Amanda Wilde (Host): Welcome to the Healthcare Executive Podcast, providing you with insightful commentary and developments in the world of healthcare leadership. To learn more, visit ache.org. I'm your host, Amanda Wilde.

In this episode, we are focusing on AI And the next era of diagnostic intelligence with Quest Diagnostics, Vice President of Digital Solutions and Interoperability, Nicole Antonson and Chief Medical Officer at Quest Diagnostics, Dr. Yuri Fesco. Quest Diagnostics is one of ACHE's premier corporate partners. Our premier corporate partners support a's vision and mission to advance healthcare leadership excellence. For more information on Quest Diagnostics, please visit the corporatepartnerspage@ache.org.

We have two fantastic experts with us today. First, I'd like to welcome back Dr. Yuri Fesco, Chief Medical Officer at Quest Diagnostics. Dr. Fesco, great to have you with us again.

Dr. Yuri Fesko: Thank you for having me on today.

Host: A pleasure. And joining us for the first-time is Vice President of Healthcare Solutions and Interoperability at Quest, Nicole Antonson. Nicole, a very warm welcome to you. Thank you for being here.

Nicole Antonson: Oh, thanks for having us.

Host: And just before we dive into our topic of AI, I'd like to ask you each to tell us a little about your role at Quest and what you focus on. Dr. Fesco, could you start?

Dr. Yuri Fesko: Absolutely. So, I'm the chief medical officer. But I oversee medical affairs, payment reimbursement strategies for the company, as well as oversight of a lot of the medical portions of everything from compliance, to medical quality, to where are we going as a company. So, that's, in a nutshell, my role.

Host: Very comprehensive. And Nicole, you have a long title. Can you explain a VP of Digital Solutions and Interoperability?

Nicole Antonson: Most definitely. Well, I partner with Yuri in many endeavors. And the team is focused on creating digital interoperable solutions. It is to ease the workflows of patients and providers. We want to be a companion to the patients and providers to help them reach their next best action, help them reach their desired outcomes. And we do that by creating our own applications as we'll talk about AI Companion later today. But we also create APIs and services that can meet the user where they are. A patient may have an app that they desire to use. A physician has an EHR that they need to use to conduct their work. And so, we will fit within those. So, the team is focused, again, on creating digital and interoperable solutions to ease workflows for patients and providers.

Host: That's so comprehensive. And the two of you integrating, I can imagine, really helps us see the big picture. We're seeing AI integrated into many facets of healthcare. So, the two of you together form a formidable team and cover all these comprehensive topics.

As we look forward into the future and looking forward into the future, AI is a really timely topic as it is developing in the moments as we speak. Let's start with the big picture. We are seeing AI integrated into many facets of healthcare. From a diagnostic standpoint, Dr. Fesco, where is AI making a tangible impact right now and moving from nice-to-have to a real baseline expectation?

Dr. Yuri Fesko: So, I think there are lots of areas that it's moving into healthcare. But I think that's probably the one that first developed and has probably matured quite a bit, that I would like to talk about is the area of digital pathology where, in the country, we're facing a shortage of pathologists. Most of our listeners probably know that the average pathologist in the country is approaching the age of 60. We have a shortage of pathology. And we have to think about ways of doing pathology differently than we did before with a sort of decentralized network as well as integrating digital solutions that incorporate AI into them.

One of the ones that I was involved in, working with our pathology network, is in prostate cancer. So, I'm a GU medical-oncologist by training. We worked with a company to basically develop a pathology solution that allowed AI to assist the physician in identifying where is the lesion in a prostate core biopsy. Making sure that we are identifying, yes, the cancer is present, giving the physician a workflow where it helps them to not make the diagnosis per se, but point them in the right area. We are also working to help with some of the grading of these.

One of the things that has been very problematic in prostate cancer is that grading is done by a Gleason score. And a lot of times, when you look at clinical trials, grading can vary from pathologist to pathologist. It's very subjective. Allowing an AI system to assist in that grading is very important. It allows some of the subjectiveness to come out so that it's leading to better outcomes for a patient. Identifying patients that are truly high-risk, who are unlikely to respond to traditional therapies or may, in fact, require more aggressive therapies. Again, think of it as more of a copilot, increasing diagnostic accuracy, but not truly giving up authority. Ultimately, it is the physician who is at the helm.

Host: Yeah, that's really important to note because that's a fear with AI, isn't it? So, this is one of the examples of the way AI is being integrated in diagnostics and even dealing with some of the shortages that we are having in medical fields. Nicole, beyond the lab, how are we seeing AI being activated in healthcare?

Nicole Antonson: As you know, we think about the complete diagnostic process and it doesn't just end with a result. But really, we think about it begins with a patient visit. And it ends with understanding and action. We were just talking about next best action. AI can help create that seamless connected experience between the patient and the provider across multiple handoffs.

So, think about, as you're managing multiple encounter types, as you're moving from acute care to chronic care, AI can help bridge the gap between those different encounters, really help a patient navigate their whole journey. But we talk about there's a patient understanding gap. And I have a single number. But what does that really mean on a report? And we need to help remove that confusion, move the patient towards empowerment, help them understand the context. So, I take that one result, or I even use AI to identify a trend over time to then inform what they need to know. Like, "What is that test? Why was I tested for that?" And then, what do I need to do next? Or what comes next? Help me bridge to that handoff, to that next step within their healthcare journey. That's the key, I think, responsibility of a diagnostic partner, that's our responsibility.

And so, we came up with the AI Companion. AI Companion is the answer to beginning to close that understanding gap. So, it's a tool. It sits securely within our MyQuest portal. We use AI to translate that complex lab result into simple, clear language, lets you interact and ask questions about that result. And then empowers the patient to ask questions; maybe, "Hey, for the last couple results, everything's been in green. But can you tell me, am I trending in the wrong direction even though I'm within range of a lab result?" And then, maybe it comes back and says, "No, actually, you're going in the wrong direction even though you still are in green." And you can prepare for your physician's visit. With that information, you can ask the AI tool, what questions should I ask my physician the next time we talk? And so, you know, it's really filling the gap of understanding what is that result about, and then forming questions to ask the physician to then get to that next best action to then drive the desired outcomes, again, creating that seamless connected experience between the patient and the provider.

Host: So, the way you describe it, it really enhances patient experience and the monitoring of information helps. The patient understands better and have more tools at their disposal, and also bridges communications with the doctor. It's clear these applications are quite broad from our discussion so far. Dr. Fesco, as we have touched on with so many AI tools, emerging trust is a major concern. In terms of data protection, how do you ensure that these advanced technologies are deployed safely and ethically?

Dr. Yuri Fesko: As Nicole said, she and I partner to do it. So, I think it breaks down into three areas, you know, expert oversight, making sure you got a really rigorous validation set and making sure you got security by design. So, what do I mean by expert oversight? That's where Nicole and I really partner— her team and I partner—to sort of, "Hey, what's best in class? What's emerging? How can we do this better? What is the question that we're looking to answer for the patient?" to basically give them some actionable insights into their laboratory and their diagnostics.

When I say rigorous validation, we have to do a continuous set of validations on it. We have to also make sure that it is matching up to clinical standards. So, Nicole mentioned before, we partner—the medical team where I have physicians and nurses who are reviewing these answers that the AI is providingl; making sure that there's nothing in there that could be less than desirable clinically, that that there's misinformation or there's something that we would not consider clinically relevant to the patient.

The other thing that I think is very important that, again, I know Nicole and I have taken very seriously and partnered on, is making sure that there's security by design. So, you are trusting somebody with your medical records, essentially; your laboratory data. So, you got to make sure that it is HIPAA compliant. That's sort of the basics, making sure that you're not sharing information, but that there's really data security here. Putting your laboratory results or your medical records up into a public database. There really isn't that degree of HIPAA protection, nor is there the data security or compliance necessarily put into those applications. Within Quest, we are obviously a medical company that does deal with diagnostics as well as laboratory, pathology. We are always thinking about making sure that it is HIPAA compliant, that it's done in a method that meets medical standards for data accuracy, that there's data privacy. And increasingly, that's a very important component of this new realm of AI in healthcare.

Host: And then, can you expand on what makes an integrated tool like the AI Companion more innovative and valuable to a health system than just telling patients to use a public search engine?

Dr. Yuri Fesko: So regarding that, Nicole and I have really partnered about like what are the data insights that we want to give the patient? We've done things like understanding whether or not some laboratory results are inpatient versus outpatient. Understanding the context that this is done in. If a person just uploads their data, there isn't necessarily that context. And so, sometimes, an AI engine does not understand the context and is just interpreting this as, "Hey, listen, there's a bunch of data here. What are the insights that I'm gleaning?" I think it's very important, and we really do try to make sure that we incorporate multiple data sources. Patients increasingly interact with a lot of different platforms that are out there, and there may be data points that are missing from those interactions.

In addition, as I said before, there may be some components that really should not be interpreted by the AI because there were circumstances that were going on that may make the AI misinterpret—I'm taking an extreme example—but somebody who is intubated and sedated. And if it is not interpreted in context, it could say, "Hey, listen, this person has severe sleep apnea." Well, you need to understand that this was an inpatient hospitalization. This patient was on a ventilator and was intubated and sedated. Those laboratory data should not be interpreted in the context of a well patient that is outpatient. It's those kinds of details that we have a team that is working on what does the next generation of this look like and how do those insights actually help the patient and the clinicians that are working with that patient find trends, find things that may affect the patient's healthcare years before they become clinically apparent.

Host: Understood. Now, Nicole, this leads to a critical decision for health system leaders. There's a real tension between trying to build these complex AI capabilities in-house versus partnering with an organization that has deep expertise. How can health systems build out desired experiences for their providers and patients?

Nicole Antonson: It really comes down to looking at key competencies and where a hospital system or a partner wants to invest their capital. And, you know, health systems own patient care. That's what they do. That's the core competency. That's why we all look to them. Diagnostics leaders excel. It's scalable, cutting-edge diagnostic intelligence. We focus on those results, those instruments, translating those results into digitizing them so that they can be interpreted and used by the end user. I think it's really where the power comes is where a health system and diagnostic leader can partner together to create the best in class solution. So, taking that expertise in patient care, taking the expertise in creating results and digitizing those results so they can be then inserted, integrated within a workflow. That's where I think the magic happens.

I think we've seen an organization's first instincts is, "Oh shoot, I want to build it on my own. I want complete control." And I think that what needs to be weighed within the pros and cons is knowledge. Like, there's challenges of building like technical infrastructure, like knowing like what Yuri was saying, like bringing all those data sources into one, ensuring that they're in a secure environment and that you're curating that user interface, that visual interface of how you may interact, with that data, the additional staffing that takes to keep up, from a design experience to the additional staffing needed to be monitoring performance. There's ongoing cost of just keeping up with the pace of AI. It changes daily, maybe hourly at this point, right?

You know, as we prepared our AI Companion, while it may look like a simple chat on the side of a screen, a lot of work naturally went into it. And just to shed some light on some of the key elements to make sure that that was a secure, safe solution is that, not only do we have a large language model That is, running behind, but we have a large language model as a judge watching over it.

We also have other tools that are looking for biased and drift, et cetera, constantly. We have tools that Ensure that performance is there. We hired a red team to try to attack, you know, the AI Companion as well. We have humans in the loop watching over the answers that come back. We also use guardrails. So, really, to Yuri's point, making sure that our responses are in line with medical recommendations. And then, turn a patient to a physician because it may not be the lab's place to provide care, right? You point it back to the physician.

So really, large language model as a judge, hiring a red team, making sure our guardrails are in line to create safe and secure answers. And then, I think the final thing that we do, as a, organization building digital solutions, we are able to do controlled rollouts. So, we can roll it out at 5%, 10%, et cetera, and monitor from a performance safe, secure answers, and from a cost perspective. Because as you're running these AI models, there's a cost component to that too.

So, I think it's a strategic and cost value evaluation that needs to go on. And I think that you choose a partner for speed to market, because we built that infrastructure there And we're watching it every day. And we have the scalability and the expertise because we built these tools over and over again, and we have the teams here at the ready. And I think it allows the partner to focus on their core mission. It frees them up to that health care systems's precious capital and their talent to focus on what they do best, and that's delivering patient care. So, I really think it's the partnership between the diagnostics lab company and the health system coming together to create a best in class type solution like the AI Companion.

Host: Nicole. With that in mind, looking forward, what do you think are the most important questions a health system leader should be asking their diagnostic partners about their AI strategy today?

Nicole Antonson: One of the main questions is Really looking beyond the lab, but how does the AI strategy empower both that clinician and the patient to create that connected, seamless experience that helps determine the next best action what is needed to reach that desired outcome that both the partnership between the patient and the provider desire to obtain? So, I think it's asking that question, how do you create that connected, seamless experience to reach desired outcomes?

I think that frames expectation of a true partner to solve. I think also some of the things that we mentioned, Yuri and I both mentioned, around making sure that that partner has the infrastructure behind it that is creating safe, secure solutions. It is really important. I think that that frames expectation around a true provider, must solve for both sides of the equation, patient and provider. But then, also, creating a scalable, secure, safe solution as well and has expertise to do so.

I think doing that helps a partner feel confident that they are not only reducing burnout and improving patient experiences, but they're doing it in a safe and secure way. Our real goal is to create these scalable, secure solutions that really help bridge that gap that we have seen between a patient understanding their information need to keep that secure to then communicate to providers so that together they can determine the next best action to reach the desired outcomes.

Host: Excellent. Dr. Fesco, anything to add to that?

Dr. Yuri Fesko: So, I would say, to our clinical partners out there, you know, how are you really putting AI to basically improve the diagnostic focus, improve clinical accuracy, and really the patient outcomes? I think what we have done at Quest Diagnostics, you know, Nicole and I, Nicole does have a healthcare background. She's a nurse by training, but has done lots of other things since then. But we do keep clinical care at the center of our focus.

I think it's very important, the things that we are working on, everything from a digital strategy. They're dealing with real-time problems that are going on as a nation, the aging pathology population. How do you do it more efficiently if you're not going to be able to staff a pathologist in every hospital? How are you going to do this? You're going to need a pathology network. It will need to have some AI capabilities to basically make those pathologists that you have as efficient while improving patient safety, improving diagnostic outcomes.

When we talk about things like our AI Companion for diagnostic results, that is really understanding the increasingly decentralized healthcare system where laboratory results may be coming back from a variety of different sources. It may be coming back from the health system, but it could be coming back through Quest Diagnostics, through a number of different partners that are out there. And increasingly, you rely on a partner like Quest Diagnostics to put those insights together, not just for the patient, but also for the clinician who's helping to care for that patient so that they get a much better picture of what is going on overarching with this patient while maintaining it within those HIPAA compliant platforms that allow data safety, and also ensure that there are quality controls in place, that the insights that you're giving have a real, connection to clinical care. And I think that's what we have to all be looking for in this journey in AI within the healthcare spectrum.

Host: This has been a very comprehensive discussion with lots to think about. Thank you both so much for sharing your expertise today in this timely discussion of AI. That was Quest Diagnostics, Vice President of Digital Solutions and Interoperability, Nicole Antonson, and Chief Medical Officer at Quest Diagnostics, Dr. Yuri Fesco.

Quest Diagnostics is one of ACHE's premier corporate partners. Our premier corporate partners support ACHE'S vision and mission to advance healthcare leadership excellence. For more information on Quest Diagnostics, please visit the corporate partners page at ache.org. And if you enjoyed this podcast, please share it on your social channels and explore our entire podcast library. This is the Healthcare Executive Podcast from the American College of Healthcare Executives. For more information about the American College of Healthcare executives, visit ache.org.