Exploring the Relationship Between Doctors Availability and Mortality in Italy: A Machine Learning Approach using Multivariate Regression Trees

We investigate how the percentages of medical specialists impact mortality rates and how they vary across different areas and time periods. The analysis is focused on Italian provinces during two different years, 2011 and 2019, to identify differences and similarities across geographical areas and time periods. We propose a new approach based on tree-based methodology that we will call \textit{Machine learning Enhanced Regression of ITAlian mortality rates (MERITA)}. It is based on a tree-based method called Multivariate RegressionTree (MRT), which is an extension of univariate regression trees. In this study, the MRT is used to identify the relationship between the percentage of active doctors by specialization per10,000 inhabitants and the Age-adjusted Mortality Rate (AMR) for specific disease and gender.We define nine models. Two focus on the overall AMR by gender. Seven are defined with specific disease-gender-AMR as response variables. All include the availability of active doctors and other socio-economics aspects as covariates. Our findings highlight that the percentages of medical specialists have a significant impact on the AMRs and how it changes across geographic areas and over time, providing insights on which specialists are most in need and where.