Assessing exposure metrics for air pollution and birthweight models in North Carolina

The link between air pollution (AP) exposure and adverse birth outcomes is of public health concern due to the relationship between poor pregnancy outcomes and the onset of childhood diseases. As personal exposure measures are difficult to obtain, proximate measures of AP exposure are traditionally used. We explore how different AP exposure metrics affect BWT regression models. We examined the impact of maternal exposure to ambient levels of PM₁₀ and PM_2.5 on BWT among infants in North Carolina (NC). We linked maternal residence to the closest monitor during pregnancy for 2000-2002 (n=350,754), restricting analyses to singleton births. We estimated exposure by averaging county level AP concentrations for the entire pregnancy and each trimester. For a finer spatially resolved metric, we calculated exposure averages for women living within 20, 10 and 5 km of a monitor. Multivariate linear regression was used to determine the association between AP exposure and BWT adjusting for standard covariates. In the county level model an interquartile increase in PM₁₀ and PM_2.5 during the entire pregnancy reduced BWT by 5.3g (95%CI=3.3-7.4) and 4.6g (95%CI=2.3-6.8) respectively. This model also reduced BWT for PM₁₀ (7.1g, 95%CI=1.0-13.2) and PM_2.5 (10.4g, 95%CI=6.4-14.4) during the third trimester. Proximity models for 20, 10 and 5 km distances showed results similar to the county level models. NC has relatively good air quality. Nevertheless, the sensitivity analysis showed that at various spatial resolutions there is still a negative association between AP and BWT, despite consistent attainment of federal air quality standards in the study area.