Abstract: SA-OR75
Personalized Patient Education on Lifestyle Modification and Nutrition Management in CKD Using ChatGPT
Session Information
- Nutritional and Lifestyle Interventions to Promote Health in CKD
November 04, 2023 | Location: Room 109, Pennsylvania Convention Center
Abstract Time: 05:33 PM - 05:42 PM
Category: Health Maintenance, Nutrition, and Metabolism
- 1500 Health Maintenance, Nutrition, and Metabolism
Authors
- Acharya, Prakrati C., Texas Tech University Health Sciences Center El Paso, El Paso, Texas, United States
- Acharya, Chirag, Texas Tech University Health Sciences Center El Paso, El Paso, Texas, United States
- Krisanapan, Pajaree, Thammasat University Hospital, Khlong Nueng, Pathum Thani, Thailand
- Tangpanithandee, Supawit, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Craici, Iasmina, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Thongprayoon, Charat, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Cheungpasitporn, Wisit, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
Background
Lifestyle modifications are critical for managing CKD. Patients struggle to access reliable information on modifying lifestyle. AI-based natural language processing technology, like ChatGPT, has the potential to provide information on lifestyle modification and nutrition. This study aimed to evaluate the effectiveness of ChatGPT in providing education by addressing questions related to lifestyle modification and nutrition in CKD.
Methods
15 frequently asked questions were generated using the Kidney Disease Improving Global Outcomes (KDIGO) guidelines "CKD evaluation and management 2012", "Diabetes in CKD 2022", and "Lipids in CKD 2013". The questions were categorized into 4 steps to evaluate the versatility of ChatGPT in generating responses. ChatGPT (March 23 Version) provided responses were evaluated by nephrologists and compared to KDIGO guidelines.
Results
Overall the response to addressing patient queries for original questions, paraphrased questions with different interrogative adverbs, paraphrased questions with incomplete sentences, and paraphrased questions with misspelled words were similar. ChatGPT provided accurate responses to 13 out of 15 FAQs across various complexity levels and paraphrasing variations. In 2 questions some statements made by ChatGPT could have been misleading to patients, such as "medications like erythropoietin stimulating agents and phosphate binders may affect the patient’s ability to engage in physical activity." Statements such as "high protein intake can result in accumulation of metabolites like ammonia in tubular cells" and "excessive protein intake can activate renin angiotensin aldosterone system" could be confusing to providers when these statements are not proven in human studies. Consultation with nephrologist was encouraged by ChatGPT in all the questions.
Conclusion
This research highlights the potential of ChatGPT as an effective tool for providing patient education. ChatGPT's response in addressing patient queries was comparable to the assessments provided by the KDIGO guidelines. Rare statements though made by ChatGPT could be misleading. Further advancements are necessary to enhance ChatGPT's performance in addressing more intricate aspects of lifestyle modification and nutrition management.