Large Language Model-Assisted Surgical Consent Forms in Non-English Language: Content Analysis and Readability...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Patient Communication

Background:

• Complex language in surgical consent forms hinders patient comprehension.
• Large language models (LLMs) offer potential for simplifying medical information.
• Limited evidence exists on LLM impact on non-English consent form content preservation.

Purpose of the Study:

• To evaluate the effect of LLM-assisted editing on the readability and content quality of Korean surgical consent forms.
• Specifically assess changes in standardized liver resection consent documents across multiple institutions.

Main Methods:

• Collected standardized liver resection consent forms from 7 South Korean institutions.
• Simplified forms using ChatGPT-4o and assessed readability (KReaD, Natmal indices).
• Evaluated content quality (risk, benefit, alternatives) via specialist review and statistical analysis.

Main Results:

• AI-assisted editing significantly improved readability scores (KReaD, Natmal) and reduced sentence length and difficult word ratios.
• Content quality analysis revealed a decline in risk description and overall impression scores.
• Statistical models confirmed significant reductions in risk disclosure and overall impression, indicating potential omissions.

Conclusions:

• LLM-assisted surgical consent forms enhance readability but may compromise content completeness, especially risk disclosure.
• A balanced approach is crucial to maintain accessibility while ensuring medical and legal accuracy.
• Further research should focus on patient comprehension, informed decision-making, and multilingual validation.