Assessing the performance of chat generative pretrained transformer (ChatGPT) in answering chronic kidney disease‐related questions

Başak Can,Esra Deniz Kahvecioğlu,Fatih Palıt,Egemen Cebeci,Mehmet Küçük,Zeynep Karaali
DOI: https://doi.org/10.1111/1744-9987.14239
2024-12-18
Therapeutic Apheresis and Dialysis
Abstract:Background Chatbots produced by artificial intelligence are frequently used in health information today. We aimed to investigate the reliability and reproducibility of the answers given by Chat Generative Pretrained Transformer (ChatGPT), one of the most used chatbots, to frequently asked questions related to chronic kidney failure. Methods We reviewed frequently asked questions related to chronic kidney disease (CKD) from social media platforms and Internet. The questions were asked to ChatGPT, and the answers were scored from 1 to 4 by two experienced nephrologists. Results Eighty‐five frequently asked questions about chronic renal failure were examined and 60 of them were included in the study after exclusion criteria. Fifty‐one (85%) of the questions received 1 point, 7 (11.7%) received 2 points and 2 (3.3%) received 3 points. The similarity rates of the answers to the repeated questions were between 80% and 100%. Conclusion ChatGPT has provided reliable responses with high reproducibility to inquiries related to CKD.
hematology,urology & nephrology
What problem does this paper attempt to address?