Prediction of in-hospital mortality in patients with acute myocardial infarction following primary percutaneous coronary intervention: A machine learning approach

Background: In-hospital mortality in patients with acute myocardial infarction (AMI) following primary percutaneous coronary intervention (pPCI) remains a significant concern. Developing a predictive model of in-hospital mortality is crucial for identifying high-risk patients, guiding clinical decisions, and preventing in-hospital mortality. Machine learning (ML) may analyze patterns in large datasets and provide accurate predictions of in-hospital mortality in AMI patients following pPCI. Objectives: To develop and validate a model for predicting in-hospital mortality in AMI patients following pPCI using ML algorithms. Methods: 1968 AMI patients after pPCI were identified from the SCIENCE database between 2019 and 2023. Four supervised ML algorithms (Random Forest, XGBoost, AdaBoost, and Logistic Regression) were used to construct the models. The performance of the models was evaluated by the area under the curve (AUC), accuracy, sensitivity, specificity, and F1-score. Results: The overall in-hospital mortality rate was 17.68 %. The model constructed using the Random Forest provided the best performance for discriminating between patients with a status of alive and dead, with an AUC of 0.976. The models constructed from the XGBoost and AdaBoost algorithms had lower discriminatory performance than the RF model, with AUCs of 0.975 and 0.974, respectively. The model based on the LR algorithm had the lowest AUC of 0.973. Conclusions: A predictive model constructed using the RF algorithm performed the best for predicting in-hospital mortality in AMI patients following pPCI. Implementing the model in a clinical setting may assist nurses/physicians in identifying high-risk patients, prioritizing them for appropriate treatment, and potentially reducing in-hospital mortality.