Development of Prediction Models for Acute Myocardial Infarction at Prehospital Stage with Machine Learning Based on...

Area of Science:

• Cardiology
• Data Science
• Public Health

Background:

• Accurate prehospital prediction of acute myocardial infarction (AMI) is crucial for timely intervention.
• Existing prediction models may not fully leverage the potential of large-scale emergency medical service (EMS) data.

Purpose of the Study:

• To develop and compare the efficacy of predictive models for AMI at the prehospital stage.
• To evaluate conventional statistical methods against machine learning algorithms using nationwide EMS registry data.

Main Methods:

• Utilized a nationwide EMS registry dataset of 184,577 patients (aged >15) from January 2016 to December 2018 in Korea.
• Developed prediction models using logistic regression, extreme gradient boosting, and multilayer perceptron.
• Compared model performance based on discriminative ability and predictive accuracy.

Main Results:

• The B-type logistic regression model, incorporating AMI-specific variables, showed superior discriminative ability (p = 0.02).
• Extreme gradient boosting and multilayer perceptron models demonstrated higher predictive performance than logistic regression.
• Different machine learning algorithms produced distinct lists of key predictive features.

Conclusions:

• Machine learning models offer improved predictive performance for prehospital AMI compared to conventional methods.
• Nationwide prehospital data, when structured appropriately, can enhance the identification of patients requiring prompt AMI management.
• The findings support the integration of advanced analytics for optimizing emergency cardiac care pathways.