Examining the Accuracy and Reproducibility of Responses to Nutrition Questions Related to Inflammatory Bowel Disease by Generative Pre-trained Transformer-4
GPT-4 comprehensively answered most questions, demonstrating the promising potential of LLMs as supplementary tools for IBD patients seeking nutrition-related information. However, 17% of responses contained incorrect information, highlighting the need for continuous refinement prior to incorporation into clinical practice. Future studies should emphasize leveraging LLMs to enhance patient outcomes and promoting patient and healthcare professional proficiency in using LLMs to maximize their efficacy.
Authors
- Yee Hui Yeo (ORCID: https://orcid.org/0000-0002-2703-5954)
- Wee Han Ng
- Kelly Issokson
- Erin Feldman (ORCID: https://orcid.org/0009-0004-5986-6076)
- Jamil S. Samaan (ORCID: https://orcid.org/0000-0002-6191-2631)
- Eric Vasiliauskas
- Nithya Rajeev (ORCID: https://orcid.org/0000-0001-5632-3739)
- Barbara Hollander
- Christina Fasulo
Institutions
- Cedars-Sinai Medical Center (US)
- Keck Hospital of USC (US)
- University of Bristol (GB)
Publication Details
- Journal
- Crohn s & Colitis 360
- Published
- 2024-12-30
- DOI
- https://doi.org/10.1093/crocol/otae077
- Citations
- 4
- Primary Topic
- Chronic Disease Management Strategies
- Type
- article
- Field-Weighted Citation Impact
- 1.55