What is the clinical evidence from this study?

Study design: Cohort. Population: Low cardiac output syndrome (LCOS) following non-isolated coronary artery bypass grafting (CABG) (n=378). Intervention: Extreme gradient boosting (XGB) machine learning model vs. Logistic regression (LR) and other machine learning models. Primary outcome: Prediction of low cardiac output syndrome (LCOS) in the testing cohort (AUC 0.868, 95% CI 0.799-0.936).

What question did this study set out to answer?

This study aims to develop a machine-learning model to predict low cardiac output syndrome (LCOS) after non-isolated coronary artery bypass grafting (CABG) surgery.

June 18, 2026Exploration of Cardiology0 citationsOpen Access

Prediction of low cardiac output syndrome in patients following non-isolated coronary artery bypass grafting surgery using machine learning

EFEmmanuel Delali Kofi FiagbeyChina Pharmaceutical University JWJ. WangNanjing University of Chinese Medicine SSS. SaadChina Pharmaceutical University

Key Result

The extreme gradient boosting (XGB) machine learning model effectively predicted low cardiac output syndrome after non-isolated CABG surgery, achieving an AUC of 0.868 in the testing cohort.

Key Points

This study aims to develop a machine-learning model to predict low cardiac output syndrome (LCOS) after non-isolated coronary artery bypass grafting (CABG) surgery.
378 patients post non-isolated CABG were assessed retrospectively.
Five machine-learning algorithms were used: logistic regression, random forest, extreme gradient boosting, light gradient boosting machine, and support vector machine.
Model performance was evaluated using area under the curve (AUC), 10-fold cross-validation, and decision curve analysis (DCA).
XGB model had superior performance with AUC of 0.933 (95% CI: 0.903–0.962) and a Brier score of 0.107.
In testing, it maintained AUC of 0.868 (95% CI: 0.799–0.936) with specificity of 0.785, accuracy of 0.781, and precision of 0.614.
Key predictors included ejection fraction, left ventricular end-systolic diameter, and lactate levels as determined by SHAP analysis.

Study Design

Type

Cohort (n=378)

Multicenter

Structured PICO

Can machine learning models accurately predict low cardiac output syndrome after non-isolated CABG?

Population

378 adult patients (median age 68, 29.9% female) who underwent non-isolated coronary artery bypass grafting surgery, retrospectively assessed for the development of low cardiac output syndrome between 1 and 48 hours postoperatively.

Exposure

Machine-learning-based prediction models (L2 regularized logistic regression, random forest, extreme gradient boosting [XGB], light gradient boosting machine, and support vector machine)

Outcome

Prediction of low cardiac output syndrome (LCOS) onset post non-isolated CABG

An extreme gradient boosting machine learning model can effectively predict low cardiac output syndrome after non-isolated CABG, potentially aiding in early risk stratification.

Main Result

Effect estimate: AUC 0.868 (95% CI 0.799-0.936)

Absolute Event Rate: 0.868% vs 0.867%

Limitations

Single-center retrospective study with a relatively modest sample size limiting generalizability
Potential introduction of heterogeneity within the LCOS cohort due to multiple diagnostic criteria
Model's calibration in extremely high-risk patients required cautious interpretation
Possible overestimation of the model's discriminatory performance due to the dual role of lactate as both a predictor and a component of the outcome definition
Risk of overfitting despite regularization strategies

Abstract

Aim: Low cardiac output syndrome (LCOS) may be improvable; hence, timely detection and intervention are essential. However, no model has been established for the prediction of LCOS onset post non-isolated coronary artery bypass grafting (CABG) surgery. Therefore, this study aimed to develop a machine-learning-based model to predict LCOS after non-isolated CABG. Methods: A total of 378 patients who underwent non-isolated CABG at Nanjing First Hospital, China, were retrospectively assessed. Five algorithms L2 regularized logistic regression (LR), random forest (RF) classifier, extreme gradient boosting (XGB), light gradient boosting machine (LGBM), and support vector machine (SVM) were employed. Model performance and clinical utility were evaluated using area under the curve (AUC), 10-fold cross-validation, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) were used to assess the model’s interpretability. A web calculator was developed. Results: XGB showed superior performance and calibration (AUC: 0.933, 95% CI: 0.903–0.962; Brier score of 0.107), with excellent specificity (0.865), accuracy (0.860), and precision (0.753). In testing, XGB maintained excellent discrimination (AUC: 0.868, 95% CI: 0.799–0.936), best specificity (0.785), accuracy (0.781), and precision (0.614). DCA confirmed clinical usefulness. SHAP analysis identified the ejection fraction, left ventricular end-systolic diameter, and lactate levels as the most influential predictors. The web calculator is accessible via https://lcos-cabg-xgb-model.streamlit.app/ Conclusions: The developed web-based XGB model effectively predicts LCOS after non-isolated CABG, aiding early risk stratification and detection.

问 AI

Bookmark

View Full Paper