News
Article
Medical Economics Journal
Author(s):
How Denton Shanks, DO, of the University of Kansas Medical System, uses artificial intelligence.
Dr. Denton Shanks started working with a virtual scribe with an AI component about five years ago. It was a combination of AI transcribing with some human oversight from the software company. The problem was that it wasn’t always the same human reviewing his notes, creating wide variation in style and quality. The human element also meant delays of a few hours before the note was ready for Shanks’ review, which increased his after-hours work because of the wait time.
In 2023, he started working with a new virtual scribe company that used AI only. “It would write my note in the time it took me to walk from the clinic back to my desktop,” Shanks says.
He went from finishing his charts between 11 p.m. and 1 a.m. after hours to leaving the clinic with the notes written.
“All I needed to do is kind of glance at them to make sure they don’t need any edits,” Shanks says. “It writes it for me almost instantaneously in less than a minute, and that truly has been a game changer.”
Internal surveys found that physicians in the health system using the AI tool were less burned out and more productive, and wrote better notes.
“Maybe now I have more time to investigate what needs to be done for this patient as opposed to just worrying about getting things down and getting the note done,” Shanks says, adding that there are other AI tools in radiology and pathology at the health system.
He says accuracy started well but often would not get medication or procedure names correct. But during the past year, he has seen a drastic improvement, getting even complicated medical terms correct, and he’s not the only one noticing the improvements. The software’s rating has gone from around a 3.3-star rating to 4.3.
“We’re at a point where you still need human oversight, and we should assume that AI is getting us to 80% or 90% of the way there, then I need to take over that responsibility to make sure the note is 100% correct,” Shanks says.
Pilot programs can often be time-consuming and costly, especially for a smaller practice, so he suggests relying on the American Medical Association and your medical society for advice on what programs are worth investing in if you can’t try a pilot project at your practice.
The health system is trying an AI tool that helps manage the in-basket from the patient portal. “The patient may send in a paragraph or a whole story, and the AI can read that and draft a response,” Shanks says. “A lot of times, it may not be me, the provider, looking at that, but a nurse [who is] actually going to use the drafted response. The goal isn’t to be 100% perfect or make it fully automated, but the goal is to get the draft that is 80% to 95% good. The nurse can add a few things to personalize it and then hit send. And by far, that AI-generated draft is most often much better than what I would have said, and it’s more polite, more sympathetic and more empathetic.
“It’s really a partnership between AI and humans that will enable us to do far more than we could previously imagine. Will AI replace doctors? The answer is no, but doctors [who] use AI will replace those [who] don’t, because the tools will allow you to do far more than someone can do without those tools.”