Neighborhood-level confounding in epidemiologic studies: unavoidable challenges, uncertain solutions.

In early contextual studies, the aim was to demonstrate overall neighborhood influences rather than dissecting such influences into their components. Researchers did not need to worry about neighborhood-level confounding. However, as our interest shifts to the exploration of specific environmental effects, failure to consider neighborhood-level confounding may result in severely biased associations. We argue that neighborhood socioeconomic position and similar area structural factors may constitute powerful sources of confounding in studies of specific environmental factors and health. Controlling for neighborhood socioeconomic position is a convenient (but imperfect) adjustment strategy. Such control entails a minimal risk of overadjustment, but conveys a non-negligible risk of collider bias. Balancing the advantages and disadvantages, we suggest that researchers should often provide complementary analyses controlling for neighborhood socioeconomic position in studies of associations between specific environmental factors and health. Researchers should provide DAG-based descriptions of plausible scenarios to explore whether any decrease in the association of interest after adjustment for neighborhood socioeconomic position is likely due to neighborhood-level confounding, indirect pathway biases, or collider bias.