Predictors of residential location change in the ARIC Study

Abstract

Background: Relatively little work has described predictors of moving residence in the context of epidemiological studies, though residential mobility could lead to selection bias and misclassification of contextual exposures. Our goal was to explore predictors of moving among participants in the Atherosclerosis Risk in Communities (ARIC) Study. Methods: We used data from ARIC visits 1 (1987-1989), 2 (1990-1992), and 3 (1993-1995). Movers were identified by a change in geocoded address coordinates between visits 2 and 3 and further defined as within-county and out-of-county moves. We further characterized moves by distance, either greater or less than the median moving distance. We compared Visit 2 sociodemographic, cognitive, psychosocial, and health-related characteristics of within and out-of-county movers to non-movers and used logistic regressions to identify independent predictors of moving within-county and out-of-county. Results: Of 12,834 participants included in analyses, we identified 390 out-of-county moves and 1,046 within-county moves. Out-of-county movers had higher cognitive test scores, were more educated, and had better self-rated health compared to non-movers (all p<0.001). Within-county movers were more likely to live alone and had worse self-rated health compared to non-movers (all p<0.001). Independent predictors of out-of-county moves included concurrent change in employment status or cohabitation status, better cognitive performance on a test of verbal fluency, higher depressive symptom burden, and fewer years living in one’s community. Independent predictors of within-county moves included younger age, being unmarried, lower income, prevalent coronary heart disease, concurrent change in cohabitation status, and fewer years living in one’s community. Conclusions: Life events, demographics, and health-related characteristics are independently associated with within-county and out-of-county moves. Understanding these associations can help explain trends in residential mobility. Our results have implications for epidemiological studies of the impact of contextual factors and the potential for selection bias due to moves outside of the study catchment area.