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

Today on This Week Health.

the most important piece of the implementation is really ensuring that as an organization, we understood the fundamentals of what was in place. But we didn't overhaul all of our opportunities in the organization before we went forward with notable. So we did not let perfection stand in the way of progress. 

All right, here we are from Noteworthy, which is a Notable conference, and I'm joined today by Justin White. Who, head of engineering, is that right? CTO. CTO. Okay. Is that not the head of engineering?

Well, so I oversee the engineering software developer team information security, and machine learning.

Yeah,

it's broader. Yeah. Got it. And Kristen Guillaume with North Kansas City Hospitals and Meritas Health. Yes. Wow, did I get that all right? You got it right. That's amazing. And what's your role at North Kansas City?

Sure, I'm the chief information officer and the executive over marketing and community health and wellness.

Fantastic. So, can I just call it NKC from now on? Yes, you can. Okay, great. first I'd love for people to get a little background on NKC and a little bit about your system, market you serve, size, those kind of things.

Sure. So, we're just north of the river, literally just off the river in Kansas City, which is a big dividing line in terms of care delivery in the Metro KC area.

And we serve about 350, 000 residents in our primary service area, Clay and Platt County. And we have a 451 bed hospital, a number of outpatient facilities and our clinics, 20 plus clinics. So, a nice size, smaller health

system. Now, the thing I love about this is we're going to talk about AI strategy, AI implementation, and I love the fact that you're a small health system.

I love the fact that you're not Intermountain, you're not, somebody with a massive budget and those kind of things, but you guys have made significant strides with regard to AI. Talk about your AI strategy.

Is there an AI strategy or are you just solving problems with the AI tool sets?

Sure. We look at tool sets. We look at tool sets, we look at platforms and we evaluate them to solve problems if you will, or to align with our strategic plan to help fuel this growth that we are seeking as an organization so, again, historically, we've been a smaller health system over the course of the past couple of years, and with new leadership, we've had a very significant focus on growth, and therefore, we needed to look at different capabilities and different platforms to help us fuel growth overall as a, as an enterprise.

So what kind of things can you do with AI as a small health

system?

Sure. So, one of the gifts that we've seen with Notable is we've been able to really become more efficient and we've been able to close the gap with regard to workforce inefficiency or the absence of workforce in some cases in some areas of the organization.

So when we had the jobs posted, we couldn't find people to work and therefore we looked at different technological capabilities to fuel that. So that's one example of our capability there. Additionally in the area of prior auth, a lesser mature area that we've been working on. Lesser mature for us in that we haven't had it deployed for so long.

We've leveraged PriorAuth where we don't have to add heads, if you will, or add head count to expedite prior authorizations to ensure that our patients are getting the care as soon as possible and they're not having to wait months at a time for PriorAuth.

It's pretty amazing. So, Justin, I want to turn to you and talk a little bit about the Notable platform.

So I just went through the demo room and it was really fascinating to me. Notable is a platform, it's an AI platform. And now some people might remember, I, interviewed Pranay long time ago, like Yeah, I think about five years ago. Wow. And he was talking about something completely different.

The company has evolved into this AI platform and there's a hey, we saw prior au we saw population health. We saw Personalized patient experiences and those kinds of things. Give us an idea of why we're calling it a platform, an AI platform.

Yeah, I mean at the core what we're really looking to do is take all of these workflows that often involve a lot of human touches that don't really need to anymore and automate that away.

Right? And so, we go and partner with amazing health systems like NKCH, Maritas Health, and we help just basically do a big process map. What are all your big, priority work streams and where can we come in and help intelligently apply automation so that we can either drastically improve the existing staff or augment the staff or really where there's just staff aren't even available.

And so we cover everything from which patients need to be reached out to for care. So there's a, under indicated on their chart Care Gap Outreach. Getting them scheduled in, once they're scheduled helping to streamline their administrative and clinical intake. And then once the visit's actually happened, to help, streamline the submission.

And then I look at this from a technical perspective, is there's a big workflow coordination problem. We need to get a bunch of data out of their existing systems of record. Do the automation of the processing on top of that. We'll reach out to the patients or to the staff where we need to, but do that as little as possible.

And then make sure that flows back in so that the rest of the system, all their existing processes can really work. And so that gives us a really nice framework where, okay, once we can ingest this data and we can generate that PDF and export it. Well , we can then work with our solutions team, not require really any additional engineering work, and we can use that to solve two, three, four additional problems,, because there's so many things that need to be automated in healthcare.

So it's really a truly platform where we can go in and configure it to match this health system specific workflows, and really use these kind of core underlying capabilities to drive that automation.

It's interesting to me because when we talk about these platforms a lot of times, health system leaders will say to me we have to get the data right first.

But one of the things I was looking at, which was really impressive, the numbers are always a little different. Some say 20%, some say 30 percent is structured within the medical record. But regardless, that means 70 percent or 80 percent is unstructured data. Which means we're not really getting the data right before we implement something like this.

First of all, how is that done? And how much of an effort is there once you engage with an organization to clean that data up at least to the point where it can be utilized.

Yeah again, we're sitting in a very fortunate time in history where all of a sudden all of that unstructured data that is locked away in the medical charts we can now structure it.

AI has gotten very good at taking that data. We don't do too much on the like... Image interpretation or clinical interpretation, but all of that text data that's locked up in the notes that tells us about hey this patient actually came in and They have indications for heart disease that's on a note about the x ray interpretation, but it isn't actually structured on the chart correctly Historically, that was very difficult to identify because a physician can word that in 500 different ways.

And so, designing a computer algorithm that can grab that robustly and accurately, get that correctly mapped into the correct structured field, historically a very difficult problem. Now with the advent of LLMs and other modern NLP techniques, we can actually ingest those charts, take that unstructured data out, structure it, and get that put into that right place using our integration framework.

And these are just brand new capabilities that, they're getting better month to month, and just an amazing, way for us to be writing together. It's,

it is pretty amazing. The I want to talk to you about the patient experience. One of the things that sort of struck me 📍 as I was going through that demo lab was We'll talk about a system.

I don't want to get there yet. I do want to talk about it. But it was interesting to me, the number of ways that we are reaching out to patients, the number of ways that we you talked about the unstructured data, the population health booth, what I was looking at was essentially they were able to identify people for outreach, but that outreach was all automated.

What we used to do is create this huge list, send it over to marketing, and marketing would like do whatever they're going to do. And that was a ton of work, but what I saw in there was essentially a a list that is not created like at one time, but it's just constantly... The outreach is just constantly happening.

Is it? Yes. Talk about the patient experience.

Sure. So it is iterative and it is ongoing and This month is Breast Cancer Awareness Month And we have had and continue to have a number of different MAMO outreaches going on to remind people to get their mammogram It is amazing what we've been able to achieve in terms of outreach with our patient population and further it is Just a wonderful thing that we're seeing all ages of our population respond to these outreaches.

So they're receiving the outreach on their email, they're receiving the outreach on their phone, and they're responding, Yes, I'd like to schedule an appointment right now. When can I come in? And then if they have to reschedule, they click reschedule, maybe the next day something comes up, and they get to reschedule.

And so the adoption has been very significant across multiple care gaps in our organization. mammo is just one of

those. It's interesting, we were just in a room with a bunch of CIOs talking about adoption, and there's concern about AI adoption, but it's seamless, isn't it? I mean, internally, are you saying, hey, this was done with AI, or is it just like...

Oh, this is really great. This sort of presents itself, and I just do these things.

Sure. No one, from a population standpoint, we haven't had that inquiry. However what we have received as a result is upwards of 97 percent satisfaction. So, we ask them about the experience at every outreach, if you will.

Did they have a good experience? And it's overwhelmingly positive. But we from a public standpoint, they're not asking, is this AI? And yet We're able to make it so efficient, and one lady said, I scheduled four appointments for my four children in less than 10 minutes because you made it so easy.

One lady said, I am legally blind, and this was the greatest, most easiest appointment scheduling experience I've had. And likewise, we carry those forward in terms of care gap outreach.

Yeah, the ADA compliance is really interesting. Built it that way from the get go which is which is amazing. I want to talk about the implementation.

Sure. Because I'm, we're sort of portraying like an easy button here. You have a smaller health system, I assume don't have a ton of resources to implement. I assume you're coming alongside and helping, because that's what you refer to either. But talk about an implementation and what that looks like.

I assume there's, just like every implementation There's sort of a learning to work with each other. There's some data that comes in that's not necessarily. Nice,

I mean, from a technical point of view getting the data in and out of the health systems existing core data systems, that's probably one of the biggest challenges, right?

Oftentimes, we're integrating with systems that have been in place for decades and, the whole entire, hospital runs around that, right? And so, wherever we can, we'll use APIs, we'll use structured interfaces and, especially with a partnership like we have with Meritas we're able to do a fair amount of that.

But unfortunately, that doesn't get you all of the way there. And, in large part that's why a lot of this innovation has been so difficult to realize in healthcare, is just how hard it is to get that data in and out. And so we've designed our platform to have a very flexible integration architecture.

This is like the foundational layer upon which everything else is built, where we'll pay, pull data out over HL7, over FHIR APIs. We'll use SFTPs. Flat file exchanges that were originally meant for call center interaction, but it, gets us access to what we need. And then where we have to, we'll fall back on RPA, robotic process automation.

And this is, again, another huge benefit of the current wave in AI, is that all of a sudden, not only can computers take that, unstructured data out of the chart, But we can give it a screenshot of the interface and we can semantically understand what we're looking at and then go and interact with it just like a human would.

And so then I'm not relying on the fact that, oh hey, this system, which very well could predate the internet, doesn't have an API for this particular data field I need to read or write from. I can go through and, automate that just like a human would. Now, admittedly, it's considerably more work for us to do it, but if that's what it takes in order to actually deliver a, truly delightful experience that allows that mom to book those appointments or allows, that, that patient to, actually get their mammogram scheduled, we're more than happy to do that.

So you're not just talking about the EHRs, you're also integrating with some payer systems, some submission systems. Yeah,

I mean, anywhere you need to get data in and out of, if that's going to be a core kind of workflow for the health systems, yeah, well, we do a ton of work on authorizations with the payer portals.

We're starting to do more and more work with various CRM providers to get data in and out of there. If they have a data warehouse and they'd 📍 like a real time data feed, out of notable into those systems, we're happy to do that. And then, of course, the, electronic health record and practice management software.

Absolutely. So, talk about the implementation from your perspective.

Sure. So, as you said, we're a small health system, and we have a few people in IT in the organization, and we didn't have the luxury of adding headcount over the course of our journey with Notable, and we still definitely haven't done that.

But what we've done is we've leveraged the individuals who really have a fundamentals understanding of our operational processes and our IT processes. And then essentially leverage their platform to close the loop there. Our implementations I would say have been weeks at a time. So, we first went and implemented, we'll use scheduling as an example at the 20 plus clinics.

Clearly primary care was, faster on the adoption than we went to specialties. But it was weeks, not months. And then we moved on to our outpatient services and outpatient facilities. And again, weeks, not months in that regard. So it's been a wonderful experience. No new IT headcount.

Again, the most important piece of the implementation is really ensuring that as an organization, we understood the fundamentals of what was in place. But we didn't overhaul all of our, maybe opportunities, if you would, in the organization before we went forward with notable. So we did not let perfection stand in the way of progress.

And it has definitely served us well.

It's, again, sitting in that room with the CIOs, they're talking about skill sets. Oh, we need these skill sets. But I assume you didn't add N L P gen ai, you didn't add those kinds of resources 'cause you really didn't need to.

We really looked to our partner to work with us in that regard.

Yes.

You've dropped a lot of technology names and I want come back to those things. . So you said NLP, you said RPA, you said Gen ai. That LLMs LLMs. So, this is really a layered approach to, it depends what the problem is. You're just bringing a different Yep. Solution set in there. Talk a little bit about the extensibility of this, because we are, we're not at the end of the race with these LLMs or any of these models.

And there's gonna be new models coming. So, talk about the extensibility. Yeah, I mean,

it comes in a couple of flavors for how to keep this extendable. One, first and foremost, like a meta optimization that we do, and I think Meritas does very well here, is that we prioritize our organization on how do we optimize for speed.

Right, so engineering velocity and our ability to rapidly iterate is probably the thing that we most jealously guard. And so making sure that when these new capabilities do come around, we have, the capability to rapidly, figure out how to adopt those, how to ingest them, and how to leverage them.

And then another big piece of that is the system architecture is designed in a flexible way where Sure, today we're leveraging Azure OpenAI APIs quite heavily for our large language models. But, while we're doing that, we're also plugging those into Anthropic and Google MedPalm and, making sure that we have, a good foundational layer where, you know, as those capabilities change, as you know, capabilities emerge we have the architecture set up so that we can rapidly shift and adopt.

And then the last piece is. We talked a lot about integration, getting data in and out, being a big challenge for healthcare. Well, just as important though, is making sure that the way we deploy this, the configuration, the actual workflow that we're automating, is NKCH's workflow. It is not how notable wishes NKCH workflow would be.

We have a lot of work that goes into our configuration layer. And that allows us to dovetail our config so it actually operates how they operate. We don't have to retrain their staff, update all the training materials, things like that. the side benefit of having a configuration engine that allows us to go in and do that is that when we go to another system, we can go in and dovetail to them.

But then even more importantly is when a new solution comes out , we already have that kind of foundation configuration layer that allows us to quickly plug in, you know, hey, here's a new output that we want to drive. And that's been really powerful for us.

Yeah, I want to ask you if there was another thing that you could do with Notable, what would it be?

And it was interesting yesterday, I was standing here with Pranay, who's the CEO. And he was showing me their Slack channels. And so when people, I don't know why this impresses me so much. But when somebody gives feedback on the experience patient feedback and that kind of stuff, it shows up in this channel.

And, you have the good ones and the bad ones. Now, they're mostly green, but even the bad ones, you can just look at it and go. We should do that. Like, they're getting that feedback almost instantly. Is there something you'd like to do with it at this point, or there's still so much you can do with it?

We have a

list. We have a very long list. We have HTCs on our mind. These are all, most of what's on our list they're working on, or they have worked on with others, and it's a matter of us getting to that particular project, but the use cases are pretty much endless.

They really are.

Yeah, I mean the ACC is really interesting because you can get very specific in terms of those outreaches. I just want to thank you. I mean, it's really interesting to me that That 📍 there's a platform that, NKCH, is is building on top of these really advanced technologies, just using a no code, front end like flow, and you can just essentially identify the way that your organization wants to work, and then build those out based on that.

Thank you. It's really fascinating. I want to thank you for your time. Justin, thank you. Thank you. Kristen, thank you. Thank you.