An XGBoost machine learning model using the top 5 important variables effectively predicted postoperative acute kidney injury in CABG patients with an AUC of 0.89.
Cohort (n=520)
No
Can machine learning models accurately predict the risk of postoperative acute kidney injury in patients undergoing coronary artery bypass grafting?
Machine learning models, particularly XGBoost and Logistic Regression, can effectively predict the risk of postoperative acute kidney injury in CABG patients using key variables like eGFR and intraoperative epinephrine use.
Effect estimate: AUC 0.89 (95% CI 0.86-0.91)
Background: Coronary artery bypass grafting (CABG) is key for severe coronary artery disease, but postoperative acute kidney injury (AKI) may increase mortality and prolong hospital stays. Reliable models for early prediction of post-CABG AKI remain lacking. Methods: Data of 520 CABG patients (September 2021-December 2024) from the Affiliated Hospital of Xuzhou Medical University were collected, and the patients were divided into a training group (70%, for model building) and a validation group (30%). Key variables were screened through Least Absolute Shrinkage and Selection Operator (LASSO) regression, followed by the construction of six machine learning models: Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Logistic Regression (LR), Light Gradient Boosting Machine (LightGBM), Softmax Regression, and Support Vector Machine (SVM). The SHapley Additive exPlanations (SHAP) was used to quantify feature importance. Results: The incidence of post-CABG AKI was 25.96%, and the median age of patients in the AKI group was significantly higher than that in the non-AKI group (66.09 ± 8.15 vs 64.32 ± 7.76, p = 0.025). In the training group, the XGBoost model using the top 5 important variables outperformed other models (Area Under the Curve AUC = 0.89, 95% Confidence Interval CI: 0.86-0.91), followed by the LightGBM model using the top 5 important variables and the RF model using the top 5 important variables (both had an AUC of 0.88; 95% CI: 0.85-0.90 and 0.85-0.91, respectively). In the validation group, the LR model using the top 15 important variables and the Softmax Regression model using the top 15 important variables maintained the highest stability (both had an AUC of 0.86, 95% CI: 0.79-0.92). SHAP analysis confirmed that estimated glomerular filtration rate (eGFR), intraoperative epinephrine use and calcium levels were the top three predictive factors. Conclusion: The machine learning models constructed in this study can effectively predict post-CABG AKI, facilitating early identification of high-risk patients.
Zhang et al. (Sat,) conducted a cohort in Coronary artery bypass grafting (CABG) (n=520). Machine learning prediction models was evaluated on Prediction of post-CABG AKI (Area Under the Curve) (AUC 0.89, 95% CI 0.86-0.91). An XGBoost machine learning model using the top 5 important variables effectively predicted postoperative acute kidney injury in CABG patients with an AUC of 0.89.