New machine learning model predicts cholangitis risk after ERCP

A novel interpretable machine learning model has been developed to predict post-ERCP cholangitis (PEC) in patients with malignant biliary obstruction (MBO). The model, utilizing data from 2831 patients, identified key risk factors including radiofrequency ablation and white blood cell count. Among various machine learning methods, the XGBoost model outperformed others, predicting PEC risk with accuracy metrics of AUC 0.7670 and AUC 0.7270 in internal and external cohorts, respectively, thus aiding clinicians in tailoring individualized treatment plans.