Towards a FHIR-Based Data Model for Coronary Angiography Observations

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Coronary heart disease is a leading global cause of death.
• Predictive models for myocardial infarction are crucial for early intervention.
• Standardizing cardiac data is essential for advanced analytics.

Purpose of the Study:

• To develop prognostic models for predicting spontaneous myocardial infarctions.
• To integrate clinical parameters with invasive coronary angiogram data.
• To establish a harmonized data model for cross-hospital federated learning.

Main Methods:

• Utilized over 30,000 coronary angiograms from Erlangen University Hospital.
• Developed a mapping for coronary angiography data based on HL7® FHIR®.
• Mapped data to ICD-10 and SNOMED CT for standardization.
• Extracted 15 patient characteristics to create HL7® FHIR® resources.

Main Results:

• A finalized coronary angiography information model was designed.
• A robust mapping approach to clinical terminologies was established.
• Data harmonization strategy for federated learning was defined.

Conclusions:

• The developed information model and mapping facilitate data integration for prognostic modeling.
• This approach supports future federated machine learning initiatives in cardiology.
• The standardized data can improve the prediction accuracy of myocardial infarctions.