Multi-Modal RAG for Early Cardiovascular Risk Assessment: Integrating EHR Records and Genetic Markers from a Nationwide Patient Database

Abstract

Cardiovascular diseases (CVD) continue to be a predominant cause of mortality worldwide, emphasizing the critical need for early risk detection and intervention. In this study, we propose a novel multi-modal retrieval-augmented generation (RAG) framework that integrates comprehensive electronic health records (EHR) with detailed genetic marker information extracted from a nationwide patient database. This integrative approach combines structured clinical data with genomic profiles to enhance predictive accuracy and uncover subtle risk patterns that traditional models may overlook. Leveraging advanced machine learning and deep learning techniques, the framework processes heterogeneous data sources efficiently while adapting to diverse patient demographics and clinical contexts. The system dynamically updates risk predictions as new data become available, facilitating a proactive and personalized strategy for cardiovascular risk management. Preliminary evaluations reveal that the multi-modal RAG model outperforms conventional risk assessment methods by accurately identifying individuals at elevated risk at earlier stages of disease progression. Furthermore, the study discusses key challenges including data heterogeneity, integration complexity, privacy concerns, and the need for transparent model interpretability. Overall, the proposed framework represents a significant step towards precision medicine in cardiology, enabling clinicians to make informed decisions and implement timely interventions.