Alcoholic hepatitis (AH) is an acute hepatic inflammatory disorder associated with excessive alcohol consumption. Without treatment, mortality rates of up to 50% have been reported in severe cases. Population-based studies on the epidemiology of AH are limited, and partially hindered by the lack of a validated algorithm to identify cases using administrative data. This study validated the coding of AH in the Discharge Abstract Database (DAD) for the Calgary region based on a review of medical records. We found the accuracy of coding for AH in this administrative database to be sub-optimal, with a positive predictive value (PPV) of 54%. Our data do not support the use of administrative data for large-scale epidemiological studies of AH. We also identified risk factors and assessed well-known prognostic models for 90-day mortality in a well-defined Canadian cohort of patients hospitalized for AH.