The performance of prognostic models as predictors of mortality in patients with acute decompensation of cirrhosis.

BACKGROUND Although several prognostic models have been proposed for cirrhotic patients listed for transplantation, the performance of these scores as predictors of mortality in patients admitted for acute decompensation of cirrhosis has not been satisfactorily investigated. AIMS To study MELD, MELD-Na, MESO, iMELD, Refit-MELD and Refit MELD-Na models as prognostic predictors in cirrhotic patients admitted for acute decompensation, and to compare their performance between admission and 48 hours of hospitalization to predict in-hospital mortality. MATERIAL AND METHODS This cohort study included cirrhotic patients admitted to hospital due to complications of the disease. Individuals were evaluated on admission and after 48 h of hospitalization, and mortality was evaluated during the present admission. RESULTS One hundred and twenty-three subjects with a mean age of 54.26 ± 10.79 years were included; 76.4% were male. Mean MELD score was 16.43 ± 7.08 and 52.0% of patients were Child-Pugh C. Twenty-seven patients (22.0%) died during hospitalization. Similar areas under the curve (AUROCs) for prognosis of mortality were observed when different models were compared on admission (P > 0.05) and after 48 h of hospitalization (P > 0.05). When models executed after 48 h of hospitalization were compared to their corresponding model calculated on admission, significantly higher AUROCs were obtained for all models (P < 0.05), except for MELD-Na (P = 0.075) and iMELD (P = 0.119). CONCLUSION The studied models showed similar accuracy as predictors of in-hospital mortality in cirrhotic patients admitted for acute decompensation. However, the performance of these models was significantly better when applied 48 h after admission when compared to their calculation on admission.