Machine-Learning Versus Traditional Scores for Predicting Outcomes After Coronary Artery Bypass Graft Surgery: A...

Area of Science:

• Cardiology
• Medical Informatics
• Data Science

Background:

• Coronary artery bypass grafting (CABG) is linked to significant patient morbidity and mortality.
• Existing risk scores (e.g., STS, EuroSCORE II) have limitations in predicting outcomes, especially for high-risk individuals.
• Machine learning (ML) may offer improved prediction by identifying complex data patterns.

Purpose of the Study:

• To systematically review and compare the predictive efficacy of ML models against traditional risk scores for outcomes following CABG.
• To synthesize evidence on the performance of ML in predicting mortality after CABG.

Main Methods:

• A comprehensive literature search was conducted across major databases (PubMed, Embase, Web of Science, Cochrane Library) up to August 14th, 2025.
• Included studies utilized ML algorithms and traditional risk scores for predicting all-cause mortality post-CABG.
• Meta-analysis employed a linear mixed-effects model, with C-statistics as the primary measure of discrimination.

Main Results:

• Twenty-six studies involving 565,063 participants were analyzed.
• ML models demonstrated a significantly higher pooled C-statistic (0.82) compared to traditional risk scores (0.73) (P < 0.0001).
• The highest-performing ML model achieved a C-statistic of 0.98; calibration data was inconsistently reported.

Conclusions:

• ML approaches exhibit superior pooled discrimination for predicting mortality after CABG compared to traditional risk scores.
• Findings suggest ML's potential but require cautious interpretation due to study limitations (heterogeneity, bias, lack of external validation).
• Prospective, externally validated studies are essential to confirm ML's clinical superiority for pre-operative risk stratification.