Many dermatologists embracing AI
Click Here to Manage Email Alerts
Key takeaways:
- An anonymous survey found a majority of respondents are using AI in their practices.
- These data suggest a need for education and best practices development.
Artificial intelligence, specifically large-language models, are being used with increasing frequency in dermatology, suggesting a need for the development of education and best practices.
Large-language models (LLMs) are trained to supply textual output in response to textual input and are trained with large datasets. Since the release of ChatGPT in 2022, the use of these forms of artificial intelligence (AI) have been increasing in popularity in many fields, including medical practices.
“As AI continues to become increasingly integrated into clinical practice, it is important for increased education on LLM strengths and shortcomings to better equip physicians to judiciously and ethically implement these technologies in their practices,” Haiwen Gui, BS, of the department of dermatology at Stanford University School of Medicine, told Healio.
Finding a balance between AI’s strengths and shortcomings is needed, especially considering ethical concerns, Gui and colleagues said in a letter to the editor published in the Journal of Investigative Dermatology.
The letter detailed the results of an anonymous 18-question electronic survey evaluating the use of LLMs in dermatology practices, their perceived accuracy and any concerns.
Of 134 respondents, 107 were post-residency practitioners and 27 were current residents. Most of the post-residency practitioners worked in urban academic settings.
A total of 87 (64.9%) respondents said they have used LLMs with 45 (51.7%) reporting daily or weekly use and 77 (88.5%) reporting they are extremely likely or somewhat likely to continue using the technology. Most reported using ChatGPT (85.1%), followed by Google Bard (25.3%).
Most reported they used LLM technology to ask questions about diagnoses and medical management, create patient handouts and to write insurance authorization letters. Concerns included accuracy (78%), nonclinicians getting medical information (55.3%) and privacy (54.5%).
Almost all respondents (97.7%) said they found the need to edit outputs from LLMs.
“To the best of our knowledge, there have been no previous studies that explored dermatologists’ usage, perspectives and concerns about large language models in real-world clinical settings,” Gui told Healio. “This widespread and rapid adoption of this new technology points to the importance of continued research and development of LLMs.”