Neighborhood Contextual Factors,Maternal Smoking,and Birth Outcomes: Multilevel Analysis of the South Carolina PRAMS Survey,2000

Abstract

Background:

Previous studies investigating relationships among neighborhood contexts, maternal smoking behaviors, and birth outcomes (low birth weight [LBW] or preterm births) have produced mixed results.

Methods:

We evaluated independent effects of neighborhood contexts on maternal smoking behaviors and risks of LBW or preterm birth outcomes among mothers participating in the South Carolina Pregnancy Risk Assessment and Monitoring System (PRAMS) survey, 2000–2003. The PRAMS data were geocoded to 2000 U.S. Census data to create a multilevel data structure. We used a multilevel regression analysis (SAS PROC GLIMMIX) to estimate odds ratios (OR) and corresponding 95% confidence intervals (CI).

Results:

In multivariable logistic regression models, high poverty, predominantly African American neighborhoods, upper quartiles of low education, and second quartile of neighborhood household crowding were significantly associated with LBW. However, only mothers resident in predominantly African American Census tract areas were statistically significantly at an increased risk of delivering preterm (OR 2.2, 95% CI 1.29-3.78). In addition, mothers resident in medium poverty neighborhoods remained modestly associated with smoking after adjustment for maternal-level covariates. The results also indicated that maternal smoking has more consistent effects on LBW than preterm births, particularly for mothers living in deprived neighborhoods.

Conclusions:

Interventions seeking to improve maternal and child health by reducing smoking during pregnancy need to engage specific community factors that encourage maternal quitting behaviors and reduce smoking relapse rates. Inclusion of maternal-level covariates in neighborhood models without careful consideration of the causal pathway might produce misleading interpretation of the results.

Introduction

There is growing evidence to suggest that disparities in health and other social outcomes across regions, states, counties, and communities are attributable to differences in neighborhood contexts rather than individual-level determinants.^1

–6 Health scientists have tried to explain these disparities by suggesting that certain health behaviors and practices are more common with certain racial/ethnic groups and populations with a lower socioeconomic status (SES).^6,7 This line of reasoning suggests, for example, socioeconomically disadvantaged women are likely to have poorer health outcomes (including low birth weight [LBW] and preterm births) wherever they live. On the contrary, the effect of the context will suggest that women with similar SES will have different health outcomes depending on where they live. Nonetheless, the potential modifying effects of individual characteristics and neighborhood contexts in determining health outcomes have been suggested.^8,9 Disentangling the relative importance of individual and contextual causal processes linked to disparities in birth outcomes is necessary to determine the targets of health interventions.^3,5,8,9

LBW and preterm births are early health indicators known to be sensitive to neighborhood contexts over and above individual-level exposures.^1,7,8 LBW (defined as birth weight <2500 g) and preterm births (defined as gestation of <37 weeks) are significant causes of infant and childhood mortality and morbidity in the United States.^10

–14 In South Carolina, preterm births are the second leading cause of infant mortality in the state.^15,16 Analyses of U.S. perinatal data for the past decade show a consistent pattern of increases in LBW and preterm births in spite of general improvement in the-social lives of the populations in terms of economic growth and greater understanding of disease causal processes.^16,17 More troubling about these statistics is the disparity among the racial groups, as an African American child born today is twice likely as a white infant of similar SES to be LBW or to be born preterm.¹⁸ Additionally, LBW and preterm births are not only key correlates of infant health and survival but also predictive of future adult health patterns and neurological impairments in later life.^19

–23

Many individual-level risk factors have been investigated as determinants of adverse birth outcomes, yet interventions built around these have failed to improve birth outcomes, particularly among the African American population. This has led to suggestions by some health researchers that a genetically based etiology associated with race/ethnicity might influence birth outcome disparities.^23,24 Other researchers, however, have shown that differences in health outcomes and behaviors do not cluster along racial dimensions but are reinforced under conditions of inequity and differential sociocultural contexts.^25
–27

Many investigators believe that of all maternal health practices that affect birth outcomes, smoking behaviors have the most consistent effects on LBW or preterm birth outcomes.^28,29 Numerous researchers have also reported positive associations between neighborhood contexts (neighborhood low education, poor neighborhoods, crowded households) and factors conducive to maternal smoking behaviors in many communities across the United States^28

–32 and other countries.³³ However, the precise causal pathway for associations between neighborhood contexts and maternal smoking behaviors is not clearly elucidated in the literature. Current support for disparities in maternal smoking behaviors and adverse birth outcomes suggests that disadvantaged people are most likely to cluster in specific neighborhoods (compositional effect).^20
–22 However, contextual effect models suggest that the presence and availability of social and improved infrastructures in the neighborhoods lead residents to adopt behaviors and practices conducive to improved health. Thus, in deprived neighborhoods, constraints imposed by the neighborhood environment, such as crime, fear of crime, drug use, incivility, and social disorder, are suggested as causing residents to adopt unhealthy behaviors as a means of coping with a harsh and stressful environment.^{18,19,34
–36} A possible interaction or confounding of neighborhood contexts and maternal characteristics in determining adverse birth outcomes has been recognized.^35,36

Very little research has focused on exploring relationships between neighborhood contexts and maternal smoking behaviors during pregnancy. Such research is needed given that maternal-level measures of SES and behaviors among the racial groups do not fully account for disparities in LBW or preterm birth outcomes.³³ Furthermore, results of previous studies on neighborhood effect on maternal smoking behaviors and birth outcomes have been inconsistent,^17,34 with some studies suggesting that maternal smoking is more likely in deprived neighborhoods,^7,33 and others showing opposite findings.^3,35

Specific research questions guiding the study were (1) Are neighborhood contexts related to LBW or preterm birth outcomes among mothers in South Carolina? (2) What are the relationships among neighborhood contexts and maternal smoking behaviors? We examined these questions by linking the geocoded Pregnancy Risk Assessment and Monitoring System (PRAMS) data for 2000–2003 to the 2000 U.S. Population and Housing Census (South Carolina Census tracts [CT]). The study protocol was approved by the Institutional Review Boards for protection of Human Subjects at the University of South Carolina and the South Carolina Department of Health and Environmental Control (DHEC).

Materials and Methods

This is a cross-sectional study of geocoded South Carolina PRAMS data for 2000–2003 with the 2000 United States Census data (for South Carolina) using the CT as the common linking variable. The process of geocoding the PRAMS data involves converting maternal address information (street address, city name, and Zip codes) provided on the surveys to latitudes and longitudinal coordinates using standard reference data files of roads and street address ranges.^37,38 To protect the identity of respondents, the final database had no street addresses. Geocoding of the PRAMS data to 2000 U.S CTs achieved a completeness of accuracy at 95%; thus, a total of 8687 of the PRAMS women were successfully linked to 670 CTs across the state. The selected variables from the 2000 Census (STF-3A) and the geocoded PRAMS datasets were thus used to create a multilevel dataset that includes maternal-level and neighborhood-level variables as two-level data files. The practical utility of this data merger for estimating neighborhood effects on health outcomes has been confirmed by other investigators.^37,38 In general, a CT provides information on populations that are homogeneous with regard to SES and other relevant indicators of well-being³⁹ and was used to define a neighborhood boundary in this study. The PRAMS survey is a systematic population-based surveillance system aimed at assessing maternal behaviors and experiences during pregnancy and immediately after a live birth.⁴⁰ The procedure used in gathering PRAMS data has been extensively described in another publication.⁴¹ Briefly, the PRAMS survey uses stratified random sampling of mothers delivering live births drawn from the birth certificate data. From about 2–6 months after delivery, the selected mothers (100–250) are sent letters to notify them of the study and later mailed a PRAMS questionnaire. Nonrespondents are contacted with a telephone call for up to three attempts. The responses of women completing the survey are linked to selected birth certificate data and weighted for sample design, nonresponse, and noncoverage to create annual PRAMS datasets. The average unweighted response rate for 2000–2003 was 72%.

In this study, respondents with records that lacked the code necessary to link them with neighborhood-level variables were excluded. Analyses were further restricted to birth weight >500 g, and gestation >20 weeks to define LBW and preterm births, respectively, in accordance with the World Health Organization (WHO) and the International Classification of Diseases (ICD-9-CM) guidelines.⁴² Finally, we restricted the analysis to only African American and white ethnic groups because of a small number of observations recorded for other racial groups and the heterogeneity of the population groups involved. These reduced the analytical sample to 8064 mothers; thus, the overall excluded sample was 7.5%.

Dependent variables

Birth outcomes. LBW and preterm birth were the main outcome variables used in measuring adverse birth outcomes in the study. Infant birth weight was assessed as a dichotomous outcome (<2500 g vs. ≥2500 g) variable. Preterm was assessed as the gestational age in completed weeks at the time of birth since a mother's last menstrual period. Preterm birth was similarly analyzed as a dichotomous (<37 vs. ≥37 weeks) variable.

Independent variables

Neighborhood-level variables. Neighborhood contexts for the study were defined by the socioeconomic and demographic characteristics of the 2000 U.S. CTs of the residents in South Carolina. To characterize neighborhood exposures, four variables were selected from the 2000 U.S. Census data based on previous studies: percent African American population, proportion of households in a CT with income <150% of the federal poverty line, proportion of CT residents with less than a high school education, and household crowding (proportion of households in a CT with more than 1 person per room). Because previous studies have shown a nonlinear relationship between neighborhood contexts and health outcomes,^{18,36
–38,43,44} the study variables were categorized as quartiles of neighborhood exposures. However, the percent of African American residents in a CT was categorized as predominantly white (<10% African American population resident in a CT), mixed majority African American population (CT with >10%–50% African American population), and predominantly African American (CT with >50%–90% African American population), which is consistent with earlier studies.^7,38,45

Categories of neighborhood poverty was categorized as low poverty (<10% of CT residents with income <150% of the federal poverty level), medium poverty (10%–19.99% of CT residents with income <150% of the federal poverty level), high poverty (≥20% of CT residents with income <150% of the federal poverty level), following the example of Datta et al.⁷ These neighborhood exposures have been shown to be the most sensitive measures of social determinants of health disparities among population groups.^1,7,8,13

Behavioral risk factor measures

Maternal smoking behaviors were based on self-reported information from the PRAMS data. Information on cigarette smoking history was obtained for three time periods: smoking at least 100 cigarettes in the past 2 years before the survey, smoking 3 months before the index pregnancy, and smoking during the last trimester. We used this information to classify maternal smoking status as: nonsmokers (mothers who never smoked in the entire 2-year period, those quitting 3 months before pregnancy, and those who never smoked during pregnancy): this is because previous studies have shown that risks for LBW or preterm births are similar for mothers who stopped smoking 3 months before pregnancy and those who never smoked^43,44; quitters (mothers who smoked during pregnancy but quit smoking 3 months before delivery); and continuous smokers (mothers who reported smoking before pregnancy or during pregnancy until delivery). Thus, maternal smoking during pregnancy was defined to reflect smoking at any time period during any of the three trimesters as continuous smokers and quitters (smokers who quit during the last trimester) irrespective of the frequency. These categories were used for our bivariate and multivariable analyses. In analyses examining neighborhood exposures and birth outcomes, however, the smoking categories were dichotomized as smokers (the combination of quitters and continuous smokers) vs. nonsmokers (never smoked before pregnancy or during pregnancy). The nonsmoking category was used as the reference group for analyses.

Maternal-level socioeconomic and demographic measures

To avoid overstating neighborhood exposure effects, we used multiple control variables associated with maternal-level exposures and also known to be related to neighborhood location. These measures, selected to control for life cycle factors related to neighborhood selection and birth outcomes, include maternal age (<20, 20–24, 25–29, 30–34, >34), race (African American, white), marital status (married, nonmarried), income (<$10,000, $10,000–$24,999, $25,000–$39,999, ≥$40,000), and education (<high school, high school, >high school).

Analytical strategy

We used the SAS system⁴⁵ for all analyses. Descriptive statistics obtained include frequency distributions of all variables, bivariate associations of neighborhood exposures with maternal smoking behaviors, and birth outcomes (LBW, preterm birth). We determined statistical significance with the chi-square test (p < 0.05) and Cochran-Armitage test for linear trend (p < 0.05). SAS PROC GLIMMIX was used to fit multilevel logistic regression models for dichotomous responses (LBW, preterm delivery), assuming a binomial distribution and a logit link function. PRAMS weight variable was included in all data analysis to reflect the population of women delivering live births across the state during 2000–2003. Unadjusted odds ratios (ORs) and adjusted ORs with 95% confidence intervals (95% CI) are reported.

First, we investigated whether maternal smoking behaviors vary by neighborhood contexts. Next, we used a series of sequential modeling processes to evaluate relationships among smoking, maternal covariates, and neighborhood exposures. In model 1, we tested separately the main exposure variables (smoking and neighborhood categories) without potential confounders. In model 2, control variables (maternal-level factors and smoking variable) were added to neighborhood exposures to determine effects on birth outcomes. In model 3, all variables were simultaneously included in the model to determine the extent of confounding. All multivariable analyses were adjusted with weight variable to account for the differences in response rates and sample selection. Probable multicollinearity of neighborhood constructs used in the study was assessed with variance inflation factor (VIF) values ≥10, and none exceeded this threshold value.

Results

Tables 1 and 2 provide brief summaries of weighted and unweighted neighborhood characteristics of mothers participating in South Carolina PRAMS surveys, 2000–2003. Almost 34% of the weighted sample of infants was born to African American mothers, and prevalences of LBW and preterm delivery were 13.16% and 14.18%, respectively (results not shown). Corresponding prevalences for white mothers were 6.53% (LBW) and 8.93% (preterm delivery). Some of these differentials might be related to the SES of the mother; for example, almost 30% of white mothers compared with about 10% of African American mothers had attained >12 years of education. Sixteen percent of African American mothers are in the lowest income category ($7,999–$11,999) compared with almost 11% for white mothers (results not shown). Almost 15% of the weighted sample of mothers delivering live births in the state smoked during the entire period of pregnancy.

Table 1.

Percentage Preterm and Low Birth Weight by Selected Neighborhood Level Characteristics for South Carolina PRAMS Mothers, 2000–2003

	Low birth weight ^a		Preterm births ^a
Selected neighborhood variables	Total (unweighted)	% prevalence ^b	Total (unweighted)	% prevalence ^b
Poverty
Low poverty	665	13.3	591	15.4
Medium poverty	1439	28.3	1238	30.5
High poverty	3236	58.4	2701	54.1
Percent African American
Predominantly white	832	27.7	723	18.4
Mixed majority black	2942	55.6	2504	57.2
Predominantly black	1665	27.8	1384	24.4
Percent low education
0.39–<14.46	1467	27.2	1287	29.0
14.46–<28.89	2545	47.8	2140	48.8
28.89–<44.59	1326	23.4	1099	20.8
44.59–55.69	98	1.5	82	1.4
Percent household crowding
0.0–3.82	3061	58.7	2665	63.3
3.82–7.71	1954	34.4	1607	29.8
7.71–11.89	364	6.0	298	6.6
11.89–15.39	47	0.9	38	0.4

Chi-square differences between low birth weight and preterm delivery for neighborhood contexts were highly significant at p < 0.0001.

Weighted prevalence to represent population of women delivering live births in South Carolina, 2000–2003. Neighborhood poverty categories were based on similar categories used by U.S. Census Bureau to classify low poverty neighborhood: <10% of families earning <$20,000, medium poverty neighborhood 10%–19.9% of families earning <$20,000, and high poverty neighborhood ≥20% of families earning <$20,000.

PRAMS, Pregnancy Risk Assessment and Monitoring System.

Table 2.

Percentage Distribution of Maternal Smoking Behaviors by Neighborhood Level Contexts for South Carolina PRAMS Mothers, 2000–2003

	Maternal smoking patterns ^a
	Nonsmokers		Quitters		Continuous smokers
Selected neighborhood variables	Total ^b	% prevalence	Total ^b	% prevalence	Total ^b	% prevalence
Poverty
Low poverty	686	14.9	92	18.2	120	12.6
Medium poverty	1111	27.9	201	35.8	305	38.8
High poverty	2365	52.2	266	46.0	502	48.6
Percent African American
Predominantly white	741	21.1	128	23.1	203	23.5
Mixed majority black	2295	55.5	326	62.2	572	61.8
Predominantly black	1184	23.4	110	14.7	161	14.7
Percent low education
0.39–<14.46	1319	35.0	192	35.1	226	25.5
14.46–<28.89	1953	45.2	259	46.9	464	52.0
28.89–<44.59	893	18.8	104	17.4	227	20.7
44.59–55.69	52	1.0	9	0.6	17	1.8
Percent household crowding
0.0–3.82	1910	63.5	254	67.4	371	70.5
3.82–7.71	1630	30.4	189	30.1	293	25.3
7.71–11.89	484	5.6	34	1.9	63	3.8
11.89–15.39	101	0.5	8	0.4	9	0.4

Chi-square differences between low birth weight and preterm delivery for neighborhood contexts were highly significant at p < 0.0001.

Weighted prevalence to represent population of women delivering live births in South Carolina, 2000–2003. Neighborhood poverty categories were based on similar categories used by U.S. Census Bureau to classify: low poverty neighborhood <10% of families earning <$20,000, medium poverty neighborhood 10%–19.9% of families earning <$20,000 and high poverty neighborhood ≥20% of families earning <$20,000.

Considering the main outcome variables, mothers resident in poor or high poverty neighborhoods had the highest recorded LBW or preterm births compared with mothers living in low poverty neighborhoods. Similarly, mothers resident in mixed majority African American neighborhoods were disproportionately represented in prevalences of LBW (≈56%) and preterm birth outcomes (≈57%). Furthermore, among mothers with less than a high school education, the greatest disparity in LBW (≈48%) and preterm births (≈49%) was in the second quartile CT (CT with 14.46%–28.89% residents with less than high school education). Within categories of neighborhood household crowding, the first quartile of household crowding exposure (CT with 0–3.82% residents having more than 1 person per room in a house) had the highest prevalence of LBW and preterm births. Mothers continuing to smoke until delivery and those quitting were likely to live in the concentrated poverty areas, low education (CT with adult populations having less than high school or college education), and mixed majority black neighborhoods (results not shown). For quitting behaviors, however, the pattern suggests that women living in the most disadvantaged neighborhoods (predominantly African American CTs, upper quartiles of CTs for household crowding, and less than high school education) have relatively lower weighted prevalences.

The first multivariable analysis examined the relationship between maternal smoking (dichotomized as smoking and nonsmoking) and CT variables (Table 3). In the crude estimates, statistically significant associations were observed between smoking and neighborhood contexts for poor neighborhoods (CT with 10%–19.99% residents with income <150% of the federal poverty), predominantly African American populations (CT with >50%–90% African American population). The results showed that residence in a poor neighborhood compared with a low poverty CT was associated with more than a 3-fold increased odds of smoking, whereas living in predominantly African American neighborhoods reduces the odds of smoking by 64%. In the adjusted estimates (with maternal-level covariates), only percent poor neighborhood (CT with 10%–19.9% residents with income <150% of the federal poverty level) showed statistical significance with maternal smoking behaviors, and residence in predominantly black neighborhoods was protective against smoking. All other neighborhood exposures were not associated with maternal smoking behaviors during pregnancy.

Table 3.

Crude and Adjusted Odds Ratios for Maternal Smoking Behavior According to Selected Neighborhood Contextual Characteristics for South Carolina PRAMS Mothers, 2000–2003

Selected neighborhood variables	Crude estimates ^a	Adjusted estimates ^b
Poverty
Low poverty	Reference	Reference
Medium poverty	3.14 (1.47-6.73)^**	2.24 (1.05-4.80)^*
High poverty	1.44 (1.72-2.88)	1.27 (0.64-2.56)
Percent African American
Predominantly white	Reference	Reference
Mixed majority black	0.93 (0.50-1.73)	1.06 (0.57-1.96)
Predominantly black	0.36 (0.17-0.75)^**	0.88 (0.42-1.85)
Percent low education
0.39–<14.46	Reference	Reference
14.46–<28.89	1.53 (0.81-2.92)	1.10 (0.58-2.09)
28.89–<44.59	1.53 (0.72-3.28)	1.09 (0.51-2.33)
44.59–55.69	5.06 (0.46-56.11)	3.73 (0.33-41.80)
Percent household crowding
0.0–3.82	Reference	Reference
3.82–7.71	0.79 (0.50-1.27)	0.96 (0.52-1.77)
7.71–11.89	0.75 (0.27-2.05)	0.99 (0.02-52.31)
11.89–15.39	1.75 (0.10-31.90)	1.18 (0.03-50.97)

Study population is 5055 singleton live births in 2000–2003 to mothers resident in 786 census tracts of South Carolina, excluding births with congenital anomalies, births <500 g, and gestation <20 weeks.

The main dependent variable is maternal smoking (recategorized as mothers smoking at any time before and during pregnancy and nonsmokers who smoked before pregnancy but quit smoking throughout pregnancy and those who never smoked).

Adjusted estimates include maternal-level covariates (race, income, education, age, and marital status).

p≤0.05; **p≤0.01.

Multivariable modeling

Tables 4 and 5 examine the unadjusted and adjusted multilevel relationships among maternal characteristics (including smoking), neighborhood contexts, and birth outcomes. In model 1, the unadjusted OR for reporting LBW for smokers compared with nonsmokers was 2.04 times (95% CI 1.89-2.19). Quitters were at 20% increased odds of LBW (OR 1.20, 95% CI 1.10-1.32). Corresponding smoking odds for preterm births were 32% (95% CI 1.23-1.41). For neighborhood exposures, the odds of LBW showed statistically significant risks for high poverty, upper quartiles of low education categories (CT with 28.84%–57.695% adult residents with less than high school education), predominantly black neighborhoods, and household crowding (CT with 3.82%–7.71% residents in a house with more than 1 person per room). Only mothers' residence in predominantly black neighborhoods was statistically significant for increased odds of preterm births. All other selected neighborhood contexts did not show statistically significant associations with preterm delivery.

Table 4.

Multilevel Model for South Carolina PRAMS Respondents Based on Two-Level Binomial Logit Model Along with Odds Ratio and 95% Confidence Interval for Low Birth Weight

	Model 1 ^a	Model 2 ^b	Model 3 ^c
Variable	OR (95% CI)	OR (95% CI)	OR (95% CI)
Smoking pattern
Nonsmokers	Reference	Reference	Reference
Continuous smokers	2.04 (1.89-2.19)^***	2.37 (2.19-2.57)^***	2.01 (1.88-2.14)^***
Quitters	1.20 (1.10-1,32)^***	0.92 (0.84-1.02)	1.11 (1.02-1.20)^*
Poverty
Low poverty	Reference	Reference	Reference
Medium poverty	1.18 (0.64-2.18)	1.12 (1.11-3.40)^*	1.09 (0.58-2.05)
High poverty	2.08 (1.19-3.65)^**	2.04 (1.16-3.58)^*	1.09 (0.52-1.62)
Percent African American
Predominantly white	Reference	Reference	Reference
Mixed majority black	1.36 (0.82-2.24)	1.37 (0.83-2.26)	0.86 (0.52-1.44)
Predominantly black	3.29 (1.82-5.93)^***	3.48 (1.84-3.48)^***	1.08 (0.59-1.97)
Percent low education
0.39–<14.46	Reference	Reference	Reference
14.46–<28.89	1.60 (0.96-2.68)	1.55 (0.92-2.61)	0.97 (0.58-1.64,)
28.89–<44.59	2.63 (1.43-4.84)^**	2.55 (1.38-4.71)^**	3.50 (1.19-10.32)^*
44.59–55.69	16.50 (2.19-124.26)^*	14.64 (1.91-112.08)^**	7.31 (0.95-56.45)
Percent household crowding
0.0–3.82	Reference	Reference	Reference
3.82–7.71	1.78 (1.09-2.92)^*	1.79 (1.09-2.94)^*	1.07 (0.65-1.76)
7.71–11.89	2.10 (0.73-6.07)	2.02 (0.75-6.31)	1.07 (0.37-3.14)
11.89–15.39	1.16 (0.06-24.30)	1.08 (0.05-23.03)	1.02 (0.05-22.34)

Model 1 is the baseline model and includes only smoking categories and selected neighborhood factors in separate models to determine effect on low birth weight.

In model 2, we included maternal-level covariates separately to smoking model and selected neighborhood contextual models to determine effect on low birth weight.

In model 3, all variables were included simultaneously to examine changes in odds ratio for low birth weight. Maternal-level covariates are age, education, income, race, and marital status.

p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.0001.

Table 5.

Multilevel Model for South Carolina PRAMS Respondents Based on Two-Level Binomial Logit Model Along with Odds Ratio and 95% Confidence Interval for Low Gestational Age (Preterm Birth)

	Model 1 ^a	Model 2 ^b	Model 3 ^c
Variable	OR (95% CI)	OR (95% CI)	OR (95% CI)
Smoking pattern
Nonsmokers	Reference	Reference	Reference
Smokers	1.32 (1.23-1.41)^**	1.43 (1.33-1.54)^**	1.32 (1.23-1.41)^**
Quitters	1.08 (0.99-1.17)	1.31 (1.21-1.42)^**	1.08 (0.99-1.17)
Poverty
Low poverty	Reference	Reference	Reference
Medium poverty	1.16 (0.66-2.03)	1.13 (0.64-2.67)	0.98 (0.56-1.67)
High poverty	1.62 (0.97-2.69)	1.60 (0.96-2.77)	1.00 (0.60-1.67)
Percent African American
Predominantly white	Reference	Reference	Reference
Mixed majority black	1.78 (1.17-2.71)	1.25 (0.79-1.98)	0.95 (0.60-1.51)
Predominantly black	2.21 (1.29-3.78)^*	2.26 (1.32-3.87)^*	1.12 (0.65-1.93)
Percent low education
0.39–<14.46	Reference	Reference	Reference
14.46–<28.89	1.25 (0.81-1.93)	1.32 (0.83-2.12)	1.01 (0.63-1.62)
28.89–<44.59	1.57 (0.93-2.63)	1.68 (0.96-2.93)	1.17 (0.67-2.06)
44.59–55.69	2.30 (0.44-12.10)	5.30 (0.86-32.29)	3.41 (0.55-21.08)
Percent household crowding
0.0–3.82	Reference	Reference	Reference
3.82–7.71	1.27 (0.82-1.99)	1.28 (0.82-2.00)	0.93 (0.59-1.45)
7.71–11.89	1.48 (0.56-3.93)	1.50 (0.57-3.97)	0.99 (0.37-2.63)
11.89–15.39	0.88 (0.06-14.02)	0.86 (0.05-13.64)	0.85 (0.05-13.58)

Model 1 is the baseline model and includes only smoking categories and selected neighborhood factors in separate models to determine effect on preterm birth.

In model 2, we included maternal-level covariates to smoking model and selected neighborhood contextual models to determine effect on low preterm births.

In Model 3, all variables were included simultaneously to examine changes in odds ratio for preterm births. Maternal-level covariates include age, education, income, race, and marital status.

p ≤ 0.01; **p ≤ 0.0001.

In Model 2 (Tables 4 and 5), the adjusted ORs for smokers (compared with nonsmokers) for both LBW and preterm births were significantly associated with increased odds for adverse birth outcomes. The results of including the smoking variable in neighborhood exposures showed that CT poverty levels (poor and high poverty neighborhoods), predominantly African American neighborhoods (CT with 50%–90% African American population), low education, and household crowding (CT with 3.82%–7.7% residents in a house with more than 1 person per room) were all significantly related to increased odds of LBW. We did not observe statistically significant changes from model 1 for preterm delivery.

Model 3 (Tables 4 and 5) results showed that smoking exposures were statistically significantly related to birth outcomes and in a direction not very different from that of model 1 (Tables 4 and 5). Only mothers resident in the third quartile of neighborhood low education (CT with 28.84%–43.27% residents with education less than high school) were at increased odds of LBW. No other statistically significant associations were observed from the analysis of neighborhood exposures and adverse birth outcomes in model 3 (Tables 4 and 5). For both birth outcomes, the exposure ORs were markedly higher among infants born LBW than infants born preterm.

Discussion

The results of the study showed a pattern consistent with a growing number of studies demonstrating that neighborhood contexts may have effects on health over and above individual-level characteristics.^29,34
–36 This study found maternal smoking to be strongly related to increased odds of LBW or preterm birth. This association remained statistically significant when maternal-level and neighborhood-level exposures were included in the smoking models, indicating that smoking has significant risks for birth outcomes. Analysis of maternal smoking behaviors across the selected neighborhood contexts indicated that mothers resident in medium poverty neighborhoods were at increased odds of smoking; however, mothers living in predominantly African American neighborhoods were somehow protected against smoking. Results of the multivariable analysis also showed that neighborhood contexts might be important in explaining observed disparities in LBW or preterm birth across the state. Our major findings were that the percent African American population in a CT, high area poverty (CT with >20% poverty), neighborhood low education (CT with 28.89%−<44.59% and 44.59%−<55.69% low education), and household crowding (CT with 3.82%–7.71% crowding) appeared to be the most important predictors of LBW in this analysis. Likewise, percent African American population in a neighborhood is the most significant predictor of preterm births across the state.

Some of these associations remained statistically significant when maternal smoking categories were included in the neighborhood models, suggesting that neighborhood contexts may be one of the multiple pathways independently affecting birth outcomes. This is consistent with the findings of Ahern et al.³ A higher percentage of the African American population in a CT appears to create greater risks for LBW or preterm birth outcomes. Among mothers, residence in a CT with a higher percentage of African Americans increases the risk for LBW or preterm births compared with residents in CTs with relatively higher proportions of whites. After adjustment for maternal-level covariates (income, age, race, education, and marital status), the association between neighborhood contexts and birth outcomes was still suggestive but not statistically significant. Overall, the analysis showed that neighborhood contexts have greater effects on LBW than on preterm births.

The exact causal mechanisms linking these contextual factors to maternal proximate factors in determining preterm delivery and LBW have been suggested as through stress effects¹⁴ and maternal exposures to other unmeasured neighborhood contexts.^{1,13,45

–48} For example, a higher proportion of African American population in a neighborhood or a higher percentage of a population with less than a high school education may be a proxy measure for some unmeasured disadvantages associated with a lack of access to quality healthcare or other social resources needed for improved pregnancy outcomes.^6,7 Alternatively, a higher percentage of African Americans in a neighborhood, even though does not measure residential racial segregation, represents the racial context of the population distribution and therefore, access to healthcare and other resources needed for improved pregnancy outcomes.^46,49 Previous investigators have suggested that the racial composition of a neighborhood is a key structural characteristic that serves as a proxy for multiple health risks, including high association with poor provision of municipal services, breakdown of social order, household crowding, crime, and increased neighborhood fear, as well as other health risks linked to poverty, unemployment, and violence.^19,21,46

Although poverty across neighborhoods potentially affects black and whites and other ethnic groups equally, through racial residential segregation, the neighborhood environment of the average white mother is likely to be much better than that of the average black mother.^46,49 Furthermore, the nature of the environmental contexts of many minority residents (particularly African American mothers) may provide significant cues for maternal increased cravings for increased tobacco use or even discourages quitting during pregnancy.^50,51 Finally, the fact that poor neighborhoods are associated with stress may elicit maternal neurobiological responses to trigger cigarette smoking and relapse among quitters.^47,48 Thus, disadvantaged neighborhoods may be unhealthful, partly due to increased exposure of residents to incessant stress and other behavioral risks that may mediate or moderate the causal pathways to LBW or preterm births. Despite reduced effects after adjustment with maternal covariates, the results of the study provide support to hypotheses that neighborhood disadvantages (low education, household crowding, and percent poverty) might be routes for adverse birth outcomes independent of smoking and other maternal-level characteristics.

The results of this study differ from those of previous studies in a variety of ways. First, previous studies examining the relationship between neighborhood contexts and birth outcomes or smoking often have focused on neighborhood disadvantages without considering how different levels of contextual exposures are related to LBW and preterm birth or maternal smoking behaviors. Results of this study suggest that it is not just living in a disadvantaged neighborhood that is associated with greater risk for LBW and preterm births or smoking but that risks for smoking and poor birth outcomes depend on whether the mother is resident in the most advantaged or disadvantaged CT quartiles. Second, previous studies have relied on birth certificate data as a source for deriving maternal-level exposures and smoking risks. Birth certificate data have limited information on maternal SES and smoking. It is most likely that previous studies may have overestimated the true neighborhood effects by minimizing maternal income and marital status as important determinants of neighborhood selection and birth outcomes. Inclusion of these variables, however, particularly maternal race, in the models without thoughtful consideration of the causal path may also lead to overadjustment of neighborhood exposure effects on birth outcomes. The validity of regarding maternal race as a mediator (not to be adjusted for) instead of a true confounder (to be adjusted for) has been elaborated in a recent study.⁵² In our analysis, inclusion of maternal race in all models led to a significant reduction of neighborhood exposure effects on LBW and preterm births. Thus, addition of variables to models without careful consideration of the causal pathway may lead to distorted interpretation of the findings. Third, the results of this study provide a reasonable basis to suggest that contextual effects may have a greater impact on LBW than on preterm delivery. Reasons for these effects are not clearly specified in the literature but are most likely related to stress effects on fetal growth associated with limited resources in disadvantaged neighborhoods.

Limitations

One important limitation of this study was the use of birth certificate data and self-reports from PRAMS to assess the effects of behavioral patterns of maternal smoking on LBW and preterm births. Smoking patterns are likely to be underreported because of the stigma associated with smoking during pregnancy and the social desirability of responses. The possibility exists that factors that were not measured in the PRAMS study might explain the associations reported. For example, information on paternal or partner smoking as secondhand smoking effects and the CT cigarette and alcohol advertising intensity, which might confound or mediate the relationship between maternal smoking and adverse birth outcomes, were not assessed by the PRAMS study. The cross-sectional nature of the study makes it impossible to attribute causal relationships. For example, it is very difficult to determine from the available data if neighborhood contexts induce mothers to smoke more during pregnancy or if other factors in the neighborhood environment precipitate biological mechanisms to affect birth outcomes. Finally, the PRAMS data did not assess the length of time mothers have been in residence at a particular neighborhood; therefore, corresponding exposures on maternal health might be different for various residents.

Conclusions

Interpretation of these findings will suggest that the program focus should target mothers as well as their communities to achieve greater improvements in maternal and child health outcomes. Specifically, public health monitoring of birth outcomes linked to potential neighborhood exposures, including maternal smoking behaviors, could provide an important measure for characterizing an area-level public health status across the state. This would provide useful indicators for planning the desired interventions to improve maternal and child health outcomes. Results of the study are generally consistent with those of previous investigations that have found an independent relationship between neighborhood contexts and birth outcomes^34,47 and between neighborhood contexts and maternal smoking behaviors.^27,28,33 Future studies should consider using longitudinal study designs by following women earlier before pregnancy, examining women's perceptions of neighborhood quality and contextual characteristics, and including the length of stay in a particular neighborhood in a comprehensive theoretical framework. In this way, the effects of neighborhood contexts on maternal smoking behaviors and birth outcomes could be estimated more accurately for policy determination and improved maternal health outcomes across places of residence in the state.

Footnotes

Acknowledgments

I am grateful to all anonymous reviewers for their thoughtful comments. I thank the South Carolina Department of Health and Environmental Control for providing the geocoded data for this study.

Disclosure Statement

The author has no conflicts of interest to report.

References

Ellen

, Turner

. Does neighborhood matter? Assessing recent evidence. Housing Policy Debate, 1997; 4:833–866.

Duncan

, Raudenbush

. Assessing effects of context in studies of child and youth development. Educ Psychol, 1999; 34:29–41.

Ahern

, Pickett

, Selvin

, Abrams

. Preterm birth among African-American and white women: A multilevel analysis of socioeconomic characteristics and cigarette smoking. J Epidemiol Community Health, 2003; 57:606–611.

Balfour

, Kaplan

. Neighborhood environment and loss of physical function in older adults: Evidence from the Alameda County study. Am J Epidemiol, 2002; 6:507–515.

Bingenheimer

, Raudenbush

. Statistical and substantive inferences in public health: Issues in the application of multilevel models. Ann Public Health, 2004; 25:53–77.

Braveman

, Taribo

. Social inequalities in health within countries: Not only an issue for affluent nations. Soc Sci Med, 2002; 54:1621–1635.

Datta

, Subramanian

, Colditz

, Kawachi

, Palmer

, Rosenberg

. Individual, neighborhood and state-level predictors of smoking among U.S. black women: A multilevel analysis. Soc Sci Med, 2006; 63:1034–1044.

Diez Roux

. Investigating neighborhood and area effects on health. Am J Public Health, 2001; 91:1783–1789.

Diez Roux

, Merkin

, Hannan

, Jacobs

, Kiefe

. Area characteristics, individual-level socioeconomic indicators, and smoking in young adults. Am J Epidemiol, 2003; 157:315–326.

10.

Demissie

, Rhoads

, Ananth

et al. Trends in preterm birth and neonatal mortality among blacks and whites in the United States from 1989 to 1997. Am J Epidemiol, 2001; 154:307–315.

11.

Ananth

, Joseph

, Oyelese

, Demissie

, Ventzileos

. Trends in preterm birth and perinatal mortality among singletons: United States, 1989 through 2000. Obstet Gynecol, 2005; 105:1084–1091.

12.

Goldenberg

, Culhane

. Low birth weight in the United States. Am J Clin Nutr, 2007; 85:584S–590S.

13.

Morenoff

. Neighborhood mechanisms and the spatial dynamics of birth weight. Am J Sociol, 2003; 108:976–1017.

14.

Mellor

, Milyo

. Individual health status and racial minority concentration in US states and counties. Am J Public Health, 2004; 4:1043–1048.

15.

March of Dimes. www.marchofdimes.org. 2006 October 15.

16.

Martin

, Hamilton

, Sutton

, Ventura

, Menacher

, Munson

. Births: Final data for 2003. National Vital Statistics Reports, Center for Disease Control and Prevention, 2005; 54:1–25.

17.

Howard

, Marshall

, Kaufman

, Savitz

. Variations in low birth weight and preterm delivery among blacks in relation to ancestry and nativity: New York City, 1998–2002. Pediatrics, 2006; 118:e1399–e1405.

18.

Huynh

, Parker

, Harper

, Pamuk

, Shoendorf

. Contextual effect of income inequality on birth outcomes. Int J Epidemiol, 2005; 34:888–895.

19.

O'Reagan

, Salsberry

. Race and ethnic differences in determinants of preterm birth in the USA: Broadening the social context. Soc Sci Med, 2005; 60:2217–2228.

20.

Janghorbani

, Stenhouse

, Millward

, Raymond

. Neighborhood deprivation and preterm birth in Plymouth, U.K. J Matern Fetal Neonat Med, 2006; 19:85–91.

21.

LeClere

, Rogers

, Peters

. Ethnicity and mortality in the United States: Individual and community correlates. Soc Forces, 1999; 76:169–198.

22.

Lynch

, Kaplan

, Salonen

. Why do poor people behave poorly? Variation in adult health behaviors and psychosocial characteristics by stages of the socioeconomic life course. Soc Sci Med, 1997; 44:809–819.

23.

Barker

DJP

, Eriksson

, Forsén

, Osmond

. Fetal origins of adult disease: Strength of effects and biological basis. Int Epidemiol Assoc, 2002; 31:1235–1239.

24.

Hao

, Wang

, Niu

et al. A candidate gene association study on preterm delivery: Application of high-throughputs genetyping technology and advanced statistical methods. Hum Mol Genetics, 2004; 13:683–691.

25.

Lucas

, Fewtrell

, Cole

. Fetal origins of adult disease—The hypothesis revisited. BMJ, 1999; 319:245–249.

26.

Kramer

, Ananth

, Platt

, Joseph

. U.S. black vs white disparities in foetal growth: Physiological or pathological? Int J Epidemiol, 2006; 35:1187–1195.

27.

Williams

, Jackson

. Social sources of racial disparities in health. Health Affairs, 2005; 24:325–334.

28.

Droungas

, Erhman

, Childress

, O'Brien

. Effect of smoking cues and cigarette availability on craving and smoking behavior. Addict Behav, 1995; 20:657–673.

29.

Pollack

, Lantz

, Frohna

. Maternal smoking and adverse birth outcomes among singletons and twins. Am J Public Health, 2000; 90:395–400.

30.

England

, Levine

, Qian

et al. Smoking before pregnancy and risk of gestational hypertension and pre-eclampsia. Am J Obstet Gynecol, 2002; 186:1035–1040.

31.

Reijneveld

. Neighborhood socioeconomic context and self-reported health and smoking: A secondary analysis of data on seven cities. J Epidemiol Community Health, 2002; 56:935–942.

32.

Ross

. Walking, exercising, and smoking: Does neighborhood matter? Soc Sci Med, 2000; 51:265–274.

33.

Pickett

, Wakschlag

, Rathouz

, Leventhal

, Abrams

. The working-class context of pregnancy smoking. Health & Places, 2002; 8:167–175.

34.

Pearl

, Braveman

, Abrams

. The relationship of neighborhood socioeconomic characteristics to birth weight among 5 ethnic groups in California. Am J Public Health, 2001; 91:1808–1814.

35.

Pickett

, Ahern

, Selvin

, Abrams

. Neighborhood socioeconomic status, maternal race and preterm delivery: A case-control study. Ann Epidemiol, 2002; 12:410–418.

36.

Pickett

, Pearl

. Multilevel analyses of neighborhood socioeconomic context and health outcomes: A critical review. J Epidemiol Community Health, 2001; 55:111–122.

37.

Krieger

, Zierler

, Hogan

et al. Geocoding and measurement of neighborhood socioeconomic position: A U.S. perspective. Kawachi

, Berkaman

. Neighborhoods and health. New York: Oxford University Press, 2003; 147–178.

38.

Mujahid

, Diez-Roux

, Morenoff

, Raghunathan

. Assessing measurement properties of neighborhood scales: From psychometrics to ecometrics. Am J Epidemiol, 2007; 8:858–867.

39.

Census Bureau. 2000. factfinder.census.gov/servlet/DatasetMainPageServlet?_lang=en

40.

South Carolina Department of Health and Environmental Control. SCAN PRAMS. scangis.dhec.sc.gov/scan/prams2/support/aboutprams.htm. 2005 December 5.

41.

Dietz

, Adams

, Kendrick

, Mathis

. PRAMS Working Group. Completeness of ascertainment of prenatal smoking using birth certificates and confidential questionnaires. Am J Epidemiol, 1998; 148:1048–1054.

42.

International classification of diseases, 9th revision. Clinical modification, 6th. World Health Organization: Geneva, 2008.

43.

Kaufman

, Dole

, Savitz

, Herring

. Modeling community-level effects on preterm births. Am J Epidemiol, 2003; 13:377–384.

44.

Subramanian

, Chen

, Rehkopf

, Waterman

, Krieger

. Comparing individual-and area-based socioeconomic measures for the surveillance of health disparities: A multilevel analysis of Massachusetts births, 1989–1991. Am J Epidemiol, 2006; 164:823–834.

45.

Stuber

, Galea

, Ahern

, Blaney

, Fuller

. The association between multiple domains of discrimination and self-assessed health: A multilevel analysis of Latinos and blacks in four low income New York City neighborhoods. Health Serv Res, 2003; 38:1735–1759.

46.

Massey

. American apartheid: Segregation and the making of the underclass. Am J Sociol, 1990; 2:329–357.

47.

Tseng

, Yeatts

, Millikan

, Newman

. Area-level characteristics and smoking in women. Am J Public Health, 2001; 91:1847–1850.

48.

Williams

, Collins

. Racial residential segregation: A fundamental cause of racial disparities in health. Public Health Rep, 2001; 116:404–416.

49.

Zenk

, Schultz

, Israel

et al. Neighborhood racial composition, neighborhood poverty, and the spatial accessibility of supermarkets in metropolitan Detroit. Am J Public Health, 2005; 95:660–667.

50.

Field

, Duka

. Cue reactivity in smokers: The effects of perceived cigarette availability and gender. Pharm Biochem Behav, 2004; 78:647–652.

51.

Geier

, Mucha

, Pauli

. Appetitive nature of drug cues confirmed with physiological measures in a model using pictures of smoking. Psychopharm, 2000; 150:283–291.

52.

Backlund

, Rowe

, Lynch

, Wolfson

, Kaplan

, Sorlie

. Income inequality and mortality: A multilevel prospective study of 521, 248 individuals in 50 U.S. states. Int J Epidemiol, 2007; 36:590–596.

Neighborhood Contextual Factors,Maternal Smoking,and Birth Outcomes: Multilevel Analysis of the South Carolina PRAMS Survey,2000–2003

Abstract

Background:

Methods:

Results:

Conclusions:

Introduction

Materials and Methods

Dependent variables

Independent variables

Behavioral risk factor measures

Maternal-level socioeconomic and demographic measures

Analytical strategy

Results

Multivariable modeling

Discussion

Limitations

Conclusions

Footnotes

Acknowledgments

Disclosure Statement

References