ML ensemble predicts post-TIPS encephalopathy and rebleeding risk at 2 and 6 years

Interpretable machine learning ensemble predicted 2-year hepatic encephalopathy after TIPS with AUC 0.960 (external 0.914) and variceal rebleeding with AUC 0.966, outperforming conventional scores (AUC 0.528–0.563). Top predictors: diabetes and recent bleeding.

• External validation for OHE prediction yielded AUCs of 0.914 at 2 years and 0.924 at 6 years.
• All prespecified subgroups achieved 2-year AUC ≥0.93, including 0.993 for OHE prediction in portal vein thrombosis.
• SHAP analysis identified diabetes as the strongest predictor for OHE and recent bleeding for variceal rebleeding.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.