Development and internal validation of a multivariable prediction model for in-hospital mortality in acute cholangitis: a retrospective cohort study incorporating comorbidities

The Tokyo Guidelines 2018 (TG18) stratify acute cholangitis severity but do not incorporate patient comorbidities. We investigated whether adding comorbidities improves prediction of in-hospital mortality and developed a multivariable risk prediction model. We conducted a single-center retrospective cohort study of consecutive admissions undergoing endoscopic retrograde cholangiopancreatography (ERCP) for TG18-defined acute cholangitis due to choledocholithiasis. The unit of analysis was the admission-level ERCP case, and in-hospital mortality was assessed at discharge for each admission; repeat admissions after discharge were treated as separate cases. Because deaths were rare, we used the Synthetic Minority Oversampling Technique (SMOTE) to create a balanced development dataset and prespecified two logistic regression models (severity-only vs. severity plus comorbidities). Internal validation used bootstrap resampling, and performance was summarized using optimism-corrected C-indices. Among 1,675 included cases, 1,616 survived and 59 died (mortality rate 3.52%). The optimism-corrected C-index was 0.77 (95% CI 0.75–0.79) for the severity model and 0.90 (95% CI 0.88–0.91) for the comorbidities model. Predictors with higher odds ratios included performance status = 4 (OR 7.96), liver cirrhosis, chronic heart failure, immunosuppressive drug therapy, malignant tumor, and hemodialysis. Predicted risks from the comorbidities model were additionally applied to the original dataset to describe observed mortality across predefined risk groups. In this single-center retrospective cohort, adding prespecified comorbidity information to TG18 severity was associated with improved discrimination for in-hospital mortality. External validation is warranted before clinical implementation.