Abstract: Objective Compare the clinical values of four scoring models, including the model for endstage liver disease (MELD) , MELD-Na, Scoring model of severe viral hepatitis (SMSVH) and Child-Turcotte-Pugh (CTP) scoring models, to predict the short-term prognosis of the patients with chronic severe hepatitis B.Methods Clinical information of 85 inpatient patients with chronic severe hepatitis B were recordedand followed up for 3 months.The patients were then divided into survival group and death group.All the patients were evaluated according to MELD, MELD-Na, SMSVH and CTP scoring models.The receiver operating characteristic curve (ROC) were used to compare the predictive ability of each model.Estimate the survival time and the survival rate by Kaplan-Meier (K-M) method and calculatethe total predictive agreement rate according to respective optimum cut-off value.Results The average scores of MELD-Na, MELD, SMSVH and CTP scoring models in the death group were greater than that in the survival group (P < 0.01) .The area under the curve (AUC) were 0.848, 0.833, 0.802 and 0.752 respectively.There were no difference of the predictive power and total predictive agreement rates of four models (χ2 = 6.64, P = 0.084) (χ2 = 2.737, P = 0.434) .The total predictive agreement rates of four models were 82.35%, 77.64%, 77.65% and 71.76% respectively.Conclusion MELD-Na, MELD, SMSVH and CTP scoring models has better effect in predicting short-term prognosis of the patients with chronic severe hepatitis B.There were no significant differences in predictive power.The clinical value of four scoring models were approximate.