Artificial Intelligence-Based Large Language Models Can Facilitate Patient Education
Bibliographic record
Abstract
Background: Artificial intelligence (AI) large language models (LLMs) are becoming increasingly popular, with patients and families more likely to utilize LLM when conducting internet-based research about scoliosis. For this reason, it is vital to understand the abilities and limitations of this technology in disseminating accurate medical information. We used an expert panel to compare LLM-generated and professional society-authored answers to frequently asked questions about pediatric scoliosis. Methods: We used three publicly available LLMs to generate answers to 15 frequently asked questions (FAQs) regarding pediatric scoliosis. The FAQs were derived from the Scoliosis Research Society, the American Academy of Orthopaedic Surgeons, and the Pediatric Spine Foundation. We gave minimal training to the LLM other than specifying the response length and requesting answers at a 5th-grade reading level. A 15-question survey was distributed to an expert panel composed of pediatric spine surgeons. To determine readability, responses were inputted into an open-source calculator. The panel members were presented with an AI and a physician-generated response to a FAQ and asked to select which they preferred. They were then asked to individually grade the accuracy of responses on a Likert scale. Results: The panel members had a mean of 8.9 years of experience post-fellowship (range: 3-23 years). The panel reported nearly equivalent agreement between AI-generated and physician-generated answers. The expert panel favored professional society-written responses for 40% of questions, AI for 40%, ranked responses equally good for 13%, and saw a tie between AI and "equally good" for 7%. For two professional society-generated and three AI-generated responses, the error bars of the expert panel mean score for accuracy and appropriateness fell below neutral, indicating a lack of consensus and mixed opinions with the response. Conclusions: Based on the expert panel review, AI delivered accurate and appropriate answers as frequently as professional society-authored FAQ answers from professional society websites. AI and professional society websites were equally likely to generate answers with which the expert panel disagreed. Key Concepts: (1)Large language models (LLMs) are increasingly used for generating medical information online, necessitating an evaluation of their accuracy and effectiveness compared with traditional sources.(2)An expert panel of physicians compared artificial intelligence (AI)-generated answers with professional society-authored answers to pediatric scoliosis frequently asked questions, finding that both types of answers were equally favored in terms of accuracy and appropriateness.(3)The panel reported a similar rate of disagreement with AI-generated and professional society-generated answers, indicating that both had areas of controversy.(4)Over half of the expert panel members felt they could distinguish between AI-generated and professional society-generated answers but this did not relate to their preferences.(5)While AI can support medical information dissemination, further research and improvements are needed to address its limitations and ensure high-quality, accessible patient education. Levels of Evidence: IV.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".