Machine Learning Risk Prediction Model for 6-Month Postoperative Prognosis after Anti-Reflux Surgery for Gastroesophageal Reflux Disease

Abstract

Background: Anti-reflux surgery is an effective treatment for GERD, but personalized prognosis tools are lacking. Methods: This prospective study included patients undergoing laparoscopic anti-reflux surgery at a single center. The outcome of interest was the 6-month postoperative prognosis. Random forest analysis was used to identify predictors, and multivariate logistic regression was used to construct a clinical prediction model. Model performance was evaluated using leave-one-out cross-validation and area under the curve (AUC) Results: Seven important variables were identified, and two independent factors, "distal contractile integral" and "proximal distance of the lower esophageal sphincter," were selected for the clinical prediction model. The model demonstrated an AUC of 0.902 (CI: 0.8458-0.9587), and the leave-one-out cross-validation yielded an ROC of 0.890, with sensitivity of 0.575, specificity of 0.920, and accuracy of 0.826. The model showed good discrimination, calibration, and clinical utility. Conclusion: A validated clinical prediction model was developed to effectively predict the risk of poor prognosis after anti-reflux surgery. Its implementation can assist patients and surgeons in making informed decisions and improving patient outcomes.