Finding the Human in the Machine — Dr. Michael Blackman on AI, Trust, and Smarter Workflows
Download MP3Well, hi, everyone. I'm Daniel Williams, senior editor at MGMA, host of the MGMA Podcast Network. We are back with a familiar guest, someone we've had on the show quite a few times, and we keep bringing him back because he has a lot to share with us about all the things that are going on with AI, with technology. So our guest today is Doctor. Michael Blackman.
Daniel Williams:He's chief medical officer at Greenway Health. We're going to be talking about looking at the human element of AI, and what that really means for medical practices. Doctor. Blackman, welcome back to the show.
Michael Blackman:Oh, thanks, Daniel. It's great to be here. Thanks for having me.
Daniel Williams:Yeah. Now, I've lost count. You've been on for years now. We just keep bringing you back because we've got a lot of great things to talk about. You've shared your background before for our audience, but just assuming we've got new listeners today, just share a little bit about your background, what got you to where you are today.
Michael Blackman:Yeah, so by background, I'm a primary care physician trained in both internal medicine and pediatrics, and relatively early in my career, after residency, the hospital I was working for was putting in an electronic health record. This goes before Meaningful Use, if people can remember back that far, and it was something that interests me and was up my alley based on things I'd done previously, and just continued to get involved from there. So now I'm at Greenway Health, switched over to the vendor side, working on the health IT side, really trying to help us create an EHR and surrounding systems that people really both love to use and see benefit from using. If we think about the history of EHRs, and this would be a different conversation on a different podcast or something else, you know, what went wrong, and why didn't EHRs meet the promise that we all thought they would have? Right.
Michael Blackman:But I think we're now, from a technology perspective, and I don't consider myself the tech guy, in a position to finally meet some of those lofty aspirations we've had for years.
Daniel Williams:Okay. I'm going to guess that most of our listeners know who Greenway Health is. However, just in case some don't, why don't you educate us on Greenway and the kind of things you guys are doing right now?
Michael Blackman:Yeah, so Greenway is an ambulatory EHR provider, EHR and surrounding system, so we mostly work with independent ambulatory practices covering the patient side, the clinical practice side, and then the revenue cycle side as well, and we're in the midst of a transformation building an AI from the ground up platform that we think is really going to change the way people interact with the EHR and really make a difference there.
Daniel Williams:Okay. We are, as I mentioned, we're going to talk about AI. It's a topic we hear a lot about these days. We're also going to talk about the human element there as well. You and I have discussed this some before, but I think it's worth bringing up yet again.
Daniel Williams:We hear AI. We hear the kind of job functions it's already doing. People go, uh-oh, what's going on here? You and I have talked about this before. There are some misconceptions about it, some truths as well, bring us up to speed about where are we with AI, and when we think about it taking on roles or just particular functions within a job that can help supplement someone at a medical practice.
Daniel Williams:Let's talk about it in that context.
Michael Blackman:Yeah, I think that's absolutely true. There are a lot of conceptions, misconceptions, and in all fairness, the conceptions and misconceptions, rather, from a couple of months ago could be different now as we go forward. I think certainly in the healthcare space, there's a lot on the back end in terms of helping automate revenue cycle and things like that, certainly working on task input. We focus more on the clinical pieces, which is where my primary focus is, the AI is an assistant, it's not a replacement. We are not taking the human out of the loop.
Michael Blackman:There are lots of things we can do. We can make suggestions. We can really improve the effectiveness of decision support, but we still need a person in the loop to review it. If you'll let me digress for an old story for a minute, when I was in medical school, and second year of medical school was a pathophysiology class, and at my med school we had long answer questions, and there was a question on the exam about something related to the pathophysiology of HIV. I freely admit at the time I didn't remember.
Michael Blackman:I finished the rest of the test. I go back. It's a test. You write something down. You don't leave it blank.
Michael Blackman:If I was seeing a patient, I would go look something up and get the right answer and everything else, but that's not where we were. So I wrote a page and a half, two pages in the Blue Book for people who remember what Blue Books are. I'm dating myself a bit. When I got the exam back, I got half credit on the answer with a note in the margin from the professor saying, This is really interesting. Where did you read this?
Michael Blackman:I never had the heart to tell her I made it up. Based on some other knowledge and whatnot, and for all I know I was onto something, but I doubt that. But that's the risk of AI, is that it pulls information together that's logical, presents it in a cohesive fashion. Is it true? So you end up with, especially from a decision support perspective, the AI falling into three buckets.
Michael Blackman:Bucket one is, Oh yeah, I recognize that. I just didn't think of it. Yep, that's right. I'll do that. Bucket three is, I recognize that it's ripe in the context of this patient.
Michael Blackman:That doesn't make any sense. Easy to dismiss. Bucket two is what I had on the exam. This is logical, but do you then look a level deeper to check the references, to make sure it's really true, and that it really makes sense? That's the risk we run, which is why it's not only keeping the human in the loop, but teaching the human how to use the AI effectively, and what to trust, and frankly when not to.
Daniel Williams:Yeah, I love that you went old school. Boy, I still remember those blue books. That's how I took my test back in the day, and I also like that you really put it into different buckets, because I've got to believe everybody who's listening, when we feed something into AI, it can give us an answer. If you look like you're almost doing an eye exam, and you look at it from a distance, sort of, it looks really good. Then when you get a little closer and a little bit closer, you start going, That's not quite it.
Daniel Williams:It's an approximate of what I was looking for. Then you start having this dialogue. This is how I do it. I have a dialogue back and forth with AI. You've given me a really nice framework, but it's not what I'm looking for right now.
Daniel Williams:Does that measure with what you've experienced?
Michael Blackman:I think that's true, and it's also a question of what's the associated risk, because if you're using AI to say, What should I do with two days in New York? It can give you great suggestions. You can ask good questions, and the worst thing that happens is you go to a restaurant that you didn't really like. You go to a show that wasn't so great. The implications of getting it wrong in medicine are appreciably higher.
Daniel Williams:Oh, yeah. Let's talk about that overreliance, own AI, and the risk associated with it. Because what I'm finding personally is it is fine tuning my analytical skills, but if we don't bring in that human oversight, we can have some real problems, some real implications here from the healthcare side. Talk about that when you don't have that human oversight, and then what the human oversight can bring to the table.
Michael Blackman:At the end of the day, when using AI, you cannot check your judgment at the door. You have to approach it with complete judgment and review and everything else, and when you don't, that's when you run the risk of being led down the wrong path inadvertently, trusting it too much, and even some of the simple things that Well, they're not so simple, but as we think about some of the AI use cases that are becoming in more widespread use, ambient documentation being one, really lets you focus on the patient, have a conversation, it writes the note for you. Does that mean you don't have to read the note and you should just stick it in the chart? No, it doesn't mean that at all. It has created a draft of the note.
Michael Blackman:So you're moving from being the author of the note to the editor of the note. At the end of the day, you still have to make a judgment about, Is the content what you want it to be? Make appropriate adjustments and then sign it. You can't simply go, Well, AI wrote the note, and I'm not responsible for it. Doesn't work that way.
Daniel Williams:Yeah. Let's talk about critical thinking. That's something that, in corresponding with you and your team leading up to this conversation, We brought that critical thinking into it. I had alluded to it earlier, where I feel like it's helping me be a more analytical thinker. I'm a creative writer, so I'm creatively writing.
Daniel Williams:Then analysis and critical thinking is coming into that, but I'm seeing more and more since AI might give me a draft that it's like a lump of clay almost. Now I'm fine tuning it. Now I'm making it look the way I want it to look, bringing in that critical eye. Talk about the critical thinking factor where that human element, the human critical thinking can come into play.
Michael Blackman:Yeah. The human critical thinking comes in with the governance around it, the monitoring of the systems, and then deciding, even before using AI, deciding where do you want to use it. What is the risk associated with it? Are we talking about something that's a low risk, repetitive decision? Well, those you can really automate, have the AI make good recommendations.
Michael Blackman:If you get to something that's sort of at a moderate risk level, perhaps you're asking the AI to provide you options, and then asking for the evidence behind those options. Putting that in a medical context, I'm seeing a patient, I have a constellation of signs and symptoms, and from that, in my head, I have developed a differential diagnosis. Perhaps I have three or four items in the differential. Well, maybe I ask AI with that same constellation of signs and symptoms, and now it gives me, in addition to the three or four that I came up with, maybe it adds another two that I didn't think of. And then you can look at them back to before, Oh, I really should consider that and make sure that's not a problem, or No, that's just not the case.
Michael Blackman:But it's not saying the patient has this, it's saying Think about this and do appropriate diligence. And then you get to high risk stuff where it's really just providing support. The human's really making the decision, perhaps you're querying it for some background and other pieces. But I think important in all of that, especially when you have questions, the same way teachers used to tell us, Well, especially in math class, Show your work. How'd you get to that answer?
Daniel Williams:Exactly.
Michael Blackman:Let's ask the AI to show its work. How did you get to the answer? What is that supporting evidence? Then more importantly, check to make sure the evidence is correct if you're not certain. I used one of the commercial AI tools recently to help prepare for a paper I was writing.
Michael Blackman:Was a place sort of like that lump of clay. It was a jump start, and then I would take it and turn it. Without exaggeration, the five references it provided, none of them were real. Zero. Oh boy.
Michael Blackman:Now, did I believe what it was saying? Did it make sense? Yes, but when it came down to this paper, this date, this title, nope, that was wrong.
Daniel Williams:Wow. Yeah. I looked it up. Today we're talking last week of June. I looked up We talked earlier back in April, and we keep talking about how AI and technology is changing so rapidly.
Daniel Williams:Just to give context, we're last week in June. Where are we with AI right now as far as the day to day in a medical practice? Where is it hitting the mark, so to speak, and then where is it still falling short?
Michael Blackman:I think the place where it's absolutely hitting the mark is what I mentioned before with ambient documentation. Okay. Using the AI to help craft the note, as I said before, you still need to read it, but it is much faster and frankly better interaction with the patient to build the note from the conversation automatically, as opposed to doing all of that, writing it, and I've had any number of physicians that I've spoken to about this say, You know something? It puts something in the note that I didn't hear, and then they go back and look at the transcript of the conversation, and sure enough, the patient said it. Because we're all human, we were thinking about our next question, we were thinking about how those things fit together, and missed the passing comment the patient made.
Michael Blackman:It catches all of those things and can really provide better care in that way. The next piece is we're starting to see a good job with summaries. You think and querying the chart in faster ways, because if you think about all the data that's in a chart of someone who's had a number of medical problems over a number of years, we're talking about tons and tons of information. At least going back to paper, when you had multiple volumes of a paper chart and they stacked up on the desk, no one really believed that A, you read all of it B, that you could effectively find anything in it. You could if you didn't file correctly.
Michael Blackman:But now that information's in the system. It's like, Well, it's in the system. You must have seen it. Well, intellectually, makes about as much sense as it did saying you knew everything that was in the entire chart, but you can query it. You can ask the right questions.
Michael Blackman:It can help surface information. You may not bring information forward based on the patient's conditions that, Hey, they also have this, or These tests have been done. So there's that piece as well. Then obviously, as we continue to improve on task management, whether that's messaging back and forth to patients, pharmacies, interoffice, etcetera, and helping organize those.
Daniel Williams:Okay. I'm pretty sure we talked about this back in April, but I think it's worth bringing up again, and that is technology overload. Because we see the tools that are being put in front of us now. They're changing and evolving so rapidly, and they're pretty remarkable. There's some really cool stuff that can also go like, I just feel like you know, there was that old ad dating ourselves again where it was a I think it was a Bose stereo system, and it showed a guy in a reclining chair, the noise was just blowing his hair back, because that the sound system was coming at him.
Daniel Williams:Sometimes technology feels like that. It is just overwhelming, but also this incredible liberating aspect to the technological advancements that are being made. How do we adopt to it? What's a good look at adoption? I know it's different for each person, but what does a good AI adoption look like, and then just this, just too much right now all at once.
Michael Blackman:Yeah, I think one, you have to pick the problem you're trying to solve. So what problem are you trying to solve? What type of AI or what appropriate AI tool or other tool is appropriate to help you solve it? It shouldn't be the reverse. It shouldn't be the tech looking for a problem, because you may end solving the wrong problem.
Michael Blackman:Second, it has to be integrated in the workflow. Yes, you're absolutely right. The tech is coming from any number of ways. There's always something new. Brand new this week is going to be old two months from now, but you need a workflow consistency.
Michael Blackman:You want to make sure these tools are embedded appropriately in the workflow so they're not distracting, so they're not sort of causing extra cognitive burden as you navigate from A to B to C, which is why here at Greenway, we're really stepping back and really trying to re envision this from the ground up, thinking about AI from the start, so it's in the right places in the workflow, and it's not an add on or a bolt on here or there.
Daniel Williams:Another thing that you've talked about previously is the impact on the care team. You've talked about healthcare being a team effort. What changes when AI becomes part of the workflow? How become does a team member, so to speak, if we can use that analogy?
Michael Blackman:Yeah, it absolutely becomes yet another member of the team, which can help improve the communication, help improve documentation, but at the end of the day, it's a new way of working, Similar if you've added new members of the team, you would have new ways of working. So you have to be cognizant of that, cognizant of how the tools work and how they don't. I know I keep coming back to the ambient documentation, but even that's new ways of working. Because, this may sound sort of obvious, but if you want the information to be in your note, you have to vocalize it. There are plenty of things that we used to do without vocalizing them.
Michael Blackman:Didn't speak through my physical exam as I did it with a patient. I just kept that in my head, then I wrote it down. Now, if there were parts I wanted to discuss with the patient, of course, but didn't routinely say all the pieces that I would put in the note. If you want it to work that way, well, and there are other options, but you need to vocalize it. So that's sort of a different experience for both the provider and the patient.
Daniel Williams:Okay. Got two more questions before we sign off today. One is on burnout incapacity. It's not a secret to you, to this audience, about the epidemic of burnout that we have in healthcare. That's one of the aspects of AI and other technologies, that they can lessen that load, that they can take some of that administrative burden.
Daniel Williams:Where are you seeing AI and the other technologies giving people time back? Where is that impact just being realized already?
Michael Blackman:It's anything that's sort of automating the routine and taking out steps, or speeding up actions. If the system, using AI or tools, can tee up information for you that you don't have to go find on your own, that adds a little bit of time back. The documentation adds some time back. The ability to appropriately prioritize incoming messages and whatnot, again, takes some time back. There's no silver bullet here, but it's incremental pieces across the board that also let people take out what some people would describe as the drudgery of their job.
Michael Blackman:I've heard plenty of people describe writing progress notes after seeing a patient as an administrative task. Don't want get rid of my administrative tasks. Personally, I don't think that's an administrative task. Core to the provision of care is writing a note. It also helps synthesize the information and put it in a way that's understandable, not only for you, but for people downstream who may read it.
Michael Blackman:So it's not administrative action, but it's still a very time consuming one. So we can continue to make that better. But at the end of the day, it's little bits at a time and then how people want to get that time back. And also, by taking some of those routine pieces out, lets people focus on perhaps the more cognitive challenging pieces of their job, sometimes the more interesting pieces. The old saying in healthcare, Let people work to the top of their license, this really helps enable that.
Daniel Williams:That flows perfectly to the last question. That's the patient experience. Ultimately, it's to provide care to those patients. How does AI impact trust, communication in any meaningful way? What are you seeing, and are you able to measure it?
Michael Blackman:I think the place where we've really seen it is some of the improvements in documentation and other pieces that make it clearer to patients when they read them as to what's happening. Even using AI to help generate patient instructions at an appropriate reading level based on the conversation. Again, still require human review, but here's a nice summary for the patient, here's your instructions summarized in a way they can really take advantage of, and that helps improve compliance and other things. Now there are people who aren't trustful of AI. They hear AI and they go, Oh wait, I don't want this involved in my care.
Michael Blackman:So I think it's incumbent upon us to talk with patients about what it is and what it isn't, and how it's being used, and make them understand and have a say in how the AI gets used, at least for now. That helps to simply build that trust that you would want to build anyway in that doctor patient relationship, and you certainly wouldn't want the AI to erode that. What I mentioned before, tell a patient, Listen, I'm going use this tool, and it lets me focus my time on you, and not on taking notes, and not on writing things down. Yeah. Most patients really like that.
Daniel Williams:They do. I know I do. When I'm in that patient chair, I do like someone to address me, know who I am, understand what's going on, and be able to communicate in that way. So for any of our listeners then who want to know more about what Greenway Health is doing in this space, do you have a particular website link, we send them straight to the main page on Greenway Health? How would they get in touch with you guys?
Michael Blackman:Yeah, the main page on Greenway Health would be perfect, which is greenwayhealth.com, and there they can see what we're doing with our new platform that we call Navarre, which is a new EHR platform built from the ground up with AI as the thought from the starting point, we have appropriate workflows and everything else, and we're really stepping back to say, Where can it be used appropriately? What can be effectively automated? And then doing that. But always, always leaving the human in the loop.
Daniel Williams:Okay. Doctor. Blackman, it's always such a pleasure to have you. I'm going to let you go for the weekend. Do you have any big plans?
Daniel Williams:You got a big hike or anything else coming up this weekend, just working? What's going on this weekend
Michael Blackman:for you? This weekend, I think, is gonna be more working than not.
Daniel Williams:Okay.
Michael Blackman:I have stuff to catch up on.
Daniel Williams:Okay. Are you into the World Cup at all? It's, we're getting into the knockout stages. Do you do anything with that?
Michael Blackman:It's it's been it's been fun, the little bit of it that I've seen.
Daniel Williams:Okay. Okay. Well, it is always a pleasure catching up with you. Everyone, we will put direct links, as doctor Blackman was talking about, in our show notes, also in an article we're gonna put on mgma.com. Until then, I wanna thank all of you for being MGMA Podcast listeners.
