Latest / Becker’s Healthcare Podcast / Building Trust and Governance in the Age of Healthcare AI with Brian Anderson

Transcript

Building Trust and Governance in the Age of Healthcare AI with Brian Anderson · Becker’s Healthcare Podcast

• 0:02
  - There's a small circle of people who truly
• 0:04
  - understand what it takes to lead a health
• 0:05
  - system right now. The decisions are enormous.
• 0:08
  - The variables are constantly shifting, and the stakes
• 0:11
  - of getting a call wrong don't stay contained.
• 0:14
  - That's not a complaint. It's a structural reality.
• 0:17
  - And it's why the most effective c suite
• 0:19
  - leaders are deliberate about where they go to
• 0:21
  - pressure test their thinking.
• 0:22
  - Becker's fourteenth annual CEO and CFO roundtable
• 0:26
  - brings together more than 1,500
• 0:28
  - executive level attendees
• 0:30
  - and nearly 400 speakers for four days built
• 0:32
  - around growth, financial sustainability,
• 0:35
  - workforce strategy, and the leadership questions that don't
• 0:38
  - have easy answers.
• 0:39
  - Join us November second through the fifth in
• 0:41
  - Chicago.
• 0:42
  - For the agenda and event details,
• 0:45
  - visit beckershospitalreview.com
• 0:47
  - and click on the events tab in the
• 0:49
  - upper right.
• 0:50
  - Hello, everyone. This is Scott King recording live
• 0:53
  - at the first ever rural health care leadership
• 0:56
  - summit put on by Becker's. I'm joined by
• 0:57
  - a very special guest, Brian Anderson, president and
• 1:00
  - chief executive officer over at CHI. Brian, how
• 1:03
  - are you doing? Thanks so much for joining
• 1:04
  - us. Scott, it's great to be here. Before
• 1:06
  - we kinda dive into some big topics in
• 1:08
  - health care, do you mind just telling us
• 1:09
  - a little about your your role and your
• 1:11
  - background? Yeah. So I'm a pediatrician by background.
• 1:14
  - Been in the AI space for about fifteen
• 1:16
  - years.
• 1:17
  - We started CHI about four years ago.
• 1:20
  - It was started in the middle of a
• 1:21
  - pandemic.
• 1:22
  - And, fundamentally, it was about bringing together private
• 1:25
  - sector leaders, health systems,
• 1:27
  - payers, technology companies developing
• 1:30
  - consensus driven best practices,
• 1:33
  - on how to develop, how to deploy,
• 1:36
  - and how to monitor and maintain AI in
• 1:38
  - a responsible way for health care. Thanks so
• 1:40
  - much. Appreciate sharing that info. And, you know,
• 1:43
  - just curious with so much going on in
• 1:45
  - health care.
• 1:46
  - What's currently commanding most of your attention right
• 1:48
  - now as a leader? Yeah. So, I mean,
• 1:50
  - in CHI, we have 3,000 organizations. About 200
• 1:53
  - of them are health systems.
• 1:55
  - And whenever I talk to health systems,
• 1:58
  - you know, the past several months, the biggest
• 2:00
  - topic has been
• 2:01
  - governance
• 2:02
  - and specifically
• 2:04
  - how in the world do we govern these
• 2:06
  - complex agents that are doing multiple tasks
• 2:10
  - and complex workflows?
• 2:11
  - How do we monitor them? How do we
• 2:13
  - observe them? And how do we make sure
• 2:14
  - that they're doing what we want them to
• 2:16
  - do? Why do you think that's been such
• 2:17
  - a big topic? Like, you saw, like, the
• 2:18
  - past month. Like, re why do you think
• 2:20
  - it's kind of been kicked into higher gear
• 2:21
  - recently? I think, you know, certainly,
• 2:24
  - a lot of the progress made by the
• 2:26
  - Frontier Labs,
• 2:28
  - in building more and more capable
• 2:30
  - models that can do complex reasoning,
• 2:33
  - that can do multiple tasks embedded in, you
• 2:36
  - know, one prompt, and then marry that with
• 2:39
  - a lot of the products that,
• 2:41
  - technology companies are building on top of those
• 2:43
  - frontier models
• 2:45
  - is opening up the aperture
• 2:47
  - of opportunities
• 2:48
  - for health systems to say, okay. I have
• 2:51
  - all these existing workflows.
• 2:52
  - Some of them
• 2:54
  - high risk, some of them low risk.
• 2:56
  - But they're, you know, varying degrees of complexity
• 2:58
  - and saying, okay. Now I have these agents
• 3:00
  - that can suddenly do things
• 3:02
  - that
• 3:02
  - people,
• 3:04
  - traditionally
• 3:04
  - were only capable of doing. Well, now I
• 3:06
  - have an agent that can do that. They
• 3:08
  - might be able to do that in a
• 3:09
  - more efficient way, in a more streamlined way.
• 3:12
  - And then, you know, health systems are grappling
• 3:14
  - with the complexity
• 3:15
  - of deploying those models
• 3:18
  - in a way that is safe, that keeps
• 3:20
  - patients safe, that doesn't, you know, have the
• 3:22
  - wheels fall off the train,
• 3:24
  - based on how complex these workflows are that
• 3:26
  - they're deploying the models in. And speaking of
• 3:28
  - complexity, you know, obviously, in health care, it's
• 3:30
  - been a year of rapid change.
• 3:31
  - What recent decision or pivot required the most
• 3:34
  - leadership conviction from you personally, you think?
• 3:38
  - You know, I mean,
• 3:40
  - probably the biggest challenge
• 3:43
  - that we've faced in CHI has been
• 3:46
  - grappling with this concept of generating evidence
• 3:50
  - around how models perform.
• 3:52
  - I sort of alluded to it when I
• 3:54
  - was describing
• 3:55
  - agents and governance. Right? Because part of governance
• 3:58
  - of of anything is kind of measuring how
• 4:00
  - it's performing.
• 4:01
  - Right.
• 4:02
  - And,
• 4:03
  - you know, I think recently,
• 4:05
  - one of the things we've seen come across,
• 4:09
  - you know, newswires
• 4:10
  - is polls and surveys that show a growing
• 4:14
  - anxiety
• 4:15
  - or fear,
• 4:17
  - broadly in society, but also in physicians and
• 4:19
  - nurses about AI. Right. And so
• 4:23
  - for a long time, we focused in Shy
• 4:26
  - about developing these best practices. But I think
• 4:28
  - one of the things that we've, like, really
• 4:30
  - begun to double down on and pivot in
• 4:33
  - is if you wanna build trust
• 4:35
  - in these products, you have to show that
• 4:37
  - they're actually working. Like, demonstrate that this thing
• 4:39
  - is actually gonna save someone's life or help
• 4:41
  - someone live a better life.
• 4:43
  - And building that trust both for physicians and
• 4:45
  - nurses,
• 4:47
  - requires being able to measure how these models
• 4:49
  - are actually performing, but also educating providers on
• 4:52
  - how to use them appropriately. So really focusing
• 4:55
  - more recently on generating the evidence Yes. And
• 4:59
  - teaching providers how to actually use these tools
• 5:01
  - in a responsible way and not be afraid
• 5:03
  - of it. Can can you talk a little
• 5:04
  - bit more about the type of evidence you
• 5:06
  - wanna see to kinda indicate that any emerging
• 5:08
  - tech you're trying out to actually work in
• 5:10
  - is gonna benefit your system? Yeah. So, I
• 5:12
  - mean,
• 5:13
  - a lot of health systems, and this is
• 5:14
  - no offense to any technology company out there
• 5:16
  - listening, when they go through a procurement process,
• 5:19
  - you know, they get solicitations
• 5:20
  - and they get all sorts of promises
• 5:22
  - from salespeople
• 5:24
  - that make all sorts of claims about how
• 5:26
  - a model is gonna perform.
• 5:28
  - And, you know, that's one kind of evidence,
• 5:30
  - right, is is what is what a, is
• 5:32
  - what a technology company is gonna tell you.
• 5:34
  - But I think what is much more compelling
• 5:36
  - to the health systems and certainly the end
• 5:38
  - users, the nurses, the providers, and the patients,
• 5:41
  - is having
• 5:42
  - other health systems
• 5:44
  - share about how these how models that are
• 5:46
  - already deployed locally at their health system, how
• 5:49
  - that model is actually doing. Right. And in
• 5:52
  - doing that, you're
• 5:53
  - building
• 5:54
  - the kind of evidence that's needed not only
• 5:57
  - to make an informed decision at the procurement
• 5:59
  - point, but you're also building the evidence on
• 6:02
  - how to make that model better. Right? Because
• 6:03
  - imagine a future where you have multiple health
• 6:06
  - systems that are sharing how ambient listening tools
• 6:09
  - are performing. Right? One of the hottest,
• 6:13
  - set of use cases for AI.
• 6:15
  - And imagine health systems that are sharing,
• 6:19
  - having different performance. Right? That begs the question,
• 6:22
  - why is one model performing really well in
• 6:25
  - in an AMC, but that same model
• 6:28
  - is performing really poorly in a rural community
• 6:31
  - and a clinic there? Yeah. And why is
• 6:33
  - that? And so it's right through those kinds
• 6:34
  - of insights that we can better train and
• 6:37
  - improve these models, but also build the kind
• 6:39
  - of trust on
• 6:41
  - what the specific uses of these models are
• 6:44
  - that are gonna,
• 6:45
  - deliver that kind of good output or that
• 6:47
  - good performance, and what are the limitations.
• 6:49
  - Yeah. And that's kind of the the feedback
• 6:51
  - in the peer to peer learning we're trying
• 6:53
  - to get these conferences. And I think we've
• 6:54
  - we've heard that's happening today is, you know,
• 6:56
  - if one system's trying emerging tech and and
• 6:58
  - x y z has worked, x y z
• 7:00
  - hasn't, that can save another system a lot
• 7:02
  - of time and and money. So Right. Yeah.
• 7:04
  - Absolutely. And when you think about all the
• 7:06
  - forces reshaping health care right now, where do
• 7:09
  - you believe leaders need to be more decisive?
• 7:12
  - You know, I think that when we look
• 7:14
  - at this agentic
• 7:17
  - space that we're in right now Mhmm. I
• 7:19
  - think one of the really exciting things about
• 7:21
  - agentic tools
• 7:23
  - is their ability
• 7:25
  - to operate
• 7:26
  - more than just a tool. Like, I think
• 7:27
  - of, like, a point solution AI tool as
• 7:30
  - sort of a tool. But an agentic,
• 7:33
  - model almost
• 7:35
  - is closer to a member of your team,
• 7:37
  - a care team member. Yeah. And when we
• 7:39
  - think about the kind of uses for agentic
• 7:42
  - tools,
• 7:44
  - it's a real challenge for any large organization
• 7:47
  - to disrupt itself,
• 7:50
  - to take existing workflows that have been human
• 7:52
  - centric
• 7:53
  - probably ever since they've that organization's been created
• 7:57
  - and reimagine
• 7:58
  - what those workflows
• 8:00
  - could be
• 8:01
  - in an agentic future. And I think I
• 8:04
  - have seen a lot of health systems,
• 8:07
  - maybe appropriately so, have a degree of trepidation
• 8:10
  - about leaning into that space. Right? You don't
• 8:12
  - wanna disrupt
• 8:13
  - too many jobs. You don't wanna disrupt
• 8:16
  - too many complex workloads.
• 8:18
  - And I think the opportunity is there to
• 8:21
  - really lean in and learn
• 8:22
  - what's working, what's not, learn how to better
• 8:24
  - orchestrate these models,
• 8:26
  - learn
• 8:27
  - how to reimagine
• 8:28
  - an AI
• 8:30
  - first set of workflows versus taking a human
• 8:33
  - workflow and plugging an AI tool into that.
• 8:36
  - Right.
• 8:37
  - I think leaders that are more decisively reimagining
• 8:40
  - how to deliver health care
• 8:42
  - are gonna be the ones, obviously, I think,
• 8:44
  - that are gonna succeed. The challenge is getting
• 8:46
  - through that disruptive phase. Right. Absolutely.
• 8:49
  - Very well put, I think. And the last
• 8:51
  - question I have for you is,
• 8:53
  - as uncertainty has become kind of the new
• 8:55
  - norm in health care, what leadership habit or
• 8:57
  - mindset has been most critical in in keeping
• 9:00
  - your organization
• 9:01
  - and your team moving forward?
• 9:03
  - Yeah. I mean, so humility
• 9:05
  - is certainly the first one and listening. You
• 9:08
  - know, I I think so many of us
• 9:10
  - in this AI space live in a bubble.
• 9:13
  - We're surrounded by
• 9:15
  - others that are optimistic, that are excited about
• 9:18
  - this,
• 9:19
  - that are building these tools, that are deploying
• 9:21
  - these tools and using these tools.
• 9:23
  - And yet, right, when you take a step
• 9:25
  - back and look at
• 9:27
  - these polls and these surveys that I had
• 9:29
  - mentioned earlier,
• 9:30
  - right, I don't I've been in the digital
• 9:33
  - health space for about twenty plus years.
• 9:35
  - I have never seen an example where you've
• 9:38
  - had a digital tool that's been created
• 9:41
  - and then
• 9:42
  - widely adopted,
• 9:44
  - and yet the excitement about it has gone
• 9:47
  - significantly down. Now you might say EHRs are
• 9:49
  - are maybe one of those that that the
• 9:51
  - excitement about it has gone down, but the
• 9:53
  - utilization of EHRs, I think, is pretty significant.
• 9:56
  - AI, when it first came out,
• 9:58
  - particularly generative AI,
• 10:00
  - a few years ago, about 60% of the
• 10:02
  - population was really excited about it. Now when
• 10:05
  - you look at some of the studies and
• 10:06
  - surveys that came out from Gallup, that's below
• 10:09
  - 20%
• 10:10
  - for some parts of our population.
• 10:12
  - That begs the question why. Yeah. And I
• 10:15
  - feel like
• 10:16
  - one of the challenges in CHI is we're
• 10:18
  - a large member driven organization, and
• 10:21
  - we can't be
• 10:23
  - so
• 10:24
  - enmeshed with just the optimists that we're not
• 10:27
  - listening
• 10:28
  - to the real concerns that a majority of
• 10:30
  - our patients that we serve,
• 10:33
  - if we're physicians or nurses,
• 10:35
  - or technology companies that are deploying and making
• 10:38
  - these models,
• 10:40
  - that they're building them for,
• 10:42
  - are really anxious and concerned about this. And
• 10:44
  - so I think having a degree of humility
• 10:46
  - about understanding what those fears are and listening,
• 10:51
  - to truly understand how to make these products
• 10:53
  - better
• 10:54
  - to serve the needs of our communities, I
• 10:56
  - think that's desperately needed right now. I I
• 11:00
  - wanna follow that. That's really interesting about the
• 11:02
  - kind of the the excitement decreasing or not
• 11:04
  - perhaps a lack of excitement now. Do you
• 11:06
  - and you said it's sounds kind of fear
• 11:08
  - based from your experience. Do you think
• 11:11
  - that's fear of not using it properly, fear
• 11:13
  - of not knowing how to use it for
• 11:15
  - various reasons? I think it's that. I think
• 11:17
  - certainly that. I mean, a lot of, you
• 11:20
  - know, the challenge with explainability,
• 11:23
  - or interpretability of these models is how the
• 11:25
  - heck did it come up with the answer
• 11:27
  - that it came up with? Yep. Right. Right.
• 11:28
  - Gonna be the new emerging capabilities that are
• 11:31
  - gonna come out. Right? Is Yeah. Mythos or
• 11:33
  - mythos class like AI models gonna destroy the
• 11:36
  - Internet and cybersecurity?
• 11:38
  - So, yeah, there's a lot of fears grounded
• 11:40
  - in that, but I think a lot of
• 11:42
  - it is also grounded in what's gonna happen
• 11:44
  - to my job. Right? We're just talking about
• 11:46
  - this agentic future
• 11:48
  - and and a model that's now more akin
• 11:50
  - to being a member of a care team.
• 11:52
  - And, you know, the challenge of getting through
• 11:54
  - that disruption
• 11:55
  - of reimagining
• 11:56
  - what your workflows and your workflows looks like.
• 11:59
  - I think there's real anxiety, appropriate anxiety that
• 12:02
  - a lot of jobs
• 12:04
  - that people have today are gonna look very
• 12:06
  - different on the other side of this agentic
• 12:08
  - future.
• 12:09
  - And so I think speaking to that, listening
• 12:12
  - truly about, like, what those fears are and
• 12:15
  - working
• 12:16
  - with our teams if we're, you know, a
• 12:18
  - leader in a health system, how do we
• 12:21
  - create the kind of upscaling
• 12:23
  - for our workforce
• 12:24
  - to orchestrate these agents, to manage and monitor
• 12:27
  - these agents, the challenges that I shared earlier?
• 12:29
  - Mhmm. Those are the things that I think
• 12:32
  - will help address a lot of that anxiety
• 12:34
  - and a lot of that fear about the
• 12:35
  - future of work.
• 12:36
  - Doctor Anderson, thanks so much for joining us
• 12:38
  - on the podcast and for speaking here at
• 12:39
  - the conference. We appreciate it. Thanks, Scott. Thanks
• 12:41
  - for having me.