The Department of Surgery’s Outcomes Center research team, including Chin Siang Ong, MBBS PhD, Nicholas Obey and Eric B. Schneider, PhD has introduced a groundbreaking framework, SurgeryLLM, designed to enhance surgical decision-making and workflow. Their study, recently published in npj Digital Medicine, explores how integrating Retrieval-Augmented Generation (RAG) with large language models (LLMs) may contribute to surgeon efficiency and patient safety.
SurgeryLLM addresses an important limitation of traditional LLMs, which may be limited in their ability to access appropriate up-to-date medical guidelines, by incorporating current context-specific patient care guidelines using RAG. SurgeryLLM demonstrates the importance and feasibility of integrating current best-practice guidelines to support the development of AI-based tools to optimize surgeon efficiency and effectiveness across multiple aspects of surgical practice.
Key findings from the study include:
- Improved Diagnostic Support: SurgeryLLM demonstrated superior accuracy in identifying abnormal lab results and missing clinical investigations, outperforming standard LLMs.
- Evidence-Based Recommendations: The framework consistently aligned its management suggestions with established guidelines, such as those from the American College of Cardiology and American Heart Association.
- Enhanced Documentation:SurgeryLLM successfully generated structured operative notes tailored to simulated patient scenarios, which may pave the way for reducing administrative burdens for surgeons.
This innovation is timely, given the increasing demands on surgical services in the face of workforce shortages. By refining tools like SurgeryLLM, the team envisions a future where AI empowers surgeons to increase efficiency in healthcare delivery while improving patient care and enhancing patient outcomes.