Evaluation of prostate cancer pathology reports generated by ChatGPT to enhance patient comprehension

Abstract

The technical language of medical reports can be difficult for laypeople to understand. Artificial intelligence (AI) tools such as OpenAI’s ChatGPT present an opportunity to bridge this gap. This quality improvement initiative explored ChatGPT’s potential to translate traditional prostate cancer pathology reports into patient-friendly versions. Using institutional templates, we created 25 prostate biopsy and 25 radical prostatectomy reports and converted them into patient-centered versions using ChatGPT-3.5. Nine urologists evaluated these reports. Additionally, 41 patients participating in the initiative were asked to provide feedback on their understanding of the biopsy reports. Feedback was collected using structured questionnaires and readability scores. The AI-generated biopsy and surgical pathology reports had higher mean Flesch Reading Ease Scores (49 ± 5; 50 ± 3) compared to the original reports (31 ± 15, p < 0.001; 41 ± 17, p = 0.02, respectively). Surgeons expressed positive opinions about AI-generated biopsy and surgical pathology reports, including patient-friendly language (77% and 95%, respectively). Patients also agreed that the AI-generated biopsy reports were well-structured and easy to follow (90%) and helpful in understanding their diagnosis (88%). 76% of patients preferred AI-generated reports as part of their care. These findings suggest that AI-developed reports may improve readability and support patient understanding.