Abstract
The OBJECTIVE was to develop, validate, and implement a clinical and laboratory model based on machine learning (ML) for the early identification and timely prevention of postoperative complications in patients who have undergone bariatric surgery.METHODS AND MATERIALS. A retrospective analysis was performed on clinical data from 2,011 patients who underwent bariatric procedures. The study included demographic characteristics, clinical symptoms, laboratory findings, imaging results, and details of the postoperative course. A logistic regression model was constructed using automatic class weighting to correct for class imbalance and improve predictive accuracy. The model’s performance was evaluated using key metrics: AUC, accuracy, sensitivity, specificity, and F1-score.RESULTS. The developed model demonstrated high predictive performance (AUC=0.975, accuracy=94.8 %, sensitivity=93.1 %). The most significant predictors of postoperative complications were elevated C-reactive protein levels, tachycardia, fever, symptom severity, and inflammatory laboratory markers.CONCLUSION. The proposed machine learning model shows substantial potential for integration into clinical practice and may serve as a foundation for developing intelligent systems for early complication warning, automated monitoring, and clinical decision support.

This publication has 13 references indexed in Scilit: