Article Impact Level: HIGH Data Quality: STRONG Summary of Clinical Cardiology, 46(3), 269–278. https://doi.org/10.1002/clc.23963 Dr. Amir Hossein Behnoush et al.
Points
- Machine learning (ML) models were developed and evaluated to predict 1-year mortality among hypertensive patients who underwent coronary artery bypass graft (CABG) surgery.
- Logistic regression (LR) exhibited the highest prediction ability with an area under the curve (AUC) of 0.82, outperforming other ML models such as random forest (RF), artificial neural network (ANN), extreme gradient boosting (XGB), and naïve Bayes (NB).
- Eleven features, including total ventilation hours and ejection fraction, were identified as the most influential predictors for mortality after CABG surgery.
- The study included a dataset of 8,493 hypertensive patients, with 303 deaths occurring within the first year after surgery.
- LR demonstrated the highest AUC in various subgroups, including different age ranges (50–59 and 80–89 years) and subgroups of hypertensive patients who were overweight, diabetic, or smokers.
Summary
This research paper applied machine learning (ML) algorithms to predict 1-year mortality among hypertensive patients who underwent coronary artery bypass graft (CABG) surgery. The study utilized a dataset from the CABG data registry of Tehran Heart Center, comprising baseline and peri-procedural characteristics and mortality data. Five ML models were developed and evaluated, namely logistic regression (LR), random forest (RF), artificial neural network (ANN), extreme gradient boosting (XGB), and naïve Bayes (NB).
The analysis involved 8,493 hypertensive patients with an average age of 68.27 ± 9.27 years, out of which 303 individuals died within the first year. The dataset identified eleven features as the most significant predictors, with total ventilation hours and ejection fraction emerging as the top indicators. Among the ML models, LR exhibited the highest prediction performance with an area under the curve (AUC) value of 0.82, while the NB model had the lowest AUC at 0.79. Notably, the LR model demonstrated the highest AUC in various subgroups, including two age ranges (50–59 and 80–89 years) and subgroups consisting of overweight, diabetic, and smoker hypertensive patients.
In summary, this study demonstrated that all the developed ML models showcased remarkable performance in predicting 1-year mortality among hypertensive patients who underwent CABG surgery. Specifically, the LR model exhibited the most accurate predictive capability, as indicated by its AUC. Utilizing these ML models can assist clinicians in evaluating the mortality risk within specific subgroups of patients, particularly those with hypertension. The identified predictors, such as total ventilation hours and ejection fraction, could aid in early identification and intervention for patients at a higher risk of mortality post-CABG surgery.
Link to the article: https://onlinelibrary.wiley.com/doi/10.1002/clc.23963
References Behnoush, A. H., Khalaji, A., Rezaee, M., Momtahen, S., Mansourian, S., Bagheri, J., Masoudkabir, F., & Hosseini, K. (2023). Machine learning‐based prediction of 1‐year mortality in hypertensive patients undergoing coronary revascularization surgery. Clinical Cardiology, 46(3), 269–278. https://doi.org/10.1002/clc.23963
