← Becker’s Healthcare Podcast
Becker’s Healthcare Podcast
Healthcare Upside Down: Lessons from Mayo Clinic and ECG on AI and Automation
00:21:06
Loading summary
Transcript25 lines
- [00:00]
A
Welcome to Healthcare Upside Down, a podcast by Becker's healthcare and ECG Management consultants, in which we'll explore the upsides and downsides of healthcare and the industry's most current trends. I'm Molly Gamble, and today I'm thrilled to be joined by two guests. Edwina Boscorin, Chief Clinical Systems and Informatics Officer with Mayo Clinic, and Henry Stockman, Principal with ECG Management Consultants. Edwina. Henry, thank you so much for joining me today.
- [00:24]
B
Nice to see you again, Molly.
- [00:25]
A
Likewise. Well, to get us started, I just shared your names and titles there, but I know there's more to your stories. Can you introduce yourself, tell us a bit more about your work in healthcare, your roles, and your organizations? Edwina, I'll turn to you first here.
- [00:38]
B
Sure. So I am the Chief Clinical Systems and Informatics Officer for Mayo Clinic. I have the privilege of being at the intersection of IT and our practice and supporting our practice in making sure that they have a voice in the technologies that we deploy and also supporting them through the implementation and rollout of the technologies as well. I've always been interested in understanding what technologies stick and what technologies perhaps don't necessarily make the cut. So this topic is near and dear to my heart.
- [01:09]
A
Great. Thanks, Edwin. I'm looking forward to hearing your perspective on that. Henry, I'll turn to you.
- [01:14]
C
Absolutely. I'm Henry Stockman. I'm a principal with ecg. I work on the revenue cycle space. At ecg. We do a lot of work across the board in the revenue cycle space, specifically, not just on the ambulatory side, but also in the acute space and also in the mid rev. And in the last few years, we've done a lot more outreach and partnership with a lot of vendors on the AI space space, working with them, forging new partnerships, and then trying to find ways in which we can integrate some of that technology into our our client work, as well as into how we can generate additional value and accretion in the output of what we're trying to accomplish for those clients.
- [01:55]
A
I mean, the AI space, there's so much enthusiasm here, so much anticipation. But then I think, Edwina, to go to your point about what sticks and what doesn't, a recent industry report shows only 30% of AI pilots in healthcare have reached full production. Maybe we can start there. I guess from your vantage point, as you see it, what really distinguishes the pilots that move forward into real operations from those that stall or plateau.
- [02:22]
B
So I think there are two different ways to look at that 30% statistic. I actually look at it both as a positive and as an opportunity. So I think from the 30% standpoint, we are still cultivating a lot of ground up innovation at Mayo Clinic. We support decentralized innovation. And so I think there is an aspect of it's okay for us to see that we're not seeing many of these opportunities come to scale because we're still learning in many ways and that's okay. But I do think as we continue to mature and evolve in our ability to leverage AI, we want to see that improve, we want to see that perhaps get to a higher percentage of those tools that can scale. And so from that standpoint, a couple of things that I think that really stick, that I see is one, how integrated is the AI tool with the existing workflow? So is it automating an existing workflow, is it advancing an existing workflow? Or is it just a bolt on or an add on that? We have to ask our clinicians and end users to do another click, do another step along their journey in the application versus something that's perhaps actually taking away something that they're already doing that's burdensome. And so what we find is those tools that can really create a new workflow or a new path and decommission something that they have worked on for years at a time and that improves what they, what they've been working on. It helps a great deal. The other thing that I will say is that from a scaling perspective, looking at how that adoption works over time is something that you can kind of see in a pilot. And so one of the, you know, one of the, one of the thoughts around adoption is where do we see that sustained? And so one of the key outcomes that I look for is in a pilot, are you able to sustain it within the first month and then the first three month mark? Because that's a good indicator for how well will that tool be able to sustain in the long run.
- [04:26]
A
And we think that's such a great point that you made. That 30% stat isn't inherently bad. It shows systems like Mayo and so many others are being quite selective about their pilots and really have some high bars that these partners and technologies need to surpass. But Henry, I'm going to turn to you here first and get your thoughts because you're working with systems as partners too. Is there, are there other criteria that maybe Edwina didn't mention that have really come to the forefront and how they're assessing what to scale?
- [04:52]
C
Yeah, absolutely. So, so I think one of the things that we're dealing a lot with, with our clients is really trying to help them decipher who the winners and losers are going to be in this entire, I'll call AI battle that's going on. Our clients, they're getting bombarded every day by tons of different vendors that are coming in, selling them all kinds of goodies and value that's going to allegedly be derived by implementing whatever their solution is. The challenge that I see in the challenge that I work on with my clients is really allowing them to try to decipher who those winners and losers are going to be. When we look at like what this is probably going to shake out to be is like, who, who is going to be the best vendor for you over the next two years, who's going to give you that value and who's going to have staying power and be able to innovate. I think that's one of the things that is really difficult to try to decipher. I don't have the answer to that. But when we look back at history and how these types of boom busts have played out before, probably 85% of these vendors are probably going to cease to exist in two years. And whatever vendor you do decide to go with, the question then, Briggs, is this a one stop shop where I have this one solution that they've implemented, but it's not going to be innovated or is it going to be someone that I can partner with that's going to continue to innovate and adopt the solution? So it's going to move along with time because anyone that you go with and you put in whatever their solution is, it's going to take time to implement, integrate and roll out. And over that time there is a cost benefit because as you've now like, you know, hitched your wagon to this vendor or going along, two years are going to go by. And as we've all seen two years ago, ChatGPT, I've and think where we are now. So think we're going to be in two more years because this isn't just ramping slowly. This is like logarithmic of how fast this is going. So picking the right vendor and having that type of rigor I think is really important. And we at ECG are trying to do a good job of trying to suss that out and provide, you know, at least the insight and direction to our clients to like hedge their bets a little bit and not just like go all in on one vendor hoping that they're going to be the one that's going to really have the solution to take us to the next level.
- [06:58]
A
Right, right. And I think until that thinning out happens that you mentioned Henry, I think it plays out in the real, real world where leaders like Edwina are managing, you know, 20 plus vendor relationships and that can also be in itself a challenge. But you mentioned this being bombarded with all the options available and then I think looking for the real results, you know, when it comes to the RPA and the BoT based solutions that are being tested right now in revenue cycle, clinical operations, data management. I'm curious Henry, what use cases that you found most promising, your partners have found most promising, what conditions or decisions have really been critical to them standing out for, for good reasons.
- [07:37]
C
So, and I don't want to sound like a curmudgeon on this but like RPA and bots is not a new thing. They just put AI in front of it. So it became new again. I've been doing RPA and bot work for like over a decade with different clients. It's just been more rudimentary and candidly not as sexy, but now it is. So I think in the rep cycle space where, where we're seeing a lot of value is basically the stuff that we've been doing for the last several years that's really just you know, removing redundant tasks off of folks work queues. So like a good example of this would be like at the end of the month when I have low dollar write offs that are just like something that's going to happen every single month because it's the x dollar threshold that's going to occur. Instead of having John and Jane sit there and transact all these little dollar I can create a bot that's going to do that for them and I free them up that they can go do other tasks now. So we're starting to see that. We're starting to also see stuff happen ever slowly on the coding front of where we're starting to see some of that automation or bot work if you want to call start to take place. I know epic for example is really trying to ramp this up and do like true automated coding. I think the proof will be in the pudding on that and I know there's other vendors that are trying to do something similar but a lot of the stuff that I'm giving from an advice standpoint with clients right now is, is taking smaller baby steps and really trying to like dip your toe into this. Because the biggest risk I see in as this is as fast as this happening in the evolution of this is like on the rep cycle side. Clients make a decision to say great, I'm going to automate XYZ components in my billing and follow up. I no longer need John and Jane anymore. Array. Well back to my comment earlier. Let's say something new comes along that's even more dynamic or, or more I would say evolutionarily from an AI standpoint that they want to maybe adopt to or maybe that company goes belly up. Now what do you do? You have to go find a new vendor, you have to hire those people back. Like there is inherent risks with this, not to mention and I'm sure we can talk about this in more detail but the inherent like risks around overall how do we write the large language models? How are they learning? There's just a lot of nuance to this which is a long winded answer to your question but basically we're trying to help clients is really take some of those smaller steps and not completely overhaul their entire org structure. And what they're trying to do from like a day to day standpoint.
- [09:56]
A
Absolutely. Especially for something as sacred in many ways is the revenue cycle. And I think your comments surely help us understand the stakes here Henry, that this isn't. These aren't easy decisions. Edwini, I'll turn to you here. Are there any use cases or anecdotes, examples you can share out of Mayo Clinic when it comes to the revenue cycle, clinical operations, data management with some of these newer solutions that I think to Henry's point may not be brand new but have been supercharged by AI in recent years. Anything that stands out that you would consider exceptionally successful or maybe a learning moment for you at a system level.
- [10:31]
B
Yeah, I would echo Henry's comments around revenue cycle. At Mayo Clinic we are fortunate to have a remarkable revenue cycle team that's continuously looking at ways that we can automate and leverage RPA and AI. And so we have 34 bots currently deployed. We've taken advantage of some of the vendor options that Henry have has identified and what we find as Henry alluded to, is that the processes that are most successful are those that are repeatable, those that have rules based logic. The other thing that I would add that are stable. And so to Henry's point, we are at a stage where we see processes evolving rather rapidly. And so as the processes evolve, while we have bots in place or automation in place, we need automation on top of that automation. And so continuously looking at those opportunities, identifying which vendors have the staying power I think are all really good points. I'd also add the other area that we see benefits for are areas where we're ingesting outside materials. Outside materials in many forms. We still get Efaxes. And so making sure that those are routed to the right appropriate offices, whether that's the referral office or a particular department, is another area that we're finding value in. And I think to Henry's point also, it's a technology that's evolving, that has been accelerated by AI, I would say. And so we're looking to see how can we perhaps potentially explore this in the clinical areas as well. We are doing that carefully, with caution in terms of what areas and processes require clear clinical decision making and which one of those areas are perhaps more rules based, like scheduling, for instance, that might be a ripe opportunity to use this type of technology.
- [12:23]
A
Edwini, I'll stay with you here because I think Henry's. One of his pieces of guidance for listeners was to take baby steps with these decisions. What else do you think leaders need to prioritize, whether it's strategically or operationally, to really make sure that they're capturing the long term value, not just the short term gains, but the long term value from AI and automation.
- [12:42]
B
Yeah, AI right now, AI and automation has captivated our industry. I would say like none, nothing that I've seen in the recent days. And so as a result, we are seeing this intense focus on how to look at AI in a different way. And so Henry cited many, many different platforms, many different opportunities. We definitely feel that at Mayo Clinic, I will say what we have strategically thought about is how do we make sure that we have sustainment power and in those technologies that we're deploying. So not just focusing on uncovering and innovating, but also how do we reinvest in some of our operational areas and staff that we need to support that to sustain these technologies that are being embedded. So to that point, I'd say one piece of advice that I would offer is to make sure that organizations are relentless about decommissioning tools that no longer are useful. And so we need to redeploy those resources, whether that's people, resources, time budget, into these newer technologies to sustain them. And so without that kind of focused attention of decommissioning those technologies that are outdated, you'll have to add additional resources to sustain the new technologies that you're implementing, which is not necessarily an easy thing to do when budgets are tight.
- [14:11]
A
Absolutely, absolutely. Henry, are systems inherently bad at decommissioning?
- [14:18]
C
You can Say yes, yes, they are. And that's just a reality. There's just a lot of other priorities that are out there. And I think it ultimately comes back to the old, the classic problem of time, people and resources. And this advent of this movement is happening so fast and it's forcing people to, to make decisions quickly, probably more quickly than they're comfortable with because I mean, for anyone that's kind of worked in this space and work specifically with like it is, whatever you want to call them, they don't make decisions very quickly usually, which there's good reason for that. And I haven't even talked about some of the security issues that we could probably have an entire podcast about that's related to this. But being able to train these models with data is, is one thing. Being able to get the data to them securely, getting that output back securely is a whole nother thing. And that's also not even factoring the fact of the number of bad actors are out there using similar, even more advanced models to break through to get the data. I mean, I think that, I think change, healthcare blow up, that happened what, a year or so ago was a good wake up call. As painful as it was for everyone. And I was on the front line to that, that was a good wake up call for like how vulnerable we can be. And I think as we're embarking on this new frontier, it's really important for us to really be mindful about what risks are we willing to take and what is the ROI and the gain on that risk and does it always make sense and is it worthwhile and do we have the people, resources and energy? That's a whole other topic we could have a podcast on. I'm teeing up a bunch for you here. That's a whole other topic, but like the energy to be able to do this, sustain it and actually scale it.
- [16:03]
A
You mentioned change. It was about, I think 20 months ago. You mentioned ChatGPT a couple years and things are moving incredibly fast. And at the same time, and we know like when you talk about this decision making timeline maybe compressing any guidance for other leaders, I imagine that can feel uncomfortable given everything that Henry just outlined for us, including the security measures. Have you had that experience where you need to move quicker now and what has been really helpful for you to keep front and center as you're maybe staying a bit more agile than you needed to before?
- [16:34]
B
Yeah, I think it's a daily activity that we're doing in terms of trying to prioritize and reprioritize which technologies we continue to invest in. It is hard, as Henry said, in terms of trying to figure out which ones we decommission. Because nine times out of 10 there is one or two use cases where you have a very passionate person who's sponsoring that and they've really had that muscle memory form as a result of using that technology. So taking it away is challenging to do so. I don't think that there's a simple, easy answer for this. I do think historical governance processes, et cetera, are being tested in a way that they haven't been tested before. And so coming up with ways by which vendors or even internal organizations can transparently share how data is being shared, what provisions we can do to make sure that the data integrity is there, it's not leaving our walls, I think is something we've invested in at Mayo Clinic to make sure that we understand where exactly our data is being used, how it's being used. It enables us to make those decisions in a bit more of an agile fashion. In addition to that, I will say again, having that clear workflow and understanding where the workflow impacts are helps accelerate. Is this a really, truly a viable option for us to sustain in the long run, or is it not so that also helps accelerate the decision making? Where we run into challenges is those tools that continue to evolve in product life cycles. And so you have to almost foresee what could this potentially look like in terms of the investment that you're making now versus perhaps how that technology will evolve in three to four years. So you also have to assess the company itself. Are they good partners? Are they committed to how you will see your practice evolving in the next three to four years or even one to two years, and will they share that journey with you? So it's multifaceted, but it is something that we are trying to continuously evolve and update on an ongoing basis, thinking.
- [18:48]
A
About a few things at once. The governance piece, in some ways going slow so you can move fast and then to your point, also anticipating where things could be headed in the next couple of years, which sounds. It sounds increasingly challenging to do that. So I want to thank each of you. I think the insights you've brought to this conversation really help me better understand the complexity of this work. I think it's so often reduced to like a hype cycle, but there's real decisions and it's affecting leaders like you, Edwina, and those you advise, Henry, and how they're navigating some of these tensions and unknowns. Is there anything I didn't ask that we should cover or make mention of for our listeners before we wrap up.
- [19:23]
C
I mean, I think I probably come off sounding a little like anti AI, but I'm not. I think that's probably something I like. That would be. The payers are investing in this at light speed, light your speed. And it is going to be paramount for the providers to be doing the same thing. The question I think ultimately is like, how and with whom and how fast. That's really the kind of the crux of this. And I, I, this is kind of funny. I was, I was with a client a few weeks ago and we were talking about something similar like this. And I, I said, you know, there's this quote, I'm a big cinephile, so there's this quote from the movie Jurassic park and Dr. Ian Malcolm sitting at the dinner table and they're all like patting themselves on the back for what they've done. And he goes, you know, your scientists were so preoccupied trying to figure out how they never stopped asked the question of if they should. And I think that has a lot of parallels back to what's happening right now is we are so preoccupied trying to figure out how we can get this up and running, how fast we can do it, where we can do it, and how many people we can get rid of by doing it that we don't too often sit back and go, should we do this and where should we do it? I think that's, that's what I like to leave my clients with, is be thinking thoughtfully about that. And how are you going to position yourself for the next five to 10 years in the new frontier in which we're embarking?
- [20:38]
A
Yeah. Yeah. To your point, the reversal of these decisions would not be easy, if even possible in many cases. So, Edwina, Henry, I just want to thank each of you. I also want to thank our podcast sponsor, ECG Management Consultants listeners. You can tune into more podcasts from Becker's Healthcare by visiting our podcast page@beckershospitalview.com and Edwina and Henry, thanks again for being my guest. Learned a lot from you in our time together.
- [21:02]
B
Thank you.
- [21:03]
A
Thanks.