Exploring Synergies Between Large Language Models and Knowledge Models in Healthcare: A Scoping Review

Large Language Models (LLMs) and knowledge models (KMs) are increasingly integrated to improve clinical decision support systems. We conducted a scoping review to examine the synergy between LLM- and KM-approaches for knowledge model construction, enrichment, and LLM optimization. The selected studies were categorized into three clusters: knowledge-grounded reasoning, LLM-driven knowledge model engineering, and hybrid approaches. Results indicate that retrieval-augmented generation (RAG) grounded in KMs improve LLM reliability, explainability, and adherence to clinical practice guidelines. Conversely, LLMs demonstrate good performance in automating key tasks for KM development, including named entity recognition, relation extraction, and knowledge fusion. Hybrid frameworks combining both paradigms further improve performance and safety. Overall, the integration of LLMs and KMs yields more reliable, scalable, and interpretable AI systems for healthcare applications.