Bridging the literacy gap for surgical consents: an AI-human expert collaborative approach

Area of Science:

• Medical Informatics
• Health Literacy
• Artificial Intelligence in Healthcare

Background:

• Informed consent is crucial but often hindered by complex language in healthcare consent forms.
• Low health literacy presents a significant barrier to patient comprehension of medical information.

Purpose of the Study:

• To investigate the efficacy of GPT-4 in simplifying surgical consent forms.
• To introduce and validate an AI-human expert collaborative approach for consent form content appropriateness.
• To assess GPT-4's capability in generating novel, procedure-specific consent forms.

Main Methods:

• Readability assessment and GPT-4 simplification of consent forms from 15 academic medical centers.
• Comparison of pre- and post-simplification readability metrics using nonparametric tests.
• Independent expert review by medical authors, a malpractice attorney, and subspecialty surgeons; evaluation of de novo forms using a standardized rubric.

Main Results:

• GPT-4 simplification significantly reduced reading time, word rarity, and passive sentences (P < 0.05).
• Readability improved from college freshman to an 8th-grade level (P = 0.004), aligning with average American literacy.
• AI-generated forms met clinical and legal standards, achieving perfect rubric scores and expert approval.

Conclusions:

• AI-human expert collaboration effectively enhances surgical consent form readability without compromising clinical detail.
• This framework offers a scalable solution for improving patient communication materials and addressing health literacy disparities.