Application of Machine Learning to Predict Postoperative Nausea and Vomiting in Laparoscopic Sleeve Gastrectomy

Abstract

Background: Postoperative nausea and vomiting (PONV) is a common complication of laparoscopic sleeve gastrectomy (LSG). This study aimed to develop and validate machine learning models to predict the risk of PONV in patients undergoing LSG.

Methods: Data from patients who underwent LSG at a tertiary hospital in China between January 2018 and March 2023 was collected for this study. The data were randomly divided into training and test cohorts in a ratio of 7:3. The boruta algorithm and multivariate logistic regression were employed to identify independent predictive factors. Various models, including random forest, extreme gradient boosting (XGB), gradient boosting machine, generalized linear models, support vector machines, neural network, and multi-layer perceptron, were developed. Model performance was assessed on the basis of area under the receiver operating characteristic curve (AUROC).

Results: A total of 860 patients were included in the analysis, of whom 473 (55%) experienced PONV. The identified risk factors for PONV were female gender, surgery duration exceeding 60 min, intraoperative remifentanil administration, and postoperative opioid use. Prophylactic administration of antiemetics during surgery was found to be a protective factor. The XGB model demonstrated superior performance, with an AUROC of 0.828 (95% CI: 0.777-0.879). Additionally, an online prediction tool based on the XGB model was developed for clinical use.

Conclusion: The XGB model demonstrated the highest predictive accuracy among the tested models. Future studies with external validation are warranted to confirm the model's generalizability across diverse populations and settings.

Keywords: Bariatric and metabolic surgery; Machine learning; Postoperative nausea and vomiting; Predictive model.