Subscribe to The Podcast by KevinMD. Watch on YouTube. Catch up on old episodes!
Urologist David Canes discusses his article, “Why AI scribes are changing medicine and the hidden risks you must know.” He shares insights from using AI scribes in over 3,000 clinical encounters, noting how they save time, reduce burnout, and enhance patient interactions. David also highlights challenges like technical glitches, over-reliance on AI, and potential documentation errors. He suggests clinicians always review AI notes and keep manual skills sharp to balance benefits and risks effectively.
Our presenting sponsor is Microsoft Dragon Copilot.
Want to streamline your clinical documentation and take advantage of customizations that put you in control? What about the ability to surface information right at the point of care or automate tasks with just a click? Now, you can.
Microsoft Dragon Copilot, your AI assistant for clinical workflow, is transforming how clinicians work. Offering an extensible AI workspace and a single, integrated platform, Dragon Copilot can help you unlock new levels of efficiency. Plus, it’s backed by a proven track record and decades of clinical expertise and it’s part of Microsoft Cloud for Healthcare–and it’s built on a foundation of trust.
Ease your administrative burdens and stay focused on what matters most with Dragon Copilot, your AI assistant for clinical workflow.
VISIT SPONSOR → https://aka.ms/kevinmd
SUBSCRIBE TO THE PODCAST → https://www.kevinmd.com/podcast
RECOMMENDED BY KEVINMD → https://www.kevinmd.com/recommended
Transcript
Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast. Today, we welcome David Canes. He is a urologist, and today’s KevinMD article is “Why AI scribes are changing medicine and the hidden risks you must know.” David, welcome to the show.
David Canes: Thank you, Kevin.
Kevin Pho: All right, so tell us a little bit about yourself and then the article that you wrote for KevinMD.
David Canes: All right, so I am—I’m about 50 years old, just to let people know where I am in my career. I’m a urologist and the first doctor in my family. I grew up mostly in New England. I always knew in the back of my mind I wanted to be a doctor ever since I went to the pediatrician as a young child, and I fell in love with surgery.
And I’ve been in and out of burnout a couple times in my life, and improving my workplace efficiency is what personally got me out of it. And so that’s why I became interested in AI scribes and adjacent topics. All right. And your KevinMD article is titled “Why AI scribes are changing medicine and the hidden risks you must know.”
Kevin Pho: Full disclosure, the presenting sponsor of this podcast is DaxCopilot, which is an AI scribe company. I just want to disclose that. But tell us what your article is about.
David Canes: So, I’m somebody who would have considered myself very efficient in clinic before I even introduced scribes. All right? I had my EHR shortcuts, dot phrases, order sets, templates dialed in. And if you’d asked me a year ago, “How efficient are you navigating around in clinic and using the computer?” I would have said, “I’m perfectly fine, thank you.” But when ChatGPT came out, I became completely enthralled. I could see it as a world-changing event, and I’m an early adopter personality.
And so I dove into AI, and the natural extension of that was, why don’t I try AI scribes? And initially, I was incredibly skeptical. My take on it was, you know what, my templates are just fine. I don’t want a fully faithful representation of my visit, and it’s not going to get the notes exactly how I want it. And so I’m all set. But after trying it—and I tried about eight to ten different companies; by the way, there are 70 companies out there offering AI scribes right now—my view on this pivoted completely. And let me tell you just briefly why.
First of all, my cognitive load in the visit went down, and I noticed this because at the end of the day I was less exhausted. I felt like I was paying more attention to the patients—really being a good listener—because there was no running subtext in my head that went something like, oh, I better remember to write that down, I want to make sure to remember this, I want to make sure to document that, where you kind of remove yourself mentally from the conversation you’re having. That went away.
Because once you try a few notes, you realize nothing is going to be missed. So I felt like I was better able to be the kind of doctor that I wanted to be in the first place, someone who listens and who’s present in the moment. And then I was saving at least two hours a day. I used to take my lunch break in order to catch up on the notes from the morning. And then I would use 5 to 6 p.m. to catch on the afternoon notes. And that documentation time went away. And, you know, were the notes exactly in my voice? Maybe not exactly, but at this point, the trade-off to my sanity, my work-life balance—at least for me personally—was very clear. This is how I’m going to start documenting. I abandoned my previous workflow completely.
And then, you know, the inspiration for writing this article was, I would just love to share my experience with as many doctors as possible in case there are others who might want to try it out and might benefit from it.
Kevin Pho: And you mentioned that some of this decrease in cognitive workload brought you back from the throes of burnout. You mentioned that earlier, right?
David Canes: Yes. So my strong feeling on this—I think fundamentally if we could go back to when we were all undergraduates and think about when you decided that you wanted to become a doctor. For me, and I think for most listeners, it was something about the patient interaction and patient care that felt really special, really meaningful, and worthwhile. So if we can shape our workflows so that we move from one meaningful patient interaction to another, I think there’s no single solution for burnout, but I think it’s a step in the right direction. It just made me enjoy my work again, focusing on patients, giving them my attention, establishing rapport, and then not scrambling afterwards to document it all. Instead, I’ve become a willing proofreader, and it’s easier to be an editor than it is to be an author.
Kevin Pho: Now, what about the experiences of your colleagues also using AI scribes? Are you an N of one? Have your colleagues also noticed a day and night difference when it comes to that decrease in cognitive load after using the AI scribe?
David Canes: Yeah. So in my health system, I sort of became—you could call me the Johnny Appleseed. I became vocal about it, and I told my colleagues, and it spread to the point where ultimately the health system decided to bring on an AI scribe for all the doctors. We happened to settle on Heidi Health. They’re all excellent, and there are a lot of great choices, but what I’ve found is there’s no universal workflow. There’s not going to be 100 percent adoption of anything. If you take something as simple as Dragon dictation, you can dictate at 160 words a minute, and most people can’t type better than 40 words a minute. And I still walk up and down the hospital hallways, and I see people pecking at the keyboard with a Dragon mic sitting there not being used. So there is no 100 percent uptake of anything. But when you try an AI scribe for the first time, and you see a note render, everybody has a sort of jaw-dropping moment like, oh my God, this is incredible. It’s pretty close to what I would have written, and guess what? I didn’t have to do it. So it’s not universal, but it’s pretty close that colleagues see that this is something that they might want to at least try. But, you know, there are some pitfalls. I think I’ve had 000 patient encounters now using an AI scribe, and so that was the inspiration for this blog: things can happen. Should we go through some of them?
Kevin Pho: Absolutely.
David Canes: All right. So there are occasionally technical failures. There’s a new paranoia that comes up, which is, oh my God, did I forget to hit record? You may occasionally go through a visit, glance at the scribe interface, and realize, oh God, I forgot to hit record. You can either let this drive you crazy or sort of take it in stride. You got to go back to the way you used to dictate for that patient or use your templates or whatever, and not let it derail your day. We’ve all been through technical failures. The EHR is glitching. There’s downtime of your EHR. It’s frustrating. I think a lot of this is a mindset: this may happen occasionally; the back end of the app is temporarily down, and I try not to let it get to me. So that’s one thing that can happen.
Another pitfall is as you start using AI scribes, you may notice that you’re barely editing the notes because they’re so good, and that’s a trap because you may occasionally decide, you know what, I’m just going to paste it into the chart and not even proofread it. And you can’t do that because occasionally it will put something in there that wasn’t exactly what happened, and you know how sometimes we filter out something that the patient says because we’re not really assessing it? It may include that, and there may be some medical-legal issues where you didn’t think it was important in your clinical judgment, but now it’s in there and you haven’t addressed it. So it requires, again, a little bit of a mindset change, and I think depending on where you are in your career, these little aspects of AI scribes may either be OK with you or a deal-breaker.
Kevin Pho: So I want to underscore that these AI scribes and large language models, they’re just not fancy dictation systems in terms of creating that note. It really synthesizes what is recording and comes up with something that is pretty close to intuitive in terms of what a physician would want. So for those who aren’t familiar with AI scribes, I just want to differentiate that from something you mentioned earlier, like a Dragon dictation system, which kind of verbatim records. This is not that, right?
David Canes: Yeah, that’s a good point. It’s not a transcript of your session; it’s not like a stenographer in a court case where it is a faithful representation of everything everyone said. That’s not what it is. It’s outputting a structured note, much in the way that you would as an abstraction from what happened in the transcript and whatever context you’ve given the scribe.
Now, some of them, including Heidi—or I should say some of them—will let you really get detailed with how you want your templates to come out, and that’s where you can bend the notes to be more exactly how you want them. But it’s not a transcript, which to me is a good thing.
Kevin Pho: You mentioned about those rare cases where the note isn’t exactly what you want and can sometimes lead to some medical-legal consequences or have inadvertent wrong information in the chart. Anecdotally, have you heard any cases whether in your system or just from your colleagues where that actually has harmed patient care?
David Canes: I haven’t. This is all theoretical. Since we’re on the medical-legal topic, I will say a lot of objections from doctors who hear about scribes is, one, I don’t want this transcript to be out there and released to the legal system. That, I think, is a far-fetched concern. None of the AI scribe companies would want that to occur, because it would ruin the industry. That, I’m not worried about.
But I’ll give you an example in urology—and again, I don’t know of any cases like this. Let’s say a patient tells me that there’s blood in the urine, and on more further questioning, it’s more that the urine was just very concentrated and they were dehydrated. So I’m not going to assess, OK, thanks, gross hematuria, because in my estimation there just wasn’t any. Now, what if in the AI scribe scenario, it puts the patient’s voice in there as having blood in the urine, and I didn’t proofread it carefully, so I didn’t notice, and I left it in there. Six years later, the patient develops a bladder tumor, and so that note now is a potential Achilles’ heel for me. To me, that would be a proofreading issue. You’ve got to proofread these notes. You’ve got to make sure that what’s in there is what you intend to be in there. We’re constantly filtering what we want to be in there and what we don’t want, and the AI scribe may not always make those choices in the same way that you would.
Kevin Pho: In your article, you talk about copy-pasting errors that can be exacerbated by scribes. So, talk more about that.
David Canes: Yeah, so most AI scribes, in my experience, at least when you first sign on, if there’s no formal integration into the EHR, you’re going to be copying and pasting from the AI scribe software into your EHR, which is, by the way, the way I do it now. I don’t find it to be difficult or onerous or annoying. It’s just part of the workflow—a couple of clicks—but it does pose an existential threat that you’re going to paste it into the wrong chart. And so you need to be aware of that. You need to put in some double checks, and it’s now part of my workflow: every time I paste, I do a pause and I check the name that’s in the note to the name that’s on the chart, just as a fail-safe to save myself from making that kind of mistake.
Kevin Pho: Do you ever worry that physicians are going to lose their manual documentation skills? I’m actually not that much older than you, just a year or two older than you, so when we were in medical school, right, we always had to handwrite the patient’s story and handwrite the assessment and plan. Then we went to electronic medical records, where we either typed it out or dictated it. And now we have these ambient AI scribes, which really capture the patient’s story without us really having to narrate it or write it down. So how do you think that’s going to affect the physician’s manual documentation skills?
David Canes: That’s a wonderful question, Kevin. Let me tell you my take on this. My answer is no, I’m not worried about this, but let me tell you why. With every major advance in the history of humankind, really—if you take the Industrial Revolution and the introduction of factories and mass manufacturing—we were afraid that people were going to lose their manual skills on the assembly line and that kind of thing. The old skill often just doesn’t matter anymore. It’s the same thing in surgery: when we go from open surgery to laparoscopy, hardly any trainee can do an open gallbladder removal now, but it doesn’t matter because it’s no longer a relevant skill.
What makes me most excited about the move toward ambient scribes and documentation this way is that it frees—what does it free you up to do instead? It frees you up to do more thinking about the patient in front of you. And actually, in my stats that are recorded in Epic, my time in the chart has gone up even though my documentation time has gone down, which means I’m going through results more carefully. I’m reading through the imaging results and the lab results and other doctors’ notes while I have the patient in front of me, and making more eye contact with them. So it’s freeing up my—again, my cognitive load is now open to do more important things. You know, a generation from now, if doctors don’t know how to write a note from scratch but they’re focusing on doctoring, I think that saves our field. I think that saves our profession.
Kevin Pho: Now, where do you see the future of AI scribes heading? Like you said, there’s kind of a Wild West out there. There are 70-plus AI scribe companies. What do you see in the foreseeable future regarding the technology itself?
David Canes: So my take on that is that in short order, AI scribes are sort of going to be table stakes. It’s like being able to electronically send a prescription. Now it’s just considered something that every EHR has to have. We don’t talk about which one does that better. I think there’s going to be a reversion to the mean where they all do a pretty decent job at the actual note writing. So the next frontier is going to be which companies can capture the downstream tasks that happen after the note.
For example, within the next couple of years, it will be, “I mentioned in the visit that I’m going to refill your prescription for metoprolol and order a CAT scan,” and as soon as I turn to the EHR, those orders are already pended for you to be proofread and signed—AI agents that then start the process of tracking the patient down to make sure they follow up on all those plans that you just instituted. So, to me, the next frontier is going to be the downstream tasks that have to be done after the note is finished. How can we make that easier for doctors?
Kevin Pho: Do you ever worry about administration not being aligned regarding that cognitive load decrease? They simply just see the numbers and say, “You’re spending two hours less documenting. Why don’t you see four more patients instead?” Do you see that there are some conflicting goals regarding the AI scribes?
David Canes: Yes, I am very worried about that. If there’s a call to action from this podcast, I would urge doctors to take action—not an aggressive stance, but more of an active role in being the ones who decide what their own workflow should look like. We’re the ones seeing the patients. We are the ones generating the revenue. And it’s our sanity at the end of the day. There are too many doctors committing suicide, leaving medicine. It’s time for us to say, you know what, this is how it should be. I need this for my own occupational distress level. I get to decide how many patients I see in a day. And if I see two more patients and it’s less work, I’m happy to do that. But you can’t break my back so that I’m back to being burnt out. I want doctors to be enabled to speak up when they feel like their workflow isn’t working right. There are too many clicks in the EHR. We’re a very deferential group of people—God bless us. We need to stand up for ourselves more.
Kevin Pho: We’re talking to David Canes. He’s a urologist, and today’s KevinMD article is “Why AI scribes are changing medicine and the hidden risks you must know.” David, let’s end with some take-home messages that you want to leave with the KevinMD audience.
David Canes: Yeah, so take-home messages for me are: I want to go back and echo what I said about time face-to-face with the patient. That time is what our purpose is, and that’s the time—the face-to-face time with the patient—that makes us feel the most fulfilled. So what I want people to do is examine their workflow and look at bottlenecks, and think of ways for it to be solved. I’ll give you one example that’s near and dear to my heart.
One of the things that I found in my practice is that I got frustrated by teaching patients the basics over and over again and never getting to the personal parts of why they were seeing me. So I started WellPrept. WellPrept is a way of sending patient education to their patients beforehand so that they walk in already up to speed. That’s just one example. AI scribes are another. Look at your workflow and think about what frustrates you, what makes you want to bang your head against the wall, and then fix it. Look for solutions. Speak up, and be an advocate for yourself. There are too many clicks in the EHR. We’re a very deferential group. We need to stand up for ourselves more.
Kevin Pho: David, thank you so much for sharing your perspective and insight.
David Canes: Thanks again for coming on the show.
