Résumé
Background: the COVID-19 problem needs a comprehensive approach to identify essential factors involved in the dispersion and severity of the disease. This study explores various determinants of COVID-19 hospitalizations in France: population health status, air pollution, meteorological and socioeconomic factors, while adjusting for the initial state of spread of the virus and population mobility.Methods: for this population-based study, we retrospectively explored these determinants at an aggregate geographic level (all 96 metropolitan French departments). We focused on the cumulative rate of patients hospitalized from 31/03/2020 to 25/05/2020. This period provides unique conditions that can no longer be met to explore the COVID-19 situation and limit bias in this type of studies: no specific immunity against the virus, restricted mobility during lockdown. After a preliminary variable selection based on Pearson’s correlations and expert opinion, we estimated multivariate linear regression models that integrate the spatial autocorrelation between departments.Findings: the best model was the spatial error model, with a significant spatial error term (p=0·024). The rate of COVID-19 hospitalizations was significantly positively correlated with the age standardized prevalence of diabetes (p=0·0008), psychiatric disorders (p=0·044), and ozone exposure (p=0·046). It was negatively correlated with exposure to ultraviolet radiation (p=0·019).Interpretation: this study provides a better understanding of the potential regional differences concerning contextual variables that could eventually lead public authorities to differentiate public health care decisions according to regional specificities. It also invites us to test the effect of the variables highlighted by this study in individual studies with a higher level of evidence.