Socioeconomic context and gastroschisis: Exploring associations at various geographic scales

Social Science & Medicine 72 (2011) 625e633

Contents lists available at ScienceDirect

Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed

Socioeconomic context and gastroschisis: Exploring associations at various geographic scales Elisabeth D. Root a, b, *, Robert E. Meyer c, d, Michael Emch e, f a

Department of Geography, University of Colorado, Guggenheim, 260 UCB, Boulder, CO 80309, USA Institute of Behavioral Science, University of Colorado, Boulder, USA c North Carolina Birth Defects Monitoring Program, State Center for Health Statistics, USA d Department of Maternal and Child Health, University of North Carolina, Chapel Hill, USA e Department of Geography, University of North Carolina, Chapel Hill, USA f Carolina Population Center, University of North Carolina, Chapel Hill, USA b

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 10 December 2010

This study examines associations between area-level socioeconomic factors and the birth defect gastroschisis in order to further our understanding of the etiology of this condition. Specifically, this study explores how measuring socioeconomic conditions at different geographic scales affect the results of statistical models. A population-based case-control study of resident live births was conducted using data from the North Carolina Birth Defect Monitoring Program and the North Carolina composite linked birth files from 1998 through 2004. Neighborhood conditions potentially related to gastroschisis (poverty, unemployment, education, and racial composition) were measured using Census 2000 data and aggregated to several geographic scales. The BrowneForsythe test of homogeneity of variance was used to select the neighborhood size by examining the effect of neighborhood size on variation in gastroschisis rates. To examine our assumptions about neighborhood size and neighborhood effects on gastroschisis, we estimated a series of logistic regression and multilevel logistic regression models. The BrowneForsythe test suggested an optimal neighborhood size with a circular radius of approximately 2500 m, which was supported by the statistical analysis. Results indicate a weak association between living in a neighborhood characterized by high poverty and unemployment and an elevated risk of a gastroschisis-affected pregnancy after adjusting for individual-level risk factors. Cross-level interactions indicate that women in low poverty neighborhoods who do not rely on Medicaid have a significantly lower risk of gastroschisis. The choice of neighborhood scale influences model results suggesting that socioeconomic processes may influence health outcomes variably at different scales. Ó 2010 Elsevier Ltd. All rights reserved.

Keywords: Gastroschisis Birth defects Neighborhood scales Spatial scale Socioeconomic Poverty factors USA

Introduction Public health research has long recognized that people residing in different geographic areas have differing health outcomes. Individuals with similar risk factors may live in the same geographic area, producing larger area-level patterns of disease. At the same time, an individual’s proximal environment exerts a variety of social, psychological and biological pressures which can directly influence health. Geographic variation in health and disease may therefore be due to differences in the kinds of people who live in these places (composition) or differences in the physical or social environment (context). The resurgence of

* Corresponding author. Department of Geography, University of Colorado, Guggenheim 110, 260 UCB, Boulder, CO 80309, USA. Tel.: þ1 303 492 4794. E-mail address: [email protected] (E.D. Root). 0277-9536/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2010.11.025

research on social determinants of health and the growing acceptance of the utility of an ecological perspective have lead to a large body of literature examining the role of “neighborhood effects” on health (Diez Roux, 2001; Kawachi & Berkman, 2003). Recent research in this area has moved beyond simply distinguishing between compositional and contextual effects and has begun to focus more on the interrelationships between characteristics of people and of places as these relate to health outcomes using a variety of modeling techniques (Jackson & Mare, 2007; Morenoff et al., 2007; Timberlake, 2007; Weden, Carpiano, & Robert, 2008). Researchers have also begun to explore the impact of socioeconomic conditions experienced at different scales (e.g., census-block vs. city unemployment) (Haynes, Daras, Reading, & Jones, 2007; Lovasi et al., 2008). While neighborhood and health studies can be used to confirm hypothesized mechanisms by which area-level characteristics

626

E.D. Root et al. / Social Science & Medicine 72 (2011) 625e633

affect health, they can also be useful in exploratory etiological research. When the cause of a disease is unknown or speculative, area-level characteristics can provide important clues about plausible social, psychological and biological mechanisms that may influence the disease. Many birth defects are thought to have complex or unknown origins and most epidemiological studies are exploratory in nature, testing possible associations between these birth defects and socioeconomic inequalities that are correlated with poor health outcomes (Brown, 1997). Studies in the United States and other countries have revealed that area-level measures of socioeconomic status (SES) are associated with several birth defects including: orofacial clefts (Carmichael, Nelson, Shaw, Wasserman, & Croen, 2003; Vrijheid et al., 2000), neural tube defects (Vrijheid et al., 2000; Wasserman, Shaw, Selvin, Gould, & Syme, 1998) and heart defects (Carmichael et al., 2003). However no studies to date examine the effect of area-level SES measures on gastroschisis, a serious abdominal wall defect that has increased in prevalence over the past several decades (Calzolari et al., 1995; Hougland, Hanna, Meyers, & Null, 2005; Penman, Fisher, Noblett, & Soothill, 1998; Williams, Kucik, Alverson, Olney, & Correa, 2005). Gastroschisis is a rare birth defect, affecting approximately 3.82 per 10,000 live births each year in the United States (NBDPN, 2007). Its causes and developmental origin are largely speculative or unknown (Curry, McKinney, Thornton, & Stringer, 2000; Feldkamp & Botto, 2008). Animal models currently indicate no specific mechanism, hindering our ability to attribute gastroschisis to specific environmental teratogens (Drongowski, Smith, Coran, & Klein, 1991; Feldkamp & Botto, 2008). Furthermore, the role of specific genetic factors is unclear. Familial occurrence of gastroschisis has been reported, as has concordance in monozygotic twins, though the reoccurrence risk appears to be low, suggesting that genetic factors are only responsible for a small fraction of all cases or the possibility of geneeenvironment interactions (Opitz & Pysher, 2008; Torfs, Christianson, Iovannisci, Shaw, & Lammer, 2006). Young maternal age has consistently been identified as a risk factor for gastroschisis. Studies show that the rate of gastroschisis among infants of mothers less than 20 years of age is between three and six times higher than the rate among mothers 25 years and older (Calzolari et al., 1995; Forrester & Merz, 1999; Hougland et al., 2005; Torfs, Curry, & Roeper, 1990; Werler, Mitchell, & Shapiro, 1992). This consistent pattern has lead many investigators to search for social or environmental factors to which younger women might more likely be exposed. Results are inconsistent, but recent studies suggest that young maternal age may be an indicator of obstetrical high risk, possibly correlated to social deprivation (Torfs, Velie, Oechsli, Bateson, & Curry, 1994; Vrijheid et al., 2000), environmental risks including malnutrition (Lam, Torfs, & Brand, 1999; Torfs, Lam, Schaffer, & Brand, 1998; Waller, et al., 2007), poor health care, increased consumption of medical or social drugs (Draper et al., 2008; Werler et al., 1992), and chemical exposures, (Dolk et al., 1998; Torfs, Katz, Bateson, Lam, & Curry, 1996). In a previous study, we reported the presence of a localized cluster of gastroschisis in North Carolina identified using spatial cluster analysis (Reference details withheld). While this analysis made no attempt to discover the cause of the cluster, we posed several hypotheses and recommended that future research explore these hypotheses in more detail. We suggested the cluster might be related to the impact of poor socioeconomic conditions in the region, since the cluster encompassed an area which has experienced a great deal of socioeconomic change due to massive downsizing of the textile industry. Such socioeconomic conditions could produce an environment in which women experience greater psychosocial stress or are exposed to poor environmental

conditions (e.g., pollutants), both which could increase risk for gastroschisis. The current study explores this hypothesis by examining the impact of individual- and neighborhood-level socioeconomic status on gastroschisis births and, as such, furthers the prior study by examining one hypothesis generated by that study. Given that the etiology of gastroschisis is largely unknown, examining the influence of neighborhood-level SES measures may assist in the generation of hypotheses about the underlying causal factors associated with socioeconomic factors. In this study we examine the relationship between gastroschisis and five areabased measures of SES after adjusting for known individual-level risk factors. Two study questions guided this research project: 1) To what extent are neighborhood-level SES variables related to the risk of a gastroschisis birth? 2) Does this relationship differ when different spatial scales are used to define neighborhoods? The second research question was introduced in order to address a major criticism that has arisen in the neighborhoods and health literature. The concept of “neighborhood” is complex and critics suggest that relevant neighborhoods need to be carefully defined and operationalized based on the underlying processes and causal mechanisms presumed to affect the health outcome being studied (Diez Roux, 2001). This means that the spatial scale at which neighborhood factors influence health may vary based on both the socioeconomic measure and health outcome used. In this study, we attempt to address these critiques by empirically defining neighborhoods and comparing model results across different geographic scales. Methods Birth defects data Birth defect and maternal characteristics were obtained from the North Carolina Birth Defects Monitoring Program (NCBDMP). The NCBDMP is a population-based active surveillance system that collects data on congenital malformations diagnosed within the first year of life among all live births in North Carolina. We conducted a retrospective case-control study of North Carolina resident live births with gastroschisis between 1/1/1999 and 12/31/ 2004. To identify infants with gastroschisis, we searched the NCBDMP database using the Centers for Disease Control and Prevention modified British Pediatric Association code for gastroschisis (756.710). Singleton infants with or without other structural malformations, and without a chromosomal abnormality were eligible for inclusion as cases in this study. All liveborn singleton infants with a birth certificate, without a structural birth defect, and born between 1/1/1999 and 12/31/2004 to North Carolina resident mothers were eligible to be controls. A random sample of all eligible infants was drawn from the North Carolina Composite Linked Birth File. Terminations of pregnancy and fetal deaths were not included in the study, as these comprise only a small fraction of gastroschisis cases in North Carolina. A total of 264 cases and 12,488 controls were selected for analysis. The data included residential address at birth, which was used to geocode cases and controls. A majority of the geocoding was completed by the Health & Spatial Analysis Unit at the NC State Center for Health Statistics (SCHS), using Geographic Data Technology (GDT) and parcel data from the NC Department of Transportation. Records not matched by the SCHS were geocoded using a multi-stage geocoding method and different web-based geocoding services (Lovasi, Weiss, Hoskins, Whitsel, Rice, Erickson et al., 2007). Using this process, we matched 242 of the 264 cases (91.7%) and 11,651 of the 12,488 controls (93.3%). Records with an invalid or unmatched address were removed from the analysis.

E.D. Root et al. / Social Science & Medicine 72 (2011) 625e633

627

Fig. 1. Example of aggregating census block group data, weighting by census block populations.

We chose to include such a large number of controls in order to accurately represent the geographic distribution of births across the state during the study period. Given that there were approximately 700,000 births over the time period of the study, our sample of controls only represents 2% of all births during this time. Had we sampled a smaller proportion of controls, the representativeness of the controls to all live births in the state might be questionable. Since we were examining a process that may be affected by population distribution, we settled on a 2% sample in order to ensure adequate geographic distribution and representativeness of the controls. The data also contained potential covariates from the linked birth files including: mother’s age, race and ethnicity, marital status, number of prior births, month prenatal care began, mother’s smoking status, and whether or not Medicaid paid for the delivery. Medicaid is a government-funded health insurance program for individuals on low incomes. These individual-level attributes are possible confounders to the neighborhood environmentegastroschisis relationship, and are reliably measured on the birth record. Descriptive statistics of unmatched versus matched records were used to examine differences between the two groups and we found some minor differences in race, parity and Medicaid status.

various birth outcomes (Carmichael et al., 2003; Krieger et al., 2003; Messer et al., 2008; Wasserman et al., 1998) six census variables were used to estimate neighborhood-level socioeconomic characteristics: percent of the population living below 100% and 200% of federal poverty level, percent of the population with less than a high school education, percent of the population unemployed, and percent of the population reporting African American race. These measures quantify several socioeconomic domains that effect health: “education”, “employment”, “poverty”, and “racial composition”. While some researchers have advocated the use of indices to measure the cumulative effects of several different measures of SES (Carmichael et al., 2003; Messer et al., 2008) others have found that estimates of effects detected using a single variable measure of poverty were similar to those based on indices or composite measures (Krieger et al., 2003). For this study, we chose to examine single variable measures because we were interested in the separate effects of each SES domain on gastroschisis outcomes. Each census measure was classified into quartiles based on the distribution among controls and each census tract, block group and block given a corresponding categorical quartile score (1 ¼ lowest, 4 ¼ highest). This was done for ease of interpretation in logistic regression models.

Socioeconomic data

Construction of neighborhood variables

Socioeconomic variables for census tracts, block groups and blocks were obtained from the 2000 Census of Population and Housing Data from the U.S. Census Bureau. Following the approach of several previous studies that examine area-level effects on

Prior research linking neighborhood SES indicators to health outcomes has relied heavily on geopolitical areas, such as counties or census tracts, to approximate “neighborhoods” or “communities”. While they may be a practical alternative given the

628

E.D. Root et al. / Social Science & Medicine 72 (2011) 625e633

availability of SES data for these geographic areas, they may not truly capture a person’s proximal environment or properly measure how that environment affects health (Diez Roux, 2001). Therefore, it is important to develop and test hypotheses regarding the precise geographic area that is relevant for a specific health outcome in order to strengthen inferences regarding area effects. At the very least, several definitions of neighborhood should be specified, and model results from each evaluated to determine sensitivity of results to changes in neighborhood definition. To address this issue, cases and controls were geocoded to neighborhoods defined in several different ways. We first assigned cases and controls to year 2000 census block groups and tracts. Next, we developed a set of “neighborhoods” by creating circular windows of various sizes around each study subject. Based on the size of the study area and the distribution of the population, we set the minimum size to a 1000-m radius neighborhood and increased the size stepwise by 500 m until at 5500 m size was reached. This resulted in 10 different neighborhood sizes from which to select the scale at which to conduct this analysis. The neighborhood level social variables were estimated by aggregating census block group data by each of the circular windows. In cases where the circular window contained only a portion of a census block group, social variables were weighted by the proportion of the population from that census block group that was encompassed by the window. The circle in Fig. 1 illustrates a 2500 m circular neighborhood while the dark black lines represent census block groups. We summed the population of the census blocks contained in the circular neighborhood (represented by light dashed lines and red dots) and divided this number by the total census block group population. This proportion was then applied to the census block group variables. Selection of neighborhood scale In order to choose the neighborhood size with which to conduct our analysis, we modified a method developed by Ali et al. (2005) which examines the variation of the variable of interest (e.g., disease incidence or known risk factors) at several different geographic scales. Ali et al applied Hartley’s test of homogeneity of variance (Fmax) to disease incidence data aggregated to various neighborhood scales. Hartley’s test is sensitive to departures from normality. Since the aggregate neighborhood populations and birth defect incidence rates were both non-normally distributed, we chose to use the more robust BrowneForsythe (FBF) test to examine the variation of birth defect rates at different scales. Whereas Hartley’s test uses the mean, the BrowneForsythe test is based on absolute deviations from the median or trimmed mean, which performs better with highly skewed or heavily tailed data (Brown & Forsythe, 1974). Similar to Hartley’s test, the underlying assumption of this method is that smaller neighborhoods will have a high variance value while larger neighborhoods will have a low variance. A high variance value means that the data are local or individualistic while a low variance means that they capture a more global process. The optimal neighborhood (within the constraints of the statistical test and the limited set of neighborhoods we are evaluating) ensures that the aggregate disease incidence data is neither local nor global, but somewhere in between. We make the assumption that the “optimal” neighborhood is the one that does not simply capture the characteristics of an individual (the local environment) but rather the characteristics of the larger area within which the individual lives. At the same time, the optimal neighborhood should not be so large as to capture the characteristics of the larger global environment that does not directly impact the health of an individual. The test statistic, FBF, was calculated as:

Pt

ni ðDi  DÞ2 ðt  1Þ 2 Pn1  Dij  D1 j¼1

i¼1

FBF ¼ P t

i¼1

ðN  tÞ where: ni ¼ number of observations from neighborhood i. N ¼ n1 þ n2 ¼ overall size of combined samples. t ¼ number of neighborhoods. yij ¼ sample observation j from neighborhood i. yi ¼ median of sample data from neighborhood i. Di ¼ average of the ni absolute deviations from neighborhood i. D ¼ average of all N absolute deviations. Dij ¼ jyij  yi j ¼ absolute deviation of observation j from treatment i median. The test assumes that the variances are equal under the null hypothesis. The critical value was calculated using an F-distribution with (te1, Net) degrees of freedom. A threshold value of a ¼ 0.05 was used to test for significance. We first compared the variance in gastroschisis rates calculated for all neighborhood sizes to rates calculated for the 5500 m neighborhoods (FBF1). Using an iterative process, we then dropped the neighborhoods in turn, starting with the smallest (from 1000 to 4500 m) ending by comparing the variances calculated for the two largest neighborhood sizes (5000 and 5500 m). Next, we compared gastroschisis rates calculated for all neighborhood sizes to the rate calculated for the 1000 neighborhoods (FBF2). Using the same iterative process, we dropped the neighborhoods in turn, starting with the largest (from 5500 to 2000 m) ending by comparing the variances between the two smallest neighborhood sizes (1000 and 1500 m). A significant value of FBF1 indicates that the neighborhood does not reveal the global structure of data; in essence each neighborhood is so small that it is only capturing disease dynamics for a small group of individuals. In contrast, a significant value in FBF2 implies that the neighborhood data are not individualistic; they are so large that local level disease dynamics are “washed out” and undetectable. The neighborhoods between the lower and the upper limits identify a spatial scale, and therefore a distribution, at which local level variation is still detectable but that also captures larger disease dynamics. There are, of course, many different ways for defining a neighborhood, some of which are distance based and others based on social connections or historical urban boundaries. The technique employed in this paper (FBF test) is one method of choosing the optimal neighborhood within the constraints of the statistical methods we employ and for the finite set of neighborhoods we evaluate. It is not, of course, the only method and may not be appropriate for choosing neighborhoods constructed using methods other than buffering.

Logistic regression To estimate the variation in risk of gastroschisis-affected pregnancy associated with differences in neighborhood SES, maximum likelihood estimates of odds ratios (OR) and 95% confidence intervals (CI) were calculated from logistic regression models. The combined influence of individual- and neighborhood-level indicators was examined to determine whether risk for women living in a lower SES neighborhood varied even after adjusting for individual characteristics. Considered as potential confounders were age (<20 years of age, 20e24 years, and 25 years or more), race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic and other), parity (no prior births vs. one or more prior births), smoking during pregnancy (yes, no) and Medicaid status (delivery paid for by Medicaid vs. other payer source). Only those covariates that

E.D. Root et al. / Social Science & Medicine 72 (2011) 625e633

showed a significant risk associated with gastroschisis during univariate and multivariate analyses were included in the neighborhood-level analysis. We estimated a set of logistic regression models which included area-level SES measures defined for six different neighborhood sizes: 2000, 2500, 3000, and 3500 m radius, census tracts and census block groups. Each SES measure was entered individually to the model, resulting in 5 separate models for each neighborhood definition. Results from these models were examined to compare our choice of neighborhood size obtained from the FBF test statistic to other neighborhood scales. For the models using census tracts and block groups as neighborhoods, we estimated multilevel logistic regression models with a fixed slope value for each predictor variable and random tract or block group intercepts. This was not necessary for models with neighborhoods defined using circular windows since these neighborhoods were created for each individual case and control. Univariate analysis, logistic regression and multilevel modeling were conducted in R v2.7.2 and WinBUGS v1.4.3 software.

Results The data variances for gastroschisis rates among the sample population show a declining trend with an increase in neighborhood size. The test results for homogeneity of variance for gastroschisis rates under various neighborhood sizes are listed in Table 1. The FBF1 test statistic at the level a ¼ 0.05 shows a neighborhood size of approximately 2000 m is optimal while the FBF2 test statistic shows a neighborhood size of approximately 3000 m is optimal. Below 2000 m neighborhood data may only capture the characteristics of each individual, while above 3000 m the neighborhoods are so large they do not capture the influence of an individual’s proximal environment. Given these results, we believe that a neighborhood size of approximately 2500 m should be used for modeling the local variation of gastroschisis. Table 2 shows the prevalence of covariates among cases and controls and risk estimates for gastroschisis as measured by odds ratios. Relative to control mothers, mothers of infants with birth defects were more likely to be young (<20 years of age), white, have no prior births (nulliparous), have smoked during pregnancy and have had their birth paid for by Medicaid. The logistic regression of individual-level covariates only, young maternal age (age <20 years) showed the strongest association with increased risk for a gastroschisis birth (OR ¼ 2.1; 95% CI ¼ 1.55e3.86) while maternal age over 25 years, black race and parity showed significant protective effects. In addition, mothers whose birth was paid for by

629

Table 2 Prevalence (%) and odds ratios and 95% confidence intervals for gastroschisis adjusting for individual-level covariates, North Carolina, 1998e2004. Prevalence Cases (%)

Multivariate Controls (%)

ORa

95% CIa

Maternal age < 20 20e24 25e29  30

101 92 37 12

(41.7) (38.0) (15.3) (5.0)

1385 3134 3177 3955

(11.9) (26.9) (27.3) (33.9)

2.10 Reference 0.46 0.14

1.55e2.86 e 0.31e0.68 0.07e0.24

Maternal race White Black Hispanic Other

162 36 37 7

(66.9) (14.9) (15.3) (2.9)

7167 2695 1356 433

(61.5) (23.1) (11.6) (3.7)

Reference 0.39 0.90 0.73

e 0.26e0.56 0.60e1.31 0.31e1.47

Birth parity No prior births 1 or more prior births

78 (32.2) 164 (67.8)

6805 (58.4) 4846 (41.6)

Reference 0.58

e 0.43e0.78

180 (74.4) 61 (25.2)

10125 (86.9) 1515 (13.1)

Reference 1.50

e 1.08e2.07

164 (67.8) 78 (32.2)

4638 (39.8) 7013 (60.2)

Reference 1.66

e 1.23e2.27

Smoker vs. nonsmoker Non-smoking Mother reported smoking Birth paid for by Other payer Medicaid

a OR ¼ Odds ratio, adjusted for other maternal characteristics; CI ¼ confidence interval.

Medicaid (a proxy for low income), showed an increased risk for a gastroschisis birth (OR ¼ 1.66; 95% CI ¼ 1.23e2.27). We estimated neighborhood SES models for all neighborhoods between 1500 and 4000 m radius and neighborhoods defined using census block groups and tracts (results not shown but included in the online Supplemental material available only with the electronic version of this paper), but only the models using the 2500 and 3000 m radius neighborhoods showed significant associations with gastroschisis. The purpose of estimating models for all neighborhoods between 1500 and 4000 was to examine our assumptions regarding the use of the BrowneForsythe test. Table 3 presents the crude and adjusted odds ratios of SES measures in relation to gastroschisis. Crude ORs for gastroschisis associated with single indicators of neighborhood-level SES were uniformly elevated. For both the 2500 and 3000 m radius neighborhood models, crude odds ratios were significantly elevated for residence in a poverty neighborhood (where at least 20% of the residents were living below 200% of the federal poverty level or 10% were living below 100% of federal poverty level), a high unemployment neighborhood (where at least 4% of

Table 1 Descriptive statistics and results for BrowneForsythe test for the gastroschisis incidence rates for various neighborhoods, North Carolina, 1998e2004. ra

1000 1500 2000 2500 3000 3500 4000 4500 5000 5500

Population size

Incidence rate/100000 births

BrowneForsythe

Min

Max

Mean

Median

Min

Max

Mean

Median

Variance

FBF1

DF1

CV

FBF2

DF2

CV

1 1 1 2 3 3 5 6 7 8

195 371 501 577 712 912 1127 1363 1543 1715

24.7 49.8 82.3 120.0 163.4 210.9 262.4 320.7 380.3 436.8

6 33 57 85 119 153 192 237 286 332

0 0 0 0 0 0 0 0 0 0

50000 25000 14285 11764 9523 6896 6060 5128 4255 4761

283.2 240.6 219.7 211.0 212.5 216.1 215.6 213.2 208.6 206.0

0 0 0 0 0 0 0 0 0 0

3338029 1205870 575775 366069 283953 224783 179831 145022 115830 100425

9.89 3.42 1.92 1.78 1.07 1.53 1.67 1.31 0.35 e

9 8 7 6 5 4 3 2 1 e

1.88 1.94 2.01 2.10 2.21 2.37 2.61 3.00 3.84 e

e 0.11 0.27 0.63 0.93 2.55 7.13 7.94 9.53 9.89

e 1 2 3 4 5 6 7 8 9

e 3.84 3.00 2.61 2.37 2.21 2.10 2.01 1.94 1.88

DF ¼ degrees of freedom. CV1 and CV2 ¼ critical values at 95% confidence level for FBF1 and FBF2. Bold figures in the FBF1 and FBF2 are the upper and lower limit of neighborhood scale, and the bold figure in “r” column is the choice of neighborhood size used in this analysis. a r ¼ size of the neighborhood radius in meters.

630

E.D. Root et al. / Social Science & Medicine 72 (2011) 625e633

Table 3 Crude odds ratios, adjusted odds ratios, and their 95% confidence intervals for SES measures in relation to gastroschisis for 2500 m radius and 3000 m radius neighborhoods, North Carolina, 1998e2004. 2500 m radius neighborhood

3000 m radius neighborhood Adjusted ORa (95% CI)

Crude OR (95% CI)

Adjusted ORa (95% CI)

100% of federal poverty level Q2 1.34 (0.88e2.06) Q3 2.26 (1.55e3.36) Q4 1.76 (1.19e2.65) b AIC

0.85 (0.56e1.33) 1.33 (0.90e2.00) 1.14 (0.74e1.76) 2092.6

1.53 (1.01e2.32) 2.03 (1.38e3.03) 1.81 (1.22e2.72)

0.98 (0.64e1.50) 1.18 (0.79e1.78) 1.17 (0.77e1.81) 2097.3

200% of federal poverty level Q2 2.00(1.29e3.18) Q3 3.34 (2.23e5.17) Q4 2.00 (1.29e3.17) AICb

1.21 (0.77e1.93) 1.85 (1.19e2.83) 1.15 (0.73e1.88) 2087.3

2.00 (1.29e3.17) 3.31 (2.21e5.11) 2.03 (1.31e3.22)

1.20 (0.76e1.93) 1.79 (1.18e2.80) 1.20 (0.76e1.95) 2088.3

Unemployment Q2 Q3 Q4 AICb

1.40 (0.92e2.15) 1.61 (1.08e2.44) 1.27 (0.82e1.98) 2098.6

2.20 (1.45e3.42) 2.58 (1.72e3.97) 2.03 (1.33e3.17)

1.81 (1.18e2.83) 1.89 (1.25e2.94) 1.50 (0.96e2.39) 2088.2

Less than a high school education Q2 1.69 (1.11e2.59) Q3 1.97 (1.32e3.00) Q4 2.25 (1.52e3.40) AICb

1.09 (0.71e1.69) 1.01 (0.66e1.57) 1.09 (0.72e1.69) 2097.6

1.47 (0.98e2.25) 1.84 (1.25e2.77) 2.05 (1.40e3.06)

0.94 (0.62e1.45) 0.94 (0.62e1.44) 0.98 (0.65e1.50) 2098.8

Percent black Q2 Q3 Q4 AICb

1.06 (0.73e1.55) 1.03 (0.71e1.50) 1.24 (0.83e1.84) 2093.1

1.10 (0.77e1.59) 1.07 (0.74e1.54) 1.07 (0.74e1.54)

1.31 (0.90e1.90) 1.15 (0.79e1.69) 1.27 (0.84e1.91) 2096.7

Crude OR (95% CI)

1.73 (1.15e2.63) 2.25 (1.53e3.38) 1.74 (1.16e2.67)

0.90 (0.63e1.30) 0.95 (0.66e1.36) 1.05 (0.74e1.49)

Each socioeconomic measure was estimated separately. Reference category ¼ 1st quartile. a Odds ratios adjusted for maternal age, race/ethnicity, parity, smoking and medicaid status. b AIC ¼ Akaike information criterion; a lower AIC score implies a better model fit. AIC scores in bold indicate the lowest score across models and, thus, the best fit model.

the residents were unemployed) and residence in a less educated neighborhood (where at least 12% of residents have less than a high school education). Adjustment for maternal age, race/ethnicity, parity smoking and Medicaid status, resulted in odds ratios that were uniformly lower than their crude counterparts. For both the 2500 and 3000 m neighborhood, residence in a neighborhood in the 3rd quartile of poverty (where at 30%e40% of the residence were living below 200% of the federal poverty level) was associated with increased odds of a gastroschisis birth, compare with residence in the 1st quartile. The 2500 m neighborhood showed the strongest association (OR ¼ 1.85; 95% CI ¼ 1.19e2.83). Residence in a neighborhood in the 2nd or 3rd quartile of unemployment (where 4%e7% of residents were unemployed) was associated with increased odds of a gastroschisis birth, compared with residence in the 1st quartile. For this SES measure, the 3000 m neighborhood showed the strongest association (OR ¼ 1.89; 95% CI ¼ 1.25e2.94). The difference in model fit between the 2500 and 3000 m neighborhoods is marginal, especially for the federal poverty model. We used the Akaike Information Criterion (AIC) to compare the fit of different models, and saw a 1 point improvement with the 2500 m neighborhood model for poverty and a 10 point improvement with the

3000 m neighborhood model for the unemployment measure. These differences are minimal, suggesting that the effect sizes are similar for the two spatial scales. To investigate the possibility that the risk of having a gastroschisis-affected pregnancy among women of lower SES differed depending on the neighborhood social condition in which they lived, we evaluated a cross-level interaction which combined individual Medicaid status with neighborhood poverty (Table 4). Crude odds ratios revealed that women whose birth was paid for by Medicaid were at highest risk for a gastroschisis-affected pregnancy, regardless of neighborhood SES. When odds ratios were adjusted for individual maternal characteristics, the 95% confidence intervals were still greater than 1 for women in the Medicaid/high poverty, Medicaid/low poverty and non-Medicaid/high poverty groups, indicating a significantly higher risk of gastroschisis for these groups compared to the reference group (non-Medicaid/low poverty). Discussion This study indicates a weak association between residence in a lower SES neighborhood, as measured by poverty and unemployment, and the risk of having a gastroschisis-affected pregnancy,

Table 4 Crude odds ratios, adjusted odds ratios, and their 95% confidence intervals for combined individual and neighborhood indicators of socioeconomic status, North Carolina, 1998e2004. Neighborhood povertya

Individual medicaid status

Cases

Controls

OR

Adjusted ORb

Low Low High High

non-Medicaid Medicaid non-Medicaid Medicaid

30 57 48 107

4228 1594 2785 3044

Reference 5.03 (3.25e7.96) 2.43 (1.54e3.88) 4.95 (3.34e7.57)

Reference 2.09 (1.30e3.43) 1.72 (1.09e2.78) 2.45 (1.57e3.91)

a b

High poverty neighborhood was defined as areas where 30% or more of the residents lived below 200% of the federal poverty level. Odds ratios adjusted for maternal age, race, parity and smoking status.

E.D. Root et al. / Social Science & Medicine 72 (2011) 625e633

even after adjusting for individual-level risk factors. In addition, if Medicaid status is considered a proxy for individual-level SES, then both individual and neighborhood measures of SES combined are associated with an increased risk for a gastroschisis birth. Adjusted odds ratios were significantly elevated for the 3rd quartile of poverty and 2nd and 3rd quartiles of unemployment measures, though not the 4th, which included neighborhoods with the highest poverty and unemployment rates. This finding is difficult to interpret, though it indicates that gastroschisis births disproportionately affect women living lower-middle class neighborhoods. In addition, cross-level interactions indicate that women whose birth was paid for by Medicaid and who lived in a high poverty neighborhood had the greatest risk of a gastroschisis-affected birth, though this elevated risk was not significantly different from women in low poverty/Medicaid and high poverty/non-Medicaid and high poverty/Medicaid groups. Thus, our results only indicate that women residing in low poverty neighborhoods who do not rely on Medicaid have a significantly lower risk of gastroschisis. Regarding the cross-level interaction and multilevel analysis, we acknowledge that this analysis only partially controls for individuallevel effects since the variable used to measure individual-level SES was not used to create the neighborhood-level SES indicators (e.g., we did not aggregate Medicaid status up to the neighborhood-level). Thus, the effect of the neighborhood-level SES variables we observed may still be due to individual-level effects that are not accounted for by the Medicaid variable. This limitation may help explain why we did not find significantly elevated risk for gastroschisis for women residing in a neighborhood the 4th quartile (highest) of poverty. The fact that we did not find a significant effect for the 4th quartile may indicate that there are unmeasured individual-level effects that compromise model fit. This study is the first to examine the association between neighborhood-level SES and the risk of gastroschisis births. Only one previous study incorporated area-level measures of socioeconomic status while simultaneously controlling for individual-level confounders and this study grouped several digestive system birth defects (Vrijheid et al., 2000). Using the Carstairs deprivation index, Vrijheid et al. found a significant increase in risk of digestive system defects in the most deprived communities compared to the most affluent communities. Very few studies have even examined the relationship between individual-level socioeconomic status and risk of gastroschisis. Only one other study has reported a significant positive association between lower individual SES (family income) and gastroschisis (Torfs et al., 1994). Our study supports prior research as we found an elevated risk of gastroschisis among women whose birth was paid for by Medicaid (a proxy for low SES) and who lived in a high poverty community. Given our findings and the results of prior research, can we begin to develop theories about plausible links between specific socioeconomic features of the neighborhood and gastroschisis? While we certainly do not have data to suggest specific causal relationships, we can begin to develop hypotheses that merit further epidemiological research. For example, while no studies have examined gastroschisis specifically, prior research on birth outcomes suggests that women who experience high levels of psychosocial stress are at greater risk for preterm and low birth weight births (for a review of this literature see Hobel, Goldstein, and Barrett (2008). While the causal mechanisms behind this are not entirely clear, some researchers suggest that chronic psychosocial stress stimulates the production of cortisol in the mother’s system which may cause the developing fetus to mount a stress response which can adversely affect fetal development (Diego et al., 2006; Field et al., 2006; Hobel et al., 2008). In our sample, women living in low poverty areas who had private insurance had significantly lower risk for a gastroschisis birth compared to other groups.

631

Women may find unfavorable neighborhood characteristic stressful (e.g., high crime rates, racial or economic discrimination or poor access to necessary health or municipal services). Perhaps women exposed to both individual-level economic stress and poor proximal environments have higher overall or chronic levels of stress. To further understand this hypothesized relationship, future research could explicitly examine psychosocial and physiological reactions to exposure to poor environments and the effect on gastroschisis and other birth defects. Another possible explanation for the relationship between lower SES neighborhoods and gastroschisis is shared physical environments that place residents at higher risk of exposure to some contaminant that may induce gastroschisis. While the association between socioeconomic inequalities and increased mortality or morbidity is well-established (Marmot & Wilkinson, 2006), the question of whether people who are socially disadvantaged live in more polluted or hazardous environments is still debated (Bowen, 2002; Brown, 1995; Morello-Frosch & Shenassa, 2006). Associations between SES and environmental pollution imply the potential for socioeconomic confounding (or effect modification) in epidemiological studies. Thus, we must be careful in interpreting results from neighborhood SES studies which examine health outcomes that may be affected by environmental contaminants when potential sources of those contaminants are not taken into consideration. Data on some environmental risk factors, such as landfill and hazardous waste sites, are publicly available and will be incorporated into future analyses of the present data. This study also demonstrates the usefulness of using the BrowneForsythe test (FBF) to evaluate the various neighborhood sizes at which to study a disease outcome. The causes of gastroschisis, like many birth defects, are complex and multifactoral and may include not only maternal characteristics and behaviors but also environmental teratogens and genetic factors (Curry et al., 2000). The FBF test statistic can assist researchers in finding the geographic scale at which neighborhood-level measures are most strongly associated with disease outcomes. Finding the scale at which such associations occur may offer new etiological clues, help to generate hypotheses about causal mechanisms and, eventually, lead to an understanding of how psychosocial stressors such as poverty and unemployment affect gastroschisis risk. While this method does not necessarily provide researchers with the exact scale at which socioeconomic factors influence or affect health outcomes, we believe it can uncover a set of geographic scales that are relevant for assessment of health variation, allowing researchers to hypothesize why and how those scales are important. We feel this is an important step in neighborhood and health research. Results from the logistic regression analysis using different sized neighborhoods appear to confirm the use of a 2500 m radius neighborhood size for studying gastroschisis, thereby validating the FBF analysis. Also of note is that census block group and census tract based socioeconomic measures did not detect significant arealevel SES effects, even when 2500 and 3000 m radius areas did. This is most likely due, in part, to the small sample size. When the number of observations in each census tract or block group is very small, not enough variance exists between these level-2 groups to produce a statistically significant contextual or group-level effect. Small sample size is often a challenge in public health research, especially when examining rare conditions. Geopolitical boundaries, though convenient and easy to use, may then not be the optimal way to measure a person’s proximal environment or properly measure how that environment affects health, especially if employing a multilevel analysis. Many studies of area or neighborhood effects on health use census boundaries and may find spurious results if sample size is small. This, in conjunction with the possible misspecification bias that can occur if an individual’s

632

E.D. Root et al. / Social Science & Medicine 72 (2011) 625e633

neighborhood is incorrectly defined may lead researchers to incorrect conclusions about the nature of contextual area-level effects. Socioeconomic processes do influence health outcomes at different scales. In our study, the poverty and unemployment measures showed the strongest association with gastroschisis risk at different neighborhood sizes, suggesting that researchers may not be able to use the same size neighborhoods to examine the effect of all area-level SES measures. One potential limitation of our study is that we use residence at the time of birth to define neighborhood-level SES measures. Studies have shown that between 25 and 30 percent of women change residence between conception and birth (Fell, Dodds, & King, 2004; Khoury et al., 1988). However, a majority of these moves appear to be local (e.g. within the same city or county), to areas with a similar socioeconomic make-up (Fell et al., 2004; Khoury et al., 1988) and the characteristics of women who move are similar to those who do not (Canfield, Ramadhani, Langlois, & Waller, 2006). Caution should be exercised when interpreting the results of geographic studies that use maternal residential address at delivery, especially when using residence to develop and assign area-level variables to study subjects. Related to the limitation above, this study uses cross-sectional data which cannot account for changes in residence or changes in the SES environment over time. We used 2000 Census data to measure SES, which were collected within the 1999e2004 timeframe from which our sample was drawn. It is our belief that the SES environment changes gradually over time and that SES data from 2000 will accurately capture the neighborhood SES environment between 1999 and 2004. In addition, cross-sectional studies are problematic because many conditions have a long time lag between effect of neighborhood and health outcome, neighborhood effects may be incorrectly attributed to current social conditions. For most birth defects the lag time between the critical period of exposure and diagnosis is relatively short so we believe a crosssectional design is adequate for examining the relationship between neighborhood exposure and health outcome. Finally, we would like to acknowledge the small sample size of this study, especially when classifying cases into neighborhood quartiles. Gastroschisis is rare and the etiology is highly uncertain, so we believe the sample size in our study is sufficient for an exploratory study of this nature. Previous studies have used much smaller sample sizes and generated compelling results that have stimulated additional research with more focused research questions. While this study has a larger sample of cases than many other studies examining the same defect, caution should be taken when interpreting the results. In summary, we have identified both individual-level and neighborhood-level socioeconomic factors that contribute to the risk of a gastroschisis-affected pregnancy. Our findings indicate that neighborhood-level socioeconomic factors exert an independent effect on gastroschisis. While gastroschisis has increased in North Carolina over the past decade and exhibits an uneven geographic distribution across the State, this the first study to explore neighborhood effects on gastroschisis outcomes. We believe these findings allow us to hypothesize plausible causal mechanisms by which proximal environment may affect birth outcomes, which may in turn inform the complex etiology of this birth defect. Acknowledgements This work was supported by the National Science Foundation Doctoral Dissertation Research Improvement grant (#0825511) and the National Science Foundation’s IGERT Predoctoral Fellowship Program (#DGE-0333193) at the Carolina Population Center at the University of North Carolina at Chapel Hill.

Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.socscimed.2010.11.025. References Ali, M., Park, J. K., Thiem, V. D., Canh, D. G., Emch, M., & Clemens, J. D. (2005). Neighborhood size and local geographic variation of health and social determinants. International Journal Health Geographics, 4(1), 12. Bowen, W. (2002). An analytical review of environmental justice research: what do we really know? Environmental Management, 29(1), 3e15. Brown, P. (1995). Race, class, and environmental health: a review and systematization of the literature. Environmental Research, 69(1), 15e30. Brown, N. (1997). Chemical teratogens: hazards, tools and clues. In P. Thorogood (Ed.), Embryos, genes and birth defects. Chichester, England; New York: John Wiley & Sons. Brown, M. B., & Forsythe, A. B. (1974). Robust tests for equality of variances. Journal of the American Statistical Association, 69, 364e367. Calzolari, E., Bianchi, F., Dolk, H., Milan, M., Lechat, M., Leurquin, P., et al. (1995). Omphalocele and gastroschisis in Europe - a survey of 3-million births 1980e1990. American Journal of Medical Genetics, 58(2), 187. Canfield, M. A., Ramadhani, T. A., Langlois, P. H., & Waller, D. K. (2006). Residential mobility patterns and exposure misclassification in epidemiologic studies of birth defects. Journal of Exposure Science Environmental Epidemiology, 16(6), 538. Carmichael, S. L., Nelson, V., Shaw, G. M., Wasserman, C. R., & Croen, L. A. (2003). Socio-economic status and risk of conotruncal heart defects and orofacial clefts. Paediatric Perinatol Epidemiology, 17(3), 264. Curry, J. I., McKinney, P., Thornton, J. G., & Stringer, M. D. (2000). The aetiology of gastroschisis. British Journal of Obstetrics and Gynaecology, 107(11), 1339. Diego, M. A., Jones, N. A., Field, T., Hernandez-Reif, M., Schanberg, S., Kuhn, C., et al. (2006). Maternal psychological distress, prenatal cortisol, and fetal weight. Psychosomatic Medicine, 68(5), 747. Diez Roux, A. V. (2001). Investigating neighborhood and area effects on health. American Journal of Public Health, 91(11), 1783. Dolk, H., Vrijheid, M., Armstrong, B., Abramsky, L., Bianchi, F., Garne, E., et al. (1998). Risk of congenital anomalies near hazardous-waste landfill sites in Europe: the EUROHAZCON study. Lancet, 352(9126), 423. Draper, E. S., Rankin, J., Tonks, A. M., Abrams, K. R., Field, D. J., Clarke, M., et al. (2008). Recreational drug use: a major risk factor for gastroschisis? American Journal of Epidemiology, 167(4), 485. Drongowski, R. A., Smith, R. K., Jr., Coran, A. G., & Klein, M. D. (1991). Contribution of demographic and environmental factors to the etiology of gastroschisis: a hypothesis. Fetal Diagnosis and Therapy, 6(1e2), 14e27. Feldkamp, M. L., & Botto, L. D. (2008). Developing a research and public health agenda for gastroschisis: how do we bridge the gap between what is known and what is not? American Journal of Medical Genetic Part C Seminar in Medical Genetics, 148C(3), 155. Fell, D. B., Dodds, L., & King, W. D. (2004). Residential mobility during pregnancy. Paediatr Perinat Epidemiol, 18(6), 408. Field, T., Hernandez-Reif, M., Diego, M., Figueiredo, B., Schanberg, S., & Kuhn, C. (2006). Prenatal cortisol, prematurity and low birthweight. Infant Behaviour and Development, 29(2), 268. Forrester, M. B., & Merz, R. D. (1999). Epidemiology of abdominal wall defects, Hawaii, 1986e1997. Teratology, 60(3), 117. Haynes, R., Daras, K., Reading, R., & Jones, A. (2007). Modifiable neighbourhood units, zone design and residents’ perceptions. Health & Place, 13(4), 812e825. Hobel, C. J., Goldstein, A., & Barrett, E. S. (2008). Psychosocial stress and pregnancy outcome. Clinical Obstetrics and Gynecology, 51(2), 333. Hougland, K. T., Hanna, A. M., Meyers, R., & Null, D. (2005). Increasing prevalence of gastroschisis in Utah. Journal of Pediatric Surgery, 40(3), 535. Jackson, M., & Mare, R. (2007). Cross-sectional and longitudinal measurements of neighborhood experience and their effects on children. Social Science Research, 36(2), 590e610. Kawachi, I., & Berkman, L. F. (2003). Neighborhoods and health. Oxford; New York: Oxford University Press. Khoury, M. J., Stewart, W., Weinstein, A., Panny, S., Lindsay, P., & Eisenberg, M. (1988). Residential mobility during pregnancy: implications for environmental teratogenesis. Journal of Clinical Epidemiology, 41(1), 15. Krieger, N., Chen, J. T., Waterman, P. D., Soobader, M.-J., Subramanian, S. V., & Carson, R. (2003). Choosing area based socioeconomic measures to monitor social inequalities in low birth weight and childhood lead poisoning: The Public Health Disparities Geocoding Project (US). Journal of Epidemiology and Community Health, 57, 186e199. Lam, P. K., Torfs, C. P., & Brand, R. J. (1999). A low pregnancy body mass index is a risk factor for an offspring with gastroschisis. Epidemiology, 10(6), 717. Lovasi, G., Moudon, A., Smith, N., Lumley, T., Larson, E., Sohn, D., et al. (2008). Evaluating options for measurement of neighborhood socioeconomic context: evidence from a myocardial infarction case-control study. Health & Place, 14(3), 453e467. Lovasi, G., Weiss, J., Hoskins, R., Whitsel, E., Rice, K., Erickson, C., et al. (2007). Comparing a single-stage geocoding method to a multi-stage geocoding method: how much and where do they disagree? International Journal of Health Geographics, 6(1), 12.

E.D. Root et al. / Social Science & Medicine 72 (2011) 625e633 Marmot, M. G., & Wilkinson, R. G. (2006). Social determinants of health. Oxford; New York: Oxford University Press. Messer, L. C., Vinikoor, L. C., Laraia, B. A., Kaufman, J. S., Eyster, J., Holzman, C., et al. (2008). Socioeconomic domains and associations with preterm birth. Social Science & Medicine, 67(8), 1247. Morello-Frosch, R., & Shenassa, E. (2006). The environmental “riskscape” and social inequality: implications for explaining maternal and child health disparities. Environ Health Perspect, 114(8), 1150e1153. Morenoff, J. D., House, J. S., Hansen, B. B., Williams, D. R., Kaplan, G. A., & Hunte, H. E. (2007). Understanding social disparities in hypertension prevalence, awareness, treatment, and control: the role of neighborhood context. Social Science & Medicine, 65(9), 1853e1866. National Birth Defects Prevention Network (NBDPN). (2007). birth defects surveillance data from selected states, 2000e2004. Birth Defects Research Part A Clinical and Molecular Teratology, 79, 874e942. Opitz, J. M., & Pysher, T. J. (2008). Invited editorial comment: further reflections on gastroschisis. American Journal of Medical Genetic Part C Seminar in Medical Genetics, 148C(3), 192. Penman, D. G., Fisher, R. M., Noblett, H. R., & Soothill, P. W. (1998). Increase in incidence of gastroschisis in the South west of England in 1995. British Journal of Obstetrics and Gynaecology, 105(3), 328. Timberlake, J. (2007). Racial and ethnic inequality in the duration of children’s exposure to neighborhood poverty and affluence. Social Problems, 54(3), 319e342. Torfs, C. P., Christianson, R. E., Iovannisci, D. M., Shaw, G. M., & Lammer, E. J. (2006). Selected gene polymorphisms and their interaction with maternal smoking, as risk factors for gastroschisis. Birth Defects Research Part A Clinical and Molecular Teratology, 76(10), 723.

633

Torfs, C. P., Curry, C., & Roeper, P. (1990). Gastroschisis. Journal of Pediatrics, 116(1), 1. Torfs, C. P., Katz, E. A., Bateson, T. F., Lam, P. K., & Curry, C. J. R. (1996). Maternal medications and environmental exposures as risk factors for gastroschisis. Teratology, 54(2), 84. Torfs, C. P., Lam, P. K., Schaffer, D. M., & Brand, R. J. (1998). Association between mothers’ nutrient intake and their offspring’s risk of gastroschisis. Teratology, 58 (6), 241. Torfs, C. P., Velie, E. M., Oechsli, F. W., Bateson, T. F., & Curry, C. J. R. (1994). A population-based study of Gastroschisis - demographic, pregnancy, and lifestyle risk-factors. Teratology, 50(1), 44. Vrijheid, M., Dolk, H., Stone, D., Abramsky, L., Alberman, E., & Scott, J. E. (2000). Socioeconomic inequalities in risk of congenital anomaly. Archieves of Disease in Childhood, 82(5), 349. Waller, D. K., Shaw, G. M., Rasmussen, S. A., Hobbs, C. A., Canfield, M. A., SiegaRiz, A. M., et al. (2007). Prepregnancy obesity as a risk factor for structural birth defects. Archieves of Pediatrics & Adolescent Medicine, 161(8), 745. Wasserman, C. R., Shaw, G. M., Selvin, S., Gould, J. B., & Syme, S. L. (1998). Socioeconomic status, neighborhood social conditions, and neural tube defects. American Journal of Public Health, 88(11), 1674. Weden, M. M., Carpiano, R. M., & Robert, S. A. (2008). Subjective and objective neighborhood characteristics and adult health. Social Science & Medicine, 66(6), 1256e1270. Werler, M. M., Mitchell, A. A., & Shapiro, S. (1992). Demographic, reproductive, medical, and environmental factors in relation to gastroschisis. Teratology, 45 (4), 353. Williams, L. J., Kucik, J. E., Alverson, C. J., Olney, R. S., & Correa, A. (2005). Epidemiology of gastroschisis in Metropolitan Atlanta, 1968 through 2000. Birth Defects Research Part A-Clinical and Molecular Teratology, 73(3), 177.