Better eye contact with patients. Less time looking at the computer during appointments. More thorough EHR notes.
These are some of the benefits Salt Lake City-based University of Utah Health has experienced since implementing an artificial intelligence "co-pilot" for clinical documentation in January.
The technology is now being used by more than 150 clinicians a month at the health system, with over 10,000 patient encounters being documented by AI in May and upward of 25,000 total since the start of 2024. The providers employ the DAX co-pilot from Microsoft subsidiary Nuance that is integrated into Epic Haiku, the EHR vendor's mobile app.
"This is, frankly, how we're going to be doing our notes in the future," Jennifer Vogt, MD, a University of Utah Health family medicine practitioner and associate director of virtual health for her primary care group, told Becker's. "Before I was on the computer trying to look through their last cardiology visit or going through their last labs — and I still do that, but I'm not feeling that pressure of having to type in the note at the same time."
She said she can also leave her clinic right when it closes to pick up her kids from school and not be under the strain of having to write her notes when she gets home.
She pointed out that the technology still hallucinates, a common problem for generative AI, but that the notes are "85-90% done" before she finalizes them. She said the co-pilot is also constantly getting better as clinicians give the developer feedback.
One barrier to wider adoption at University of Utah Health has been the price. The co-pilot costs about $400 a month per clinician — an amount the health system splits 50/50 with provider groups — and some providers have been skeptical about the return on investment, especially if they're only in a clinic a couple days a week. Others don't want to learn a new technology; it's a new way of documentation, where clinicians have to audibly state everything that happens during a visit so the microphone can pick it up.
On the flipside, some providers have reported spending half as much time on notes as they did before, while others say they're now able to see more patients in a day because of that time savings. One physician said it's helping prevent carpal tunnel syndrome among providers.
"We've heard from patients that the physicians in the practices using this had better eye contact with them and felt like, 'Somebody's really talking to me,'" said University of Utah Health CIO Donna Roach. "Also, in an academic setting we might see better adoption because the physician already is explaining to the medical student and resident what they're doing and that kind of cadence is what the Nuance co-pilot is looking for."
University of Utah Health providers have also started a chat page where they share feedback and pointers for how to more effectively use the technology.
"It can be a game-changer," said Lori Petersen, associate director of IT training for University of Utah Health. "We've got providers saying, 'This was the first time that I've never had to work on the weekend closing my charts.' And I'm like, 'What more could you want than that?'"
As for the price, University of Utah Health applications chief Travis Gregory said health systems are willing to pay a premium for DAX because it has the deepest integration with Epic of any solution of its kind. But increased competition in the years ahead could change that equation, he noted; healthcare AI company Abridge is starting to incorporate its technology into Epic as well.
University of Utah Health trialed the technology for behavioral health but found it ineffective because of the length and complexity of those visits. But the health system is developing an AI co-pilot model for hospitalists.
"I am somebody who was really efficient at my notes before, but they were very short and they were really just for me," Dr. Vogt said. "And now I actually step away from the computer. I'm better with my patients as far as communication and eye contact.
"And then I think my note is much more accurate to what the patient says in the HPI [history of present illness] because instead of me typing one line for their symptoms, it's capturing what they said. And I think the plan is actually easier for them to read. Because the whole point of a patient being able to read their note on MyChart is for them to understand it. So this could actually could be really empowering for patients."
Dr. Vogt said she's also looking forward to an upcoming Spanish version of DAX since a quarter of her patients only speak that language.
"We're really excited about AI in general," she said. "It can hopefully function almost like a search tool in the EMR for those things where I'm going back and forth trying to figure out what's going on. It could tell, say, which of my patients with uncontrolled diabetes I haven't seen in the past three months."
"The rate at which it's improving is just incredible," she added. "It's just really exciting."