Observation time, temperature diurnality, and weather variable selection influence heat-related mortality

Mortality associated with extreme heat is the main weather-related cause of death in the United States. Although models that explore temperature/mortality relationships have become quite sophisticated, the influence of observation time and lag of the weather variables has not been thoroughly examined. If the temporal sampling of the independent variable influences heat-mortality relationships, it is important to know which temperature variables to incorporate into models and retrospective analyses.

Mutli-decadal time series of daily mortality counts and hourly temperature for seven U.S. cities with different climates were examined via a generalized additive model. The temperature effect was modeled separately for each hour of the day (with up to three-day lags) in addition to several different methods of determining daily maximum, minimum, and mean temperature.

In three northern U.S. cities that are typically considered to have “heat-sensitive” populations (Boston, Philadelphia, and Seattle), the heat-mortality response is diurnal, with the strongest relationships for afternoon or maximum temperature at lag zero (day of death) or lag one (afternoon and evening). The temperature-mortality relationships are weaker in the warmer, southern U.S. cities, where slightly stronger relationships to morning temperatures are evident. The strongest temperature-mortality relationships are associated with maximum temperature, though the results using mean temperature are comparable.

The results for the three northern cities, in which heat-mortality relationships developed using afternoon/maximum temperatures are stronger than those using morning/minimum temperatures, suggest that cooler mornings may provide a protective effect from heat-related mortality in these climates. In general, choice of observation time, lag, and variable calculation method often impacts the robustness of the model results. In general, heat-related mortality is most closely coupled to afternoon and maximum temperatures in most cities, particularly those that are typically heat-sensitive.