Prediction of Postoperative Pulmonary Edema Risk Using Machine Learning

Area of Science:

• Anesthesiology and Perioperative Medicine
• Medical Informatics and Machine Learning

Background:

• Postoperative pulmonary edema (PPE) is a recognized complication following surgical procedures.
• Accurate prediction of PPE risk is crucial for optimizing patient management and outcomes.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting postoperative pulmonary edema (PPE) risk.
• To identify key pre- and intraoperative factors contributing to PPE development.

Main Methods:

• Retrospective analysis of over 250,000 patient records from five South Korean hospitals (January 2011 - November 2021).
• Utilized multiple ML algorithms including extreme gradient boosting, light-gradient boosting machine, multilayer perceptron, logistic regression, and balanced random forest (BRF).
• Model performance was assessed using AUC, precision, recall, F1 score, and accuracy.

Main Results:

• The balanced random forest (BRF) model demonstrated the highest predictive performance with an AUC of 0.91.
• Key predictors for PPE included arterial line monitoring, American Society of Anesthesiologists physical status, urine output, age, and Foley catheter status.
• Despite strong AUC, the BRF model's precision and F1 scores indicated areas for further refinement.

Conclusions:

• Machine learning models, particularly BRF, show significant potential in predicting postoperative pulmonary edema risk.
• Integration of these ML tools can enhance clinical decision-making and improve postoperative care strategies.
• Further research may focus on improving model precision and generalizability for broader clinical application.