Substantial discrepancies are evident in the clinical features and natural history of coronavirus disease 2019 (COVID-19), among different countries, so we aimed to evaluate the fatality rate and to identify predictors of mortality in a cohort of hospitalized patients. We performed a retrospective multicenter cohort study on medical records of patients admitted between 1st March and 28th April 2020, involving three hospitals in Northern Italy. We included 1697 patients older than 18 years of age and with a confirmed diagnosis of SARS-CoV-2 infection by reverse-transcriptase polymerase chain reaction (RT-PCR). During the study period we observed 504 deaths, with a CFR of 29.7%. We further looked for predictors of mortality in a subset of 486 patients (239 males, 59%); the median age of the study population was 71 [58-80] years. Among demographic and clinical variables considered, age, a diagnosis of cancer, obesity and current smoking independently predicted mortality. When laboratory data were added to the model, age, the diagnosis of cancer, and the baseline PaO2/FiO2 ratio were independent predictors of mortality. During the COVID-19 outbreak, the CFR of hospitalized patients in Northern Italy was high, approaching 30%. The identification of mortality predictors might contribute to better stratification of individual patient risk.