Vulnerability patterns to heat and cold across Europe: a spatial two stage analysis of temperature-mortality risks in 800 urban areas

Abstract

Background: Despite efforts of previous multi-location studies to characterize temperature vulnerability patterns across large regions, evidence on spatial variation of risks and its drivers remains patchy, as geographical coverage was limited to locations with available observed health data. This study offers a comprehensive geographical assessment of the differences in vulnerability to non-optimal temperature across a representative sample of 800 European cities. Methods: We firstly estimated the exposure-response associations in 150 cities of 14 European countries for whichobserved daily temperature-mortality data was collected in the MCC Collaborative Research Network. Second, we developed a novel spatial multivariate meta-regression model that characterizes risk in terms measurable meta-predictors and residual spatial structures. The meta-predictors were represented by city-level factors available in the European Union (EU) database Urban Data Platform (UDP), including environmental, socio-economic, and infrastructural characteristics. We then used the selected meta-regression model to predict temperature-mortality associations in the 800 cities, with a full coverage of the EU region.Results: The predicted mortality risks showed large geographical heterogeneity across Europe, with relative risks in the range 1.00-2.30 for heat and 1.22-2.07 for cold, respectively. Heat-related mortality displayed a clear geographical distribution, with an increasing North-South gradient and a high-risk area in Central Europe, while a more diffuse pattern was observed for cold. The geographical variation is largely explained by a set of meta-predictors, specifically gross domestic product, annual levels of particulate matter, and access to green areas.Conclusions: This first EU-wide analysis on temperature-related mortality provided a comprehensive picture of the geographical distribution of vulnerability to heat and cold. We found clear geographical differences in risks associated with non-optimal temperature, with the identification of areas of high vulnerability. These differences seem partly explained by measurable urban characteristics. These findings can contribute to the design and implementation of EU-wide public health and climate policies.