OBJECTIVE: To explore the function of the baseline model for end-stage liver disease (MELD) scores, MELD-Na scores and iMELD scores in short-term prognosis in the initial treatment of hepatitis B virus (HBV) related acute-on-chronic liver failure (ACLF) patients. METHODS: 232 HBV-related ACLF patients who received initial treatment in 302 Military Hospital of China from January 2011 to January 2013 were enrolled in this prospective clinical follow-up. The relationship between the baseline MELD scores, MELD-Na scores, iMELD scores and clinical outcomes were analyzed, and the value of these three models for short term prognosis was assessed. RESULTS: Finally the 12-week clinical follow-up was completed in 191 patients, with the completion rate of 82.33%. Eighty-five patients died, with the fatality rate of 44.50%. Compared with the survival group, in non-survival group, the baseline of MELD scores (26.65 ± 7.75 vs. 21.19 ± 5.42, t=-5.720, P=0.000), MELD-Na scores (29.16 ± 11.35 vs. 21.72 ± 6.33, t=-5.729, P=0.000), iMELD scores (47.19 ± 10.96 vs. 38.02 ± 7.01, t=-7.011, P=0.000), total bilirubin (TBil: 374.3 ± 150.1 µmol/L vs. 305.5 ± 147.1 µmol/L, t=-3.182, P=0.002), creatinine (Cr: 110.7 ± 90.1 µmol/L vs. 71.1 ± 35.1 µmol/L, t=-4.157, P=0.000) and international normalized ratio (INR: 2.3 ± 0.9 vs. 2.0 ± 0.6, t=-2.754, P=0.006) were significantly increased, but the baseline of serum Na⁺ (132.8 ± 6.1 mmol/L vs. 136.7 ± 5.1 mmol/L, t=4.861, P=0.000) was significantly lowered. It was shown by Spearman correlation analysis that the baseline MELD scores, MELD-Na scores and iMELD scores all had positive correlation with the short-term prognosis of patients (r value was 0.398, 0.404, and 0.470, respectively, all P=0.000), the baseline of serum Na⁺ had a negative correlation with the short-term prognosis of patients (r=-0.365, P=0.000). It was shown by receiver operating characteristic curve (ROC curve) that the cut-off scores of the baseline of MELD scores, MELD-Na scores and iMELD scores were 25.07, 25.43 and 43.11 respectively, and the area under ROC curve (AUC) of the baseline of MELD scores, MELD-Na scores and iMELD scores were 0.731, 0.735 and 0.773, respectively. The sensitivity of the three models was 55.3%, 57.7%, 63.5%, and the specificity was 84.9%, 84.0%, 84.9% respectively. The value of the three models had no difference in short-term prognostic prediction. According to the respective cut-off score, the three prediction models were divided into four groups, and all of them had differences in fatality rate on the whole (χ² for MELD scores was 34.740, P=0.000; χ² for MELD-Na scores was 36.861, P=0.000; χ² for iMELD scores was 50.127, P=0.000). The mortality was elevated gradually as the equation scores increased. CONCLUSIONS: The baseline of MELD scores, MELD-Na scores and iMELD scores can predict well the short-term prognosis of the initial treatment in HBV-related ACLF patients, and have relatively good clinical value for guiding therapy.