Improving spatial microsimulation estimates of health outcomes by including geographic indicators of health behaviour: The example of problem gambling

- The distribution of gambling losses can be estimated for small areas (R2=0.78).
- This can be used to improve estimates of problem gambling prevalence in small areas.
- We present empirically-calibrated local estimates of problem gambling prevalence.
- Including data on behaviours improves spatial microsimulation estimates of outcomes.

Gambling is an important public health issue, with recent estimates ranking it as the third largest contributor of disability adjusted life years lost to ill-health. However, no studies to date have estimated the spatial distribution of gambling-related harm in small areas on the basis of surveys of problem gambling. This study extends spatial microsimulation approaches to include a spatially-referenced measure of health behaviour as a constraint variable in order to better estimate the spatial distribution of problem gambling. Specifically, this study allocates georeferenced electronic gaming machine expenditure data to small residential areas using a Huff model. This study demonstrates how the incorporation of auxiliary spatial data on health behaviours such as gambling expenditure can improve spatial microsimulation estimates of health outcomes like problem gambling.