Contextual determinants of cardiovascular diseases: Overcoming the residential trap by accounting for non-residential context and duration of exposure

Multilevel neighbourhood analyses rarely account for (1) non-residential exposures and (2) duration of exposure, which have the potential to improve contextual level variance explained, model fit and strength of associations. Using cross-classified logistic regressions, we evaluate the impact of socio-environmental factors at work and home on cardiovascular disease risk for 1626 adults in Toronto-Canada. In the fully-adjusted model, increased CVD risk was associated with poor food environments, lack of parks/recreational facilities, home and work proximity to a major road and noise, and working in a low-SES neighbourhood (p<0.05). Adjusting for exposure duration improved model fit and the strength of associations.