Prediction of venous thromboembolism after metabolic and bariatric surgery using machine learning approach: a MBSAQIP study.

backgroundSeveral studies have developed predictive models for venous thromboembolism (VTE) after metabolic and bariatric surgery (MBS), but none have incorporated complications that increase VTE risk. This study aimed to evaluate the risk factors for VTE following MBS using machine learning (ML) algorithms, using models that either excluded or included time-aware postoperative complication variables-these were engineered to include only complications that occurred before the VTE event.

methodsWe used the MBSAQIP database from 2020 to 2023. XGBoost, random forest classifiers, support vector machines (SVMs), artificial neural networks (ANNs), and logistic regression models were used to predict VTE after MBS in models with and without considering other postoperative complications that occurred before VTE occurrence.

resultsThe study included 2198 and 698,284 VTE and non-VTE cases, respectively. For models excluding complications, the area under the curve (AUC) values were as follows: XGBoost (0.668), RandomForestClassifier (0.660), SVM (0.649), ANN (0.657), and logistic regression (0.669). Including time-adjusted complications improved all model performances, with AUCs of XGBoost (0.680), RandomForestClassifier (0.664), SVM (0.666), ANN (0.697), and logistic regression (0.690). The strongest postoperative predictor of VTE was the need for reoperation. Other postoperative predictors included prolonged length of stay, ICU admission, reintervention, organ-space surgical site infection, sepsis, and anastomotic leak.

conclusionPostoperative complications and prolonged length of stay were strongly associated with VTE after MBS, suggesting that patients with these risk factors may benefit from enhanced prophylactic strategies.