Evaluating large language models for simplifying non-English medical consent with clinician involvement - PubMed
18 hours ago
- #large language models
- #human-AI collaboration
- #medical consent
- The study assesses large language models (LLMs) for simplifying surgical consent forms in Chinese to improve patient comprehension.
- LLM-simplified versions increased readability and layperson comprehension but compromised content quality, especially in risk details.
- Clinician revision of the LLM outputs restored information accuracy while preserving clarity, achieving the highest comprehension scores.
- Research utilized original, LLM-simplified, and clinician-revised forms from nine hospitals, analyzing structure, readability, and content via metrics and expert ratings.
- Findings underscore the potential of LLMs in patient communication and highlight the importance of human-AI collaboration for optimal outcomes.