We sought to assess prevalence, and utility of discriminant function (DF) and MELD score in predicting septic events (SE), type 1 hepatorenal syndrome (HRS), and short-term mortality in severe alcoholic hepatitis (AH). Charts of patients with AH (group 1) and cirrhosis without AH (group 2) were retrospectively reviewed. Severe AH, discriminant function (DF) >or= 32 was treated with pentoxifylline. One hundred ninety-five patients were enrolled in the study and divided into 2 groups: group 1, n=99, and group 2, n=96. Of those with AH, 82% had a DF >or= 32 at presentation. Group 1 patients had a higher prevalence of SE (38% versus 25%, P=.04), type 1 HRS (30% versus 9%, P=.0003), and short-term mortality (28% versus 7%, P=.0001). In patients with AH, a MELD score >or=20 (but not a DF >or= 32) at presentation was an independent predictor of a SE (odds ratio [OR] 2.8 [1.0-7.9], P=.04), HRS (OR 4.0, 95% confidence interval [CI] 1.0-16.6, P=0.05), and short-term mortality (OR 6.4, 95% CI 1.1-37.6, P=.03). Kaplan-Meier survival curves confirmed that that a MELD >or= 20 but not a DF >or= 32 was associated with a poorer survival (P = .005 and .5, respectively). In conclusion, patients with severe AH have higher prevalence of SE, HRS, and short-term mortality compared to those with cirrhosis without AH. A MELD score >or=20 at presentation is an independent predictor of these adverse events in patients with AH who have been treated with pentoxifylline.