Changing Etiological Patterns and Predictors of In-Hospital Mortality in Decompensated Cirrhosis: A 10-Year Cohort Study Before and After COVID-19

Background: Decompensated cirrhosis is associated with high in-hospital mortality, influenced by disease severity and underlying etiology. The COVID-19 pandemic may have altered both the etiological spectrum and clinical presentation of hospitalized patients. This study aimed to assess longitudinal changes in etiology and identify predictors of in-hospital mortality over a 10-year period. Methods: We conducted a retrospective cohort study including 812 patients hospitalized with decompensated liver cirrhosis between 2015 and 2025. Patients were grouped into pre-COVID-19 (2015–2019), COVID-19 (2020–2021) and post-COVID-19 (2022–2025) periods. Etiological factors and mortality rates were compared using chi-square tests. Independent predictors were identified through multivariate analysis. A clinical risk score based on Child–Pugh stage, platelet count and age was developed and evaluated using ROC analysis. Results: Alcohol-related cirrhosis increased significantly from 51.2% (pre-COVID-19) to 90.4% (COVID-19) and remained high post-COVID-19 (86.3%) (p < 0.001), while HCV decreased from 34.4% to 13.5% and stabilized at 14.8% (p < 0.001). HBV showed no significant variation. All-cause in-hospital mortality increased from 19.7% pre-COVID-19 to 42.3% during COVID-19 and remained elevated post-COVID-19 at 34.5% (p < 0.001). Independent predictors of all-cause in-hospital mortality included advanced Child–Pugh stage, thrombocytopenia and age above 70 years. The risk score (0–7 points) showed good discrimination (AUC = 0.752), with mortality rates of 2.8%, 24.0% and 45.7% across increasing risk categories. A score <5 had a negative predictive value of 84.3%. Conclusions: A significant etiological shift from HCV to alcohol was observed, accompanied by persistently increased mortality after COVID-19. Thrombocytopenia remains an important predictor of mortality. The proposed score enables simple and effective risk stratification.