All right, we're going to go ahead and get started. Thanks for joining us this afternoon. You're in the AI session, if you didn't already know that. Let me introduce us really briefly, and then we'll get started. With us today is Dr. Bob Murray. He's the chief medical officer for NextGen Healthcare and a practicing family practice provider. Dr. Keith Nord is an orthopedic hand surgeon from Sports Orthopedic and Spine in Jackson, Tennessee. And I'm Molly Van Oort, specialty solutions director at NextGen Healthcare. So thanks for joining us today. We're here to pose the question, is AI the new easy button for the orthopedic practice? So let's discuss how AI can help ease you into summer. And today is the first day of summer, and you've got sunglasses to kind of represent that today. So future is bright. All right. I think I'm kicking things off right, Molly? Yes, you are, sir. Okay. So thank you for coming, everybody. This is a great turnout. Hopefully, we show you some cool stuff today, give you some things to think about for your practices. But the main use of AI that we're going to talk about is to kind of alleviate what is one of the biggest problems in orthopedic surgery, and actually throughout all of medicine, and that's physician burnout, the burden that using the EHR and doing all the documentation, both as an outpatient and as an inpatient, puts on the physicians and the other health care providers. And AI actually has massive potential. I think this is going to be one of the first kind of slam dunks use of this technology. You'll see in a moment how helpful it can be. Even if it doesn't do everything, it does so much that it's already for sure useful for a wide variety of physicians in different contexts. But we did kind of kick things off. We asked AI to generate a frustrated physician stuck in the office, his family's outside, or some family's outside. Although, I don't know who the beach ball with the legs is. And like most AI fixed photos, if you look closely, there's creepy stuff in there. Anyway, so this is a typical, right? The provider's stuck there with bizarre stuff going on, and he cannot get out. And what he really wants, of course, is the next position. And so this is kind of what we're trying to offer physicians that are using this kind of ambient listening technology. Awesome. So we're going to start right off. It's a little panel discussion. We'll definitely slide you guys into this, definitely from a QA perspective. Dr. Nord, how do you introduce ambient assist to your patients? Honestly, it's really a non-issue. It's no different than if you brought a scribe into the room to go meet a patient. You bring your iPad in, or your phone, or whatever device you're using, and it's just listening in the background. I've had maybe one or two patients in the nearly 5,000 encounters I've had. I've had maybe one or two patients in the nearly 5,000 encounters I've had over the past year that have been really interested in it and actually been looking at it because they can see the text going on the screen. But for the most part, patients don't mind it at all. I think, honestly, they enjoy more face-to-face time with me instead of me focusing on taking notes. And at the end of this session, we have some statistics from Dr. Nord's practice. Dr. Murray, you're using next-gen ambient assist in your family practice. What unexpected benefits are you experiencing? Yeah, that's a great question. And let's just take a moment to back up a step, because it's certainly possible that folks don't really understand what we're talking about when we're saying ambient listening. There's not a completely standard terminology for this. I certainly think ambient listening is good. But the basic idea is the provider goes in the exam room, introduces the patient, puts their phone down. The phone is just a microphone, but it just listens to that conversation. Whatever conversation is normally happening in the room between the provider and the patient, the description of the procedure, the nurse, the children, whoever, it listens to that conversation. It transcribes it. And then using AI, large language models, it turns that into a medical note for that encounter. And it sounds futuristic, but it works like a charm. And so again, the note might not be perfect, just like the AI picture is not perfect, but it is a very good start to things. So in terms of benefits from family practice, I see a lot of patients with multiple chronic conditions. So if you put it in the right settings, it can handle the complexity of those visits fairly well. But I find that I'm really able to pay attention to the patient, because I don't have to be in the computer getting all those details down, because I don't have time to do it after the visit, or furiously writing notes, because the simple stuff about the interactions, as you'll see in a second when we show you, the AI does a great job of capturing those. And the other thing is that the AI is really, really good with patient plans. If I say out loud, this is what I think you have, and this is what I don't think you have, and this is what we're going to do, and this is what we hope to learn with what we're going to do, and this is what we'll do next, depending on that, it nails that stuff. And then as you'll see, the NextGen solution is fully integrated with the EHR. And so that plan can immediately show up in the patient after visit summary, which we call the patient plan. And so there's other benefits that we can get into as well. But those are the main two. It's the capturing of the patient plans, and then the ability to just be present with the patient, and let some of the complexity kind of get captured by the machine. Perfect. So, Dr. Murray, thanks for laying out the concept here. Do you guys want to see it? Might make it easier to understand. So do I have an audience member that's willing to participate in a little visit? Yes. What we need is a patient. Dr. Nord's going to be the doctor. It'll be a simple orthopedic visit. You can make up whatever you want. But it's perfect. Perfect. Come on up. Thank you. And what's your name, sir? Charles Roberts. Charles. Wonderful. All right, guys. If you could come up on stage, just so you're not back to the audience, but it doesn't really matter. Bear with me. We're going to do an actual live demonstration. This is off the actual iPad that I use in my office and clinic. So my practice is a little bit different from yours. I'm a hand surgeon. So it's more fast-paced, quick encounters. I'm trying to get through 45 to 55 patients in a day. And it's sometimes a little bit of a rat race. And that's honestly what drove me into doing this, is I was spending such a large amount of time on documentation each day. I think it's almost coming up. Let's see. There's no signal there. No signal. Yeah. Give it up. Everything worked, of course, when you tested out ahead of time. Yeah. Two seconds ago, it worked perfectly. Of course. Sure. There it goes. But anyway, just from a... There we go. There we go. Okay. So I had my staff today, my nurse was so kind to put a bunch of test patients in here for me. But this is actually the app that I use for day-to-day for seeing around 50 patients a day. So what I'll do, it auto-tracks where I am on the schedule, and it'll kind of open up straight to where I am. We'll go to the test patient, Christine, here. Thank you, Christine, for being our guest. And this is kind of the landing page you see here, where it's got the phase sheet. I can look into medications or different charges or add medications and write prescriptions if I want to. But mainly, I'm going to today's visit. And then I just go straight into the ambient button. And this is where it just starts listening. So typically, I'd walk into the room, I'd set my iPad down on the counter, and I'd introduce myself and I'd say, hi, I'm Dr. So-and-so. How are you? What can I do for you today? Say you have a wrist problem or any problem you have. So I'm Christine. Nice to meet you. I was at jiu-jitsu where I work out, and this guy did an arm bar maneuver on me, and I just... I can't fully extend my arm, and it hurts like heck. How long ago did this happen? About a week ago. About a week ago. What have you tried for it so far? I don't know. Some ibuprofen, maybe. That's about it. Yeah. Have you heard anything else besides the elbow? If I did, it's not... I mean, not really. Not that I'm aware of. Okay. Do you have any allergies? Are you allergic to anything? I'm allergic to sulfa drugs. Any other past medical history or any other medical problems you've had? I've had some depression in the past. Took some Zoloft for that. But I don't know. I don't know. I don't know. And I had some reflux for a short period of time, but I lost weight. That's gone. So... Okay. Well, first things first. I'd like you to try some physical therapy. The x-rays that we got today, they were normal. We got two views of x-rays, and... Excuse me. I'm going to use a little bit of doctor speak. But two views, x-rays of the elbow were normal, AP and lateral views. There were no fractures. There was slight effusion about the joint. So was anything broken? Nothing was broken. Okay. Yeah. And I frequently will dictate this before I walk into the room, and we'll get into kind of the tips and tricks. I treat it just like I would an in-person scribe. I dictate my exam findings as I leave the room. So at the end of the encounter, I may tell you, you know, we're going to try some physical therapy. We're going to see if it gets better. If it doesn't, we may order an MRI next time I see you in about three to four weeks. Okay. If we don't have any improvement. And then typically, I may end the encounter there. And as I walk out of the room, I'll then say, physical exam, left elbow was examined. He has negative milking maneuver. He has full range of motion, pain, tenderness over the lateral epicondyle, pain with resisted extension of the wrist, distally neuro intact, no numbness, no tingling. And then I'm leaving the room, and I hit end. Thank you, Christine. Thank you. So this is real time. It's actually sending this to my server back in the office and generating the note right now. In the meantime, I'm going back to my desktop, which I'm also running while I'm running my iPad. In this time, I'm either putting if I did an injection or doing my, you know, my billing. So I'm clicking through that, adding a diagnosis. This really usually only takes about 30 seconds. And then by the time this is generated, I'm clicking submit. Boom. I've hit submit on the computer. My charges are in. My injections are in. Everything is done. This has taken my total patient computer time down to less than 90 seconds per patient, which is huge. But this is just the rough draft of our discussion we just had now. You can see it does a pretty good job. It gives you nice bullet points. It says 44-year-old female experienced an armbar maneuver during jiu-jitsu a week ago. A lot of times when I'm seeing 45 patients, I miss these details. This catches it, which is kind of amazing. Sometimes it actually generates better notes than I would have done by myself. You can see it has no other injuries. Left elbow. It gets the exam in there. When I dictated it in normal physician speak and technical terms, it adds all that in there. Pain over the lateral epicondyle, pain with resisted extension of the wrist. No fractures. It puts a diagnostic interpretation for me so I don't have to dictate a separate x-ray report. So the time savings here are incredible. You can see it does a good job with the plan. If no improvement, three to four weeks, MRI. So this already is a good-to-go note. If I wanted to make a change, say I wanted a little more detail on my x-ray, I can go here and just tap. X-ray finds a very subtle lucency over the radial head. Some concern may correlate with further imaging in the future. I can do voice-to-text and augment my note very quickly. Once I'm happy with it, I could type on here freehand if I want also. I just hit send to EHR. Boom. It's done. The note's in the chart. This actually merges with one of my templates that I have preset so I can hit a level five for every visit if I need to. It has my standard physical exam and those things. Not everyone is billed as a level five, obviously, but it has all the data points I need. And so it makes my job as a physician really easy. It's click the button and go. And so I get to focus more on the patients. We were talking about earlier how this has changed my practice. Seeing 45 patients a day, that only gives me about 10 minutes per patient for a full eight-hour day total for computer work and patients. That split used to be about five minutes for a computer, five minutes for the patient. Now we've dropped the computer part for me down to 90 seconds. That gives the patient so much more time. That gives me so much more time. I've been able to add five to six extra patients each day just because I'm not spending so much time at the computer. You can see on the live demonstration there's not too much to it. It's really just click the button, walk in, and have a normal conversation. So this is just kind of an example of how AI is really revolutionizing the experience for a provider. I do use macros, especially for operative notes, which I also use their mobile app for operative notes. And that's coming to the Ambient, I believe. That's a feature that's being worked on right now. It's a part of the mobile app for the voice-to-text already. And so I already use it for that, but yeah, it's a great question. I check each note, but you can see that process is very quick. It's right there as soon as you're done. You know, you're not waiting for, say, an old-school dictation to come over the phone that you then have to log back in and review and sign. It's immediate. It's by the time I've sat down at my computer, it's ready for me, and I can change it and send it. Yes, as with all dictation stuff, just like the picture of the AI-generated physician, you've got to watch for something. Like, I've had once where it decided randomly to put a 50-year-old female as a 600-month-old female. Yes, it makes sense to a computer. But for the most part, it is shockingly accurate. I mean, it honestly, a lot of times, surpasses my own notes as a busy orthopedic surgeon. Orthopedic surgeons get a bad rap for abbreviating everything. And I try to make a progress note for the hospitalist that's as few lines as possible And so this actually captures a lot more detail than I routinely put in my notes. And I agree. I mean, I've been using this for over 18 months because I got the early version as the chief medical officer. I've never seen it hallucinate. The only thing I've ever seen it do, I mean, it'll miss words, although it's obviously very sophisticated medical speech-to-text. The only time I've seen it gets confused is if the patient describes the physical exam that they think that they saw. You know, I looked in the mirror, and I thought my throat was red. I thought my throat was swollen. You know, I looked in the mirror, and I thought my throat was red. You've got to watch to make sure that it doesn't think that you said that and then put that in as the physical exam. So it's those kinds of subtle things. But it's never given me a spine-shivering experience. So that sounds a little scary. Yeah, so this particular one, I mean, this is unique to NextGen. It's called Ambient Assist. And so it's not only is the engine that's behind it proprietary, but the way that it integrates to the EHR is as well. There is similar technology for some of the others out there. Of course, I'm quite biased, and I don't think it's anywhere near as good for a couple reasons. The quality of the note is going to depend a lot on the situation of the provider. But the integration, the fact that it comes back so fast and really invites you to read it right away and edit it right away, I think is a workflow that's very, very useful. Because we're not to the point where you can rely 100% on the AI. But if you saw 15 hands in a row, or for me, five coughs in a row, if I don't fix those details immediately, you know, it's gone. Yeah, it's a great, great series of questions. And this is one of the ones that I've said that usually the flu of questions come. So privacy, security, regulatory. All right? So let's hit each of those in turn. So in terms of privacy, the way that this is implemented, and this would definitely be a question to ask if you're thinking about implementing another solution with another EHR. But for us, the transcription that's coming back, of course, all the communication back and forth to the cloud is encrypted. Once the note is sent to the EHR, everything is wiped off the phone, everything is wiped off the cloud. So there's the only record of the whole conversation is the note itself, because that's what you're responsible for. The audio is gone, everything. So it's all encrypted. That makes it HIPAA-compliant, the whole process is HIPAA-compliant. You can, if you get interrupted or whatnot, it'll store it on your phone long enough until it reaches that send a note. But even if you let too much time pass, and that can depend on the practice, it'll wipe it even at that point. And even on the phone, of course, everything is encrypted at rest and serviced the same way. So that's privacy. Security is also relatively tight, because the app has its own security, and the cloud servers are not accessible to the outside as well. So that's usually not an issue. Regulatory is a very interesting field. So there's a lot of regulatory activity around AI, as I'm sure you've heard, at the federal level, the state level, the local levels. So we know that regulation is coming. We obviously, as an industry, the EHR industry, is advocating that the regulators try to keep the benefits of AI while paying attention to the risks, which can be significant. So to my knowledge, at this point, the only state, the only district, or the only jurisdiction that actually regulated medical AI is Colorado, and it has nothing to do with this. If you're using an AI chatbot in Colorado, you've got to tell the patient that it's an AI. It's not a person. So but your individual compliance departments and lawyers are going to get into this and whatnot. And I think as executives, my advice to you would be, for many physicians, maybe not all, many physicians, this is profoundly changing the time that they have to do all day long. And especially for orthopedics, that's not your favorite part of the job, is writing your note, right? And so I would really try to not let the lawyers and the compliance people kill an initiative to at least get this into their hands to try, see if they like it, see how useful it is, and then handle things. And in terms of introducing it to the patient, and this would be easy to make a macro, I just say, hey, it's great to see you again. I'm going to use my smartphone and a little AI to help me write my note. Is that OK? And they say yes, and we're good to go. And I've never had anybody refuse. I've never had a patient complain about it. Most are very interested in it. They're very kind of like, wow, that's cool. And I think it's also common practice for any dictation software. A lot of physicians will put, there may be typos. This was generated by voice-to-text software. In my mind, this is no different. There may be typos or errors. This was a AI-generated document. Yeah, I think, I mean, we'll maybe towards the end get into some other kinds of thoughts on AI. I think this one is primarily just designed to help physician burnout, physician efficiency, removing some of the administrative burden. The unfortunate truth is that that administrative burden, if you're a physician, you know, practice owner is going to get replaced with some other crap, right? So that AI can't help you with. But I do think if you're running physicians and trying to recruit them and maintain them, to be able to offer a system that is as painless as possible to do the administrative work is the goal. And I do think there are other areas where AI can intersect and help remove some of that burden. Yeah, that's a common one. I think patient education is a big one in ortho where, you know, you could set up your own. Because patients, they ought to have a zillion questions about this fact that you're going to cut them open and do stuff to them, right? And it's a better way to offer them reliable answers to those questions as opposed to Dr. Google. Other thoughts on AI? I think just from, you know, for my example, like I'm in a small private physician to own practice, right? So cost is a very big concern for us. We're always looking at how do we do more with less resources? Part of that is just physician time, right? If I spend five minutes a day per patient doing charges, dictation, everything, which is not an unreasonable estimate for 45 patients, that's three hours and 45 minutes a day. When I first saw that number, I was like, that can't be right. I'm not spending nearly four hours of my day dictating, but I used to dictate through lunch every single day. I used to eat as I dictated. I used to stay for an hour after clinic every single day. I don't do either of those now, and I'm seeing five to seven more patients every day just because the sheer time saving. So and we're saving on our transcription costs. We're not paying a scribe in person anymore. I love the co-pilot phrase. That's how I envision this and like it's perfect adaptation as a co-pilot that is with me in the office that just does all the burdensome tasks for me. I walk into a room, tap a button on my wrist that I'm seeing this patient, the notes, the billing, everything is done. I click a button, it's over. I go to the OR. I'm trying to get the perfect angle on an implant or do something. It's looking at the camera. It says, hey, raise your hand five degrees. That's where I envision it going is just being the co-pilot assistant. You wouldn't have an AI fly a plane by itself yet, but you would have a pilot with an AI saying, hey, you're too low. Pull up. The wind's coming from a cross angle at 12 degrees. You need to accelerate another whatever. I see it being no different than that in medicine. And from my standpoint as the technologist, this is where, again, bias admitted, but where the technology that's deeply integrated with the EHR gives the opportunity to look at all of the challenges, right? Not just writing the note, but getting the orders and talking to the patient and educating the patient and getting the consent and all the other, like everything that slows people down or is friction in that patient experience and in the provision of good care is up for grabs if you're fully integrated. So it's nice. It is. Again, this is the recording is erased as soon as the transcription is done, which is you see that is basically within seconds of the sound stopping. And that's flushed. And then the transcription, which is used to bake the note, is also flushed and deleted as soon as the note is sent. But you're right. You don't want the recording around because, well, there's a variety of reasons. We're going to have to get into it. But that's been the general practice. This, it's still the physician who is signing off on the note and who hopefully has read and edited the note so that it does reflect an accurate capture of that encounter. So ambient listening as a technology is being used in some larger practices. There's a few articles out there that are not just straight from the vendor that you can see. And if you want to give us your card, we can provide those references. But it is certainly on the cusp of being able to. And it's like all technology. The way I like to describe the way I see where we are in this is it's like the early days of Uber. And in that way, I mean there's a hesitancy among physicians to adopt this. I mean some are like, I love technology. This is obviously because they're all over it. I'm like Dr. Norton, which is great. Others are like, I don't know. I'm going to let some other people get into strangers' cars before I do, right? But I do think as soon as people start to see the advantage, and just like with a lot of other industries, there's going to be a separation between people that embrace AI, learn its weaknesses, learn its strengths, and those that don't. And in terms of productivity, happiness, work-life balance, whatever it may be, I think that separation. And organizations are going to realize that. Physicians are going to move from organizations that are hesitant to those that are more embracing, in my opinion. Yeah. We were actually talking about that, like what other unexpected benefits have come from this. Besides just me getting home early, actually eating lunch, seeing more patients. I mean patients are waiting less. My wait times have gone down in the office. So more people are cycling through, waiting less time to be seen. I'm spending more time per patient and still saving time. So patients are happier because I don't feel rushed to get out of the room because I've got to go, go, go. When I used to get slammed, I used to stop dictating. I used to just scribble little notes and I'd have this massive stack of paper. Now I don't even have to worry about that because it's just happening in the background. So those days are gone. Even if I don't have time to edit the note immediately, it's going to save that note until the end of the day. It's there. If I'm really behind, I'll just send it to the EHR unedited and realize I'll come back and edit it in the EHR and then it becomes part of the note. So it's a flexible kind of workflow. I think it's certainly possible for some scenarios. I mean it may pre-populate things. I really see it, especially for family practice and those kind of settings, of staying on top of your guidelines. Hey, this patient's overdue for this. You may consider this for this encounter. Kind of going back to that co-pilot scheme that we talked about for all professions, it'll help you with those guidelines and the best practices so that you're staying in that sweet spot of doing the best thing for the patient. Yeah, I completely agree. And I think if we especially consider that the patient may be interacting with your practice online, providing information before you see the patient. So if they're starting to answer questionnaires or they record themselves talking about why they're there to the doctor, you can see AI using these things. And AI summaries of medical visits is still further behind this, but we're working very hard on that too. As a family doctor, there's 100 pages of hospital records. I'm like, please summarize this for me. But also to just summarize the intake questions and be able to pull out what it's learned over time, the provider cares about for that. And then from there, it's not too big a step to say, hey, you didn't seem to do this hand exam, but they had symptoms that were suggestive of whatever. Not that you would do a hand exam, but you know what I mean. And so yeah, I don't think we're too far from that. But I do think it's going to be a while before people trust the AI to make the diagnosis. And the other thing is that depending on your specialty, the diagnosis is not always the reason why the patient is there, especially in family practice. And so there's a lot of other healing that goes on that I do think it's going to be difficult for the AI to take over in the near future. It's its own large language model. We have a dedicated large language model. Now, we do have a partner in this. So it's their large language model, but it's not a publicly available model. No, the partner is called NABLA, N-A-B-L-A, but it's not open source. It's completely proprietary, closed, secure, et cetera. Each section has a copy forward button. I use it all the time. You just go, boop, and it just pulls it in. And frequently, I'll merge the copy forward with a new note. So it'll put the old stuff on the bottom and the new stuff right on top. And you end up with little sequential paragraphs doing it that way. So that's the text of the old visit that's copied into the text of the new visit in the note. But you can bet that we are working like gangbusters to get the meaning of the old visit into the AI as though it was said out loud and then begin to suggest. Because that highly complex internal medicine family practice type of care, there's just a lot of stuff that's missed. At least our volunteer patient was on target. But in my exam room, the patients are all over the place. And it can be trouble just to get them to rein in long enough to go over everything. But again, that gets into AI summarization and beginning to suggest, hey, you forgot this and whatnot. So coming fairly soon for ours, we're doing easy suggestions like, what's the ICD-10 codes that we think we heard out loud? What are some of the lab and radiology orders? Those are all ready to go. If you approve them, it'll push that stuff to your EHR. You don't even have to do that parts in the EHR. And there's more of that coming. But the content of having the AI help with longitudinal care is still researched at this point. But I do think that we're going to crack that nut. So we, as NextGen, have not thought about that yet. But if you ask me what is an excellent use of AI in health care, that is definitely one of them. Because if the patient is carrying a smartphone in their pocket as they are moving around, there is an extraordinary amount of data about their gait and their spine and their back and their sleep and everything else that you can learn from that. And to mirror that with formal patient-reported assessments and whatnot, there's no doubt in my mind that people are going to turn that into useful monitoring of the patient's results after a procedure or before a procedure. Yeah, I didn't mean to be provocative. I just meant that many times the diagnosis is already known, you know, and it's the interactions with the provider. Because a lot of family, I mean, I would say that 30 percent of what comes through my door as a family doctor is behavioral health. It's mental health, right? It's presenting in a physical way, but it's not that. So that's all I meant by that. And so the AI, you know, telling me that the patient has depression, it's not helpful. I mean, the patient knows the patient has depression, right? And so that was all I meant by that. But it's important. It's extremely important in family medicine, absolutely. If you don't address the whole person, then you are not going to get good outcomes on their diabetes or their hypertension or their heart failure because it all is one, we're all one people. So I've been using this, like I said, for 18 months. And that's since basically CHAT-GBT-4 came out. Again, our model is proprietary, but it's the same generation as that model. Even since then, it's extraordinarily better. And another thing is that our patient interaction, they kind of stayed on target, but if you know your patients, you're going to talk about their kids and your vacation and it just completely ignores all of that because it knows that that's not relevant to the conversation. And its ability to distinguish what is important and not important is getting better. And as the feedback, because you can individually give feedback on certain notes in there and it doesn't save the note, so you can't, you know, there's no PHI in there, but you can tell it it was good or it was bad in this regard. And as that begins to influence the models at scale, that's also going to be another, you know, leap forward in terms of their utility, if not their accuracy. Because if you let five doctors, you know, listen to a recorded conversation, you're going to get six different versions of that conversation, right? So there's no right answer for what the AI ought to be producing as the note. What really matters is what does Dr. Nord want to get him out of the office, but also capture what he needs to capture.

ambient AI technology

orthopedic

family practice

physician burnout

medical notes

EHR system

patient care