Development of a Machine-Learning Model for Predicting Postoperative Complication Occurrence After Radical Gastrectomy Using Electronic Medical Records.

This study aims to develop a machine-learning model using electronic medical records to predict postoperative complications after radical gastrectomy for gastric cancer. Data from 4892 patients who underwent the procedure between January 2012 and March 2022 were analyzed. Complications were defined as grade 1 or higher according to the Clavien-Dindo classification. Thirty-two predictors identified from previous research were matched to electronic medical records. After preprocessing, 5 machine-learning models-logistic regression, random forest, extreme gradient boosting, CatBoost, and multilayer perceptron-were developed. Model performance was assessed using the F1 score, accuracy, and area under the precision-recall curve. Key predictors were identified using permutation-based feature importance. The random forest model performed best, with an F1 score of 0.86, an accuracy of 0.95, and an area under the precision-recall curve of 0.90. Important predictors included late fever on postoperative days 4-7, postoperative pain management, operating time, age, and perioperative blood transfusion. The model demonstrated strong performance in predicting postoperative complications by incorporating insights of nurses throughout factor selection and model development. This approach may assist nursing decision-making by identifying high-risk patients for early, personalized interventions. Further validation through prospective and external studies is required.