Development and validation of a machine learning prediction model for postoperative arrhythmia after left atrial appendage occlusion.

This study aimed to develop an effective predictive model for postoperative arrhythmia following left atrial appendage occlusion using a machine learning (ML) algorithm. The objective was to identify risk factors of postoperative arrhythmia and provide accurate predictions for patients undergoing left atrial appendage occlusion (LAAO). Patients who underwent LAAO at the Seventh People's Hospital of Zhengzhou City between June 2016 and June 2018 were included. A comprehensive set of preoperative data, including demographic characteristics, medical history, preoperative scores, and surgical details, was collected. A simulated annealing feature selection algorithm was applied to identify the most relevant variables. Eight ML algorithms were evaluated using receiver operating characteristic- area under the curve analysis and clinical decision curve analysis. The SHapley Additive exPlanations algorithm was used to interpret the predictions of the optimal model. A total of 322 patients were included in the study. Among the 8 ML algorithms, eXtreme Gradient Boosting demonstrated an overall favorable performance, with a receiver operating characteristic - area under the curve of 0.7153 (95% CI: 0.6498-0.7808) in the testing set. The model also exhibited good calibration, with a Brier score of 0.114. Decision curve analysis indicated that the model provided net clinical benefit, with the highest net benefit observed within a threshold range of 0% to 25%. Key predictive variables included occluder size, gender, height, weight, and a history of hypertension. This study developed an eXtreme Gradient Boosting-based model that effectively predicts postoperative arrhythmia following LAAO. Further validation across larger, diverse cohorts is needed to optimize its generalizability and clinical applicability.