Subscribe to The Podcast by KevinMD. Watch on YouTube. Catch up on old episodes!

Join Robert Murry, chief medical officer at NextGen Healthcare, to delve into the transformative power of artificial intelligence in rural medical practices. Robert, also known as a country doctor practicing in the picturesque “horse country” of New Jersey, shares his firsthand experience of integrating AI-driven ambient listening into his workflow. Discover how this innovative technology is revolutionizing patient care, streamlining documentation, and combating physician burnout in rural health ccare settings.

Robert Murry is chief medical officer, NextGen Healthcare.

He discusses the KevinMD article, “How this country doctor embraced artificial intelligence.”

SUBSCRIBE TO THE PODCAST → https://kevinmd.com/podcast

RECOMMENDED BY KEVINMD → https://kevinmd.com/recommended

GET CME FOR THIS EPISODE → https://earnc.me/ev0lY1

Powered by CMEfy.

Transcript

Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast, and get CME for this episode by clicking on the CME link in the show notes. Today we welcome Robert Murry. He is chief medical officer at NextGen Healthcare. Today’s KevinMD article is “How this country doctor embraced artificial intelligence.” Bob, welcome to the show.

Robert Murry: Thank you so much for having me, Kevin.

Kevin Pho: So we’ll talk about your article in a little bit. Let’s start by briefly sharing your story and journey.

Robert Murry: Sure. So as Kevin mentioned, I am the chief medical officer for NextGen Healthcare. We’re of course a vendor of electronic health records to practices across the country. But I’m also a family physician, and I still practice in New Jersey. It’s a rural area. New Jersey is obviously a highly populated state, but I tell people to picture the area that I live in as more like Vermont than, say, The Sopranos. I’ve been practicing medicine for 20 years, and I’ve been involved in clinical informatics and health IT for that whole time.

Kevin Pho: All right, so tell me how you blend those two worlds, as a practicing primary care physician and a physician executive.

Robert Murry: Yeah, there are actually a lot of us out there. You can be board certified in clinical informatics, which I am, much like I’m board certified in family medicine. And usually physicians who do this start out with some kind of EHR implementation, and their role becomes to help all of the physicians or providers who are learning how to use that system use it better. Making tweaks to the system, making training materials, and then analyzing the data that’s coming from that system, both for quality metrics and to see that folks are using it to the fullest.

Kevin Pho: All right, so today we’re going to talk about artificial intelligence, and you don’t need me to tell you it’s been making huge waves in pretty much everything that we touch in health care, but specifically when it comes to health IT, and we talk a lot about this on my show and on my site. Your KevinMD article is titled “How this country doctor embraced artificial intelligence.” Now, for those who didn’t get a chance to read your article, tell us what it’s about.

Robert Murry: So it’s about one particular use of AI in medicine, and it’s one that I really think is a slam dunk. I mean, there’s all kinds of experimentation with AI that’s going on. I think some of that’s going to pan out and some of it’s not, but this one I think is a win-win really, both for providers and patients.

It’s called ambient listening, and the basic concept is that somehow the conversation in the exam room is listened to by the AI, in our case just through an app on a smartphone, and that conversation is then turned into the start of the note. It really allows the provider to pay a lot of attention to the patient, as opposed to having their face buried in the computer, because they don’t really have to take notes. As that conversation’s happening, that information, at least the medically relevant stuff, will be captured for them automatically. And it can really dramatically change the level of engagement that the physician and the patient have, and it has a number of other benefits as well.

Kevin Pho: All right, so just give us a scenario of how you yourself use it in your primary care practice.

Robert Murry: Sure. So I’ll be seeing patients, I’ll walk into the room, I’ll put down my phone and I’ll say, I’m going to use my phone and voice recognition to help me write my note, is that OK with you? Patients have universally said yes up until this point. And then I just start. If it’s an acute visit, a chronic care visit, whatever it may be, we just go through our normal visit.

Typically I don’t need to use the computer or look at the computer during the encounter. I do look at the computer beforehand and kind of make my agenda for whatever I want the visit to be about. As the visit is wrapping up, I hit end visit on the app on the phone, and maybe 30 seconds later the start of my clinical note shows up in my EHR. If I want to, I can edit it before it goes there, on the phone, which I do sometimes, or I can edit it in the EHR.

And it’s capturing a lot of the little details of the conversation that are still medically relevant, but I didn’t have to take notes or type or anything. Sometimes it nails the entire visit note and I don’t need to do anything. Other times I need to add information into that note. But both times it’s saving me a lot of time and allowing me to really pay attention to the patient.

Kevin Pho: And how does the app, or the language model that you use, differentiate between what you’re saying and what the patient’s saying?

Robert Murry: So it turns the conversation into voice to text in real time, and then AI behind the scenes can easily differentiate between speakers. In fact, I’m not even 100 percent sure that’s really needed. There aren’t that many times where the AI needs to tell the difference between what the doctor’s saying and what the patient is saying.

But in any case, it takes that voice to text transcript, and then, if you think about it, what it’s been trained on is zillions of clinical notes. And so it tries to recast the conversation as one of those notes. Those who are familiar with how large language models like ChatGPT-4 work know they’re actually really good at that. And so that’s how it’s automatically ignoring the chitchat, and how your kids are, and who’s in college, and focusing on the types of information that it has seen in other clinical notes that it’s been trained on. It’s pretty cool.

Kevin Pho: And has that changed the dynamic of when you’re in the room with a patient? Do you have to introduce it to your patient and say, hey, my phone is listening to this conversation and it’s going to turn our conversation into a note? So tell us what kind of expectations you have to set for the patient with the ambient listening in the background.

Robert Murry: Yeah, I mean, this is going to vary by the people who are using it. When I do it, I’d say just what I just said. I’ll say, I’m going to use my smartphone and voice recognition to help me write my note today, is that OK with you? Patient says yes, universally, like I said, and that’s it. And then from then on they don’t notice. The app is working, of course, but they’re not involved in the process of me creating my note anyway, and so they don’t notice that part. So if they notice anything at all, it might just be that I’m paying more attention to them, as opposed to being a doctor whose head was buried in the computer while they’re trying to see the patient.

Kevin Pho: So give us a before and after picture. What was life like before this ambient AI app, and what’s life like now, when it comes to your workflow and how that’s impacted your documentation routine?

Robert Murry: Yeah, for me personally a lot of it is similar, but it’s even faster. So prior to using ambient, which I’ve been using for about a year now, I was using a different way to get into our EHR, which of course is NextGen since I’m the chief medical officer of NextGen. I would use my voice to dictate sections of the note, and those would automatically flow into the EHR. But if I got behind in my schedule, I found it hard to dictate those little snippets between patients, and so I ended up doing a lot of that at the end of the day. I would still typically get out on time or nearly on time, but it was a little bit more disjointed.

With ambient, since I know the majority of the note is already there, I do try to take a couple of seconds, clean it up, add whatever I need, and be done with it. So I’d say the fraction of notes that I’m finished with before I see the next patient has definitely gone up, and just the overall time spent documenting has decreased.

Of course, we’re rolling this out throughout the country, and so we have lots of different types of physicians and different specialties and practice settings that are trying to use this. And again, this is just our particular version of ambient listening. But a lot of them who were struggling with documentation, or finding their EHR difficult to use, are noticing dramatic drops in the amount of documentation time. Almost every physician that we give this to to try out says, OK, I want this immediately, and you’re going to be prying it from my cold dead hand, so to speak.

Because it’s really just incredibly useful, both for doing the scut work of writing your clinical note, as well as really allowing a different level of patient engagement, because you’re not having to capture every little detail during the encounter.

Kevin Pho: So you mentioned that the large language models, like the one that you use, are trained on however many hundreds of thousands of patient notes. So talk about some of the privacy and security concerns that sometimes arise when large language models use patient notes to train on.

Robert Murry: That’s a great question. So we are using a third party to run that large language model for us. It’s a company called Nabla that we think does a particularly good job of it, and so I’m actually not 100 percent sure where they get all of the materials that they’ve trained on.

There are AI algorithms out there that are pretty good at de-identifying notes. They’ll catch every proper name within a note and put redacted or whatever, and so that is typically what is done when training these large language models, because you don’t want PHI, obviously, to ever be coming out of the model.

Perhaps also interestingly, we get asked by physicians who are thinking about using this, well, what about the privacy and security concerns as you’re using it with the patient? And so for our implementation of it, as soon as that note makes it back into the EHR, everything is deleted. The audio files that were used to create the voice to text transcript, and all of the materials used to do the notes. So we’re not using anything that is used or created for patient care to come back and train the large language model. Nothing is stored. Of course everything’s encrypted at rest and in transit and everything.

So it’s good to ask about security and privacy concerns with AI, but in our particular implementation of it, I think that they’re very well taken care of.

Kevin Pho: Now, in your year of using ambient AI, have you noticed any challenges that you face in terms of adapting to this new workflow?

Robert Murry: Yeah, it’s a great question. I think the key to physicians being successful with this, and I really think that this is going to change the way that a lot of doctors practice medicine because it is so incredibly useful, but the key thing going in is to not expect that the note is going to be perfect, or that the note is going to be complete. You have to go into this thinking about it as, this is a tool that helps speed you up, it helps you do your job, it doesn’t do your job.

So if you go into it with the expectation that every single time you’re going to need to read it quickly, to make changes and to edit the note, then most people find that a very satisfying experience.

Ambient is really good at capturing little details. How long did you have that cough, did you have a fever? Because if you’re a primary care physician, I might see 10 coughs in a row, and if I don’t jot down notes I’m going to get them all mixed up. With ambient, though, it’s going to capture all of that. But it’s not smart. It doesn’t really know what I’m thinking, and so I need to expect to add that kind of thought process, or to make the note polished for the next person who’s going to read it. Those sorts of things need to be done.

The other thing that I’ve learned over the course of using it is that, being a good doctor, I’m a little bit lazy. And so I know that if I say something out loud, it’s going to capture it. One thing that ambient listening is really, really good at is actually the patient plan. What do you think is going on, and what are you going to do?

And so I’ve purposely been making a point of calling those out during the patient encounter. You can do it along the way, you don’t actually have to do things in the same order that the note is in, so you don’t have to do it at the end. But I’ll say, this is what I think you have, this is what I don’t think you have, these are some tests that we can do and this is what I expect to learn from them, here’s stuff you can do to make yourself healthy, here are the prescriptions. I just lay out the prescriptions with the dose and the sig and everything, and I know that all that stuff’s going to get captured.

And it’s very good, because the patient hears all that and immediately has the chance to ask questions, or say, I’m not sure I’m going to be able to do that, or I can’t afford that medication, or whatever it may be. You can talk about any kind of barriers to that treatment plan right then and there.

And also just the way that you formally summarize it. I mean, I’ve been practicing 20 years, I thought I was a good communicator, but to say, this is this, they actually really appreciate that, and they remember more. And then the ambient returns so fast that you can put all of that stuff in the patient summary as well, so it’s like an immediate version of the plan that you just said out loud.

So all of that promotes, I think, good patient engagement. It’s ironic, right? It’s the AI that’s encouraging the human interaction in the room. So it’s kind of funny.

Kevin Pho: So you mentioned that there’s some degree of post-production that needs to be done on some of these notes. How much time do you spend doing that?

Robert Murry: It’s going to vary. I mean, again, some notes are perfect as is. For me in primary care, typically if you’re there for almost any acute complaint that I would see, that note is perfect. But if you’re there for multiple chronic conditions and I know you well, which I do with a lot of my patients, having been practicing a long time, there’s a lot that’s just not said. Did you do this thing I said last time? I don’t know. Oh really, was it this reason? It’s just snippets of conversation between people who know each other, and so of course it doesn’t do a great job of that. So those notes for me are the ones where I typically need to do the most editing.

But even given that, I know the plan will be nailed, so I don’t have to try to immediately enter all the orders, for instance, because I know that they’re there and I can refer back to that. And the editing that I do is, at most I’d say two or three minutes per note, versus if I needed to obviously type it, or use my EHR accelerators to do that, or even use my voice and then double check the voice to text. It’s still overall a time savings for me in almost every case.

Kevin Pho: Now, ambient AI, as you said, is an exploding field, and we’ve had several ambient AI solutions, and in full transparency one of your competitors is a partner of mine and a sponsor of my site. In general, when physicians and practices are looking at ambient AI solutions, what separates one solution from another?

Robert Murry: It’s a good question. I think, in addition to the technology itself, which does vary somewhat between vendors, and I think over time that variation will go down and everyone will kind of, rising tide, hit a pretty good level of the actual conversion of the conversation into the note.

The other big differentiator is exactly how this interfaces with the EHR. Because again, I would set expectations for people using this that, even though you don’t have to sometimes, you’re going to need to edit most of these notes, add something just so you sound smart if nothing else. And so how easy is it to do that? How smooth is it to get it into the EHR? How much is coming in once it gets to the EHR? How much is automatically recognized? Does it go to the right place? What do you have to do once it’s there? Those are key differentiators in the marketplace.

The other is of course just making sure that the solution that you’re looking at has the privacy and security and regulatory compliance that you think you will need, based on your local or state laws.

Kevin Pho: We’re talking to Robert Murry. He is chief medical officer at NextGen Healthcare. Today’s KevinMD article is “How this country doctor embraced artificial intelligence.” Robert, we’ll end with some of your take-home messages to the KevinMD audience.

Robert Murry: I think the take-home is, number one, that this type of ambient listening is a good type of AI. There are a lot of really scary possible uses of AI, there’s a lot of experimentation out there, but this one I think is one that helps the provider do their job faster and helps the patient, because the provider can take care of them.

So I’d really strongly encourage people to look into this. In one sense it can be expensive, although the ROI in almost every case, for any busy provider, makes it very much worthwhile once they look at it.

Don’t think that it’s going to hit a home run every time. First of all, the tool will get better over time, but also, like other things, it is something to help you do your job, it’s not going to do your job for you. So use it for what it’s good for. Experiment with it. You will find your own style of using it, and what types of encounters you expect it to do really well, and what types of encounters you’re going to need to add to. And make sure that you’re trying to maximize its benefits, not getting frustrated when it makes mistakes.

But I think overall, if you give it a try, you play with it for a little bit, see if it adapts to your practice style, you will save time. You’ll get out of there a little faster. You won’t have this kind of task hanging over you, that you have to finish all your notes for the day or all that kind of stuff. A lot of that, it eases the pressure, and it’s really beneficial.

So I would encourage every physician, every provider out there, to at least try it. And then if they don’t like it, maybe try it in six months or a year, because the technology is evolving unbelievably rapidly.

Kevin Pho: Robert, thank you so much for sharing your perspective and insight, and thanks for coming on the show.

Robert Murry: Thank you so much for having me, Kevin. Have a good one.