Italy was harshly hit by COVID-19, registering more than 35,000 deaths between February and July, 2020. The virus spread unequally across the country, with northern regions witnessing more cases and deaths than those in the centre and south. We investigate demographic and socio-economic factors that contributed to the diverse regional impact of the virus in Italy. Within a smoothing framework, we divide regions into three well-defined groups of High, Middle and Low mortality by cluster analysis. Extending the Poisson regression model to account for regional clusters, we find that COVID-mortality is positively associated with the share of ICU utilization, GDP per capita, proportion of the older population and the number of COVID-19 positive cases, while it is negatively associated with the delay of region-specific outbreaks and the number of tests performed. Our results have relevant policy implications for potential resurgence of COVID-19 infections in Italy and across the world.