Significance: People who experience homelessness endure extreme adversity, putting them at up to a five-fold increased risk of early death compared to the general population in the US. Existing, but limited data, on homeless mortality suggests that death rates vary substantially by location. However, only a handful of states collect data on homeless deaths, and between state comparisons are limited by differences in definitions and estimation strategies. Thus, a better understanding of state-level geographic variation in homeless mortality rates is drastically needed.

Aim: We estimate the first comparable and complete estimates of state-level homeless mortality rates in the US.

Data & Methods: The decennial US Census includes individuals experiencing homelessness (n=423,000 in 2010) who are enumerated at emergency and transitional shelters as well as unsheltered locations such as soup kitchens, mobile food vans, vehicles, and tent encampments. We link individual-level records of people experiencing homelessness at the time of the 2010 Census to the US Census Bureau Numerical Identification file (Numident), which contains full-population death records from the Social Security Administration. We then use discrete time-survival analysis to assess mortality during the 5-year period following the 2010 Census and estimate age-adjusted homeless mortality rates for each US state.

Preliminary Results: We have linked the data, established our sample of people experiencing homelessness, and begun estimating homeless mortality rates. These preliminary estimates show substantial variation by state. However, we are unable to share numbers or comment on patterns as these results have not yet been approved for sharing by the US Census Bureau.

Next Steps: Before the IAPHS conference, we will finalize our state-level homeless mortality estimates and go through the required disclosure avoidance review process to get Census Bureau approval for presenting the results.