*Purpose: With the rising proportion of patients receiving liver transplantation for alcoholic hepatitis, it is important to identify methods of ascertaining hospitalizations with alcoholic hepatitis to facilitate research using large administrative databases.

*Methods: This was a retrospective study using data from patients with cirrhosis identified between 2008 and 2016 in the Veterans Health Administration. Hospitalization data including admission laboratory data and administrative codes were obtained for each patient. We randomly sampled charts from patient hospitalizations with abnormal liver tests or codes for alcoholic liver damage or acute hepatitis. Charts were manually adjudicated for the presence or absence of alcoholic hepatitis as the gold standard using published criteria. Several a priori algorithms were then evaluated to determine the positive predictive value (PPV) of each, as well as the 95% confidence interval (CI).

*Results: A total 355 randomly sampled patient hospitalizations were manually adjudicated, 36% or which were determined to have alcoholic hepatitis. The median age was 57 years (interquartile range [IQR] 53, 62), 96.9% male, with 78.3% alcohol or alcohol plus hepatitis C virus as the primary etiologies of liver disease. Evaluating four series of algorithms (Table), a combination of a diagnostic code for alcoholic hepatitis and supportive laboratory findings yielded ≥90% PPV, with >95% PPV using optimized laboratory cut points.

*Conclusions: We have identified algorithms to identify alcoholic hepatitis hospitalizations with extremely high PPV. These may be used in retrospective studies incorporating large administrative datasets.

« Back to 2021 American Transplant Congress