Independent effects of long and short-term exposures to non-optimal increased temperature on mortality.

Although the short-term heat effects are well-established, longer-term effects, beyond those, have recently received attention, in the context of climate change. Our study aims to investigate the potential effects of long-term exposure to non-optimal warm period temperatures on all-cause mortality in four large regions in the UK, Norway, Italy, and Greece. Daily all-cause mortality counts from 1996 to 2018 for four European NUTS-2 regions including 52-662 small areas were collected and associated with spatiotemporal temperature estimates. A NUTS-2 region-specific mixed quasi-Poisson over-dispersed model, with a random intercept per small area within NUTS-2 regions, was applied to investigate the association between long-term temperature exposure and mortality during the warm period (May to September), adjusting for short-term temperature, seasonality, long-term trends, and small-area population characteristics. As long-term temperature exposure indices per small area, we considered: 1) the warm period annual average temperature, 2) the annual standard deviation (SD) of warm period temperature, and 3) the coefficient of variation of warm period temperature (CV). We found consistent results following short-term temperature exposure on mortality, with higher effects in southern areas. Results on the shape of the long-term association between average temperature and mortality differed by country, while the different temperature metrics produced inconsistent findings. Increased mortality was associated with increased annual warm season temperature, lower SD and increased CV in Greece, with higher SD and decreased CV in Italy and with decreased annual temperature and CV in Norway. Effects in the UK did not reach the nominal level of statistical significance. Although our study implies an impact on mortality resulting from longer-term temperature exposure, its direction varied across areas and on the temperature metric used. Further research is warranted, applying non-ecological study designs and covering various geographical areas to capture the impact of individual and area-specific characteristics.