HIMSS24: How Epic is building out AI, ambient technology for clinicians

ORLANDO, Florida—A year ago, at HIMSS23 in Las Vegas, Epic announced it was working with Microsoft to integrate large language model tools and AI into its electronic health record software.

Fast-forward 12 months, the health IT giant now has 60 use cases in development, according to Epic executives.

AI technology in healthcare is rapidly evolving, and the potential for generative AI to free up time to focus more on patients is getting clinicians excited about technology, said Jacqueline Gerhart, M.D., Epic clinician and vice president of clinical informatics.

"We are collaborating with customers, we are co-developing with customers, we are fine-tuning with customers. We see it as an advantage that we have a good network of people who want to put this technology in. This has never happened before. They're coming to us. And they're saying, 'When can we get this? When can we do it?' From a clinician perspective, I haven't seen clinicians more excited," Gerhart, who also is a family physician at UW Health in Wisconsin, said in an interview during the HIMSS24 conference this week.

Through Epic's partnership with Microsoft, the two companies have combined the Azure OpenAI Service with Epic’s EHR software. A year ago, several health systems, including UC San Diego Health, UW Health in Madison, Wisconsin, and Stanford Health Care signed up to deploy one of the initial solutions that automatically drafts responses to patient messages.

The company also is working with Nuance, Microsoft's speech recognition subsidiary, to integrate DAX Copilot into its EHR software. Nuance's ambient AI technology automates clinical documentation by "listening" to the physician-patient medical visit and taking notes, allowing the doctor to be "hands-free" when speaking with patients. With the addition of GPT-4, DAX Copilot automatically creates draft clinical notes in seconds after the patient visit and is available to the clinician for immediate review.

In the past year, there have been over 150,000 notes drafted in Epic's EHR software using ambient technology, according to Seth Hain, senior vice president of R&D at Epic. "There have been multiple millions of In Basket message draft responses created," he said, referring to Epic's electronic messaging system.

The integration of AI and ambient technology into Epic's EHR is changing how clinicians use technology, making it "approachable and fun," Gerhart said.

"Clinicians often first think about technology as outside of how they practice. I think this is the first time that I've seen the technology so integrated that it makes the process more joyful. When I'm using ambient, I can see my patient, I can hear what it is that they're saying, I'm not looking at my computer. I'm not necessarily taking time out of my direct contact with them, so I can focus on my higher-level activities like diagnosis and therapeutics, rather than trying to think what I'm putting into the system," Gerhart explained.

She added, "And we're thrilled because in the future with ambient, we're not just going to be writing notes, we're going to have orders placed and multiple other activities. The decrease in burden, the way I feel when I use ambient, it brings me closer to my patient. I find that the patients feel that they're heard more. And, truly they are; it's getting their exact voice into the record. What better way to empower patients, rather than asking them individual questions and their history, for example, but actually having them tell their story."

One of the largest EHR vendors, Epic currently has the largest share of acute care hospitals in the U.S. market, at 35.9%, with 48% of U.S. hospital beds, according to KLAS Research's report on 2022 U.S. hospital EMR market share.

Epic approaches AI innovation with a strategy of building it into the core of its application, Hain said.

"Our key focus, whether it's in the traditional predictive analytics space or in the generative AI space, has been putting the technology, in this case AI, into the workflow directly so that it's running on the latest information and is able to help drive forward good decision making," Hain said. "I'm a developer at heart, so having a consistent way to monitor and assess the quality of the AI and have transparency for the health systems that are responsibly deploying it, also is critical."

He added, "The reason why you see 60-plus use cases from the clinical to the bedside to the back office is because we've integrated generative AI into our core development processes. It's just part of the standard approach, where predictive analytics tended to be very point solution and it just wasn't a part of how a developer thinks."

Having AI solutions seamlessly integrated into clinicians' workflow is key to reducing administrative and documentation burdens and enhancing patient-doctor communication, Gerhart said.

"Everything we're doing is in workflow. I think a lot of other AI solutions may be added on. We work to make sure that if you're in your In Basket responding to a patient message, for example, you don't need to go anywhere else; it's just right there to help you draft a response. Or if I'm in my note in front of my patient, I simply put my phone between you and me, and I turn on ambient and it flows right through my Haiku app on my phone," she said.

She added, "We're focused on the clinician and on use cases that are going to be the least risky and the most beneficial for clinicians and patients. When we use the draft AI messages, we have icons and terminology so that it's transparent to the clinician that they are using AI. They can choose whether they'd like to start with the draft response that's blank, or whether they'd like to start with a draft response that was generated by AI."

In many ways, AI and ambient technology are changing how EHRs work.

"There is an increasingly general capability to adapt the software to the context of the circumstances, so it is drafting a response to the message that came from the patient, direct to what they said, including empathy, and aspects of what was in that note, rather than just a form letter," Hain said. "It's that type of contextual awareness that gets inserted into the software. What we're finding is that it opens up a number of doors for new types of experiences in the exam room and at the bedside. And I think ambient is a perfect example of that where you now have technology playing an entirely different role in the room than you did before."

Beyond automating clinical notetaking, advances in AI technology can also help clinicians pull together relevant information from disparate sources to aid in clinical decision-making at the point of care, Gerhart noted.

"In the future, it's going to help augment me in being another tool in my tool belt so that I can work at the top of my license," she said.

Mayo Clinic kicked off a project in early 2023 to use generative AI technology to respond to In Basket messages from patients. Mayo started with a small pool of physicians piloting AI-enhanced messaging, and now nurses across departments are using AI-enhanced message drafts daily.

Initial pilots showed that it saved nurses around 30 seconds per message and drafts more empathetic responses. Mayo Clinic now plans to expand access to all licensed practical nurses and registered nurses by mid-2024, which could save 1,500 hours per month.

"Generative AI is different than many others that we used to take months or years to figure out how to deploy. This is one of those technologies where our patients and the physicians are likely already using it, so it's not foreign. It makes it easier and more comfortable and it just speeds it up even more," Gerhart noted.

As generative AI evolves, models are being developed that are more multi-modal to incorporate text as well as images and videos, Hain said.

Agent-based workflows represent another opportunity, he noted. "You use the generative AI capabilities to help plan out a series of actions or steps that the software might take. This particularly is applicable at the front desk and in the back office, helping folks schedule appointments more effectively, optimize the schedule for the day, and in the back office, help coordinate everything from coding to communications and bill pay," he noted. "Those agent-based frameworks I think we'll see surfacing over the next 12 to 18 months fairly heavily."

As a physician working in a primary care practice with 2,500 patients, Gerhart is optimistic about how AI can enhance patient care.

"It used to be I would only connect with individuals that came into the office, but using the Health Grid, using telemedicine and being able to leverage AI in In Basket responses, I can get back to my patient more quickly. I am providing the service to my community and my patients that is above and beyond what I could provide before," she noted. "I truly believe that they're going to be able to get care more efficiently and more effectively because of using AI."

The health IT company continues to innovate to give clinicians new ways to access patient medical records. Its EHR software is now available to clinicians on the Mac App Store. Atlanta-based Emory Healthcare is the first health system in the U.S. to implement Epic on a large scale on Mac platforms, the health system said in a news release earlier this month.