Exploration of multimorbidity patterns of chronic diseases in patients with atrial fibrillation or flutter(AFF) using BEHRT and BERTopic: A feasibility study

Abstract

Drawing from the extensive dataset of the National Health Insurance Corporation (2002-2019), we employed the BEHRT model and BERTopic analysis to predict and understand complex disease trajectories. We identified detailed disease associations using BEHRT optimal architecture with a 6-layer design, achieving notable metrics in predicting Atrial Fibrillation and Flutter (AFF). Our findings underscore the transformative potential of AI in enhancing health informatics and promoting patient-centric care. ©2024 IEEE.