How health systems are winning over late adopters of AI scribes

Encouraging AI use among late adopters

On the panel, Michael Attilio, M.D., chief medical information officer at Rome Health, a health system in upstate New York, noted that AI adoption challenges have persisted two years into the organization's rollout as some clinicians remain hesitant to adopt new technology and workflows.

Health system staff would rather hold on to their dictation and typing workflows rather than start using AI, he shared, which led to the slow uptake in adoption.

"I'm finding that we're getting to this phase where the super users are going 100 miles an hour, and they have been since day one," Attilio said.

"Now we're having to ask ourselves the question: How do we get the late adopters on board?" he said, adding that more effort and resources will be necessary than he anticipated.

When it comes to getting late adopters started with Heidi, Attilio compared the process to teaching someone how to ride a bike -- it involves small steps. He first provides simple instructions for these clinicians to get started with the AI tool before they move on to more detailed steps. For example, he writes some of the AI prompts for the physicians so they know what type of response to expect when they ask the AI tool a question in a certain way.

"[We] try to even give them just a little bit more structure or framework around how they interact with AI, so they get a feel for it," Attilio told Health IT and EHR following the session.

Another strategy for increasing AI usage is to take advantage of the flexibility that Heidi brings in customization compared with the toggle switches in the ambient AI platform the health system previously used, according to Attilio.

"Where our old product simply had toggles like 'paragraphs or bullets,' we can give Heidi whatever instructions we want," Attilio said. "So because it can look exactly the way our clinicians want, adoption was better."

In the interview, Attilio said that as a health technology leader, he likes to understand the inner workings of AI, but he realized that late adopters of technology "just want it to work."

He recalled a conversation with a doctor at a meeting who said, regarding AI, "I just want it to work like my Tesla and my iPhone."

Attilio explained that building templates for clinicians can spur late adopters to use the AI tools. He compared building templates in an AI tool to building something with a bucket of Lego bricks.

"Some do better if given instructions to follow or a model to copy," he said. "Others simply say, 'You go ahead and build it. I'll play with it when you're done.' That last group is often our late adopters."

With Heidi, for example, physicians can have the tool write notes in their voice. However, he noted that it takes time for physicians to get familiar with this functionality. So, Attilio often helps those physicians by writing some of the prompts.

At Hawse Health, a nonprofit federally qualified health center in rural West Virginia, Johnathan Lyon, assistant director of behavioral health services, noticed some surprising findings in who tended to adopt Heidi more quickly at the clinic. Lyon noted that behavioral health clinicians were particularly averse to technology and AI, but that resistance melted once the clinicians began using the tool.

"The people that were most opposed to Heidi in the beginning ended up being the ones who use it the most now," Lyon told the audience.

He added that clinicians' opinions about the AI platform improved within six months of the implementation.

What clinicians gain by adopting AI

Panelists shared observations on how AI tools can reduce burnout among providers and improve their workflows, despite the hesitancy of late adopters to use the technology.

Nancy Cibotti, M.D., a physician-owner at MD^2, a primary care practice in Boston, noted how AI can ease workflow for providers by gathering data about patients before a visit. This allows physicians to be more present when meeting with patients.

"I want to be able to look at that information in a very succinct way, so that when I go into the room and make that eye contact [with a patient], I'm not looking in the computer to find that information," Cibotti, who is also regional chief medical information officer at Heidi, told the audience. "So that's a huge time-saving."

The tool has also helped save her time on after-hours work. She recalls a period before using AI when she would spend Friday nights writing her clinical notes and Sunday nights prepping her charts.

"I no longer have to do that because I can gather that information [with AI], and then it kind of grows from there," Cibotti said.

At Hawse, Lyon said he is seeing a 10% increase in productivity from AI use department-wide, and clinicians have reported that they are saving two hours a day from using AI.

According to Lyon, AI has eased documentation by allowing providers to complete psychosocial evaluations much closer to the date of service -- either the same day of treatment or within 72 hours.

Lyon described psychosocial evaluations as the most documentation-intensive service that outpatient therapists provide, equaling the amount of time spent with patients.

Before AI, therapists would complete the documentation on weekends, resulting in a pileup of more than 20 psychosocial evaluations in Lyon's inbox on Monday mornings, which he needs to review and cosign.

Now, therapists are able to complete the evaluations more consistently throughout the week, he added.

"With the addition of [AI], it allows the information gleaned from the interview as well as information from provided records or administered psychometrics to be included in the output, which the therapist then reviews and changes anything if necessary," Lyon told Health IT and EHR.

Adoption of the AI scribe has also improved clinician workflows at Hawse by increasing patient encounters from 100 per month per full-time employee to 115 six months after implementation, Lyon shared during the panel.

And, echoing Cibotti, Lyon emphasized that AI helps improve the quality of those provider-patient interactions.

"Anything that increases the space for the clinician to be a person with the patient is going to be hugely valuable, and I think AI definitely does this," he said.

Brian T. Horowitz started covering health IT news in 2010 and the tech beat overall in 1996.