Studies of associations between environmental exposures and health outcomes require reliable assessment of the exposures, often assigned based on residential addresses from one data source at a single time point. In large-scale epidemiology studies, residential addresses of study participants commonly originate from administrative data and are used under the assumption that they are up-to-date. Yet, if residential mobility is not reported, inaccurate addresses may introduce exposure estimation errors that might affect the study results.

Studies assessing the associations between prenatal air pollution exposures and birth outcomes commonly use maternal residential addresses at the time of delivery as a proxy for the mothers' residency throughout pregnancy. This is supported by several studies that found that only a small portion of mothers change their address between conception and delivery. These changes are of a relatively short distance, and accordingly, they do not represent large differences in exposure estimations (Bell and Belanger, 2012; Chen et al., 2010; Hodgson et al., 2015; Lupo et al., 2010; Pennington et al., 2017; Pereira et al., 2016; Warren et al., 2018). To examine changes in residential addresses during pregnancy, researchers used complete address histories from birth certificates, surveys, phone or in-person interviews, and self-reported addresses. Some of these studies further examined how exposure estimation errors from such changes of address impact the associations between prenatal air pollution and birth outcomes (Pennington et al., 2017; Pereira et al., 2016; Warren et al., 2018).

In large-scale studies, which utilize mandatory data collected nationally for administrative reasons and not for research purposes, examining the mobility of the participants is challenging, as complete address histories are commonly unavailable. In addition, the residential addresses in most of these studies originate from a single administrative source, typically with very large sample size. Therefore, conducting follow-up or in-person interviews is not feasible. Accordingly, the validity of such large-scale single-source addresses and their consequences to epidemiological association measures are sometimes questionable.

This study aimed to validate the use of official addresses originating from population registries and reported by the citizens, and to evaluate the impact of inaccurate addresses on air pollution exposure estimates and epidemiological association measures. For this, we compared the Israeli population registry addresses for a national neonate population with addresses from hospital records, which are reported at the hospital at the time of birth and directly by the parents. Citizens are required to report changes in their official residential address yet this may not always be the case (e.g. students or young couples who move to a temporary address). Thus, official addresses may differ from the addresses reported at the hospital. Consequently, we assume that the address from hospital records is the most updated and represent, with higher validity, the residential addresses during pregnancy. We had no data on residential mobility of the mothers during pregnancy.

We further used data from a recent study on the associations between prenatal air pollution exposure and congenital hypothyroidism (CHT) (Harari Kremer et al., 2021), and examined the effect of the exposure errors on the measures of association.