This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

I'm Bill Russell, creator of this Week Health, where our mission is to transform healthcare, one connection at a time. This is an executive interview

quick powerful Conversations with Leaders Driving Change. So let's get started.

All right. It's an executive interview today, and today I'm joined by Yaw Fellin Wolters.

Kluwer Health Senior vp. And General Manager, UpToDate, clinical Decision Support and Provider Solutions and Proud Penn State graduate. Uh, maybe not this week, but, in prior weeks, uh, it, it was, um, you, I want to thank you for coming on the show. There's a lot happening in your space uh, these days, clinical decision support, generative ai you know, what's what's top of mind, what's going on in the industry that you wanna talk about today?

Yeah, so first Bill, just really appreciate you having me on top of mind is somewhat easy. I think generative AI is the thing that's top of mind, both, you know, within our world and our business, and then also within the clinical decision support space. And so I would say right now our primary focus is really on how do you find reliable.

Safe ways to apply the promise of this technology to clinical decisions that are being made at scale with hospitals and health systems?

Well, I mean, one of the things, you guys have so much history and there was a big to do when OpenAI put out ChatGPT 5 and they decided to put healthcare right in the middle, and it's like, Hey, you've got this diagnosis, go ahead and talk to ChatGPT and whatever.

But I mean, you guys have curated. Expert content for decades and chatGPT is just sort of saying, yeah, ask us, let's see how it, let's see how it goes. and physicians were kind of taken aback by, by the fact that a tech company was saying, oh yeah. You know, just as good as all the other stuff that's out there.

Yeah. You know, in some ways. I I think there's a history of this. You know, you'll see a lot of the big tech players, whether it was Google when it first came out, or, you know, apple has long had a presence that continues to evolve into the healthcare space. So in some ways, I would say, you know, not shocking.

But you know, for a company like us that spends all day, every day. Thinking about how do you make the best clinical decisions? I would say that's mission number one. And then importantly, really across the last decade, we've been investing a ton in how do those decisions become harmonized across key players?

So think of your physician recommendation being harmonized with the information that a patient would receive as they're being discharged or as they're being cared for. And then you think of. You know, ChatGPT entering the space. You know, it, it does add a lot of new complexity to that dynamic and in some ways you know, I think we all stand back and marvel at what the large language models can do in other ways.

You, you know, you now have patients who are, you know, really having conversations with ChatGPT uh, about their symptoms uh, and. You know, it, it almost takes that mission or that objective to harmonize content to a new extreme just in terms of both importance and complexity.

So you've spent an awful lot of time integrating this into the workflow over the years.

But now you're starting to go in the direction of generative AI on top of your data set. what makes that special in the industry? Are there guardrails? I mean, how does, what does, how does it work?

I do. Fundamentally feel like we've got a different and kind approach to the application of generative ai.

And I think some of the limitations and downsides of generative AI are well known, right? It can tend to hallucinate, so it can make up information it can be very authoritative, which is you know, obviously a challenge when. You need to make nuanced clinical decisions. And then also I think it's critical just to.

You know, remind ourselves this is a very new technology and it's a new space and it's evolving rapidly. So if you think of our approach, what's unique about it is, first and foremost, we're grounding all of the work that we're doing, not on the broad based, you know, content that's available across the wide web.

And that has varying degrees of quality, but we're grounding it on UpToDate content and UpToDate content is. Trusted. It's trusted because we have 7,600 authors, contributors, editors that make sure it is pristine and accurate and always up to date, hence the name. So I would say that's one component of it, but it's also proprietary in the sense that, you know, a lot of this.

Work gets to the proliferation of medical research that's out there and the amount of studies that are published every day, week, month. One of the things that's unique about us is we take the time to go through all that information to prioritize which of those studies are important, and then actually to author an opinion if those studies happen to conflict.

So I would say first piece. Is just that it, it starts with grounding in a trusted, validated, authoritative source. The second piece is we've been focused on really getting our experts in the loop of these AI systems. So Bill, I'd be curious to your perspective on this, but I had an aha the other day that

the way that we start to feel safer about AI systems is we say that there is a human in the loop, right? Typically when we deploy that statement, people feel like something is safer and. You actually can't just have any human in the loop. My aha moment was I would be a horrible human to be in the loop of a clinical decision support AI system.

Not because I'm not a well-meaning, I'm raising three kids, I think, relatively successfully, but because I've never trained as a clinician and so I cannot just based on my background, arbitrates, so. We're working extensively to get those experts in the loop in the right way.

And then we're designing for enterprises. I think that's the third piece. So large hospitals and health systems that are deploying this, to, tens of thousands of doctors and that need to ensure that the systems that they're putting out there are reliable and trustworthy and can scale.

You asked me a question, which is dangerous 'cause because I answer questions. you know, I think about this, one of the things in the AI era that I keep harping on is quality data is infinitely more valuable than quantity of data. That qualified data becomes you know, just.

Becomes how you how you educate these models in a way and you keep the guardrails on them to say, Hey, stay within this framework. Don't go out to you know, to Google search or don't go out to whatever, like stay within this framework of studies and whatnot.

The other thing is, as much as I want human in the loop, I want AI in the loop. I've seen the power of it, being able to interrogate you know, 15 years of medical history of me The thing I want to ask you about is that context. Conversation, which is when, when you have a chat GPT or a Gemini or whatever coming in, they say, look, we can consume everything that's in the EHR. And that's what, that's where our value is.

I'm curious how you guys are approaching, providing context and then hitting your curated dataset.

You know, first I would just say I think getting the right context in at the right moment is the true opportunity in this space, right? So, you know, from a mission or objective perspective, we view that as, you know, on the critical path to delivering the most value to, you know, customers, patients, communities, et cetera.

Getting context in is hard. You know, in some ways it's getting easier. So the most basic way that people would enter context into our systems is just by typing it into a query. But generally, I think what you and others are pushing on is how do you take all of the context that exists? In the systems of record today.

So that's typically the electronic medical records. And or I think the other frontier for context is, there's this simultaneous wave around adoption of ambient technologies. And ambient technologies are also gathering, a tremendous amount of context. So, I would say first and foremost, it's on the critical path.

I think. Ideal state and we've got, actually got a lot of work in this space. We've been working on context aware clinical decision support for the last decade and certainly, you know, for the last, you know, four and a half, five years that I've been here at Wolters Kluwer. So historically we actually have some good approaches to doing this.

I would say they're more rules based and they leverage. Deep integration into electronic medical records. So we actually have a pilot site going right now. A couple weeks ago, they demonstrated tremendous outcomes and what they're doing is they were taking different patient characteristics.

They were comparing those to our clinical decision support pathways, and then they were prompting. Improved or better, you know, a actions in partnership with the clinician. We're working on hypertension, so these happen to be around, improving blood pressure screening or, you know, just some of the, you know, kind of key quality metrics associated with seeing those patients.

But the new frontier is how do you start to do this with generative technologies I think that opens a lot of doors. So some of the previous friction around. which data do you grab and how do you process it and how structured is it and how much data can you process? You mentioned earlier kind of this I idea of there's a history of bill that's out there to be consumed and if you can appropriately process that and get it into the application, that's tremendous context.

So, so yes, we are following that path. I think the challenge there is. Probabilistic technologies do have some wiggle room and variants associated with them. And so what we're working on is how do you design the systems so that you can be certain in the outcomes or range of outcomes that you might be prompting around the context of your care.

Given the kind of the pros and cons of these technologies.

we see chat, GPT, we hear about this, you know, people standing up. Chat GPT instances within their clinical setting. I don't suppose that's a major competitor at this point, but we are hearing open evidence talked about it's good and it's bad, quite frankly, at our 2 29 meetings.

I'm curious how you position up to date versus an open evidence at this point.

A couple thoughts, right? One is they're very well funded, and so there's certainly a lot of, you know, discussion around that these days.

Two hits, hits to your previous point. I do think folks are experimenting, whether it's with chat GPT or open evidence or, Gemini models or what have you. You know, physicians are scientists, our key customers or scientists by nature and, I think this is a tremendous time for the scientific process where people are kind of putting these, you know, through, through the paces.

I do think there are some key differences in the approach that, open evidence and others are taking one as I noted, we're heavily focused on having experts in the loop. So, , actually getting our 7,600 experts and these are not just any clinician, these are truly the top, thinkers and leaders in their field that contribute to UpToDate.

There really is a belief in big tech that we're just gonna be able to somehow grab this information from somewhere and make high quality decisions. You know, I use generative AI every day for programming and I will say it's getting better, but it's still a really good intern. Yes.

Yeah.

No, I it's such a great way of thinking about it. I mean, I'll digress for a minute. One of the common benchmarks that people put out there on these AI systems, at least in the healthcare space, is how did they do on the US MLE, right? the MLE is the test that you take.

Basically as you to move into your residency period, and I was having, you know, obviously I'm not a clinician, but I was having conversations with our clinician peers and customers and. What are the tasks that you might ask your resident to do? You might ask your resident to go do research but you're probably not asking you know, a resident for advice in the same way that you would ask, the top expert that has been practicing medicine around, you know, any given.

Disease or you know, condition for the last 30 years and has that depth of practical, real world experience. So, I think the analogy, you know, holds up for sure.

This can be a fascinating space to watch and a fascinating space for you to be in because it's moving so rapidly at this point. But I like the fact that they're using scientific methods across the board. They're measuring results, they're measuring outcomes based on that.

It'll be, you know, science-based outcomes.

And Bill, if I could put two other things to your question. I think second difference between us and some of the others out there is the approach and the business model. So when you look at some of these companies, it is a advertising based business model, which I think in a lot of ways conflicts with the mission of, providing, just the absolute.

You know, evidence-based answer at the point of care. So I think that is a difference that needs to be considered. And then the third piece is really around you know, designing for hospitals and health systems at an enterprise grade and an enterprise level. And I know, you've got a background , as a.

CIO and a leader within hospitals and health systems. And so, again, maybe you could probably answer this part better than me, but those requirements, you know, feel very different than. What it takes to serve a consumer or an individual professional that's purchasing a subscription.

I'll tell you what, in Southern California we could not employ the physician groups.

And so we had all sorts of EHRs and some of the EHRs were actually that business model you were talking about pharma putting ads in there and whatnot. And later that went to a court case and that EHR was fined significantly for contributing to the opioid crisis. And I'm like, yeah, I'm not sure that business, that business model, well, maybe there's a different way to do it today, but generally it has not worked out real well for healthcare.

Well, hey, I want to thank you for your time. I wanna thank you for coming on the show and the work that you're doing. We will definitely keep in contact with you as this pro progresses. I've never seen the pace of technology move this quite this fast. It is really something

I agree.

I, and I do think that's the opportunity you know, we're. Everybody is moving much faster, particularly us at Wolters Kluwer and up to date. So we'd love to, you know, come back. I really appreciate you having me on. We'd also love to come back and just, you know, keep you updated as we make progress.

  📍 Thanks for joining us for this executive interview with me, bill Russell. Every healthcare leader needs a community they can lean on and learn from. Subscribe at this week, health.com/subscribe and share this conversation with your team. Together we're 📍 transforming healthcare.

Thanks for listening. That's all for now.

📍