Evolution of Sub-National Longevity and Causes of Death Composition Using Data on Italian Provinces

Amerigo Novaro , University of Padua
Davide Benussi, Università degli studi di Padova
Emanuele Aliverti, Università degli studi di Padova
Stefano Mazzuco, University of Padua

Understanding how cause-specific mortality rates evolve across space and time is crucial for assessing health disparities within and across countries. At sub-national scales, however, analyses are not trivial since stratifying death counts for causes and age groups leads to sparse tables with several zero death counts. In this work we focus on Italian provinces and suggest a flexible framework to estimate and interpret cause specific mortality trends. We propose a model for death counts as a function of the known population exposures and the latent smooth and spatially-varying latent mortality trajectories. Such rates are decomposed into a national annual trend, province-specific components that evolve over years, and a latent factor structure that captures dependencies between age groups and causes of death in a parsimonious and interpretable way. Spatial dependence is accounted for by the province-specific effects considering a Gaussian Process prior with distance-based covariance kernels, while the smooth temporal evolution is characterized with a polynomial expansion. The estimated provincial trajectories are then used as inputs for a model based clustering procedure, facilitating the identification of latent groups with similar mortality dynamics; ideally, this approach allows one to distinguish among provinces that show increasing mortality and others that show improvements, accounting for the different age composition and prevalence of the causes of death. Our results show interesting connections among provincial trends and the geographical location, suggesting that local health inequalities remain relevant and require constant attention.

See extended abstract

 Presented in Session 52. Flash Session Advances in Subnational and Small-Area Population Analysis