AIM: To compare the ability of model for end-stage liver disease (MELD)-Na and Maddrey discrimination function index (DFI) to predict mortality at 30 and 90 d in patients with alcoholic hepatitis (AH). METHODS: We prospectively assessed 52 patients with AH. Demographic, clinical and laboratory parameters were obtained. MELD-Na and Maddrey DFI were calculated on admission. Short-term mortality was assessed at 30 and 90 d. Receiver operating characteristic curve analysis was performed. RESULTS: Thirty-day and 90-d mortality was 44% and 58%, respectively. In the univariate analysis, sodium levels was associated with mortality at 30 and 90 d (P = 0.001 and P = 0.03). Child stage, encephalopathy, ascites, or types of treatment were not associated with mortality. MELD-Na was the only predictive factor for mortality at 90 d. For 30-d mortality area under the curve (AUC) was 0.763 (95%CI: 0.63-0.89) for Maddrey DFI and 0.784 for MELD-Na (95%CI: 0.65-0.91, P = 0.82). For 90-d mortality AUC was 0.685 (95%CI: 0.54-0.83) for Maddrey DFI and 0.8710 for MELD-Na (95%CI: 0.76-0.97, P = 0.041). CONCLUSION: AH is associated with high short-term mortality. Our results show that MELD-Na is a more valuable model than DFI to predict short-term mortality.