Enhancing Readability of Lay Abstracts and Summaries for Urologic Oncology Literature Using Generative Artificial...

Area of Science:

• Urologic Oncology
• Artificial Intelligence in Medicine
• Scientific Communication

Background:

• Original abstracts (OAs) in urologic oncology can be challenging for patients and caregivers to understand.
• Effective communication of research findings is vital for patient engagement and informed decision-making.

Purpose of the Study:

• To assess a generative artificial intelligence (GAI) framework for generating accurate, clear, and readable lay abstracts and summaries (LASs) of urologic oncology research.
• To evaluate patient and caregiver comprehension and perception of GAI-generated LASs compared to OAs.

Main Methods:

• Forty urologic oncology OAs were used to generate LASs with a GAI tool.
• Readability metrics compared LASs and OAs; independent reviewers assessed accuracy, completeness, and clarity.
• A pilot study involved 277 patients/caregivers randomly assigned to read either OAs or LASs, followed by comprehension and perception assessments.

Main Results:

• GAI-generated LASs were created in under 10 seconds with high quality (85-100%) and minimal hallucinations (1%).
• LASs demonstrated significantly better readability (68.9 vs 25.3) and grade level compared to OAs.
• Patients and caregivers receiving LASs showed significantly improved comprehension and perception (P < .001).

Conclusions:

• GAI-generated LASs are highly readable and maintain research quality for urologic oncology.
• LASs significantly enhance patient and caregiver comprehension and perception compared to original abstracts.
• Human oversight is essential to ensure the accuracy and completeness of GAI-generated scientific summaries.