Impact of extreme temperatures on mortality in nine French cities: evolution between 2001 and 2009

Abstract

Background and aims:In France, past major heat waves have been characterized by an excess mortality. A national prevention plan was implemented in 2004, and its contribution to the reduction of the health impact is still to be determined. As the plan includes measures targeted on very warm days, we assessed if the mortality response during these very warm days has changed over time. Methods: The analysis was performed in nine large French urban areas, for the period 2001-2009. August 2003 was excluded from the analysis. The associations between temperature and non-accidental mortality were investigated using a generalized additive Poisson regression model. The mean temperatures at lag 0 and 1-7 were introduced as natural splines with 3 degrees of freedom. We assessed the percent increase in mortality during the 1% warmest days compared to the 10% warmest days for the period 2001-2003 with no heat prevention plan, 2004-2006 (first period of the heat prevention plan) and 2007-2009 (second period of the heat prevention plan). A similar analysis compared the 1% coldest days to the 10% coldest days. City specific results were combined using random effect models. Results: Mean temperatures in the nine French cities were similar over the three periods, with less important heat waves during 2007-2009. A mean difference of 4.6°C was observed between very warm and warm days (5.4°C between very cold and cold days). During 2001-2003 and 2004-2006, the mortality differences between very warm and warm days were similar: 7% [4:11] and 7% [5:9]. It decreased to 3% [0:6] during 2007-2009. A 3% difference was observed between very cold and cold days during the first periods. It increased to 6% [3:10] during 2007-2009. Conclusions: Despite the heat prevention plan, mortality remains higher during very warm days compared to warm days. A decrease was observed in 2007-2009 maybe due to the absence of sustained heat episodes. The analysis will be extended to a longer period and 7 more cities to better reflect the geographic and climatic variability.