Machine Learning Based Risk Prediction for Major Adverse Cardiovascular Events for ELGA-Authorized Clinics1.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Artificial Intelligence (AI) and machine learning (ML) are increasingly utilized in medical diagnostics to support clinical decision-making.
• ML models can predict patient prognoses, aiding healthcare professionals.

Purpose of the Study:

• To address challenges in training ML models on data from one hospital for use in another.
• To identify effective data filtering techniques for cross-hospital ML model training.

Main Methods:

• Applied data analysis to determine necessary data filters for electronic health record (EHR) data.
• Conducted experiments using real-world data from Austria's ELGA health record system and KAGes healthcare provider network.

Main Results:

• Trained a prediction model using KAGes data for ELGA-authorized health service providers.
• Demonstrated the impact of data filtering on model performance in a cross-institutional scenario.

Conclusions:

• Feature and value selection significantly enhances the classification performance of ML models trained for different healthcare systems.
• Optimized data filtering is crucial for successful deployment of predictive models across hospitals.