Bridging the Gap: A Pilot Study Using Artificial Intelligence to Make Plastic Surgery Research Accessible.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Health Communication

Background:

• Many Americans struggle with health literacy, hindering understanding of complex medical information.
• Plastic and reconstructive surgery literature often exceeds recommended readability levels, impeding patient education.
• Disparities-focused research in plastic surgery faces additional barriers due to complex language.

Purpose of the Study:

• To evaluate the ability of large language models (LLMs) to generate accurate and patient-accessible summaries of plastic surgery research.
• To assess if LLMs can bridge the gap in understanding complex medical literature for diverse patient populations.

Main Methods:

• Eight disparities-related plastic surgery articles were processed by four LLMs: ChatGPT-4o, Gemini 1.5 Pro, Grok 3, and DeepSeek-V3.
• A standardized prompt was used to simplify the text to a sixth- to eighth-grade reading level.
• Summaries were evaluated using readability metrics (Flesch Reading Ease, Flesch-Kincaid Grade Level) and physician review for accuracy.

Main Results:

• Grok 3 produced the most readable summaries, with average grade levels between 7th and 9th grade.
• Grok 3 significantly outperformed other models in readability metrics (p < 0.05).
• ChatGPT, Gemini, and DeepSeek showed minor improvements but did not reach statistically significant readability levels (10th-12th grade average).

Conclusions:

• Grok 3 demonstrated superior readability and accuracy, meeting health literacy benchmarks for patient-accessible summaries.
• Current AI tools, except Grok 3, fall short in making complex surgical literature understandable.
• Improving LLM performance can enhance patient access to information and empower diverse populations through better health communication.