Longitudinal exposure assessments of neighbourhood effects in health research: What can be learned from people's residential histories?

Abstract

Health research into neighbourhood effects has generally examined neighbourhoods cross-sectionally, ignoring the fact that neighbourhood exposures might accumulate over people's lives and affect health outcomes later in life. Using longitudinal Dutch register data with complete 15-year residential address histories, we examined whether health effects of neighbourhood socioeconomic characteristics differ between cumulative and current exposures. We illustrated these differences between exposure assessments using suicide mortality among middle-aged adults. All suicides aged 40-64 years between 2012 and 2016 were matched with 10 random controls in a nested case-control design. We measured neighbourhood exposures longitudinally for circular buffers around residential addresses at the current address and through three accumulative measures, each incorporating the residential address history with increasing detail. Covariate-adjusted conditional logistic regressions were used to assess associations between suicide and neighbourhood social fragmentation, population density and unemployment rate. Our results showed that total and male suicide mortality was significantly lower in highly fragmented neighbourhoods when using accumulative exposures, but not when using the current residential address. However, we observed few differences in coefficients between exposures assessments for neighbourhood urbanicity and unemployment rate. None of the neighbourhood characteristics showed evidence that detailed cumulative exposures were a stronger predictor of suicide compared to more crude measures. Our findings provide little evidence that socioeconomic neighbourhood characteristics measured cumulatively along people's residential histories are stronger predictors of suicide mortality than cross-sectional exposures.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 17 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.