A group of doctors at UW Health is trying out a tool that uses artificial intelligence to take notes on their visits so they can focus more on patients and less on screens. Proponents say the approach could not only improve the patient experience but also reduce burnout among overworked doctors.

The pilot program began in July with 20 physicians in several UW Health specialty and primary care clinics. It uses a feature from Verona-based healthcare data company Epic Systems that allows its health data software to integrate with a variety of third-party AI tools, such as Abridge and Nuance’s DAX Copilot.

These tools can record doctor conversations and create summaries that doctors edit and add to the patient’s record. Health systems pay for these tools, but do not pay extra for integration with their Epic software. UW Health would not publicly disclose which of these tools it uses, citing concerns about future negotiations with the vendor.

The technology is designed to solve a common complaint among patients and doctors alike: Patients go to their doctor, but the doctor often spends much of the visit staring at the computer and taking notes. Doctors who forego the screen to give their patients more face-to-face time may have to catch up on their notes at home, long after the workday is over.

With the new tool, the doctor asks the patient for permission to use AI to do the recording. “If the patient doesn’t agree, that’s it. We don’t use the tool,” said Chero Goswami, chief information officer at UW Health.

If the patient consents, the doctor activates the technology and turns it off at the end of the visit. Within a few minutes, a summary of the conversation is generated. The doctor reviews and edits it, then adds it to the patient’s medical record with the click of a mouse.

UW Health leaders say the tool has received overwhelmingly positive feedback from participating physicians. Among them is internist Dr. Jason Dambach, who calls it “a huge step forward” from his previous way of taking notes. “I was already pretty efficient, and this is still a huge step up from what I was doing before for my documentation.”

Still, UW Health said it will continue testing the tool before deciding to make it available to all 1,800-plus physicians at its more than 100 locations in Wisconsin, Illinois and Michigan.

More than 180 medical centers and hospitals use Ambient technology through Epic’s software integration, said Garrett Adams, one of Epic’s lead software developers for AI projects. Epic software also uses generative AI in about 100 other scenarios, from helping nurses compose messages to patients to summarizing a patient’s medical history.

As AI becomes more widely adopted, Epic and UW Health convened a panel of experts earlier this summer to advise lawmakers on how this technology should be used and regulated in healthcare.

A better experience for doctors and patients?

In the month since the pilot began, Dambach has treated about 80 patients, almost all of whom have agreed to participate. In the age of ChatGPT, many patients are familiar with the concept. “They are very happy to hear that we have tools, and some patients have told me that they feel like they got more out of the visit and were able to ask more questions,” Dambach said.

He also notices a change. “I’m more responsive to patients’ concerns and ask them more often during the visit: ‘Is there anything else you need?'” Dambach said.

The benefits aren’t limited to the visit, he said. The time saved on notes means more time for following up on lab results, X-rays and patient messages, work he would otherwise have saved for the evening or weekend. For patients, that means faster results.

The notes generated by the AI ​​usually require only minor edits, Dambach said. One time, a patient’s family member mentioned something in their family history, but the person’s facial expression told Dambach they weren’t quite sure. He decided to remove the comment from the notes.

Dambach, who also helps oversee software training for UW Health medical students and doctors, said colleagues have told him the tool saves them many hours each week.

Goswami has heard the same thing. “The provider doesn’t have to go home and write notes after dinner instead of spending time with their kids or partners. They actually get time back for their personal life to do other things,” he said.

At a time when 40% of U.S. physicians are considering leaving the profession, this is significant.

“One of the other (participating) doctors was quoted as saying, ‘I would be devastated if you took it away from me,’ and I can only agree with that,” Dambach said.

Pilot project to be expanded

Dambach is not worried about the new technology. “I continue to look for use cases where this product fails, and I haven’t found one yet,” he said.

While some tech tools come with a steep learning curve, this isn’t one of them, said developer Adams. In fact, UW Health intentionally selected a mix of more tech-savvy and less tech-savvy physicians to test the tool.

“I think the ease of use of this tool means that there is a pretty level playing field between the different users who are using it,” Adams said. “Those who are using it are using it right.”

In the coming months, UW Health leaders plan to expand the pilot project to about 50 and then 100 physicians.

Still, they want to wait to roll out the technology systemwide until they’ve thoroughly studied how it will impact the billing process. Health systems bill for services based on specific codes for procedures and diagnoses noted in a patient’s medical record. If the new tool makes it harder to determine the correct codes, that could be a dealbreaker, Goswami said.

“We are very cautious about coding and today we are investing additional manual effort in coding to measure and ensure accuracy,” Goswami said. “Until we reach the high bar we have set for ourselves, we will probably take very small steps.”

In the meantime, Epic is working to expand the tool’s capabilities, Adams said. Developers are currently working on features that will, for example, use recorded conversations to determine what instructions to send to the lab, and they are creating a version of the tool for nurses.