Publications Search - Abstract View

Title: Impact of residential mobility on estimated environmental exposures in a prospective cohort of older women.
Authors: Medgyesi DN,  Fisher JA,  Cervi MM,  Weyer PJ,  Patel DM,  Sampson JN,  Ward MH,  Jones RR
Journal: Environ Epidemiol
Date: 2020 Oct
Branches: BB, OEEB
PubMed ID: 33154988
PMC ID: PMC7595244
Abstract: Longitudinal studies of environmental hazards often rely on exposure estimated at the participant's enrollment residence. This could lead to exposure misclassification if participants move over time. Methods: We evaluated residential mobility in the Iowa Women's Health Study (age 55-69 years) over 19 years of follow-up (1986-2004). We assessed several environmental exposures of varying spatial scales at enrollment and follow-up addresses. Exposures included average nitrate concentrations in public water supplies, percent of agricultural land (row crops and pasture/hay) within 750 m, and the presence of concentrated animal feeding operations within 5 km. In comparison to gold standard duration-based exposures averaged across all residences, we evaluated the sensitivity and specificity of exposure metrics and attenuation bias for a hypothetical nested case-control study of cancer, which assumed participants did not move from their enrollment residence. Results: Among 41,650 participants, 32% moved at least once during follow-up. Mobility was predicted by working outside the home, being a former/current smoker, having a higher education level, using a public drinking water supply, and town size of previous residence. Compared with duration-based exposures, the sensitivity and specificity of exposures at enrollment ranged from 94% to 99% and 97% to 99%, respectively. A hypothetical true odds ratio of 2.0 was attenuated 8% for nitrate, 9%-10% for agricultural land, and 6% for concentrated animal feeding operation exposures. Conclusions: Overall, we found low rates of mobility and mobility-related exposure misclassification in the Iowa Women's Health Study. Misclassification and attenuation of hypothetical risk estimates differed by spatial variability and exposure prevalence.