None of the statistical models we used suggested that 2016 should have been a year with an elevated count of heat-associated deaths in Maricopa County, despite the anomalous warmth. All combinations of exposure variables and model parameterizations we chose suggested that the number of heat-associated deaths in 2016 should have fallen near or below the historical average. This finding suggests that non-meteorological factors were responsible for the spike in heat-associated deaths in 2016. Among the plausible explanations is a change in the social service programs that provide shelter for homeless individuals. The number of heat-associated deaths among homeless individuals and the share of heat-associated deaths that those cases represented were both anomalously high in 2016. This pattern may be related to the observed 25% increase in the number of unsheltered homeless individuals in the community between 2015 and 2016. Non-meteorological factors may play a more important role in driving year-to-year and long-term trends in heat-associated mortality than commonly modeled or understood. Our findings suggest that heat-health risk assessments should strive to include temporal analysis of social vulnerability factors to be more informative.