Using ChatGPT-4 for Lay Summarization in Prostate Cancer Research to Advance Patient-Centered Communication:...

Area of Science:

• Biomedical literature analysis
• Artificial intelligence in scientific publishing
• Health communication

Background:

• Biomedical literature's complexity challenges public understanding.
• Lay summaries aim to improve health literacy and public trust.
• Current lay summaries often lack clarity due to limited training in plain language.

Purpose of the Study:

• Assess ChatGPT-4's effectiveness in generating prostate cancer lay summaries.
• Evaluate how prompt design impacts the quality of AI-generated summaries.
• Provide guidance for using generative AI in scientific publishing.

Main Methods:

• Extracted 204 prostate cancer articles from an oncology journal.
• Generated lay summaries using ChatGPT-4 with simple and extended prompts.
• Evaluated summaries on readability (FKRE), factual accuracy (expert rating), and word count.

Main Results:

• ChatGPT-4 summaries showed improved readability and factual accuracy over originals.
• Extended prompts yielded higher readability scores than simple prompts.
• High-quality ratings were significantly higher for AI-generated summaries (79.4% extended prompt) compared to originals (5.4%).

Conclusions:

• Optimized prompts enable ChatGPT-4 to produce high-quality lay summaries.
• Generative AI integration can enhance science communication for nonexpert audiences.
• Further research should include layperson evaluation and broader clinical domains.