Tag
This paper presents a hybrid framework that combines structured clinical data with LLM-generated narratives for coronary artery disease prediction, achieving high fidelity in variable extraction and comparing ML models with LLM-based zero-shot and few-shot classification.