Evaluation of retrieval-augmented generation and large language models in clinical guidelines for degenerative spine conditions.

Abstract

Degenerative spinal diseases often require complex, patient-specific treatment, presenting a compelling challenge for artificial intelligence (AI) integration into clinical practice. While existing literature has focused on ChatGPT-4o performance in individual spine conditions, this study compares ChatGPT-4o, a traditional large language model (LLM), against NotebookLM, a novel retrieval-augmented model (RAG-LLM) supplemented with North American Spine Society (NASS) guidelines, for concordance with all five published NASS guidelines for degenerative spinal diseases. A total of 118 questions from NASS guidelines regarding five degenerative spinal conditions were presented to ChatGPT-4o and NotebookLM. All responses were scored based on accuracy, evidence-based conclusions, supplementary and complete information. Overall, NotebookLM provided significantly more accurate responses (98.3% vs. 40.7%, p < 0.05), more evidence-based conclusions (99.1% vs. 40.7%, p < 0.05), and more complete information (94.1% vs. 79.7%, p < 0.05), while ChatGPT-4o provided more supplementary information (98.3% vs. 67.8%, p < 0.05). These discrepancies became most prominent in nonsurgical and surgical interventions, wherein ChatGPT often produced recommendations with unsubstantiated certainty. While RAG-LLMs are a promising tool for clinical decision-making assistance and show significant improvement from prior models, physicians should remain cautious when integrating AI into patient care, especially in the context of nuanced medical scenarios.