Machine learning-based risk prediction of overt hepatic encephalopathy after transjugular intrahepatic portosystemic shunt in patients with cirrhosis: a cohort study

Background: Overt hepatic encephalopathy (OHE) is a frequent complication after transjugular intrahepatic portosystemic shunt (TIPS) in patients with cirrhosis and can markedly impair quality of life and prognosis. This study aimed to develop and validate a machine learning-based risk prediction model to determine the most effective model and key predictive factors. Methods: This retrospective study included 297 patients with cirrhosis who underwent TIPS at the First Hospital of Shanxi Medical University from 2019 to 2024, among whom 89 developed postoperative OHE. Preoperative clinical characteristics and procedure-related variables were compared between the OHE and non-OHE groups using univariate analyses. Feature selection was conducted using least absolute shrinkage and selection operator regression and random forest. The selected variables were then used to develop five machine learning models: logistic regression, support vector machine, random forest, extreme gradient boosting (XGBoost), and artificial neural network. Model performance was evaluated in terms of discrimination, calibration, and clinical utility. The optimal model was further interpreted using Shapley additive explanations to identify key predictors. Results: The final predictors incorporated into the model were BUN, GGT, Age, FIB, and the portal vein puncture site. Model comparisons indicated some variation in predictive performance across models in the test cohort. XGBoost achieved an AUC of 0.792 (95% CI: 0.671-0.914) and showed relatively stable performance in calibration and clinical net benefit. Conclusion: The XGBoost model was developed using routine clinical indicators and procedure-related factors. It demonstrates potential utility in estimating the risk of post-TIPS OHE and may serve as an adjunct in preoperative risk assessment.