Spatial Analysis of Birth Defects in Massachusetts to Investigate Social and Environmental Risk Factors

Abstract

The distributions of birth defects vary geographically with higher risk observed in regions of lower socioeconomic status. Spatio-temporal exposures such as PM2.5 may also influence patterns of risk. To explore the spatial distribution of cardiac birth defects in relation to social and environmental risk factors, we analyzed individual level data from the Massachusetts Birth Defects Registry (2001-2009). Our analyses included 3,683 cardiac defects and 9,000 controls randomly selected from all births. We used geocoded birth addresses and remote sensing data to estimate exposure to particulate matter up to 2.5 micrometers in size (PM2.5) during pregnancy. Covariates included education, race, smoking, drinking, season of conception, prenatal care, and infant sex. We generated smoothed maps of crude and adjusted risk for cardiac birth defects using generalized additive models (GAMs). GAMs allow for the analysis of binary outcome data while simultaneously smoothing location and controlling for confounders. We compared spatial models with and without PM2.5 exposure to explore the impact of PM2.5 on patterns of risk. The adjusted maps show that risk of certain cardiac birth defects varies geographically after accounting for known risk factors. Notably, we found that the risk of Patent Ductus Arteriosus (PDA) was spatially confounded by maternal race, season of conception, and infant sex. We also noted the presence of reverse confounding due to maternal age, masking areas of increased PDA risk in the crude analysis (max OR= 1.45). Adding PM2.5 to the model attenuated the risk magnitude (max OR= 1.25)but did not completely explain an area of increased PDA risk north of Boston, with location remaining statistically significant (p= 0.03). The observed spatial patterns could be a result of residual confounding from individual level socio-demographic or other environmental variables.