Development and validation of a predictive model for post-endoscopic retrograde cholangiopancreatography cholangitis:...

Research on risk factors and predictive models for post-endoscopic retrograde cholangiopancreatography (post-ERCP) cholangitis remains limited. This study aimed to identify key risk factors for post-ERCP cholangitis and develop a clinical risk prediction model to enhance risk assessment accuracy and provide a reliable basis for clinical decision-making. The study cohort was randomly divided into a training set and an validation set at a ratio of 7:3. Within the training set, independent risk factors for post-ERCP cholangitis were screened through univariate analysis, LASSO regression, and subsequent multivariate logistic regression. A predictive nomogram was then constructed based on the identified risk factors. The performance of this nomogram was assessed in both the training and validation sets by receiver operating characteristic curves, calibration curves, the area under the curve, and the Hosmer-Lemeshow (H-L) test. Furthermore, its potential clinical utility was evaluated using decision curve analysis and clinical impact curve. A predictive nomogram was developed incorporating the following independent risk factors: history of diabetes mellitus (OR 3.698, 95% CI [1.734-7.962], P <.001), previous ERCP (OR 2.451, 95% CI [1.079-5.484], P = .03), malignant biliary obstruction (OR 2.750, 95% CI [1.185-6.375], P = .018), high biliary obstruction (OR 3.343, 95% CI [1.394-7.987], P = .006), and albumin <35g/L (OR 5.499, 95% CI 2.493-12.496], P <.001). The prediction models based on these factors achieved an area under the curve of 0.903 (95%CI:0.85-0.947) in the training set and 0.884 (95% CI:0.803-0.966) in the validation set. The calibration curve shows strong alignment between model predictions and actual outcomes. The P-values from the H-L test were .845 for the training set and .121 for the validation set, indicating no significant deviation between predicted and actual values. decision curve analysis demonstrated that the model offered a significant net clinical benefit across a broad range of risk thresholds. Clinical impact curve showed that at specific thresholds, the model's identification of high-risk patients was highly consistency with actual outcomes, confirming its clinical utility. The nomogram for predicting post-ERCP cholangitis, based on identified risk factors, demonstrated excellent predictive performance and clinical utility. Therefore, it can assist clinicians in early identification of high-risk patients and in developing personalized intervention strategies.