ePoster
Presentation Description
Institution: Princess Alexandra Hospital - Queensland, Australia
Purpose
Chat-GPT is a prominent large language model from Open AI that is openly available to the public for use. Skin grafts are a common procedure done by surgeons as means of wound closure. Post operative care is pivotal to ensure success of graft take. This study aims to explore the extent of knowledge Chat-GPT has on skin graft management.
Methodology
Five questions relevant to the skin grafting were provided to ChatGPT for answering. In addition, ChatGPT was asked to provide five high level evidence references to support its answers. Responses were evaluated based on their accuracy, detail, and coherence. A stimulated doctor – patient consultation was also conducted to assess whether ChatGPT could offer safe and accurate advice for management of skin grafts.
Results
ChatGPT supplied relevant information pertaining to skin graft care at a mostly superficial level. With regards to stimulated doctor-patient interaction, it was able to provide generalised advice on management and able to identify red flag symptoms. Despite being able to retain information from previous interactions, it produced irrelevant references to specific skin graft care.
Conclusion
ChatGPT has potential to provide information to skin graft management to the general public, but advice needs to be taken with a caveat due to limitations of reliable references to support advice. These large language models cannot currently replace the expertise of healthcare professionals and require further specialised date set training from experts to increase utility in medicine.
Speakers
Authors
Authors
Dr Daphne Wang - , Dr Sheramya Vigneswaran - , Dr Atul Ingle -