OR WAIT null SECS
AI Use in Ophthalmic Interdisciplinary Communication, with Prashant Tailor, MD
Tailor discusses the future of AI in ophthalmology and beyond, and some of the challenges it may bring.
Advancements in AI technology have led to bridges being built between specialists in countless industries. By quickly and efficiently translating technical jargon, large language models (LLMs) can efficiently decode even the most scientific writing for transfer between different medical fields. Recently, an LLM was tested by having it transcribe notes from several ophthalmology appointments.
In an interview with HCPLive, Prashant D. Tailor, MD, a retina fellow at the Jules Stein Eye Institute and Department of Ophthalmology, David Geffen School of Medicine at UCLA, explained the potential applications for and impediments to the widespread use of AI summaries to transcribe ophthalmology notes.
"These new protocols could be applied to any niche field, especially subspecialty medicine, in the retina clinics or the glaucoma clinics or the cornea subspecialty clinics...you're diving down not just into ophthalmology, you're diving down into a niche field," Tailor told HCPLive.
The results of Tailor’s study showed a significant level of acceptance, with attending physicians reporting significantly higher overall satisfaction (82.8% vs 72.7). Additionally, while errors were noted in 26% of PLSs, the vast majority took less than 1 minute to correct. Most of these errors occurred in the planning section of the summary. The majority (83.9%) of all errors were considered low risk.
However, attending physicians did report longer correction times and greater perceived burden (26.9% vs 2.5%) than with trainees. Tailor and colleagues noted that, with a 15% overread error rate, the influence of both AI and human error are compounded. The team therefore sees multilayered human oversight as the next method of filtering out the LLM’s mistakes.
“Patients look and refer to [our notes] asking ‘what do I need to look at or know about?’ And if there’s a summary that’s wrong, or an AI generated thing is incorrect, that can cause a lot of confusion,” Tailor told HCPLive. “I think it needs to be vetted by multiple steps.”
Overall, however, the results of the study showed that LLM-generated summaries were associated with greater satisfaction and comprehension among non-ophthalmology clinicians and professionals, while simultaneously reducing time burden for ophthalmologists. PLSs, the team therefore indicates, have the potential to bridge additional disciplines outside of ophthalmology alone.
“I think the team based approach in healthcare is the most important thing about taking care of patients, especially as they get more complex. I think AI is just a supplement there, a tool," Tailor told HCPLive. "If we can help communicate better in different ways so everyone’s on the same page, then we can better take care of our patients, too.”
Tailor reports no relevant disclosures.
