Using machine learning to predict adverse events in acute coronary syndrome: A retrospective study

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Acute coronary syndrome (ACS) poses significant mortality risks, with up to 30% of patients experiencing adverse events such as renal failure and myocardial infarction (MI).
• Accurate prognostication is crucial for improving patient outcomes and guiding clinical management strategies.

Purpose of the Study:

• To evaluate the efficacy of various machine learning (ML) models in predicting adverse events in ACS patients.
• To identify the optimal ML model for predicting acute kidney injury (AKI), in-hospital MI, and 1-year all-cause mortality.

Main Methods:

• A cohort of 5240 ACS patients undergoing percutaneous coronary intervention (PCI) was analyzed over a 1-year follow-up period.
• Eight ML models, including Support Vector Machine, Extreme Gradient Boosting, and Linear Discriminant Analysis (LDA), were developed and validated using 10-fold cross-validation.
• Feature selection was performed using XGBoost based on Shapley Additive exPlanations scores, and model performance was assessed using AUC, F1 score, accuracy, and precision/recall curves.

Main Results:

• The Linear Discriminant Analysis (LDA) model demonstrated superior performance compared to other ML models.
• LDA achieved an AUC of 0.83 for predicting 1-year all-cause mortality, with an accuracy of 0.85 and F1 score of 0.90.
• The LDA model also exhibited good predictive capacity for AKI and MI, with an AUC of 0.74.

Conclusions:

• The LDA model, utilizing readily available in-hospital patient data, shows significant potential for predicting adverse events and mortality within one year post-PCI in ACS patients.
• This predictive capability can aid clinicians in risk stratification and personalized treatment planning.