Enhancing one-year mortality prediction in STEMI patients post-PCI: an interpretable machine learning model with risk...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• ST-elevation myocardial infarction (STEMI) is a major cause of mortality and disability worldwide.
• Percutaneous coronary intervention (PCI) has improved in-hospital survival, emphasizing the need for effective post-discharge care.
• Machine learning (ML) shows potential for predicting adverse outcomes in STEMI patients, but simple, accurate models are needed.

Purpose of the Study:

• To develop and validate a machine learning model for predicting one-year mortality in STEMI patients post-PCI.
• To identify key predictors of mortality and assess the impact of risk stratification on model performance.
• To create an accurate and interpretable model for improved long-term outcome prediction.

Main Methods:

• Retrospective analysis of 1,274 STEMI patients undergoing PCI.
• Application of data processing and ML algorithms (Random Forest) to predict one-year mortality.
• Evaluation of model performance using AUROC, AUPRC, accuracy, sensitivity, precision, and F1-score.

Main Results:

• The Random Forest model achieved an AUROC of 0.94 and AUPRC of 0.44.
• Key predictors identified: cardiogenic shock, creatinine, NT-proBNP, diastolic blood pressure, and left ventricular ejection fraction.
• Integrating risk stratification improved performance to AUROC 0.97 and AUPRC 0.74.

Conclusions:

• Accurate and interpretable ML models can be built using a minimal set of predictors for STEMI patients.
• Risk stratification enhances the predictive power of ML models for long-term outcomes.
• This approach facilitates improved post-discharge management and outcome prediction for STEMI survivors.