Predicting atrial fibrillation and flutter using BEHRT and identifying multimorbidity patterns using BERTopic.

Area of Science:

• Cardiology
• Artificial Intelligence
• Public Health

Background:

• Atrial fibrillation and flutter are complex heart rhythm disorders often co-occurring with other chronic conditions.
• Effective management requires optimized care, highlighting the need for proactive and personalized healthcare strategies.
• Understanding pre-diagnosis comorbidity patterns is crucial for early intervention.

Purpose of the Study:

• To analyze pre-atrial fibrillation and flutter comorbidity patterns using a population-based dataset.
• To develop and evaluate an AI-driven model (BEHRT) for predicting atrial fibrillation and flutter.
• To identify sex-specific multimorbidity patterns associated with atrial fibrillation and flutter using BERTopic.

Main Methods:

• A nested case-control study utilized Korean National Health Insurance Corporation data (2002-2019).
• A transformer-based model (BEHRT) predicted atrial fibrillation and flutter using 5-year disease histories.
• BERTopic identified sex-specific multimorbidity patterns in 600,030 participants (8,661 cases, 591,369 controls).

Main Results:

• BEHRT achieved an AUC of 0.80 in predicting atrial fibrillation and flutter.
• Distinct sex-specific multimorbidity patterns were identified: males showed aortic aneurysm, hypertensive heart disease, COPD; females showed Alzheimer's, Parkinson's, rheumatic heart disease.
• The combined AI approach successfully predicted atrial fibrillation and flutter based on multimorbid histories.

Conclusions:

• The integration of BEHRT and BERTopic effectively predicts atrial fibrillation and flutter and reveals unique sex-specific disease associations.
• These findings demonstrate the potential of artificial intelligence in enhancing personalized healthcare.
• AI-driven insights can optimize prevention and management strategies for chronic conditions like atrial fibrillation and flutter.