RESUMO
Using an extensive database of every resident death in Virginia from 2005 to 2020, climate-mortality relationships are examined for 12 climatically homogeneous regions within the Commonwealth. Each region is represented by a first-order weather station from which archived temperature and humidity data are used to generate a variety of biometeorologically relevant indices. Using these indices and other variables (such as air quality and heat and cold waves), daily mortality and climate relationships are modeled for each region over a 21-day lag period utilizing generalized additive models and distributed lag non-linear models. Optimal models are identified for each region, and a consensus model was also run based on maximum temperature to facilitate inter-regional comparisons. The relative risk of mortality varies markedly as a function of climate between regions, with U-shaped, J-shaped, and inverse linear relationships evident. Cold mortality exceeds heat mortality across most of Virginia (typical relative risks are 1.10 for cold and 1.03 for heat), with cold risks strongest at lags 3 to 10. Low temperatures (or low humidity) are protective at lags 0-2 days except in the colder, western parts of state. Heat mortality occurs at short lags (0-2 days) for three-fourths of the stations, but the spatial pattern is random. Mortality displacement is evident for most regions for several days following the heat-related spike. Although the use of region-specific models is justified, the simple consensus model based on a consistent set of predictors provides similar results.