Objective:To study the accuracy of prognosis evaluation of patients underwent liver transplantation with two acute physiology and chronic health evaluation Ⅱ (APACHE Ⅱ) models retrospectively. Methods:The clinical data of 74 cases of liver transplantation performed between March 2004 and August 2005 in Ruijin Hospital of Medical School of Shanghai Jiaotong University was analyzed retrospectively.All the cases were divided into survival group (69 cases) and died group (5 cases).The gender,age,etiology,preoperative childPugh classification and operative pattern were compared between two groups.APACHE Ⅱ score,risk of death,predicted hospital mortality calculated by original and modified APACHE Ⅱ models were compared also.The calibration and discrimination of two models were tested with HosmerLemeshow goodness-of-it and receiver operating characteristic curve statistic analysis respectively. Results:There was not magnificent difference in gender,age,etiology,preoperative childPugh classification and operative pattern.But score and risk of death had a significant difference between two groups.Original model overestimated hospital mortality.The calibration and discrimination of modified model were better than those of the original model. Conclusion:Original model overestimates the mortality.The modified model can provide an closely accurate prognosis evaluation of patients performed liver transplantation.