Latest / Becker’s Healthcare Digital Health + Health IT / Designing Health Systems for Better Outcomes

Transcript

Designing Health Systems for Better Outcomes · Becker’s Healthcare Digital Health + Health IT

• 0:00
  Hi, everyone. This is Laura Dierda with Becker's
• 0:02
  Healthcare. Thank you for tuning in to the
• 0:04
  Becker's Healthcare podcast series. We are thrilled today
• 0:06
  to be joined by doctor Holly Urban, vice
• 0:08
  president of business development strategy and clinical effectiveness
• 0:12
  at Walter's Kloer Health. Doctor Urban, it's a
• 0:14
  pleasure to have you on the podcast today,
• 0:15
  and thank you so much for joining us.
• 0:17
  We are excited to dive into your journey
• 0:19
  and experience.
• 0:21
  Thanks so much, Laura. I really appreciate the
• 0:23
  opportunity
• 0:23
  to, speak with you all.
• 0:25
  Fantastic. Well, let's dive right in. Doctor Urban,
• 0:28
  could you introduce yourself and your work in
• 0:31
  the health care to our listeners?
• 0:33
  Yeah. Absolutely. I'm Holly Urban. I am a
• 0:35
  pediatrician by training.
• 0:37
  I practiced in primary care pediatrics for several
• 0:40
  years before starting a second career in health
• 0:43
  care IT, and I've been in health care
• 0:44
  IT for about 15 years.
• 0:47
  In multiple leadership roles, I was vice president
• 0:50
  of product management at McKesson Horizon Clinicals,
• 0:53
  and then recently was CMIO at Oracle Cerner
• 0:56
  for their federal business.
• 0:58
  And I started working at Wolters Kluwer this
• 1:00
  year in,
• 1:02
  and my role in business development
• 1:04
  is really focused on innovation opportunities and how
• 1:07
  we do innovation,
• 1:09
  with some of Walter Scott's most influential and
• 1:12
  important products, which are our,
• 1:14
  up to date product,
• 1:16
  our up to date patient engagement, and, up
• 1:19
  to date Lexi drug content, formerly known as
• 1:21
  Lexi Comp.
• 1:22
  People who don't know the UpToDate product, it
• 1:25
  is probably the most used clinical decision support
• 1:27
  product used by physicians and other clinical users
• 1:30
  across the US. We have about 2,400,000,
• 1:34
  clinicians who use UpToDate,
• 1:36
  and clinicians really trust the content. They know
• 1:39
  that our 74100
• 1:41
  medical experts
• 1:43
  who are leaders in the field,
• 1:45
  they curate the content. They author the content
• 1:48
  as well as it's being updated
• 1:50
  every business day. During COVID, it was every
• 1:52
  day.
• 1:53
  So it's constantly being updated with the most
• 1:56
  important, most recent evidence. And so the our
• 1:58
  customers who use the content have a lot
• 2:00
  of trust in the content.
• 2:02
  So we did a big study of of
• 2:03
  providers
• 2:04
  this year that said, how do you think
• 2:06
  about generative AI tools? Since I'm in innovation
• 2:09
  with UpToDate, this is something I think about
• 2:11
  a lot.
• 2:12
  And, you know, we did learn that over
• 2:14
  40% of these users are are ready to
• 2:17
  use generative AI. And that's been a change
• 2:19
  over the past year. So that's been increased
• 2:21
  quite a bit, 68%
• 2:22
  over they're they have changed their views on
• 2:24
  generative AI since last year.
• 2:26
  But 91%
• 2:28
  said that any generative AI tool, they would
• 2:30
  wanna know where the materials were sourced and
• 2:33
  and know that they were created by physicians
• 2:35
  and medical experts,
• 2:37
  and they'd also want their vendors of j
• 2:39
  Gen AI tools to be transparent where the
• 2:42
  information came from.
• 2:43
  So
• 2:44
  as we think on up to date clinical
• 2:47
  decision support,
• 2:48
  we we know that there's some limitations. And
• 2:50
  and I'm and I'm not talking about,
• 2:53
  some of the really amazing tools out there
• 2:55
  that are all about physician efficiency. So, like,
• 2:58
  you know, some of the ambient listening or
• 3:00
  chart summarization tools. What I'm really talking about
• 3:02
  is
• 3:03
  AI tools that are focused on, clinical decision
• 3:06
  support, like diagnostics
• 3:08
  decision support or therapeutic decision support. And, you
• 3:11
  know, to be totally frank, we see a
• 3:12
  lot of limitations.
• 3:14
  You know, hallucinations are 1. I mean,
• 3:17
  even today, it continues to hallucinate. And when
• 3:20
  you're talking about making care decisions for a
• 3:22
  patient,
• 3:23
  that level of risk is just unacceptable for
• 3:25
  most clinicians. Again, it comes back to trust.
• 3:28
  So if if you're if you're concerned that
• 3:30
  there may be a hallucination, how can you
• 3:32
  possibly trust what the what the LM is
• 3:35
  telling you? The other is the inconsistency
• 3:37
  of the response. You ask the same question
• 3:40
  3 days in a row, you could get
• 3:41
  different answers from an LLM, which again,
• 3:44
  I think would impact trust. Right? So it
• 3:47
  how can you trust what they're really telling
• 3:48
  you if you're not getting a consistent response?
• 3:51
  And then the final, and this is something
• 3:53
  that I think about a lot is,
• 3:55
  is patient context. So
• 3:58
  one example is so let's say you,
• 4:00
  put it into an LLM,
• 4:03
  there's a patient who has a urinary tract
• 4:04
  infection, a bladder infection,
• 4:06
  and they're allergic to penicillin, so what antibiotic
• 4:08
  should I use?
• 4:10
  And you'll get a response, it's probably a
• 4:12
  fluoroquinolone.
• 4:13
  Great. I've got my answer. I'm gonna go
• 4:15
  write a prescription for a fluoroquinolone
• 4:17
  for my patient.
• 4:18
  Well, the challenge there is you don't have
• 4:20
  the L and M didn't have all the
• 4:21
  patient context. What if it's a pregnant patient?
• 4:24
  Patient is pregnant. You do not wanna be
• 4:26
  prescribing a fluoroquinolone.
• 4:28
  What if it's a child?
• 4:29
  Again, you do not wanna be prescribing fluoroquinolone.
• 4:32
  What if it's a recurrent bladder infection? What
• 4:34
  if it's a complicated bladder infection? There's just
• 4:37
  a lot of patient context that's there that
• 4:39
  you don't get in that LLM response,
• 4:42
  and that makes it really challenging again to
• 4:44
  trust what the response you're getting. And that's
• 4:46
  why at UpToDate is we're looking at some
• 4:49
  of these models is we're also really gauging
• 4:52
  how do we make sure that
• 4:54
  no hallucinations at 0 risk, so it's really
• 4:56
  grounded in the up to date content,
• 4:59
  as well as giving those nudges or giving
• 5:02
  that guidance to say,
• 5:04
  have you really thought about all the patient
• 5:05
  contextual factors?
• 5:07
  Well, it's fascinating to hear. You know, there's
• 5:09
  some really impressive results and things you're talking
• 5:11
  about that make a big difference in the
• 5:13
  clinical space. So
• 5:15
  excited for what's ahead, but I know as
• 5:16
  you mentioned, still things to work through. Yes.
• 5:19
  More hospitals and health systems right now are
• 5:21
  also getting serious about value based care. As
• 5:24
  a former care provider for pediatric patients, how
• 5:26
  do you view this shift,
• 5:28
  toward more proactive preventative care? Yeah. It's it's
• 5:31
  really interesting to me. So as a pediatrician,
• 5:33
  I mean, you're sort of grounded
• 5:35
  in proactive preventative care from the beginning. It's
• 5:38
  really the core of pediatric medicine.
• 5:41
  I mean, we
• 5:42
  do growth charts where you measure a patient's
• 5:44
  growth against the growth chart. And if they
• 5:46
  fall off the growth chart, that may be
• 5:47
  a sign of a problem. We do developmental
• 5:50
  screening to make sure kids are hitting their
• 5:52
  develop developmental milestones.
• 5:55
  We do anticipatory guidance to say, hey. Have
• 5:59
  you thought of you know, are you doing
• 6:00
  the right diet nutrition for your patients? And,
• 6:02
  of course, we do immunizations
• 6:04
  to make sure that you're preventing disease. So,
• 6:07
  so having that sort of proactive
• 6:10
  preventative mindset is very core to what I
• 6:13
  always did and what all pediatricians do. And
• 6:15
  so when you think about the broader scale
• 6:17
  of health care, you know, most physicians aren't
• 6:19
  trained in that model. It's really more, I'm
• 6:21
  gonna treat the patient that's sitting across from
• 6:23
  me in my office or that's in the
• 6:24
  hospital
• 6:26
  bed. So you're not thinking about the patient
• 6:28
  the other
• 6:29
  364 days of a year when you're not
• 6:31
  seeing them because there aren't economic drivers to
• 6:33
  do that, frankly.
• 6:34
  So I really that's why I've been an
• 6:36
  early proponent of moving toward the more value
• 6:39
  based care to really understand how do what
• 6:42
  do we need to do for our patients
• 6:43
  in those other 364
• 6:45
  days so that they aren't consuming health care
• 6:47
  services, so they are staying healthy, so they
• 6:50
  don't have to come see me. And that's
• 6:51
  what value based care is all about.
• 6:54
  That makes a lot of sense. You know,
• 6:55
  it is amazing to hear a great vision.
• 6:57
  And I I love the idea of pediatrics
• 6:59
  being so proactive,
• 7:00
  from the get go and then instilling that
• 7:02
  within other areas of the health care system.
• 7:05
  How can health care
• 7:06
  leverage systems thinking and design systems that facilitate
• 7:10
  the delivery of the highest quality care to
• 7:12
  generate better outcomes,
• 7:13
  increase safety, and improve efficiency,
• 7:15
  especially in this move towards value based care?
• 7:18
  Yeah. Those things these things are very intertwined.
• 7:20
  So what I would argue is that in
• 7:23
  how we call our health care system,
• 7:25
  it isn't a system at all.
• 7:27
  It's a very broken, fragmented,
• 7:30
  siloed
• 7:32
  with very, sort of perverse economic incentives.
• 7:35
  And and the data shows that the system
• 7:37
  isn't working. You know, if you look at
• 7:39
  the Commonwealth Fund,
• 7:41
  every year they publish that we spend
• 7:44
  3 to 4 times more in our health
• 7:46
  care dollars
• 7:47
  than other developed countries. And yet our outcomes
• 7:50
  are significantly worse. I mean, it's very, very
• 7:53
  sobering
• 7:53
  that we are just we are just not
• 7:55
  getting value for the health care dollars that
• 7:57
  we spend.
• 7:58
  And so that's why I've been talking a
• 7:59
  lot about how systems thinking needs to be
• 8:02
  addressed. Because if you if you throw in
• 8:05
  other issues like, you know, workforce burnout
• 8:08
  and,
• 8:09
  and bias, like, if you don't think about
• 8:12
  those issues with the systems thinking approach, it's
• 8:14
  almost impossible, because they're systemic issues. So how
• 8:17
  do you fix it without taking a systematic
• 8:18
  approach into a broken system? So
• 8:21
  what systems thinking is, it's all about designing
• 8:25
  a system to meet the outcomes you're looking
• 8:27
  for.
• 8:28
  So that's like, in health care, obviously, it's
• 8:31
  better care at lower costs.
• 8:33
  And, you know, so many health care systems
• 8:36
  do process improvement or quality improvement
• 8:38
  where the focus is really on fixing people.
• 8:41
  Like, how do you change their mindsets or
• 8:43
  how do you change their workflow? But, again,
• 8:45
  I'd argue that if the system that they're
• 8:47
  working in isn't designed to meet the outcome
• 8:49
  you want, just trying to fix the people
• 8:51
  is not gonna lead to sustainable or systemic
• 8:54
  change. So that's why I really advocate for
• 8:55
  this broader systems thinking to make sure that
• 8:58
  you're designing the entire system and all of
• 9:00
  the people in it and all the processes
• 9:02
  and all the technology
• 9:04
  to meet your intended outcomes. And there's an
• 9:06
  interesting,
• 9:07
  so the leaders at the IHI published an
• 9:10
  article in the Harvard Business Review in July,
• 9:13
  that talks about this. So if anybody wants
• 9:15
  to read more about it, I highly recommend
• 9:17
  the article. It's really well done. And they
• 9:19
  call out, James Clear's book, Atomic Habits, and
• 9:22
  he says, you do not rise to the
• 9:24
  level of your goals. You fall to the
• 9:26
  level of your systems.
• 9:28
  That's fascinating. What a poignant,
• 9:31
  thing to keep in mind in the health
• 9:32
  care space and truly
• 9:34
  overall just a really important,
• 9:36
  lesson, to think about on a regular and
• 9:39
  daily basis. Systems are definitely make a difference.
• 9:42
  Now I know you're currently working on a
• 9:43
  publication that's going to highlight the generative AI's
• 9:46
  potential in health care and practical applications for
• 9:49
  clinicians.
• 9:50
  Can you share your insights on how AI
• 9:52
  can disrupt the traditional health care models
• 9:55
  and what challenges health care organizations should really
• 9:57
  be aware of? Yeah. Absolutely. So the, the
• 10:00
  textbook is Digital Health AI and Generative AI,
• 10:03
  a concise practical guide for clinicians. It's gonna
• 10:06
  come out,
• 10:07
  Springer Nature's publishing it edited by Terri Adiram
• 10:10
  in in 2025.
• 10:12
  And my chapter is on implementing
• 10:14
  IA tools.
• 10:16
  Again, you have to take a systems thinking
• 10:18
  approach when you're implementing these tools. It's not
• 10:21
  a very simple plug and play, like, I'm
• 10:23
  gonna plug in an EKG machine,
• 10:25
  get a patient the leads on, and I'll
• 10:26
  get an EKG out of it. It's far
• 10:28
  more complex than that. You really need to
• 10:30
  think about all of the people and processes
• 10:32
  that are gonna be involved with the tool
• 10:34
  itself.
• 10:35
  So just to take an example, let's talk
• 10:37
  about,
• 10:39
  a a patient inbox where you're where you're
• 10:41
  using AI to do triage of patient messages.
• 10:44
  So I would say taking a systems thinking
• 10:47
  approach means you start with what metrics are
• 10:49
  you gonna follow to know that you're going
• 10:52
  to,
• 10:53
  meet the outcomes you're looking for, and who's
• 10:56
  tracking them, and who's owning them, and who's
• 10:58
  reporting out on those metrics,
• 11:00
  what resource
• 11:01
  is managing those,
• 11:03
  triage messages,
• 11:05
  how are messages
• 11:06
  rerouted to other clinical users when needed, and
• 11:09
  what are the criteria to reroute,
• 11:11
  and and auditing? So are there any messages
• 11:14
  falling through the crack? That's that's to cover
• 11:17
  risks.
• 11:18
  Is there AI drift? I mean, we know
• 11:20
  a lot about these AI models that they're
• 11:22
  trained in, you know, with the vendor side,
• 11:24
  but once they get into patient population models,
• 11:27
  that they can drift and they can change.
• 11:29
  So who the only way to really
• 11:31
  see that happening is through audits. And then
• 11:34
  finally, addressing issues of bias. I mean, doing
• 11:37
  subgroup population analysis to make sure that you're
• 11:40
  not seeing any bias in the in the
• 11:42
  AI response. And then finally, the most important
• 11:45
  is governance. Right? Who has governance over all
• 11:47
  of that? So you really need to be
• 11:49
  thinking people process and technology when you're implementing
• 11:51
  these tools.
• 11:53
  That's fascinating to hear and certainly
• 11:55
  a very poignant reminder in terms of how
• 11:58
  you can really
• 12:00
  most effectively implement these tools, get them out
• 12:02
  to the broader system, and make sure that
• 12:05
  they're used and optimized correctly. I know that
• 12:07
  can be a big challenge, and a lot
• 12:09
  goes into it. So
• 12:10
  fascinating to hear. Is there anything else you'd
• 12:13
  like to share with our audience that we
• 12:14
  haven't already touched on? No. I think that's
• 12:16
  great. Thank you again for the opportunity to
• 12:18
  chat today, and, I enjoyed the conversation.
• 12:21
  Doctor Urban, thank you so much for joining
• 12:23
  me on the podcast today. It has been
• 12:24
  such a pleasure speaking with you. I'd also
• 12:26
  like to thank our podcast sponsor, Wolters Kluwer
• 12:29
  Health. You can tune in to more podcasts
• 12:30
  from Becker's Healthcare by visiting the podcast page
• 12:33
  at beckershospitalreview.com.