Latest / The Clinical Realist / The Other Side of the Table: What Academic Health Systems Actually Need From Health Tech

Transcript

The Other Side of the Table: What Academic Health Systems Actually Need From Health Tech · The Clinical Realist

• 0:01
  SPEAKER_00: If you've ever sat across the table from a health tech founder pitching their AI product to a major academic medical center, you know the look.
• 0:08
  SPEAKER_00: They came in believing that the hardest part was already behind them.
• 0:11
  SPEAKER_00: The science checked in them, the demo was clean, and then the meeting ended and nothing absolutely happened.
• 0:16
  SPEAKER_00: So I wanted to bring someone on who's on the other side of that table.
• 0:19
  SPEAKER_00: Not a critic, not an observer, someone who's actually running one of the largest academic medical centers in the county.
• 0:25
  SPEAKER_00: Actually the country, and deciding every day which bets are worth making.
• 0:28
  SPEAKER_00: So Dr.
• 0:29
  SPEAKER_00: Marshall Rungi has been the previous Dean of the University of Michigan Medical School and previous CEO of Michigan Medicine.
• 0:36
  SPEAKER_00: And today he's actually doing a lot of great work in cardiology, including work with Indian Health Services and the VA.
• 0:43
  SPEAKER_00: So Michigan Medicine treats more than five million patients across one of the most complex health systems in the country, with a research enterprise, a medical education mission, and a clinical delivery mandate that has to coexist.
• 0:55
  SPEAKER_00: And sometimes they actually could be every single day.
• 0:58
  SPEAKER_00: So Dr.
• 0:58
  SPEAKER_00: Rungi is a cardiologist by training, has published hundreds of papers, a novel, which I just found out about, a nonfiction book, and he has led major institutions through the kind of tech transformation that most consultants write about from the outside.
• 1:12
  SPEAKER_00: So today we're going to talk about what academic health systems actually need from health tech and what most vendors get completely wrong before even get to the contract and where the honest gaps really are too.
• 1:24
  SPEAKER_00: So we're going to talk about in adoption, in physician authority, and who actually gets reached.
• 1:29
  SPEAKER_00: So, Marshall, welcome to the clinical realist.
• 1:32
  SPEAKER_00: Thank you, Sarah.
• 1:33
  SPEAKER_00: It's great to be on your show.
• 1:34
  SPEAKER_00: Great to have you.
• 1:36
  SPEAKER_00: It's such a treat.
• 1:37
  SPEAKER_00: So today I want to start with something that I think will be useful to everyone listening who's ever legitimately trying to get a product or a service into a system as complex as Michigan Medicine.
• 1:48
  SPEAKER_00: So what do you want to tell health tech founders who think they just need to get into a major academic medical center?
• 1:54
  SPEAKER_00: What do they actually not understand about how these institutions work?
• 2:00
  SPEAKER_01: Well, uh, first of all, uh let me just say it's an exciting time with uh very innovative people, very entrepreneurial and innovative people who are uh potentially bringing AI solutions uh to healthcare.
• 2:13
  SPEAKER_01: Uh what and I talked to so many of these folks in the last, particularly in the last several years, uh some were our faculty, but many were not, most were not.
• 2:24
  SPEAKER_01: And uh they brought some really fascinating uh potential uses of AI.
• 2:31
  SPEAKER_01: What they did not realize is while they were developing that, I mean this is my opinion, but what they didn't realize is that um it's almost impossible to do bolt-on programs.
• 2:43
  SPEAKER_01: And what I mean by that I understand that.
• 2:45
  SPEAKER_01: You know, we're we're run by we're we're owned by Epic right now, and Epic uh fiercely uh prevents any adoption of any external programs.
• 2:56
  SPEAKER_01: And so we can't, we we just can't adopt anything without ourselves having to maintain it.
• 3:04
  SPEAKER_01: Uh and that we we we went through that phase more than 10 years ago, uh before way before AI, and found it was just hopelessly expensive and complicated because in a year what was perfect a year ago is no longer fitting quite as well, and so it requires tremendous programming support.
• 3:22
  SPEAKER_01: And uh we spent a lot awful lot of money on that and and time, and and it was frustrating.
• 3:28
  SPEAKER_01: And so people will come with these great ideas and and I tell them that, and they're like, Yeah, but you don't understand my idea, it's so great.
• 3:34
  SPEAKER_01: And I say, I understand your idea, it is great, but you need to understand that uh you need to partner up with somebody before you come and give them a pitch and say how how can we work with you?
• 3:47
  SPEAKER_01: And generally with us, uh there are two major uh EMRs.
• 3:52
  SPEAKER_01: We use Epic.
• 3:53
  SPEAKER_01: Uh Epic is uh difficult to uh get to adopt external vendors.
• 3:58
  SPEAKER_01: Uh Oracle is very different.
• 4:00
  SPEAKER_01: I mean they they they're come all, bring all, and uh I think it might be easier for some of those folks to work with people that use Oracle as their primary uh electronic medical medical record than Epic.
• 4:13
  SPEAKER_01: That's where I come from.
• 4:15
  SPEAKER_01: And so I appreciate that, yes.
• 4:16
  SPEAKER_01: So disappointing to them because I mean they really do have some ingenu ingenuous products, but I'll bet oh if if we took one out of a hundred that were pitched to us, I'm I'm I'm trying to think what that one was.
• 4:31
  SPEAKER_01: Uh it just it's just legit and and the IT folks, I mean they don't want to do it either.
• 4:36
  SPEAKER_01: Um so so it's it's that that's the big hurdle, I think.
• 4:41
  SPEAKER_00: I appreciate that.
• 4:42
  SPEAKER_00: Having been on both sides myself and having been at Oracle when we acquired Cerner and seeing kind of this.
• 4:48
  SPEAKER_00: Oh, really?
• 4:50
  SPEAKER_00: So yeah, I assisted with the acquisition and seeing kind of the spaghetti on the back end, um, I know is one of the major pieces that Oracle Health had done was they wanted to be that open um organization.
• 5:02
  SPEAKER_00: But again, a lot of people have Epic and it's been in their system for decades.
• 5:06
  SPEAKER_00: And once it's in, it's very difficult.
• 5:08
  SPEAKER_00: You are you're married to it for sure.
• 5:11
  SPEAKER_00: Um, this kind of brings up something I've been thinking about a lot.
• 5:14
  SPEAKER_00: So when you think about academic medical centers, they kind of have three different jobs simultaneously.
• 5:20
  SPEAKER_00: So you've got teaching, you've got research, and you have to actually deliver care.
• 5:24
  SPEAKER_00: And I think a lot of vendors think they're selling to the clinical delivery mission, but those three missions kind of share the same doctors, the same infrastructure, the same budget.
• 5:34
  SPEAKER_00: So when some of those vendors kind of show up, assuming that they're gonna work with your Epic or CERN or system, whichever you have, and they're pitching those clinical tools, which of those three missions is the most likely to get shortchanged?
• 5:46
  SPEAKER_00: And does the vendor even know it?
• 5:49
  SPEAKER_01: Um, another great question.
• 5:50
  SPEAKER_01: And I and I think um the one it's ironic.
• 5:56
  SPEAKER_01: The one that gets shortchanged the most is the educational mission.
• 6:00
  SPEAKER_01: The reason I say it's ironic is that's the easiest one to get into.
• 6:03
  SPEAKER_01: I know, right?
• 6:05
  SPEAKER_01: Research has its own boundaries, clinical care has its own boundaries.
• 6:08
  SPEAKER_01: Uh there's no uh NIH for uh education.
• 6:13
  SPEAKER_01: Uh there's no uh necessity that those things have to interface with the electronic medical record.
• 6:20
  SPEAKER_01: So it's it's interesting because then I'll tell I'll say, well, you know, we might be able to use this in our educational programs, and they're like, eh, not interested.
• 6:28
  SPEAKER_01: Because that's not where the money is, and it's not where the impact is.
• 6:31
  SPEAKER_01: The the other comment I want to make is uh one that you'll understand perfectly, and I hope your listeners do, uh, which is from a clinical perspective, and uh it has to do with AI advances, and I'm gonna give you one example uh around uh digital imaging.
• 6:49
  SPEAKER_01: So it's been long felt that that's a great strength of AI.
• 6:53
  SPEAKER_01: And I do think it's a great strength of AI, but I'm I think the and so we get hit with that all the time.
• 6:58
  SPEAKER_01: Well, look, we could help you with digital imaging, and and I think where AI is really effective is in uh defining normal.
• 7:07
  SPEAKER_01: Uh where it gets more difficult is defining abnormal.
• 7:11
  SPEAKER_01: And the example I'll give you is oh, about um a year ago, uh there were several papers published in in big time national press about AI and mammography, uh, showing how AI was better than radiologists.
• 7:27
  SPEAKER_01: There's one one study that compared uh uh mammographers in the United States where a single person interprets versus uh in uh one of the European countries where two people interpret, then AI made it even better.
• 7:40
  SPEAKER_01: And that's good, and I like that, but none of those studies that I have ever found actually have a clinical correlation.
• 7:46
  SPEAKER_01: So you'll get this.
• 7:48
  SPEAKER_01: Uh so great, you know, they they find some some abnormalities.
• 7:53
  SPEAKER_01: Is it did it turn out to be biopsy positive?
• 7:56
  SPEAKER_01: Or were half of them biop biopsy negative, and women who had those things biopsied were exposed to risk with no real benefit.
• 8:04
  SPEAKER_01: So great on the negative side, uh, I'm less convinced with even with digital imaging, uh using mammography as an example, that uh the correlation is quite there yet in the in the diagnostic side.
• 8:17
  SPEAKER_01: Because you gotta have more than just an image, you have to have some way to verify that that image is is valuable.
• 8:24
  SPEAKER_00: I like that.
• 8:24
  SPEAKER_00: So and I've seen this too.
• 8:27
  SPEAKER_00: The AI is showing something, and they're pitching some sort of value, but sometimes the economics are off.
• 8:35
  SPEAKER_00: So is it value for the hospital, value for the imaging company, or is it actually outcome value for the patient?
• 8:40
  SPEAKER_00: And sometimes those are harder to find.
• 8:42
  SPEAKER_00: But I do appreciate that because c being clinically significant is different than working as designed.
• 8:50
  SPEAKER_01: Yeah, and and uh you you hit something uh exactly, you hit the nail exactly on the head.
• 8:56
  SPEAKER_01: I mean, at the end of the day, and this is what you say when you speak and in your book, uh, it's about the patient.
• 9:03
  SPEAKER_01: I mean, yes, of course, healthcare is whatever it is today, 18% of the GNP.
• 9:10
  SPEAKER_01: And healthcare systems that don't make money go out of business or get bought.
• 9:14
  SPEAKER_01: So that's a reality.
• 9:15
  SPEAKER_01: We all understand that.
• 9:16
  SPEAKER_01: But on the other hand, uh we want to do things that both meet our financial goals, but uh absolutely are good for patients.
• 9:27
  SPEAKER_00: So let's kind of zoom out for a second.
• 9:29
  SPEAKER_00: Michigan Medicine, Ann Arbor, Michigan.
• 9:33
  SPEAKER_00: The patients who need the most complex, most specialized care often live nowhere near Ann Arbor.
• 9:40
  SPEAKER_00: So there's communities that are 50, 100, 200 miles away that are never gonna get a cardiologist like you on staff.
• 9:46
  SPEAKER_00: So what's the actual clinical infrastructure being built to reach those communities?
• 9:51
  SPEAKER_00: And where does AI actually fit into that picture?
• 9:54
  SPEAKER_00: And honestly, not aspirationally.
• 9:58
  SPEAKER_01: Well, uh this was something we started facing way way before AI was on the scene, and that is we were a huge uh uh academic medical center in a tiny town.
• 10:08
  SPEAKER_01: Um and a thousand beds and five million outpatient visits, as you mentioned, a year, five million patients, it uh two million outpatient visits, but five million patients we cared for, and many were from across the state.
• 10:23
  SPEAKER_01: So our first goal was to develop partnerships uh across the state, and that was helpful.
• 10:29
  SPEAKER_01: And we own a couple of other health systems, one in Lansing and around Lansing, uh, and one uh in uh Grand Rapids and around Grand Rapids, but that didn't really serve everybody either.
• 10:42
  SPEAKER_01: And so what what we have been moving toward, and this is where there's some products that I think have great value.
• 10:49
  SPEAKER_01: Uh I haven't seen the best one yet, uh, I hope.
• 10:52
  SPEAKER_01: Um, but this is the uh the integration of technology and telehealth.
• 10:58
  SPEAKER_01: So telehealth, I mean, I think most of us have that down.
• 11:02
  SPEAKER_01: And most patients, when they um they they're very happy with telehealth.
• 11:09
  SPEAKER_01: Uh if they don't need to have a physical examination, uh they're they're very happy because they don't have to drive, find parking, uh, all that sort of thing.
• 11:17
  SPEAKER_01: And they can have really almost uh next day access.
• 11:21
  SPEAKER_01: Uh the the challenge is that how do you get some baseline information?
• 11:26
  SPEAKER_01: If you have to have a a uh extender there, uh even a medical assistant, but a nurse or nurse practitioner or PA, uh that introduces uh probably waste of time and a lot of expense.
• 11:38
  SPEAKER_01: So I have seen uh demos, and as I said, I I hope there's a better one out there.
• 11:44
  SPEAKER_00: There better be, right?
• 11:45
  SPEAKER_00: There better be.
• 11:46
  SPEAKER_01: Um these stations, which are inexpensive.
• 11:50
  SPEAKER_01: They can be located in rural areas, and the patient goes in and and you get real-time vital signs.
• 11:56
  SPEAKER_01: Uh, you can get other real-time uh information, clinical information.
• 12:00
  SPEAKER_01: The vital signs, I think they've mastered that, they get blood pressure, heart rate, etc.
• 12:03
  SPEAKER_01: But um uh getting uh additional uh clinical information, I think, is just on the verge.
• 12:11
  SPEAKER_01: And so what do I mean by that?
• 12:12
  SPEAKER_01: Well, you you can't examine you're a surgeon, you can't examine the abdomen uh that I know of with uh any way, but uh from my standpoint, what's important is the uh cardiac exam, uh the pulses, uh, the pulmonary exam.
• 12:28
  SPEAKER_01: Those things, there there's technology to uh provide information on those.
• 12:34
  SPEAKER_01: Uh weirdly, really weirdly.
• 12:37
  SPEAKER_01: Uh about now uh twenty-five years ago, I worked in Galveston, Texas, which cared for the state prisoners across the state of Texas, huge state.
• 12:47
  SPEAKER_01: And they had developed this technology.
• 12:49
  SPEAKER_01: It wasn't great, but it was like a digital stethoscope and a digital uh ophthalmoscope.
• 12:55
  SPEAKER_01: Uh ophthalmology's gotten pretty good at it.
• 12:57
  SPEAKER_01: Yes, it has.
• 12:58
  SPEAKER_01: So I think if that if somebody can come up with a cost-effective way to do that, and I I think it's it's like all these technology barriers, it's expensive to develop, but then probably not as expensive to uh to expand, uh I think that would be a big win-win because Michigan's it's half rural.
• 13:18
  SPEAKER_01: And uh you mentioned uh when we were talking before the show that uh you work in a a pretty remote uh clinical setting.
• 13:26
  SPEAKER_01: And uh you can get specialists in this way, you can get uh all the things you need.
• 13:32
  SPEAKER_01: So so I uh short short of uh being able to actually put hands on a patient.
• 13:39
  SPEAKER_01: So I I think that's a great opportunity.
• 13:42
  SPEAKER_01: Uh needed and the technology's there, but somebody's got to put it all together.
• 13:46
  SPEAKER_01: And it's more complicated than saying, well, here's an app, just put it on your phone.
• 13:50
  SPEAKER_00: Yes.
• 13:50
  SPEAKER_00: And I think also, you know, I'm actually giving a talk tomorrow, and this is one of the main points that I'm putting out there, is that every healthcare system actually has a telehealth component already in there.
• 14:00
  SPEAKER_00: Right.
• 14:01
  SPEAKER_00: It's live.
• 14:01
  SPEAKER_00: You know, in 2020, we got these things rolling in a matter of weeks or months.
• 14:07
  SPEAKER_00: And at the time we had upwards of 30% of patients utilizing telehealth on a regular basis, you know, on average over the United States.
• 14:15
  SPEAKER_00: And now that we are more in-person, those numbers have gone down, down, down.
• 14:19
  SPEAKER_00: However, I think it's because the systems just aren't pushing them as hard.
• 14:24
  SPEAKER_00: Um, they're already paid for.
• 14:26
  SPEAKER_00: There's people who know how to utilize the systems, it's already integrated for infrastructure.
• 14:30
  SPEAKER_00: I also think this is an area where we have the massive amount of, I'd say, reach for patients, like you just mentioned, with or without the additional devices that we'd love to have.
• 14:41
  SPEAKER_00: Um, and it's something that could really decrease the pressure of our in-person clinics dramatically.
• 14:47
  SPEAKER_01: I I couldn't agree more.
• 14:49
  SPEAKER_01: I mean, we can't thank COVID for very much, but we can thank it for telehealth.
• 14:52
  SPEAKER_00: We'll thank it for that.
• 14:53
  SPEAKER_00: That's the only thing I'm gonna give it.
• 14:55
  SPEAKER_01: And uh yeah, it helps with access.
• 14:58
  SPEAKER_01: Access is a real challenge for in many settings.
• 15:02
  SPEAKER_01: Uh you know, it could be used locally to reduce uh ER visits.
• 15:07
  SPEAKER_01: RER is jam-packed because people don't have anywhere else to go.
• 15:11
  SPEAKER_01: And people, particularly people who don't have, you know, readily available physicians.
• 15:15
  SPEAKER_01: So they come to the ER for their routine care.
• 15:17
  SPEAKER_01: In our ER, probably 20% of the people need to be admitted, and 80% there are there for some kind of uh wellness check.
• 15:24
  SPEAKER_01: So, yeah, I I think that uh that's a uh a great example of uh where technology can be helpful to us.
• 15:34
  SPEAKER_00: Well, let's talk about something I think some people in the room are gonna get squirmy about, but let's see if we can save some people a lot of time.
• 15:43
  SPEAKER_00: You and I have seen hundreds of pitches.
• 15:45
  SPEAKER_00: I'm sure you get a random email almost every other day.
• 15:47
  SPEAKER_00: I know that's how it is for me.
• 15:49
  SPEAKER_00: And I've sat through a gazillion demos, I'm sure you have too, and I know we've both probably signed some agreements and walked away from a lot more.
• 15:59
  SPEAKER_00: So when a vendor comes to pitch their newest clinical AI, their newest gadget, when you're looking at them, what are the two or three things they do in the first meeting that immediately tells you they don't understand how a hospital operates?
• 16:16
  SPEAKER_01: Well, um a profound question.
• 16:21
  SPEAKER_01: I'm I'm just I'm just thinking about it for a moment.
• 16:24
  SPEAKER_01: I I think that the the biggest mistake they make is to uh perseverate when when you say, well, we have to understand how this is gonna work with our systems, they perseverate with the belief that if I if you just if the person they're selling to understood how great it is, then they would figure out the solution.
• 16:47
  SPEAKER_01: And and big systems just aren't gonna try.
• 16:50
  SPEAKER_01: So so I think that's that's the first misstep they take.
• 16:53
  SPEAKER_01: When when they hear we we n which was kind of the first thing always out of my mouth, well how how are we gonna make this work in our system?
• 17:02
  SPEAKER_01: They should pause for a second and say, okay, let's think through this together, because then you think, well, maybe I have a partner that I can work with here.
• 17:11
  SPEAKER_01: So so I think they need to understand it's just not all about how great their idea is or their technology.
• 17:19
  SPEAKER_01: I think it wouldn't be a bad thing, and I have sometimes told people who just keep going on and on, I've said, look, we get these pitches every single day, and we rarely take one, and it's because of this reason.
• 17:31
  SPEAKER_01: Can you help me with that?
• 17:32
  SPEAKER_01: And then they one more time go through how great it's gonna be how transformational it's gonna be.
• 17:37
  SPEAKER_00: I can't tell you how many founders, even on teams I've been on, I have arm wrestled with because even if you don't have existing integration with the system, you at least have to have a plan.
• 17:49
  SPEAKER_00: So if you have a plan to integrate with Epic, Sterner, whatever the clinical piece is, I will listen for five more minutes.
• 17:57
  SPEAKER_00: But if you have no plan, then I know that you don't understand exactly what you just mentioned, the importance of it being part of the workflow and an area where doctors, nurses, and other organizational members can actually access it in stuff they already know today.
• 18:11
  SPEAKER_00: Because if it's something I have to log into over here or go to a different device for, I'm just gonna say no.
• 18:16
  SPEAKER_00: Because my doctors already are upset with the administration.
• 18:21
  SPEAKER_00: My doctors are already exhausted in seeing way too many patients.
• 18:24
  SPEAKER_00: If I add one more thing, it's gonna be the straw on the camel's back, if you know what I mean.
• 18:30
  SPEAKER_01: Yeah, and if this is maybe a a a poor example, but when they come up with something really great that does work, uh is uh in my leadership roles, I mean it it was a great day if 51% of the people agreed with the decision I made and 49 didn't.
• 18:47
  SPEAKER_01: Uh yeah, but but with uh um with ambient AI, so with uh either uh a bridge or uh Dax the Microsoft approach, where uh the and and I've never had a patient say, no, it's not okay to do that.
• 19:06
  SPEAKER_01: I always ask them, okay, if AI listens and we'll get your report and you'll get to see it.
• 19:11
  SPEAKER_01: Uh that has been that I have to say, I think 75% or 80% of the our physician population, you know, were ecstatic once they had a chance to try that, because it saved them uh time.
• 19:26
  SPEAKER_01: So the reason I bring up that example, maybe there will never be another example like that, but if they find things that can save the staff, not just physicians, save the other staff time.
• 19:36
  SPEAKER_01: And what you mentioned was just the opposite, which is often true.
• 19:39
  SPEAKER_01: Well, you gotta log in through a different system.
• 19:41
  SPEAKER_01: I mean, golly, in in cardiology, uh, we're still getting back almost over the days where the ecosystem's one, the cath lab's another.
• 19:49
  SPEAKER_01: Yeah.
• 19:49
  SPEAKER_01: And and you have to log into them separately.
• 19:52
  SPEAKER_01: Apollo, you know, you're like messing around with that thing forever.
• 19:55
  SPEAKER_01: So um so that that's another uh I hadn't thought of it in that way, but that's a great tip to give entrepreneurs.
• 20:03
  SPEAKER_01: Uh figure out how it's gonna save people time.
• 20:06
  SPEAKER_00: Well, let's talk about that time.
• 20:07
  SPEAKER_00: When we think about nursing and physician burnout, I also love ambient listening because it's again completely in the workload, um, in the flow, and theoretically giving providers more time.
• 20:20
  SPEAKER_00: More time to make eye contact, more time to not be up all night making notes.
• 20:26
  SPEAKER_00: But a lot of times in practices, that time is not given back to the provider.
• 20:32
  SPEAKER_00: Instead, it's actually to see more patients.
• 20:36
  SPEAKER_00: Now, again, this helps with our access problem.
• 20:38
  SPEAKER_00: We can bring more patients in, but from a burnout perspective, my concern is that it's actually making things harder.
• 20:44
  SPEAKER_00: You've seen this at scale.
• 20:46
  SPEAKER_00: What have you seen that's working or not?
• 20:50
  SPEAKER_01: Well, uh, in our system, it's a little bit easier to control because we have centralized scheduling, which is not all that great, but and and actually I think AI can help tremendously with uh centralized scheduling.
• 21:03
  SPEAKER_01: But we uh we have a uh strong practice plan uh that looks at uh um expectations, and so we have set expectations for different kinds of clinical practice, and dermatologists may see 40 patients a day, a uh uh psychiatrist or a neurologist may see 10 patients a day.
• 21:26
  SPEAKER_01: And so those expectations expectations there, and uh I don't know how often this is violated, probably all the time, but the expectation of the practice plan is if you want to adopt a new tool and then want to uh increase demands, and it's always around RVUs.
• 21:46
  SPEAKER_01: Um if you want to increase demands, then that has to be approved by the practice plan.
• 21:50
  SPEAKER_01: Which I think, you know, I'm sure chairs have figured out how to get around that.
• 21:54
  SPEAKER_01: But but it it does help with this problem.
• 21:57
  SPEAKER_01: But I've heard that problem many, many times.
• 22:00
  SPEAKER_01: Some from our own physicians, but particularly from our affiliated physicians, they say, Well, what good did that do me?
• 22:06
  SPEAKER_01: I uh now they just piled over patients.
• 22:09
  SPEAKER_01: And and they everyone understands that access is a problem, but they don't want it to be solely on their back to fix the access problem.
• 22:18
  SPEAKER_01: And I do think, you know, the other comment I'll make, although I have not done this, but we have some pretty good success stories of people who've used AI as a uh extender, uh, not a replacement, but they're they they're using AI uh in ways that uh are available through Epic, I think, uh to gather information just as you would with uh it's like an assistant.
• 22:46
  SPEAKER_01: Um does it take the place of a medical assistant, but it uh enables the physicians to work more quickly.
• 22:52
  SPEAKER_00: I like that.
• 22:53
  SPEAKER_00: So we think about we've said yes to a pilot, we think that maybe they understand healthcare, um, we think it's a promising technology.
• 23:02
  SPEAKER_00: When we deploy an AI tool and the pilot actually goes well, the numbers look good, the vendor writes a case study with us, wonderful, and six months later the tool is quietly not being used anymore.
• 23:16
  SPEAKER_00: What happened?
• 23:17
  SPEAKER_00: Not the press release version, but what really kills a promising pilot even after the initial deployment goes well?
• 23:25
  SPEAKER_01: Well, uh I I uh I'll complain about our own system.
• 23:30
  SPEAKER_01: We're very slow to make change, and we do pilots on our pilots, and uh That's a good way to put it it's just uh it's just hard.
• 23:40
  SPEAKER_01: But uh I think the biggest uh it's almost like thinking about uh the political landscape in healthcare.
• 23:49
  SPEAKER_01: Uh who's gonna change it?
• 23:50
  SPEAKER_01: It's not it's not the it's not the two two polarized political parties, it's the people.
• 23:57
  SPEAKER_01: And uh so I'm I'm a believer, I'm I'm a hopeful believer that if the people finally stand up and say, hey, this is not acceptable, what's happening in healthcare, uh, you know, that that could pressure uh politicians to make change.
• 24:10
  SPEAKER_01: The the corollary is in healthcare, if the providers, if if they reach a I think the pilot has to be big enough, that they reach a group of enthusiastic, highly enthusiastic clinician pilot members, and then that spreads.
• 24:25
  SPEAKER_01: That spreads like wildfire.
• 24:27
  SPEAKER_01: If it's a little small group, people say, I know that person, you know, they're always wanting to thinking only of themselves.
• 24:36
  SPEAKER_01: And so so they need to have a large enough pilot that it will have some oomph, and then that's what carries it.
• 24:41
  SPEAKER_01: So we we've seen both situations that you described.
• 24:44
  SPEAKER_01: Pilots that went great and were quickly adopted.
• 24:47
  SPEAKER_01: Um again, hearkening back to the ambient AI, that that was one.
• 24:51
  SPEAKER_01: We did a pilot of about 100 people.
• 24:53
  SPEAKER_01: We have like uh 3,000 faculty, uh, but those hundred were so enthusiastic it just sipped through.
• 25:01
  SPEAKER_01: We've had others where you know it's a group of 50 or 100 uh who have they tend to have a much more focused interest.
• 25:09
  SPEAKER_01: And so maybe there are 50 people who are interested in in the GI system.
• 25:16
  SPEAKER_01: Um it it it's just not gonna have that impact.
• 25:18
  SPEAKER_01: Everybody says everyone's gonna say, well, how does that impact what I do?
• 25:22
  SPEAKER_01: So so I think they need to think about their audience and think about the size of the pilot.
• 25:26
  SPEAKER_01: Of course, it's harder to get a big pilot going than it is a smaller pilot.
• 25:29
  SPEAKER_01: True.
• 25:30
  SPEAKER_01: But I think it's uh it is a pathway to success, so they don't just get stalled out.
• 25:34
  SPEAKER_01: But I think what happens when they haven't heard anything for six months is there although there was a group of people, I'm gonna my light went off here, um, that uh were uh enthusiastic, it just wasn't a big enough group to carry the day.
• 25:50
  SPEAKER_00: That's interesting.
• 25:50
  SPEAKER_00: All right, so a pilot has to be big enough so that the clinical champions are not just one person, but a small coalition so that the success of the pilot and the enthusiasm can be spread more wildly.
• 26:03
  SPEAKER_00: Yes.
• 26:04
  SPEAKER_01: By by people that are there.
• 26:06
  SPEAKER_00: So respected people that are there.
• 26:08
  SPEAKER_00: Yes.
• 26:09
  SPEAKER_01: You know, it's one thing to have a glitzy publication and press release, it's another to have people say, Well, hey, wow, I tried this.
• 26:14
  SPEAKER_01: It worked.
• 26:16
  SPEAKER_00: So that's interesting.
• 26:17
  SPEAKER_00: So one of the things I see is again, the very enthusiastic clinicians are part of the pilot, and then it goes to the other department or the other hospital or the other building, and it's people who are a little bit not so excited about it.
• 26:31
  SPEAKER_00: And when we think about digital health deployment, there's often times where physician clinical judgment kind of runs into someone else's decision, either IT governance, legal, finance, C-suite discussions, procurement, compliance.
• 26:45
  SPEAKER_00: So where does the physician's authority actually end in a digital health deployment?
• 26:50
  SPEAKER_00: When should it end in when does it get overridden in ways that kind of create a real clinical risk?
• 26:56
  SPEAKER_00: What do you think?
• 26:58
  SPEAKER_01: Well, I think that physicians have to, they they must be involved.
• 27:02
  SPEAKER_01: And they must help identify pros and cons of the technology.
• 27:07
  SPEAKER_01: Now at the end of the day, they don't make the decision typically.
• 27:12
  SPEAKER_01: Uh, I do like, obviously, I was in this role, I like physicians who are in the leadership role because I think they interact differently with uh physicians uh that uh than people do who don't really have a strong clinical background.
• 27:26
  SPEAKER_01: So I think that's one thing.
• 27:28
  SPEAKER_01: I I will you you brought up something that triggered a thought in me.
• 27:31
  SPEAKER_01: So with ambient AI, the system that we bought in uh acquired, bonded with, whatever, uh in Western Michigan and Grand Rapids was called Metro Health.
• 27:42
  SPEAKER_01: Metro Health had been using Ambient AI uh for two years before we adopted here.
• 27:49
  SPEAKER_01: And they had great experience, and I think it was largely their push that got our pretty risk-averse IT group to say, well, let's take a look at this.
• 28:01
  SPEAKER_01: And so the reason I mentioned that is maybe where there are large systems, maybe that pilot could take place in one of the smaller components because those smaller components tend to be much more flexible.
• 28:13
  SPEAKER_01: True.
• 28:13
  SPEAKER_01: So something to think about for entrepreneurs.
• 28:17
  SPEAKER_00: Alright, so now it's time for us to say the quiet part out loud.
• 28:20
  SPEAKER_00: So I want you to think about something, and I think it's really this is an interesting one.
• 28:25
  SPEAKER_00: So we can both name the things that academic health systems are supposedly not prepared for.
• 28:31
  SPEAKER_00: So interop, staff shortages, payment model transition, that's the usual list.
• 28:37
  SPEAKER_00: What are academic health systems genuinely not prepared for?
• 28:40
  SPEAKER_00: You don't have to call up Michigan, but in general in the next three to five years, and not the obvious answer, but the one that actually keeps you up at night.
• 29:01
  SPEAKER_01: And not just in AI, but particularly in AI, but in AI in devices and AI and everything.
• 29:07
  SPEAKER_01: Uh that technology is roaring along.
• 29:10
  SPEAKER_01: And uh as with much technology, there's a balance.
• 29:15
  SPEAKER_01: So if you jump out and and hook on to the first one that's out of the gate, you know, that's probably not the best one.
• 29:20
  SPEAKER_01: But it is the glit glitsiest one.
• 29:22
  SPEAKER_01: You can you know you can be in Becker's healthcare uh talking about what you're doing.
• 29:26
  SPEAKER_00: You can get in Becker's for a lot of things, not necessarily always positive.
• 29:29
  SPEAKER_01: Yeah, maybe that's not but anyway, you can you can get uh a lot of street reputation for doing that.
• 29:36
  SPEAKER_01: Um on the other hand, if you put off technology as as we have done several times, uh you're just putting yourself behind.
• 29:44
  SPEAKER_01: And so let's take uh a practical example of AI, uh, and this has to do with the revenue cycle.
• 29:50
  SPEAKER_01: So we have we live in a very difficult commercial insurance environment with a very dominant single commercial insurance company, uh Blue Cross Blue Shield of Michigan, that owns about 80% of the commercial insurance in the state.
• 30:03
  SPEAKER_01: So they don't they have no real competition.
• 30:06
  SPEAKER_01: And negotiating with them, and we just completed the negotiation, I didn't thank goodness have to participate in this.
• 30:12
  SPEAKER_01: Um, but uh, you know, it's always about rates, and uh but a big part of it's about denials, about uh all the things that go through the process of getting things approved, pre-approval, etc.
• 30:28
  SPEAKER_01: It turns out, um I don't I didn't learn this until about a year or and a half ago, but Blue Cross Bushield, Michigan, like many insurance companies, commercial insurance companies, probably like like CMS, I don't know, uh they use AI.
• 30:43
  SPEAKER_01: And they use it for uh prior approval and for claims negotiations.
• 30:49
  SPEAKER_01: Uh it takes like a couple of minutes.
• 30:51
  SPEAKER_01: And then they send it back.
• 30:53
  SPEAKER_01: And so we were we should have known, uh, and I don't know if we knew that or not, if the our IT folks knew that or not, but we're just now starting to use AI, so it's like battling computers.
• 31:05
  SPEAKER_01: Um had we done that a year and a half ago, I I'll bet we would have saved tens, if not hundreds, of millions of dollars in just churn time and uh revenue cycle and in human time.
• 31:22
  SPEAKER_01: Uh so I'm giving that as an example of uh a technology that uh you wouldn't think of off the top of your head, but one that uh emphasizes this point of don't adopt too early, but but be ready.
• 31:36
  SPEAKER_01: When it when it's ready, be ready to implement it.
• 31:38
  SPEAKER_01: And I think particularly at large health systems, particularly large academic health systems, have trouble with that implementation because in a way you have many, many uh points of of disapproval.
• 31:55
  SPEAKER_01: Uh so you have to convince those people, uh, whether they're clinicians or in the business side of things or whatever.
• 32:02
  SPEAKER_00: So one of the things that keeps me up at night is our aging population.
• 32:07
  SPEAKER_00: So if we look at countries like Japan with a super aged population, you know, they're many years ahead of us, and they're truly having a hard time taking care of so many older individuals with comorbidities.
• 32:18
  SPEAKER_00: And I think there's a space for technology to help us scale, but at the same time with increased numbers of comorbidities in patients, increasing age, less healthcare workers to do the job, and a general decrease in the trust of the healthcare system at large.
• 32:37
  SPEAKER_00: My concern and what I'm seeing is more and more complex patients, less and less people to take care of them, more and more people on Medicaid, Medicare, and reimbursement being terribly low.
• 32:50
  SPEAKER_00: So the margins are even tighter, and the incentives for people to go into medicine I feel are getting maybe not as shiny as they once were.
• 33:01
  SPEAKER_01: You did a great job of putting that together.
• 33:03
  SPEAKER_01: I'm not sure I can add too much to it, but I would drop that mic and walk away, right, Marcel?
• 33:09
  SPEAKER_01: Uh well I I uh you identified, I mean, there's a big mismatch of providers in the aging population.
• 33:18
  SPEAKER_01: Uh there is uh an increasing mismatch of uh a terrible mismatch in primary care and the needs.
• 33:29
  SPEAKER_01: So as I was studying this uh a year or two ago when I was working on a uh book about health care, uh the uh you look at health care costs in the United States, I mean we're way high.
• 33:41
  SPEAKER_01: About$13,000 per capita.
• 33:44
  SPEAKER_01: Every person in the United States just provide our health care.
• 33:46
  SPEAKER_01: That's at least twice as high as most places, three times as high as others.
• 33:50
  SPEAKER_01: But when you look at the number of primary care providers, we're at the bottom of the bucket in terms of how many primary care providers we have per capita.
• 33:58
  SPEAKER_01: And that gets to a point that you mentioned uh primary care doesn't pay very well, and it's not likely to start paying well if we continue what we're doing because with the our payment models, um in addition to poor payers, uh in the state of Michigan uh and nationally, there's a significant number of people who don't have any insurance.
• 34:21
  SPEAKER_01: And we're we're gonna take care of them.
• 34:24
  SPEAKER_01: You you'll take care of them when they come to your clinic, we'll take care of them.
• 34:27
  SPEAKER_01: And uh, but you can't pay for that.
• 34:30
  SPEAKER_01: So in my dream world, which is truly a dream world, um, I think that we need to appreciate that the RPU system was developed for Medicare back in the late 60s.
• 34:42
  SPEAKER_01: It's not been tinkered with hardly at all since then.
• 34:45
  SPEAKER_01: I mean some diagnoses and uh subdiagnoses have been added, but you know, it's a it's a 50-year-old system.
• 34:52
  SPEAKER_01: Um and I think the salvation, which is probably not going to occur in my lifetime, is to rethink healthcare and how we provide it and how we um pay for it.
• 35:04
  SPEAKER_01: It is true that healthcare in the United States is much more of a capitalistic approach than anywhere else in the world, and we have to address that head up.
• 35:14
  SPEAKER_01: Um I was talking to somebody about this the other day, and they said, and I haven't done this, maybe maybe you've done this, but I I'm I'm gonna do it one of these days.
• 35:21
  SPEAKER_01: Well, take your favorite AI engine and say, Here are all the problems.
• 35:25
  SPEAKER_01: If you were gonna redesign healthcare in the United States, what would you do?
• 35:29
  SPEAKER_00: Alright, I'm gonna do it right now while we're on the line together.
• 35:31
  SPEAKER_00: So I'm gonna open my Claude.
• 35:32
  SPEAKER_00: I'm not gonna share my screen, but I'm gonna say, given all the healthcare problems in the US today, what's the first one you would fix?
• 35:47
  SPEAKER_01: And how.
• 35:48
  SPEAKER_00: And oh how, oh my god, and how.
• 35:52
  SPEAKER_00: Alright, it's chugging away.
• 35:54
  SPEAKER_00: Hold on.
• 35:54
  SPEAKER_00: I just dropped my earbud.
• 35:57
  SPEAKER_00: And I say this basically on every single talk I'd ever give.
• 36:01
  SPEAKER_00: I would love to just burn up the whole system, burn it down, and start from scratch, but we know we can't because people need care today and yesterday, and they need it tomorrow.
• 36:13
  SPEAKER_00: Alright, so I can't wait to hear this.
• 36:15
  SPEAKER_00: I know, right?
• 36:16
  SPEAKER_00: So my Claude version has said harder authorization, not because it's the sexiest answer, but because it's the one upstream lever that simultaneously kills patients, burns out physicians, and wastes billions, all while prioritizing and producing zero clinical value.
• 36:32
  SPEAKER_00: And that wasn't for me.
• 36:33
  SPEAKER_00: Yeah.
• 36:34
  SPEAKER_00: It's merely what I read.
• 36:36
  SPEAKER_00: Here's how it said it would fix it.
• 36:38
  SPEAKER_00: So first, mandate gold carding nationally.
• 36:41
  SPEAKER_00: Any physician with a 90% approval rate for a given procedure over 12 months gets automatic approval going forward.
• 36:46
  SPEAKER_00: No review done.
• 36:48
  SPEAKER_00: Interesting.
• 36:49
  SPEAKER_00: Second, cap approval window is at 24 hours for urgent requests, 72 hours for standard.
• 36:54
  SPEAKER_00: Okay.
• 36:55
  SPEAKER_00: And third, require payers to publish denial rates and overturn rates by procedure and physician category publicly.
• 37:03
  SPEAKER_00: Awesome.
• 37:04
  SPEAKER_00: Right?
• 37:05
  SPEAKER_01: You know, I think actually, Sarah, this is what I've I only all I know about what's going on in DC is what I read.
• 37:12
  SPEAKER_01: But this is a bipartisan, there's bipartisan support for uh two things.
• 37:16
  SPEAKER_01: One is prior authorization, the other is uh killing uh PPMs, uh which are just you know a money grubbing, uh worthless not uh let me commercial, you're gonna get us banned on YouTube.
• 37:30
  SPEAKER_00: You gotta stop it.
• 37:31
  SPEAKER_00: Hold on.
• 37:33
  SPEAKER_01: You can go back and edit, right?
• 37:34
  SPEAKER_01: So uh maybe but you know the the value proposition of pharmacy-based uh pharmacy benefit managers is is not in favor of healthcare, in my opinion.
• 37:47
  SPEAKER_00: It's tough.
• 37:48
  SPEAKER_00: There's a lot of different layers to this for sure.
• 37:50
  SPEAKER_00: For sure.
• 37:51
  SPEAKER_01: But I love that answer from Claude.
• 37:52
  SPEAKER_01: That's great.
• 37:53
  SPEAKER_00: That's not bad.
• 37:54
  SPEAKER_00: Thank you, Claude.
• 37:55
  SPEAKER_00: Um so here's the same two questions I ask everyone at the end of every episode.
• 37:59
  SPEAKER_00: So here they are.
• 38:01
  SPEAKER_00: The first one is what is the clinical reality that most people in health tech get completely wrong today?
• 38:08
  SPEAKER_01: Um It's I think it's the reality of how complicated uh the comp how complicated healthcare itself and the administration of it is, but also how complicated patients are.
• 38:22
  SPEAKER_01: So you you mentioned earlier aging population, more medications, more chronic diseases.
• 38:28
  SPEAKER_01: If they have a a uh a program that's focused on a specific area, it's just not going to be applicable for lots of people.
• 38:36
  SPEAKER_01: So I think it's both the organizational complexity, but also the complexity of our aging population.
• 38:42
  SPEAKER_00: I like that.
• 38:43
  SPEAKER_00: And then last question.
• 38:45
  SPEAKER_00: What is the last what is the last time you changed your mind about something in healthcare?
• 38:50
  SPEAKER_00: Now, I'm laughing because as we got on this call, you said, oh, it was five minutes ago.
• 38:55
  SPEAKER_00: But but truly, when was the last time you changed your mind?
• 38:59
  SPEAKER_01: Oh, I uh it it happened all the time.
• 39:06
  SPEAKER_01: I'm trying to think of a really great example.
• 39:09
  SPEAKER_01: Um But one one example that was toward the end of my tenure was uh there was a lot of interest in building more facilities.
• 39:20
  SPEAKER_01: And as we look at it though, uh if we succeed in the ways we're trying to succeed, and and I think at the end of the day, we need to focus much more on health than health care.
• 39:31
  SPEAKER_01: And if we get if we get our population healthier, then there's not as much need for health care.
• 39:35
  SPEAKER_01: And so I'm I'm a believer that we'll move in that direction.
• 39:38
  SPEAKER_01: So the idea uh that we need to build more ambulatory, that that's the solution to waiting is the facility, I think is not correct.
• 39:48
  SPEAKER_01: I don't I don't know about you, but when I I was just driving um uh across a a sort of suburban area or s outside of a suburban area where there were these big buildings that were where insurance companies had been housed or where uh uh various things unrelated to health care.
• 40:09
  SPEAKER_01: They they're empty.
• 40:10
  SPEAKER_01: And whereas the you live close to New York City, I mean I don't know what it stands like now, but after the pandemic, they had all these empty buildings.
• 40:19
  SPEAKER_01: And so the idea that we might spend fifty million dollars building a building that uh we don't need originally it sounded pretty good.
• 40:27
  SPEAKER_01: I thought, hey, this this might work.
• 40:29
  SPEAKER_01: And then uh, and this was this was toward the end of my tenure uh in my uh CEO role.
• 40:35
  SPEAKER_01: Uh then although I'd been with it a long time, I said finally had to pull the plug and say, we're just not gonna do that.
• 40:44
  SPEAKER_00: Understood.
• 40:44
  SPEAKER_00: Well, Dr.
• 40:45
  SPEAKER_00: Marsha Rangi, this has been exactly what I wanted.
• 40:48
  SPEAKER_00: The institutional reality said plainly from someone who has actually run it.
• 40:52
  SPEAKER_00: So thank you very much for making the time.
• 40:55
  SPEAKER_00: Um, for everyone listening, if what we talked about today, the gap between what vendors pitch and what health systems actually need, the accountability question, who gets reached and who doesn't, if any of that landed for you, subscribe to the clinical realist wherever you get your podcast.
• 41:08
  SPEAKER_00: And if you're working through a real clinical AI question, definitely contact me.
• 41:11
  SPEAKER_00: I'm Dr.
• 41:11
  SPEAKER_00: Sarah Matt, and this is the Clinical Realist.
• 41:14
  SPEAKER_00: See you next week.
• 41:15
  SPEAKER_01: Thank you, Sarah.