a self-administered questionnaire to obtain demographic information and information about their current pregnancy, reproductive history, health behaviors, and social and psychosocial experiences. Covariates Several covariates were considered in this study, and information on these variables was collected from the questionnaire. These variables were selected to reflect those commonly seen in the literature as potentially confounding the association between neighborhood conditions and PTD. Maternal age at delivery was modeled as a continuous variable (range: 15-48 years old). Race/ethnicity was initially comprised of 6 categories: non-Hispanic White; non-Hispanic Black; Asian; Hispanic; Native American; Other. Maternal education level considered both education and age. The categories were: <20 years old and