LLMs for Cardiovascular Risk Prediction from Structured Clinical Data
Researchers have developed a hybrid framework to predict cardiovascular risk by combining structured clinical data with natural language narratives generated by large language models. Using a dataset of 1,190 patient records, they converted structured variables into both interpretable representations and synthetic clinical narratives. While traditional models like Random Forest achieved higher accuracy, the LLM approach offers enhanced patient data privacy by operating directly on natural language descriptions. AI
IMPACT LLM-generated narratives could enable privacy-preserving clinical prediction systems, complementing traditional models.