Objectives: To determine whether a widely used measure of neighborhood-level socioeconomic deprivation was associated with unintended pregnancy, delayed/no prenatal care, low birth weight (LBW), and not breastfeeding, after adjusting for a more comprehensive set of individual-level socioeconomic factors than previously reported.
Methods: Data from CDC's Pregnancy Risk Assessment Monitoring System (PRAMS) in Washington and Florida (1997-1998) were linked with census tract-level data (2000) based on birth certificate addresses. For each state, logistic regression models were estimated for associations between neighborhood deprivation measured by the Townsend Index and each dependent variable, unadjusted and then adjusting for maternal age, parity, racial/ethnic group, and education; paternal education; and family income. Similar models were estimated for each racial/ethnic group separately.
Results: Despite significant unadjusted associations between neighborhood deprivation and all dependent variables except LBW in Washington, few statistically significant associations were found in the adjusted models overall. In stratified models, African American women in low-deprivation Florida neighborhoods had higher odds of delayed/no prenatal care compared with their moderate-deprivation counterparts, and only among European American women were high-deprivation neighborhoods associated with increased odds of delayed/no prenatal care.
Conclusions: These results, which may not be generalizable beyond Florida and Washington or to other health indicators, suggest that some previously reported associations between neighborhood characteristics and the selected health indicators may reflect residual confounding by individual-level socioeconomic status/position. Until methodological and conceptual challenges regarding mediation ("over-controlling") and measurement of neighborhood exposure are overcome, conclusions regarding independent neighborhood associations should be made with caution.