Floods are among the most frequent and damaging natural hazards, affecting millions of people worldwide, and the risk of catastrophic losses due to flooding is expected to increase as a result of climate change. The possibility of predicting and estimating the expected fatalities in flood-prone regions is among the top priorities of decision-makers in flood risk management. Thus, predicting the conditions leading to loss of life is crucial for assessing the risk to the population. Here we focus on the Po River District in Northern Italy which covers the largest Italian hydrographic basin. We demonstrate that the occurrence of flood-related fatalities can be estimated by utilizing a random forest (RF) algorithm applied to a dataset of fatalities that occurred in this area from 1970 to 2019. This method relies on nine explanatory variables that describe the hazard intensity, and the environmental and sociodemographic conditions leading to fatalities. The proposed model is a primary attempt to estimate the probability of flood-related fatalities in the Italian context, and it provides a proxy for the quantitative estimation of flood risk to the population.