A Machine Learning Approach to Predict Mortality in Patients Enrolled in Cardiac Rehabilitation After Percutaneous Coronary Intervention for Acute Coronary Syndrome
AI · 02.11.2026 · 7 (11) · 450
DOI: 10.3390/ai7110450 ↗Kurz gefasst
Fachartikel in der Zeitschrift AI, online erschienen am 02.11.2026. Verfasst von Dimitris Gatsios und 8 weiteren Autor:innen. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: European Union, EU-CaRE.
Abstract (Originalsprache)
Background and Aims: Patients with acute coronary syndrome (ACS) treated with percutaneous coronary intervention (PCI) have a residual risk of adverse events and all-cause mortality. Machine learning (ML) approaches may help integrate clinical, functional, and lifestyle information for risk assessment. This study aimed to develop and externally evaluate models for predicting all-cause mortality in patients enrolled in cardiac rehabilitation after PCI for ACS. Methods and Results: Data from 1261 patients (mean age 61 ± 11.1 years, 16.6% women) enrolled in cardiac rehabilitation after PCI for ACS were analysed using seven prediction algorithms and repeated stratified cross-validation. External validation was performed using the EU-CaRE dataset of 1386 patients and predictions were explained using the model-agnostic SHapley Additive exPlanation (SHAP) methodology. LightGBM and Random Forest achieved the highest AUC and balanced accuracy (mean AUC > 0.8). SVM also exceeded logistic regression on AUC, though logistic regression remained competitive. LightGBM achieved a balanced accuracy of 0.87, sensitivity of 0.86, and specificity of 0.88. External evaluation yielded sensitivity of 0.78 and specificity of 0.73 in the PCI-treated cohort; specificity was 0.60 in the broader CAD cohort. After omission of predictors unavailable in EU-CaRE, the AUC was 0.816 with seven and 0.812 with six predictors. The highest SHAP values for impact on the model were detected for the variables “diabetes”, “physical exercise capacity (METs achieved during stress test)”, “average daily sitting duration”, “statin use”, “history of family heart disease”, “number of affected vessels”, “revascularization history”, “serum creatinine levels”, and “chronic hepatopathy”. The model was rated as “neutral” using the independent validation screening tool. Conclusions: ML models showed potential for predicting all-cause mortality in patients enrolled in cardiac rehabilitation after PCI for ACS. After prospective validation and assessment of clinical utility, the model may support clinicians in identifying patients at higher risk who could benefit from intensified treatment and lifestyle interventions. The small number of deaths and limitations of the evaluation procedure warrant cautious interpretation.
Förderung
- European Union
- EU-CaRE
Zitieren
APA
Gatsios, D., Sestayo-Fernández, M., Pappas, N., Kluiver, E. D., Hof, A. W. J. V., Mooren, F. C., Tsakanikas, V., Pena-Gil, C., & Schmitz, B. (2026). A Machine Learning Approach to Predict Mortality in Patients Enrolled in Cardiac Rehabilitation After Percutaneous Coronary Intervention for Acute Coronary Syndrome. AI, 7(11), 450. https://doi.org/10.3390/ai7110450
Vancouver
Gatsios D, Sestayo-Fernández M, Pappas N, Kluiver ED, Hof AWJV, Mooren FC, et al. A Machine Learning Approach to Predict Mortality in Patients Enrolled in Cardiac Rehabilitation After Percutaneous Coronary Intervention for Acute Coronary Syndrome. AI. 2026;7(11):450. doi:10.3390/ai7110450
BibTeX
@article{Gatsios2026,
author = {Gatsios, Dimitris and Sestayo-Fernández, Manuela and Pappas, Nikos and Kluiver, Ed De and Hof, Arnoud W. J. Van’t and Mooren, Frank C. and Tsakanikas, Vassilis and Pena-Gil, Carlos and Schmitz, Boris},
title = {A Machine Learning Approach to Predict Mortality in Patients Enrolled in Cardiac Rehabilitation After Percutaneous Coronary Intervention for Acute Coronary Syndrome},
journal = {AI},
year = {2026},
volume = {7},
number = {11},
pages = {450},
doi = {10.3390/ai7110450}
}
EUVIMED