Benjamin Aretz, Gabriele Doblhammer, Fanny Janssen

This study evaluated how (changing) exposures to the living environment are associated with health at baseline and, most importantly, with changes in health over time among both, movers and non-movers. Longitudinal data from 1999 to 2014 were obtained from the German Socio-Economic Panel. Our analysis sample consists of 4,373 individuals aged 50 and above, and included both, longitudinal information on living environment and health. Physical health was measured by the Physical Composite Summary (PCS), which is one of the two main dimensions of 12-Item Short Form Survey (SF-12 v2). We combined this with time-varying characteristics of the living environment between two waves on a five-year basis. The focus was on the four dimensions (1) perceived infrastructure, (2) perceived environmental pollution (3) housing condition and (4) social contacts to neighbours. Linear regression models with robust standard errors were used to explore the associations between the living environment and PCS at baseline (Level Model). Associations between the environmental exposures and PCS declines over time were assessed with Generalized Equation Modeling (Change Model). To tackle the issue of potential reverse causation we strictly imposed a time order between cause, i.e. the living environment, and outcome, i.e. PCS. The issue of same source bias was addressed by controlling for time-varying subjective health simultaneously to the time-varying living environmental conditions. We also included important time-varying individual characteristics to account for altered life circumstances that may influence PCS. Among both, men and women worsened environmental pollution and bad housing conditions were significantly associated with worse PCS at baseline. Additionally, we found different results for men than for women. Infrastructure was just associated with women’s health, namely at baseline. Men who experienced stable worst or worsened environmental pollution had a significantly stronger declining PCS from baseline onwards (ref. stable best).