Determinants and spatio-temporal changes in morbidity curves of Italian population

Understanding the mechanisms that drive the changes in morbidity curve is of paramount importance to improving life quality and designing interventions and policies. In this study, we analyze data from the Italian behavioral risk factor surveillance system (PASSI) and propose a pseudo-panel approach to study the spatio-temporal changes in Italian local health authorities (ASLs). We develop a Bayesian logistic hierarchical model, in which unit-specific covariates (e.g. age, sex, socio-economic status, etc) explain the observed variations by means of regression coefficients varying in space (ASLs) and time (cohorts). We leverage a state space formulation of the model where temporal changes are driven by correlated impulses and the degree of correlation is determined by weighting available external information at the ASL level (e.g. region, social habits, pollution, etc). By using out-of-sample predictive inference, we show how our method outperform other standard approaches, allowing for interpretable results that highlight how different social and environmental factors influence the shape of the morbidity curves.