PRELIMINARY EVALUATION OF MACHINE LEARNING MODELS FOR PREDICTING CARDIOGENIC SHOCK IN PATIENTS WITH ACUTE MYOCARDIAL INFARCTION
Main Article Content
Abstract
Background: Cardiogenic shock (CS) is the most severe complication of acute myocardial infarction (AMI), associated with high in-hospital mortality. Early prediction of CS at the time of admission is critical for guiding timely treatment. In recent years, machine learning (ML) models have been increasingly applied to predict major cardiovascular events.
Objective: To develop and evaluate the performance of various machine learning models in predicting in-hospital CS among patients with AMI, and to identify the model with the highest clinical application potential.
Methods: This was a cross-sectional descriptive study with longitudinal follow-up on 316 patients diagnosed with AMI. Patients were classified into two groups: with and without CS during hospitalization. Data were collected at admission, including medical history, clinical features, laboratory results, ECG, and echocardiography. After preprocessing, the dataset was randomly split into training (70%) and testing (30%) sets. The SMOTE-Tomek technique was applied to the training set to address class imbalance. Feature selection was performed using SHAP. Eight ML models were developed and their predictive performance was compared on the testing set.
Results: Among the 316 patients with AMI, the incidence of in-hospital CS was 10.1%. The AdaBoost model demonstrated the best predictive performance with an AUC of 0.924, recall of 0.60, and F1-score of 0.571, indicating a good balance between sensitivity and overall accuracy. Logistic Regression and LASSO Logistic also showed strong performance, with high AUCs of 0.902 and 0.912, respectively; both models achieved a recall of 0.60 and F1-scores of 0.600 and 0.571. All machine learning models outperformed traditional risk scores such as GRACE (AUC = 0.755) and the simplified ORBI score (AUC = 0.780).
Conclusion: The three machine learning models—AdaBoost, Logistic Regression, and LASSO Logistic—demonstrated high performance in the early prediction of in-hospital CS among patients with AMI. These findings highlight the potential for integrating machine learning models into clinical decision support systems to optimize care for high-risk patients.
Keywords
Cardiogenic shock, acute myocardial infarction, machine learning, early prediction, AdaBoost
Article Details
References
2. Auffret V, Cottin Y, Leurent G, et al. Predicting the development of in-hospital cardiogenic shock in patients with ST-segment elevation myocardial infarction treated by primary percutaneous coronary intervention: the ORBI risk score. European heart journal. 2018;39(22):2090-2102.
3. Bommu R. Machine Learning Applications in Cardiology: A Viable Practical Solution for Developing Countries. The Metascience. 2024;2(1):10-20.
4. Aleman R, Patel S, Sleiman J, et al. Cardiogenic shock and machine learning: a systematic review on prediction through clinical decision support softwares. Journal of Cardiac Surgery. 2021;36(11):4153-4159.
5. Or B. Improving Requirements Classification with SMOTE-Tomek Preprocessing. arXiv preprint arXiv:250106491. 2025;
6. Bai Z, Hu S, Wang Y, et al. Development of a machine learning model to predict the risk of late cardiogenic shock in patients with ST-segment elevation myocardial infarction. Annals of Translational Medicine. 2021;9(14):1162.
7. Böhm A, Segev A, Jajcay N, et al. Machine learning-based scoring system to predict cardiogenic shock in acute coronary syndrome. European Heart Journal-Digital Health. 2025:ztaf002.