Bobbi Pritt, M.D. (00:04):

Hello, I'm Dr. Bobbi Pritt, a clinical microbiologist and laboratory leader at Mayo Clinic, and your host for today's episode. I'm excited to be here today with Mayo Clinic Laboratories' CEO and president, Dr. Bill Morice, to get the latest news. And then I'll be speaking with Dr. Chris Garcia about digital innovation during the deep dive. So let's start with what's new and noteworthy in clinical diagnostics. Bill, welcome back to discuss this hot topic with us.

William Morice II, M.D., Ph.D. (00:31):

Yeah, it's a topic that remains hot, so always happy to come back.

Bobbi Pritt, M.D. (00:36):

Absolutely. I know, and I've been seeing a lot of headlines about AI, digital technology. We all are. And since Dr. Garcia is joining us today for the deep dive on digital innovation, I thought it would be fun to get your take on some of the related news in this area. And the thing that really caught my eye is that Microsoft recently launched Copilot Health. And this is a great topic, because people have been using, for good or for bad, these generative AI systems for healthcare advice. So, do you want to tell us what your thoughts are on this?

William Morice II, M.D., Ph.D. (01:09):

Yeah, well it'll be interesting, but we can go a little bit further on this specific announcement as well and how it's maybe a little bit different from some of the other things that we've seen. And you're right, I mean, certainly the desire has been there right from the start. I remember one of the first use cases I saw of generative AI and its potential, when someone was showing me where they had uploaded their dog's symptoms, I think it was, and it turned out, it pointed them in direction of what their dog's ailment was. I think that was published in like the lay press. And so, yes. And now, like 40% or more individuals say they've actually uploaded either their test results or their doctor's notes or other healthcare information into a ChatGPT-like AI engine to get some advisement. And so, clearly the demand is there. The desire is there. I think you and I might have even discussed before that this is even more common in rural communities, areas where it's more difficult to access care. So yeah, it's certainly timely. It certainly the desire is there, and it now looks like Microsoft is trying to take more of a leadership role in this space.

Bobbi Pritt, M.D. (02:12):

Yeah. And I can tell all the listeners about this. So Microsoft launched Copilot Health. It's an AI-powered tool. It's designed to help users interpret their personal health data. And importantly, as you mentioned, Bill, this one's a little different because it's a secure space within Microsoft's Copilot platform, and it's designed for health. It will aggregate your health records, it will even aggregate and interpret your wearable device data and look for trends and your personal history. And then it will generate these personalized insights. So it helps patients really prepare for conversations with their physician.

William Morice II, M.D., Ph.D. (02:51):

Yeah. Well I think one of the key things you pointed out early on there was, it's secure. That's one of the things that's been a real concern around people using other publicly available AI engines, is that sometimes that information, there's no guarantee actually. And some of them even have disclaimers I think saying that you have to be careful about what you upload. So, it's secure. And it's not surprising that Microsoft would step into the lead here. Because I've had a chance actually to work with their head of science, I think now, was their head of research, a gentleman by the name of Dr. Peter Lee, who I believe might actually be a Mayo Clinic trustee now. But he was actually very active during COVID, understanding what Microsoft could do in terms of using AI to understand what was happening during COVID. So, he's been thinking about the diagnostic space and the healthcare space.

I don't know if he's related to this effort or not, but probably reflects just something culturally within Microsoft, where they've been interested in going this direction. Having it secure is a big step forward. Also very intriguing: that it can pull in the wearable data that you spoke to. Because I think it was mentioned after I visited San Francisco for the JP Morgan Healthcare Investor Conference, there is a big push from the wearable companies that are creating wearable devices to kind of push into the healthcare space, be it Whoop or Oura or others, because they're measuring a lot of things. They're getting more sophisticated on what they can measure. So having a safe, if you will, a trustworthy platform that can pull in this data and give you advisement and using other information around your health is quite a compelling offering.

Bobbi Pritt, M.D. (04:16):

It really is. And as we learn more about these wearables and their reliability, I think that clinicians are gonna be more willing to trust some of the data, at least to prompt additional testing. And, you and I have spoken in the past about, how do you get that last mile from the wearables and all the information it generates, into the actual healthcare record? So, how do you transfer that? And yet to be seen, but it's possible that Copilot could actually generate something that could be integrated into the healthcare record.

William Morice II, M.D., Ph.D. (04:46):

Yeah, I mean that to me, that's one of the last miles. It's still right now, going to be, you could also ask yourself, the futurist could say, "Wow, this is great. This is going to change how healthcare is delivered, how patients experience healthcare." The Luddite, if you will, or the person that was more of the cynic, maybe, would say, how's this that much different from WebMD, when the person would search on WebMD or MayoClinic.org or any of these sites, and come in with a bunch of paperwork about what they think they have to their doctor and to the physician or care provider, if it's not a physician. It takes the wind out of the sails, because you have all this information to try and get through. So, I think to your point, one of the last miles will be when that can actually cross over into an electronic healthcare record, so when you go into your physician or your care provider, they could pull it up as part of, you know, just your kind of cockpit, if you will, of healthcare information. Or even better, that it could proactively flag. I think that's what some people are really hoping for, is that these things and wearables that are continuously measuring could actually flag for a physician, a provider, when one of the people that they're caring for actually needs an intervention. That's all the upside futurist stuff, which would be great.

The other thing, which in the back of my mind, is the other last mile of cross, which is equally potentially filled with potholes, is the regulatory one. A lot of the devices right now don't have FDA claims around the information they're producing because they're not considered clinical grade utilities, right? So if that changes, well how will the FDA step in on this and start to think about regulating this space?

Bobbi Pritt, M.D. (06:18):

You raised two very good points and one thing that we really do need to think about is, what are our healthcare providers going to think about this, having all this additional data or these reports generated from a generative AI system? It's one more thing for them to look at, in addition to all the other things the patients may have gotten on the internet. I guess the optimist in me thinks that perhaps this will be a little bit more accurate if it's actually using wearable data. It could generate insights based on real data, not just some random search on the internet. But the cynic in me said, well, it's yet to be seen. We'll have to keep an eye on this one.

William Morice II, M.D., Ph.D. (06:56):

Keep an eye on this one. I do think though that, I don't know how fast, but clearly change is coming in healthcare with these tools being introduced, people introducing them in ways that they are actually safe in terms of just compliance with healthcare data security, in this case. And most importantly, with what we started with, how many people, even in the absence of those things, are already using AI for this application. It just speaks to the fact that we don't know when, we don't know how fast, but certainly think that the landscape is changing for sure.

Bobbi Pritt, M.D. (07:25):

It's coming, for sure. Well, thanks again for joining me, Bill, for a great discussion as always.

William Morice II, M.D., Ph.D. (07:31):

Yeah, as always, I really enjoyed it and look forward to coming back.

Bobbi Pritt, M.D. (07:43):

Welcome to the deep dive. We're going beyond the headlines today with Dr. Chris Garcia, Mayo Clinic Laboratories’ chief digital innovation officer. He leads our digital transformation initiatives, and he's here today to talk about what advancements he anticipates will shape clinical diagnostics in the coming year. Dr. Garcia, thank you so much for joining us today.

Chris Garcia, M.D. (08:05):

Oh, thank you so much, Bobbi. It's a pleasure to be here.

Bobbi Pritt, M.D. (08:07):

Yeah, it's always great having you back. And so I have a few questions for you. Let's start with a basic one. From your perspective, how has the way we think about digital innovation and diagnostics evolved? It's such a quick-changing field.

Chris Garcia, M.D. (08:22):

It really is. You know what's so interesting about the lab is that we didn't choose digital, initially. It was a necessary change. So, I mean, the lab is so high-volume, there was really no way to do all of the diagnostics and reporting that we needed. So, you know, the lab information system is digital and is one of the earliest digital systems. And then if we think about the science that has really pushed diagnostics forward, the science demanded it. Next-generation sequencing, mass spectrometry, really there was no way to provide that kind of information without the digital transformation there. And I think when it changed, is like whole-slide imaging. There wasn't something in the science that said, “You've got to have a digital slide.” But I think we're seeing the capabilities that are introduced as we elect to digitize. So like you see in parasitology, the wonderful system that your lab uses, and as we're seeing in precision medicine, there are now new digital assays on whole-slide images that just, now you need to have a digital image to get an answer on whether a patient can get access to that drug or not.

And so, the first phase was really, I think by necessity or by choice. It’s that whole physical –to-digital transformation. And then the next bit was bioinformatics. OK. So, you've got a bunch of information that comes out, you're trying to say, what does this mean? And then I think where we're really headed is the data-driven diagnostics, which is looking at patterns not only in that one dateset, but in patient outcomes and multiple test results and different modalities, like imaging in radiology, our lab testing, omics testing, and then really pulling those together and providing new insights into what this can mean for the patient. So, we kind of started at the routine basics, not really digital diagnostics, but digitized to diagnostics. And now we're moving towards insights for patients that just are not possible without that whole chain working, and quite a bit of different pieces.

I think it's that shift from bioinformatics to data-driven diagnostics that we're feeling so much, because that's the last decade. I think the hard part now isn't really the algorithms, because there's just so much. And you feel it. I feel it. Look at where the large language models have gone in the last two to three years, to where OpenAI and ChatGPT were kind of a novelty, and now they're changing workflows and changing industries. And so the hard part is not the application of the science is important on the data side, but it's the building of the trust, from, “We see a signal,” to “This matters for patient care and I'm going to make decisions, and my patients are going to make decisions, based off this information.”

Bobbi Pritt, M.D. (11:20):

I love all of this, Chris, how you've taken us through the history and we've gone back from the necessity of just managing our workflows, our laboratory information systems, our data, bioinformatics, next-gen sequencing. A human can't easily compute all of that. And you really want to have a bioinformatics system. But now you've taken us up to the present, and you're right, when large language models first came out, they were just a fun tool for me. Like generative AI. I would create cartoon pictures of worms, you know, but now we're using them every day. And in fact, that's what we're hearing from our CEO, is we should be using these tools in a responsible, thoughtful way every single day for patient care.

Chris Garcia, M.D. (12:05):

Yeah, it is quite the change. And while we should be exploring these and trying these, it is still us who's making the decision. I work with a lot of really sharp people, yourself included. And there's a colleague of mine who's a CTO of one of a smaller diagnostic company, and he was saying that he feels that the world is undergoing a reorg with these kinds of tools and capabilities. And in the past, a lot of the work would be done by us manually, in spreadsheets and aggregating it. And now, I think what he meant by reorg, is now everybody who has those tools is having to learn to act and direct as basically an executive. "Hey, I need you to pull this and pull this and synthesize the information,” but it needs to stop there." The decision-making still rests in the experts and those who can feel the patterns as opposed to deferring that judgment. And I think that's going to be the interesting thing that we see now is, there's a lot of concern from teachers and education, that students are going to be outsourcing their decision-making process to AI. And that's the last thing we want to do.

Bobbi Pritt, M.D. (13:09):

Yeah. Oh wow, this is really great discussion. And you know, it goes right into my next question I have for you, which is, what are the challenges you see? We can focus in on diagnostics. You're talking about like the broad impacts to society, but narrowing down on diagnostics, and what are those challenges that the digital solutions might be particularly well suited to help address?

Chris Garcia, M.D. (13:30):

Oh, that's a great question. You know, the digital solutions are really powerful when you are decoupling expertise from proximity. So you can see that with what we're doing right now over video and having a conference, a teleconference, together. Having AI support on top of that is key. I know I use it in Teams all the time, and it lets us make joint decision-making faster and come to decisions faster with our colleagues for diagnostic work. And really AI, I probably sound like a broken record, should be thought of as a colleague in that case. It is collaborating with you to make that decision. I think one of the major challenges in that part is that most labs don't have the pipes yet to connect this rich data information. It's really hard to send a whole-slide image from one lab to another lab. Our information systems aren't built to send orders that way.

So, I think that's one of the issues that we're facing in diagnostics, is we're not all connected, we're not all on the same bandwidth and able to take advantage of these tools in our workflow and how we communicate at this time. Now, I think the technology is getting there, the economics are still kind of hard to figure out to make it work, but just having us be AI-ready and being digitally part of that network is still something that we're struggling with. Another thing I would say is these digital solutions are really, really good at making complex processes repeatable at scale. So that could be flow cytometry. The more colors we get, the more complex it gets, the more help we need to be able to understand it. Like you said, we can only keep so many things in our head at the same time.

People say that if you can keep seven variables in your head at the same time when making a decision, you're like at expert level. Most of us come in at like three or five. So, there's this magic number of five plus-or-minus two when it comes to cognitive capabilities. And so, having that complex process for people skills great for digitization. The hard part is also not letting that AI or that solution make the decisions for you. You still have to know what's going on under the hood, still have to know when is it doing its job, when is it not? And I think that's one of the challenges. We're learning how to apply these technologies to the crucial work that we do in the lab. We're also trying to learn, how do we manage this appropriately? And I love what we're doing here, because at Mayo there's great people who are learning at the bench and learning at the laboratory and medical director level, and we're sharing that.

So nobody has all the answers, we're all learning this together. But I think that's a big part. We're trying to do more and more complex work. How do we make sure that we're still doing it at the same level of quality, same level of oversight, and same level of competence in the calls that we're making? And the last one, I think, kind of goes back to the case that I was mentioning before, which was, like in digital pathology, there's a Trop-2 assay now that's out there for a companion diagnostic. And so, it's this new lung cancer drug, and basically the only way to score it to say whether the patient would be a good candidate for it is with AI. And so it kind of goes back to, it's really fascinating. It identifies the cell, looks at how much color is in the cell, creates a ratio, and does it amongst as many cells in the tumor as you can.

So that's just something that we can't, like again, that's more than seven variables. That's a lot of different types. And so the system can do it really well, but now how do you do that and make sure that it's repeatable across many performing labs? How do you do that and make sure that it's repeatable within your own experts in your own performing lab? So, handling that complexity and these powerful tools in a way that is still appropriate and reliable and doing the intended purpose that we have for it, I think is really key for us. So clinical trust isn't a software problem, it's something that we're learning how to manage and how to grow.

Bobbi Pritt, M.D. (17:44):

I love these examples. And the last one really goes towards personalized and precise care. And, as we see digital innovation advancing our clinical diagnostic capabilities, what gives you the most optimism for our ability to support this more precise and personalized care?

Chris Garcia, M.D. (18:02):

I'm generally an optimist.

Bobbi Pritt, M.D. (18:05):

Same here. Same here.

Chris Garcia, M.D. (18:06):

I really love the potential, and the use cases that we're seeing, where these digital diagnostics are really democratizing expertise. It's really hard to see so many rare cases and edge cases in the world. And I think at our laboratory and at other specialized laboratories, we're very rare and fortunate to be able to do that and to provide that expertise out. But there aren't enough Bobbi Pritts in the world, and we can't clone you. So how do we use these technologies to provide that to underserved populations and expand and scale the service that we provide? So I really am not just an optimist in the concept, but I love the cases when they are coming up. "Hey, I didn't have an answer to this. This information was able to get to an expert who really knew this and provided an answer after years of struggle."

That's what really drives me, is that these are happening every day. You know, I can talk to our colleagues here at Mayo, and they provide lots of these, but also talking to groups around the country, whether they're at an academic institution or a local regional hospital or even, like in federally qualified health center. And seeing how these tools are providing answers that normally people couldn't get access to.That is, I think, one of those major things that I'm really excited about. The other is exactly in the precision medicine. We're able to get answers now or provide reasonable suggestions at a level to oncologists, to treating physicians in different areas, to help them make decisions that they didn't have any support for on beforehand. So some of the cancer treatments, it would be, "Well, do I give them a six-month course or do I give them a three-month course?"

And we know that longer treatments of chemotherapy can have adverse effects and there's different things to manage. And now there are, on the market available digital assays that are even starting to be integrated into clinical guidelines for professional groups that help answer and stratify these risks for patients and their treating physicians. And so, seeing these come from, "Oh, wouldn't it be cool if this was the case?" to seeing them integrated into clinical guidelines, is really exciting because it moves the technology, the approach, and this new concept, or I would even say just realm of diagnostics, from conceptual phase to, it's making daily patient impact. So, it's no longer curiosity. It is part of daily decision-making. That makes me very optimistic.

Bobbi Pritt, M.D. (20:51):

Yeah. You know, speaking as a fellow optimist, we really are seeing these tools starting to come into our daily work in a positive way. At the same time, you know, these digital tools are expanding very quickly, as we said, and it's hard to keep up with all of the literature and the new technologies and what we should pursue. So, what guidance would you offer lab leaders to help them select tools that would actually give them that meaningful impact?

Chris Garcia, M.D. (21:18):

My advice would be very practical. You are still the expert, and you have the best possible judgment on what's going to make a difference for your patients and the physicians and the whole care team. Come back one step. Don't think of it as a tool, don't think of it as a shiny must-have toy. Which, I am not saying that we are, but look at, what is the real problem? Does this really solve it? And then, come back to, is it financially sustainable? And I think those three are the major things that our teams can address. And the clinicians and laboratory scientists are the great ones at phrasing, "What's the real problem and does this really get us to where we need to be?" And then working with your team to say, "Is it financially sustainable? Is it flight-ready?" Those are all great questions that you can vet. Because some things are really exciting, have a lot of flash, and you can look under the hood and go, "Oh, it's not quite built yet." That's normal. Don't think that you've been swindled. That's not the case. That's just the pace at which is going, but if you really believe, "Hey, this is going to solve a problem," stick with that group. Develop it to make it what you need it to be. That's really the suggestions that I would give to people looking to this space.

Bobbi Pritt, M.D. (22:37):

I was asked this question, oh, a few months ago, when I was at a digital workshop as well, and I said the same thing. You know, we already have this experience. It's really very similar to bringing in a new instrument, a new test. But you have to determine, is it financially feasible? Is it solving a problem that you need? And then, what does success look like? How are you going to measure to show that you're actually achieving what you hope with this new tool? So, whether you need to do like, pre- and post-analytics, so you can look at data trends over time. You know, that's the sort of thing that we as laboratorians are used to doing. We do this all the time. And AI shouldn't be any different, or a new digital tool.

Chris Garcia, M.D. (23:18):

I think that's really well said. I remember being a trainee out and having a mentor, Sterling Bennett, out in Intermountain Health, and we were talking about these tools and how they're adopting, and he shared how important it is to not leave common sense behind. And so when you're looking at any of these solutions, one of the major things that he was thinking of, let's say we go through thinking about it as any tool, but it still feels a little foreign. An important thing that he shared and I still live by is how do you know when it's working and when it's not working? And that's one of the hardest things. There's not enough time for you and me to go and get Ph.D.s in data science and all the experience that it goes to, but you know the signals and your team knows the signals to say, "It's not working here. It is working here." And can you require that of your very eager, very tech-savvy vendors to make sure that they build that in for you? I think that's just, that's the extra thing I would add, is make sure that you are given the information you need when you have these, along with, you know, like the data performance, that you know when you're able to trust it and when you're not. I think that's one of those extra things that we shouldn't be shy in asking or demanding. We wouldn't shy away from doing that with any other laboratory equipment.

Bobbi Pritt, M.D. (24:34):

Right.

Chris Garcia, M.D. (24:35):

So why should we accept that from anything when it comes to data science and AI?

Bobbi Pritt, M.D. (24:41):

Really well said. I like that. Well, Chris, it's been wonderful to have you on as always, and I'm sure we'll have you back. This is, you know, such a hot topic. It's changing so quickly, as I said, and so there'll be more things for us to talk about in the future.

Chris Garcia, M.D. (24:54):

Bobbi, thank you so much for having me, and I look forward to it.

Bobbi Pritt, M.D. (25:02):

Let's wrap up with the top takeaways and how to learn even more on the topics we discussed. Dr. Morice first joined us to cover recent news, including Microsoft's launch of Copilot Health and how more AI-powered tools are emerging to support different users and needs at key points along the care journey. And then in the deep dive, Dr. Chris Garcia joined us to discuss digital transformation underway in clinical diagnostics. He shared how things have advanced in recent years and what he predicts is next for our industry. In the show notes, you'll find more information and links about the content from Dr. Garcia and the use of AI at Mayo Clinic Laboratories. Thank you for joining us today. If you haven't already, be sure to subscribe so you never miss an episode. Coming up next, I'll be joined by Dr. Julia Lehman, a dermatologist at Mayo Clinic, to discuss innovative tests used to diagnose autoimmune diseases affecting the skin. I hope you'll join us.