Exposure errors due to inaccurate residential addresses and their impact on epidemiological associations: Evidence from a national neonate dataset.

Abstract

Studies assessing the associations between prenatal air pollution exposures and birth outcomes commonly use maternal addresses at the time of delivery as a proxy for residency throughout pregnancy. Yet, in large-scale epidemiology studies, maternal addresses commonly originate from an administrative source. This study aimed to examine the use of population registry addresses to assign exposure estimations and to evaluate the impact of inaccurate addresses on exposure estimates and association measures of prenatal exposures with congenital hypothyroidism. We used morbidity data for congenital hypothyroidism from the national program for neonatal screening for 2009-2015 and address data from two sources: population registry and hospital records. We selected neonates with geocoded addresses from both sources (N=685,491) and developed a comparison algorithm for these addresses. Next, we assigned neonates with exposures from ambient air pollution of PM and NO₂/NO_X, evaluated exposure assessment differences, and used multivariable logistic regression models to assess the impact that these differences have on association measures. We found that most of the exposure differences between neonates with addresses from both sources were around zero and had a leptokurtic distribution density, with most values being zero. Additionally, associations between exposure and congenital hypothyroidism were comparable, regardless of address source and when we limited the model to neonates with identical addresses. We found that ignoring residential inaccuracies results in only a small bias of the associations towards the null. These results strengthen the validity of addresses from population registries for exposure assessment, when detailed residential data during pregnancy are not available.