Development and external validation of a pre-drainage prediction model for moderate-to-severe gallstone-related acute cholangitis

Background Acute cholangitis may rapidly progress to sepsis and organ dysfunction, making early severity assessment essential. However, few practical tools are currently available based on routine pre-drainage data. This study aimed to develop and externally validate a pre-drainage model for identifying moderate-to-severe gallstone-related acute cholangitis. Methods This two-center retrospective study included 159 patients in the training cohort and 119 in the external validation cohort. Tokyo Guidelines 2018 Grades II–III defined moderate-to-severe disease. Candidate variables measured within 24 h of admission and before drainage were selected using least absolute shrinkage and selection operator regression and refitted using multivariable logistic regression. Performance was assessed by bootstrap resampling and external validation. Results The four-predictor model included the direct-to-total bilirubin ratio, ln-transformed systemic inflammatory response index, red cell distribution width, and serum calcium. Apparent and optimism-corrected areas under the curve were 0.823 and 0.777, respectively. The external area under the curve was 0.871 (95% confidence interval, 0.804–0.937), but the model overestimated absolute risk (calibration intercept, − 0.828). Exploratory intercept-only recalibration reduced the Brier score from 0.158 to 0.137. Decision curve analysis suggested net benefit across selected threshold probabilities. Conclusions The four-variable model showed good external discrimination using routinely available pre-drainage laboratory variables. However, model-selection uncertainty and systematic external overprediction indicate that recalibration and further prospective evaluation are warranted before clinical implementation.