June 14: Today on TownHall, Brian Young, a System Physician Informaticist at CommonSpirit Health talks with Bill about analytics, AI, and the communication between his team and physicians in a 138 hospital system. Where does Brian see analytics and predictive modeling moving into the future? What is a top of mind problem that he is currently working to solve? How does he introduce new systems to physicians?
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.
Do you have to turn that note into some, semblance of discreet data in order to start playing with it?
There are both vendor side and open source mappings for medical terminology. I like to use the term. It used to be heavy metal was the example, it means something different in medicine than it does in music.
Welcome to This Week Health Community. This is TownHall a show hosted by leaders on the front lines with interviews of people making things happen in healthcare with technology. My name is Bill Russell, the creator of This Week Health, a set of channels designed to amplify great thinking to propel healthcare forward. We want to thank our show sponsors Olive, Rubrik, Trellix, Medigate and F5 in partnership with Sirius Healthcare for investing in our mission to develop the next generation of health leaders. Now 📍 onto our show.
Here we are from the Scottsdale Institute. I'm here with Dr. Brian Young w the enterprise physician information, maybe not the, but an enterprise physician infer Madison's with common spirit health. Brian, nice to nice to catch up. We just had lunch together.
It's been good to talk to you.
This is like a follow-up conversation to our lunch conversation.
So what what's top of mind right now for you in your role at common spirit?
Well, I think as you mentioned in our lunch conversation the scale of the new merged organization is pretty large and wow.
How big is chemistry?
It's changed it a little bit with the divestiture of the Iowa facilities that was announced recently, but I think we're somewhere in the high, a hundred and thirteen, a hundred and thirty eight hundred thirty nine hospitals.
Now we have 68 obstetric centers, one in every 40 births in America occur.
I tell people you're not at 16 hospital system. Like, wow. That was a hundred and thirties just at that scale is, is pretty huge. So you end up focusing on system issues or do you focus on specific?
Yeah, I think, how different health systems rolled with it, analytics and data science back several years back was there were some bottom up type projects and some top down.
Bottom-up tend to be solutions that don't scale. Well, so our CMIO wanted us to start with the things that were aligned with enterprise strategy. And so we're under the office of the chief medical officer. We aligned with board goals and try to do analytics to support execution
on the board.
So where's big data.
These might be two different questions where is big data, where's predictive models where where's analytics going.
Well, I think it's going to solve the. Need for a longitudinal notion, if you will, some indication of a, of a human being the patient over time, as opposed to kind of transactional snapshots or glimpses.
And you need that to what I was speaking with you earlier about to do a real-time analytics to bring relevant information as things, Flux and change
How do we do that Well, we, we hear so much here that speakers will get up and talk about the 20, 80, 20% of your health is related to healthcare and 80% is related to other factors outside of healthcare.
Do you try to look at that problem or is that something, or are you mostly focused on the 20 making the 20. Better the, the time when we're actually in front of the clinician
Well, yeah, I think we were talking and I was talking with you about the advent of cloud server architecture and scalable compute and scalable storage to try to get to real time.
So if you're At a level of capability and maturity in your organization to move toward the cloud and be able to do that. Then you start looking at the 80% and how to do, that if not, you have to kind of leverage what you can get your hands on. We've been doing more leveraging what we got our hands on. When we were dignity health with a 38 hospital system, we had at least span in the inpatient arena over, what was happening in the Cerner EMR and associated systems.
So that was, pretty good. But now that we're larger scale and we've got four EMRs and we need to mung notes from a lot of sectors and try to execute on genomics and all this stuff. We need to kind of move to the next level. That's where the, the
top of mind problem that is in front of you, that you're trying.
In the short run, you, everybody talks about artificial intelligence. And I did a slide once that had a little Jeep guy in a Jeep at the bottom, and it's titled it's a hill climb and it kind of moves up the path from a structured semi-structured unstructured data on, up into imaging and video and down near kind of the first toe in the water for artificial intelligence is natural language process.
And you can find vendors. That'll do that for you, but it's fairly expensive. If you can execute on that within the walls of your organization, we've done some. That dipping the toe water, trying to leverage. I don't know if you've heard in your past, but predictive models of any kind, usually creatively undergo some lift when you ingest note based information, that's the dark data.
So going after that I think has been a little bit of a subtle to medium priority to try to leverage what's in the notes that we have. It's it's water going over Niagara falls right now at the scale we're at. But if we can dip into that and do some things as far as
Do you have to turn that note into some semblance of discreet data in order to start playing with it.
Yes. There are both vendor side and open source mappings for medical terminology. And so, some hybrid of, you know, based on what you can or can't do within your organization of, of leveraging a vendor to do that, to turn it into a discreet medical concept. I like to use the term. It used to be heavy metal was the example it means something different in medicine.
than it does In music, a funnier one to me was we talk about getting mucus up out of the lungs is pulmonary toilet and pulmonary toilet. Those words outside of the medical record. You know, Who knows what that would be, but when it means something in healthcare that has to be mapped and, made sense of, and those mapping systems would do that for us.
When we turn the, the note text into a
hundred and he, and come spirits grew through acquisition. You talk about the 38, that sounds like the core. That's where you were before the, I want to say each guy's saying. C H I C H Y mixing those letters up. So Chi you probably ended up with a lot of different systems and a lot of implementation of some of these models.
If you're looking at a sepsis model or something, does that become a lot more complex or is it just, is it not complex because of the technologies is complex because of the governance and the scale.
A little bit governance and scale more from the standpoint of the unsolved interoperability problem in mapping between multitenant, EMR.
The theory of doing, asepsis predictive and all that I think is less of a hill climb than actually doing it. If you get the okay to do it in epic and Cerner and Allscripts in Meditech. And
so you don't get much pushback from physicians. I mean, if you have a model that you can show them, Hey, here's, here's how we're coming up.
They look at it and go, yeah, that will help us in the process.
Yeah. True. If, and particularly, fortunately, we're going to team that brings data. We try to bring, the validation of that, that it's, that it's good data. And then it's telling them the right things they need to know once you do that it gets pretty hard for folks to slip out from under, like, why would you not want to use that?
It's automated and ready to go,
but that's, that's one of the pushbacks we're getting on AI models is the black box. We're not just going to put this stuff in here. We need to understand how did it come to this? Even if it's a recommendation, not a, not a diagnosis. How did they come to this recommendation?
Because otherwise I, almost have to ignore it because I don't understand what it's doing.
Right, right. And so I think that's where it's important to one solution is to start with small pilots where you have a very kind of tight structure in, around the trust and provenance of not only the data, but what the.
model is Suggesting a physician to do you also want to build solutions like that that allow the physician to object and opt out? They don't feel relegated as in the model told me to, and I didn't feel like I had any choice to do anything different. You want to create a. culture Around that there are some things, afoot to try to make the black box, less black boxy, but by nature of the build and how those deep neural networks work, it's just a node with weighting supplied at each node.
And you don't know, what's out of 30,000 to a million variables is being weighted in that node. And then you move to the next node and guess what? It's different stuff being weighted. It's hard to,
I want to go back to where you started reporting to the CMO, take your direction really from system initiatives and objectives.
I think some people are listening to this or they're hearing that, and they're going, that's interesting because , where data science and all those things sort of grew up in the company, they ended up starting in different spots, maybe stayed in those spots and whatnot. It sounds more relevant and more sound to be at a strategic level and let, let the data sort of take you to, Hey, here's, what's going to have an impact.
Here's what's going to impact the physician's life. Here's what's going to impact the patient's life.
Right. It gets to that whole system of belief around if you're, you can't change what you can't measure, if you can measure it, predict it, put some analytics and technology in the. You have to be prepared to execute on it.
And there has to be a kind of longitudinal engagement with that a lot of times to get the full delivered value. And so, our team structure was I think very strategic and shrewd when it was first built to have us start at that top layer. And then if we need to do some point solutions or whatever, we've got experience in chops and maybe we could justify it, but if you can't.
The meaningful stuff that the, direction the battleships heading in. If you can't align with that, it's, it's really hard to get out up out of the starting ball.
How does the frontline physician bubble up something? If they say, Hey, here's an idea that would really impact the organization?
Well, as you'd mentioned, I mean, if there's, if there's governance structure around it, usually there's an entry path for physicians to come. If they want to, and, pitch an idea, pitch some. We're certainly open on our team. We're still , fairly small, even though the organization is large to hearing and listening and trying to determine if something has the capability to scale and helping develop and build something like that.
So I wouldn't say there's lots of limits to at least the idea of being, brought forward. Yeah. It turns into a project or not would be subject to a lot of things. A lot of other things. Yeah. Yeah. Yeah.
Brian, I want to thank you. Thank you. Thank you for lunch.
And no, I really.
I love this show. I love hearing from people on the front lines. I love hearing from these leaders and we want to thank our hosts who continue to support the community by developing this great content. We also want to thank our show sponsors Olive, Rubrik, Trellix, Medigate and F5 in partnership with Sirius Healthcare for investing in our mission to develop the next generation of health leaders. If you want to support the show, let someone know about our shows. They all start with This Week Health and you can find them wherever you listen to podcasts. Keynote, TownHall, Newsroom and Academy. Check them out today. And thanks for listening. That's all for now.