In a recent online seminar at Vanderbilt University Medical Center (VUMC), Judy Faulkner, Founder of Epic, was the guest of honor with moderator Dara Mize, MD, assistant professor of Biomedical Informatics. This was part of a weekly Department of Biomedical Informatics seminar they hold according to a VUMC announcement. In the seminar, Judy shared some interesting perspectives on what we can expect from Epic.
The biggest piece of what was shared by Judy was Epic’s work to create an EHR AI Assistant and how it is still at least 2 years out. Here’s the summary of what she said:
Faulkner said the company is still at least two years away from releasing an artificial intelligence component that could listen in on the conversation between the clinician and patient and then draft clinician orders and a plan of care.
She observed that U.S. clinicians in part bring the documentation burden on themselves, writing notes in the record that are far longer than those written by their foreign counterparts, often needlessly repeating information in the note that is already captured and presented elsewhere in the record.
It will come as no surprise to any readers here that Epic is working an EHR AI Assistant or as we more often call it, an ambient clinical voice solution. As is described above, Epic and Judy know that the documentation burden on clinicians is overwhelming. In fact, Epic and their EHR often get the blame for this documentation burden. Ambient clinical voice is one possible solution to this problem they face.
What’s more interesting is that I’ve seen some people suggest that the above comments from Judy indicate that Epic is building their own EHR AI assistant. While I can’t say for sure that Epic is not working on something on their own, it is clear that Epic is working collaboratively with both Nuance (Nuance DAX) and M*Modal (now part of 3M) on making their ambient clinical voice solutions work with Epic. Given both of these efforts, I’d be surprised if Epic was trying to build an EHR AI assistant separate from these two efforts. Especially because Epic needs the voice recognition, NLP, and other related technologies that these companies already have in their portfolio. I don’t see Epic building or going out and buying these technologies on their own.
I do think it’s smart that Epic is working with the two leaders in this area. This gives Judy and her team a unique opportunity to see what both are doing. If they cut tail with one of them and go all in with just one, that will definitely say something so we’ll keep an eye out for that announcement. However, I could also see them supporting both solutions since many Epic customers are strong users of either Nuance or M*Modal’s voice recognition. Epic may want to be able to offer both solutions to its customers since those customers already have relationships with one of the two.
It is funny in the VUMC announcement above that they said “Everything Epic sells is developed in-house.” All of us know that Intersystems provides the MUMPS database that powers Epic. Plus, we reported previously that Epic was using the Twilio API to roll out their telehealth solution. They integrate with a wide variety of clinical decision support vendors. Plus, Epic has hundreds of apps in their App Orchard. It’s a little bit semantics here and probably doesn’t matter, but it’s fascinating to see that Epic is still pushing the narrative that they develop everything in house.
Ironically, MEDITECH had that claim to fame first. They literally developed their own phone system and email system back in the day (Check out our interview with MEDITECH CEO Howard Messing for that story). Probably more interesting is to note Epic’s lack of acquiring companies vs developing everything in house. In that respect they’ve never had to integrate a separate acquired software into their product, but they do integrate outside software into their product and in fact they should embrace this mentality even more. However, when Epic does announce an EHR AI Assistant, I won’t be surprised if we don’t hear much about Nuance or M*Modal powering it. According to Judy, we’ll have to wait 2 years to find out. As someone following this area closely, 2 years seems like a reasonable timeframe for ambient clinical voice to fully mature.
Another interesting comment that VUMC highlighted in their announcement was around Epic MyChart:
With the federal government soon to require that health care providers make the full EHR immediately available to patients, Faulkner said Epic has no plans to reprogram the Epic patient portal, My Chart, to allow patients to selectively turn off or delay notifications for things like cancer pathology results.
My guess is that this comment is a little disingenuous too since I would be shocked if Epic chose not to provide a pathway for their customers to meet regulatory obligations. Either Judy believes that MyChart in its current form meets those obligations or there’s an App Orchard partner that will allow an Epic user to comply with the April 5, 2021 Information Blocking and health data sharing requirements of the 21st Century Cures act. Either way, I’m sure there’s a pathway to meet these requirements or Epic and Judy will be making changes to their patient portal to be able to meet these requirements.
Judy also commented to VUMC that the best way academic medicine can work with vendors like Epic is to “know the software and utilize it well.” Plus, she added “Nobody utilizes the software as well as it should be done.” This is an age old software question on who is responsible for software utilization and upgrades. Plus, it’s a commentary on a true multi-tenant vs single tenant EHR software architecture where users can choose to remain on old software as long as they want versus being automatically upgraded to the latest software on a multi-tenant EHR setup (similar to how Gmail upgrades everyone). However, she’s right. A lot of their customers are on old versions of Epic and/or don’t use all of Epic’s functionality. She can only take partial blame for that. I’ve been told that Epic has an “on by default” policy to try to encourage adoption of things like patient access to data in the portal. However, Epic can’t and shouldn’t stop a user from turning it off. They’re the user after all.
A number of fascinating insights into the world of Epic. What do you think of these comments from Judy? Are you seeing or hearing something different from what I described above? Do you think 2 years is the right time for an EHR AI assistant in Epic?
