Predicting short-term mortality and long-term survival for hospitalized US patients with alcoholic hepatitis

Background: No study has evaluated current scoring systems for their accuracy in predicting short and long-term outcome of alcoholic hepatitis in a US population. Methods: We reviewed electronic records for patients with alcoholic liver disease (ALD) admitted to Parkland Memorial Hospital between January 2002 and August 2005. Data and outcomes for 148 of 1,761 admissions meeting pre-defined criteria were collected. The discriminant function (DF) was revised (INRdf) to account for changes in prothrombin time reagents that could potentially affect identification of risk using the previous DF threshold of >32. Admission and theoretical peak scores were calculated by use of the Model for End-stage Liver Disease (MELD). Analysis models compared five different scoring systems. Results: INRdf was closely correlated with the old DF (r ² = 0.95). Multivariate analysis of the data showed that survival for 28 days was significantly associated with a scoring system using a combination of age, bilirubin, coagulation status, and creatinine (p < 0.001), and an elevated ammonia result within two days of admission (p = 0.012). When peak values for MELD were included, they were the most significant predictor of short-term mortality (p < 0.001), followed by INRdf (p = 0.006). Conclusion: On admission, two scoring systems that identify a subset of patients with severe alcoholic liver disease are able to predict >50 % mortality at four weeks and >80 % mortality at six months without specific treatment.