Objective: Comparison of the predictive performance of MELD 3.0, MELD-Na, and MELD scores in assessing the 3-month and 1-year survival prognosis of patients with decompensated cirrhosis. Methods: We conducted a retrospective analysis of clinical data from 438 patients diagnosed with decompensated cirrhosis who received either outpatient or inpatient care from January 2013 to December 2022. Based on their survival status at 3 months and 1 year, the patients were categorized into survival and death groups. The predictive value of MELD 3.0, MELD-Na, and MELD scores for 3-month and 1-year mortality was compared using receiver operating characteristic (ROC) curves, the area under the curve (AUC), calibration curves, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision curve analysis (DCA). Results: At the 3-month and 1-year follow-up points, 15.53% and 26.26% of patients had died, respectively. The area under the curve (AUC) for predicting 3-month and 1-year mortality for the MELD 3.0, MELD-Na, and MELD scores were 0.859 (0.805-0.913), 0.857 (0.802-0.912), 0.856 (0.800-0.911) and 0.841 (0.796-0.886), 0.832 (0.785-0.880), 0.830 (0.782-0.878), respectively. However, the differences were not statistically significant (P >0.05). The MELD 3.0 score can increase the score range for 18.0% of patients with MELD-Na scores, and 21.5% of patients with MELD scores are reclassified into a higher score category. On the calibration curve, all three models showed similar trends between predicted and actual probabilities at both time points. In terms of predicting 3-month mortality, the NRI for MELD 3.0 compared to MELD was 0.240 (0.009-0.401), a statistically significant difference (P = 0.032). For predicting 1-year mortality, the NRI and IDI for MELD 3.0 compared to MELD were 0.201 (0.079-0.401) and 0.032 (0.006-0.057), respectively, both of which were statistically significant (P < 0.05). In subgroup analyses, including male and female patients, viral and non-viral causes, and Child-Pugh Class B and C patients, no significant statistical differences were observed among the three models (P > 0.05). Conclusion: Compared to MELD or MELD-Na, MELD 3.0 demonstrates improved predictive performance, making it more effective for precise risk stratification in patients with decompensated cirrhosis.