Prediction model for myocardial injury after non-cardiac surgery using machine learning
Ah Ran Oh1,2, Jungchan Park1, Seo Jeong Shin3
1Department of Anesthesiology and Pain Medicine, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea.
Scientific Reports
|January 26, 2023
Summary
Researchers developed machine learning models to predict Myocardial Injury after Non-cardiac Surgery (MINS). These models identify key risk factors, aiding in better patient outcomes following surgery.
Area of Science:
- Cardiology
- Medical Informatics
- Perioperative Medicine
Background:
- Myocardial Injury after Non-cardiac Surgery (MINS) significantly impacts postoperative outcomes.
- Accurate prediction of MINS is crucial for patient management and risk stratification.
Purpose of the Study:
- To develop and validate prediction models for Myocardial Injury after Non-cardiac Surgery (MINS) using machine learning.
- To identify key predictive variables associated with MINS development in patients undergoing non-cardiac surgery.
Main Methods:
- Utilized machine learning, specifically an extreme gradient boosting algorithm, on data from 6811 patients.
- Developed two prediction models based on the top 12 and top 6 identified risk variables.
- Evaluated model performance using Area Under the Receiver Operating Characteristic Curve (AUC) and accuracy.
Main Results:
- MINS occurred in 22.0% of patients (1499 out of 6811).
- Top predictors included preoperative cardiac troponin (cTn) levels, intraoperative inotropic drug infusion, operation duration, and emergency surgery.
- Both 12-variable and 6-variable models demonstrated high accuracy (0.97) and strong predictive power (AUCs of 0.78 and 0.77).
Conclusions:
- Machine learning effectively generated prediction models for Myocardial Injury after Non-cardiac Surgery (MINS).
- The developed models, available online, can aid in predicting MINS risk.
- Further validation in diverse populations is recommended for broader clinical application.
