Roupen Odabashian: Can ChatGPT 3.5 assist oncologists in clinical decision-making CMD?
Roupen Odabashian, Hematology and Oncology Fellow at Karmanos Cancer Institute, shared on LinkedIn:
“Can ChatGPT 3.5 assist oncologists in clinical decision-making CMD? Not quite yet.
In this study, we assessed the performance of ChatGPT in answering oncology-related multiple-choice questions and compared it to the performance of an oncologist. The accuracy rate was 56%.
A strength of our article is that the questions used are unpublished and have not been discussed online. This differs from other studies that utilized USMLE-type questions, where the answers were discussed on Reddit forums.
However, we acknowledge several limitations, including instances where ChatGPT chose two answers that, while potentially correct in real-life scenarios, were considered incorrect for our study.
Stay tuned for follow-up studies assessing the use of AI tools in clinical decision support for oncologists. However, I am very excited about the development of more tools that will assist oncologists in everyday clinical decision-making.”
