Objective To evaluate the prognostic value of the model for end-stage liver disease (MELD) and △MELD in liver failure patients infected with hepatitis B virus.Methods Based on prospective study design,98 hospitalized cases were studied and followed up for 24 weeks.The clinical data were recorded.We calculated the score of MELD and △MELD,and also compare the score between the survival group and death group.Using ROC curve plotting obtained the better decisive threshold.The case fatality rate were compared at different time points which the patients were classified by the best critical value of MELD and △MELD.We draw the Kaplan-Meier survival curve of different group and analyse the change of survival rate by log-rank analysis.Results 52 of 97 patients died and 46 survive during 24 weeks of followup.There was significant difference between the two groups for MELD and △MELD (P ＜ 0.01 ).The case fatality rate in group which MELD ≥ 23 was obviously higher than in that MELD ＜ 23. The rate in group which △MELD ＞4.5 was obviously higher than in that △MELD ＜ 4.5 (P ＜ 0.001 ).The area under curve (AUC) for the twelfth and 24th week' s prognosis judgment of △MELD(0.823,0.815) was larger than that of MELD ( 0.680,0.684 ) ( P ＜ 0.05 ).Survival analyses (Kaplan-Meier) indicated that there were significant differences in cumulative survival rates among the groups which were grouped by optimization critical value(P =0.000).Conclusions The scoring system of MELD also applied to the forecasting of prognosis for severe hepatitis B patients in China.The accuracy of △MELD to predict the prognosis was higher than that of MELD.The combination of MELD and △MELD showed good clinical practical value.