PHOENIX — Mayo Clinic researchers have developed and evaluated MedEduChat, an electronic health record (EHR) that works with a large language model to provide accurate, patient-specific prostate cancer education.
The findings are published in Nature Portfolio Digital Medicine and highlight a new approach to delivering timely, individualized guidance for people navigating a prostate cancer diagnosis.
Cancer patients often face uncertainty as they process complex information about their diagnosis and treatment options. Limited time with clinicians can make it difficult for patients to receive the detailed answers they need to understand decisions that shape their care.
This MedEduChat study demonstrates how advanced AI, grounded in Mayo-validated clinical data, can help bridge these gaps by delivering clear, conversational explanations based on each patient's own health record.
Understanding the patient experience using AI
Fifteen prostate cancer patients interacted with MedEduChat for 20 to 30 minutes as part of a mixed-method usability study conducted at Mayo Clinic campuses in Arizona and Minnesota.
Patients reported higher confidence after using the tool, with Health Confidence Scores increasing from 9.9 to 13.9 on a 16-point scale. Usability scores were also high; average survey responses ranked MedEduChat 83.7 out of 100.
Patients noted that MedEduChat helped them understand their diagnosis in a more accessible way. The tool provided relief by explaining unfamiliar or complex terms in a clear and concise manner.
The conversational format helped participants replace incorrect assumptions with medically accurate information derived from their own EHR.
Clinician-evaluated accuracy and safety
Wei Liu, Ph.D. , a radiation oncology medical physicist, and three Mayo Clinic clinicians independently reviewed 85 anonymized question-and-response pairs. They rated MedEduChat's answers as highly correct (2.9 out of 3), complete (2.7 out of 3) and safe (2.7 out of 3).
Clinicians also noted strong patient-readiness and moderate personalization, reflecting MedEduChat's ability to tailor explanations to each person's age, treatment history and cancer stage.
Although MedEduChat delivered accurate and clinically aligned information, clinicians emphasized the importance of ongoing monitoring to prevent errors that could arise from incomplete or inconsistently documented EHR data.
The research team incorporated a multilayer approach to address these concerns and guide future system enhancements.
Combining patient-centered education with AI
MedEduChat was designed with a structured educational model that combines closed-domain clinical data, semi-structured guidance and personalized interaction.
Patients can explore diagnosis details, learn about treatment options and side effects, and review lifestyle considerations and follow-up expectations. The tool draws only from validated sources, such as Mayo Clinic materials and National Comprehensive Cancer Network guidelines.
"This research demonstrates how large language models can be safely and effectively integrated into real clinical systems to improve cancer education," according to Dr. Liu. "By combining advanced AI with Mayo Clinic's electronic health records, MedEduChat delivers personalized, accurate and easy-to-understand explanations tailored to each patient's medical history."
Expanding AI cancer research
The study team plans to translate this work into clinical use across all three Mayo Clinic campuses in Arizona, Florida and Minnesota. Next steps include expanding MedEduChat beyond radiation oncology to additional cancer specialties. These efforts aim to make personalized AI-assisted education a routine part of cancer care.
This research was funded by the National Cancer Institute, the Eric and Wendy Schmidt Fund for AI Research and Innovation, the Fred C. and Katherine B. Andersen Foundation, and the Kemper Marley Foundation.
