Generative AI Is Not Ready for Clinical Use in Patient Education for Lower Back Pain Patients, Even With Retrieval-Augmented Generation
Project Overview
The document explores the application of generative AI, particularly Retrieval-Augmented Generation (RAG) in conjunction with large language models (LLMs), in the realm of education, focusing on creating personalized educational materials for patients suffering from low back pain (LBP). It showcases the effectiveness of RAG-based models in producing accurate, comprehensive, and easily understandable content tailored to individual needs. Despite these promising outcomes, the study emphasizes that the current limitations of AI models, particularly in maintaining the relevance and specificity of the generated information, render them unsuitable for direct clinical application at this stage. The findings suggest that while generative AI holds significant potential for enhancing educational resources, further advancements and refinements are necessary before it can be reliably implemented in clinical settings.
Key Applications
Generative AI using Retrieval-Augmented Generation for personalized patient education materials
Context: Patient education for individuals with low back pain (LBP)
Implementation: Utilizing large language models (LLMs) combined with Retrieval-Augmented Generation (RAG) to generate tailored educational materials based on user queries.
Outcomes: RAG-based models produced more accurate, complete, and readable materials compared to traditional models; however, they are not ready for clinical use.
Challenges: The generated materials often lacked clinical specificity and relevance, leading to generic advice that may not cater to individual patient needs.
Implementation Barriers
Technical Barrier
The generative AI models produce materials that are not yet suitable for clinical use due to a lack of specificity and relevance.
Proposed Solutions: Future studies should focus on using real patient data and creating a more comprehensive knowledge base to enhance the accuracy of generated content.
Content Quality Barrier
Generated materials often provide general advice rather than specific clinical recommendations, limiting their effectiveness.
Proposed Solutions: Incorporate multi-clinician input to broaden the knowledge base and ensure diverse patient needs are addressed.
Project Team
Yi-Fei Zhao
Researcher
Allyn Bove
Researcher
David Thompson
Researcher
James Hill
Researcher
Yi Xu
Researcher
Yufan Ren
Researcher
Andrea Hassman
Researcher
Leming Zhou
Researcher
Yanshan Wang
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Yi-Fei Zhao, Allyn Bove, David Thompson, James Hill, Yi Xu, Yufan Ren, Andrea Hassman, Leming Zhou, Yanshan Wang
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai