Data-Driven Analysis of Mortality and Survival Outcomes in Patients with Liver Cirrhosis: A Global Healthcare Analytics Study

Abstract: Liver cirrhosis is a progressive chronic disease associated with high morbidity and mortality, requiring accurate risk stratification to support clinical decision-making. This study presents a data-driven analytical framework integrating survival analysis, machine learning, and uncertainty quantification to improve mortality prediction in cirrhosis patients. A retrospective dataset of 418 patients was analyzed using Kaplan–Meier estimation and Cox proportional hazards modeling to evaluate survival patterns and identify significant predictors. For predictive modeling, multiple supervised learning algorithms—including Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Gradient Boosting, and Extreme Gradient Boosting (XGBoost)—were implemented to classify patients into high- and low-risk groups. Model performance was assessed using accuracy, recall, F1-score, and ROC-AUC, with cross-validation employed to ensure robustness. To enhance reliability, conformal prediction was applied to quantify predictive uncertainty at a predefined confidence level. Results indicate that disease stage, age, bilirubin, and albumin are significant predictors of mortality. Ensemble models demonstrated superior predictive performance, with XGBoost achieving the highest recall and strong discrimination. Conformal prediction provided well-calibrated uncertainty estimates, improving the interpretability and trustworthiness of model outputs. The findings demonstrate that integrating statistical and machine learning approaches enhances mortality risk stratification and supports the development of reliable clinical decision-support systems.

Keywords: Liver Cirrhosis; Survival Analysis; Machine Learning; Mortality Prediction; Cox Proportional Hazards; XGBoost; Risk Stratification; Conformal Prediction