Latest / Becker’s Healthcare Digital Health + Health IT / Dr. Philip Payne, inaugural Chief Health AI Officer at Washington University Medicine, and Deborah O'Dell, Chief Data and Analytics Officer at BJC Health System

Transcript

Dr. Philip Payne, inaugural Chief Health AI Officer at Washington University Medicine, and Deborah O'Dell, Chief Data and Analytics Officer at BJC Health System · Becker’s Healthcare Digital Health + Health IT

• 0:00
  This is Laura Dirdo with the Becker's Health
• 0:02
  Care podcast. I'm thrilled today to be joined
• 0:04
  by doctor Philip Payne, inaugural chief health AI
• 0:07
  officer at Washington Medicine and BJC Health System,
• 0:10
  as well as Deborah O'Dell, chief data and
• 0:12
  analytics officer at BJC Health System. Doctor Payne,
• 0:15
  Deborah, it's a pleasure to have you on
• 0:16
  the podcast today.
• 0:19
  Great. Thanks for having us.
• 0:21
  I'm excited for our conversation today because, you
• 0:23
  know, we're really gonna zero in on a
• 0:25
  collaboration or partnership between you both as well
• 0:27
  as the Washington University and BJC Health System,
• 0:31
  which created a new center for health AI.
• 0:34
  How do these strengths of each organization complement
• 0:37
  each other in advancing AI in health care,
• 0:39
  and how will this collaboration evolve over time?
• 0:42
  I would love to hear from your perspectives
• 0:43
  what's really important in the development process and
• 0:47
  and how you plan to continue to grow,
• 0:49
  the partnership.
• 0:51
  So maybe I'll start.
• 0:52
  You know, I think there's this unique opportunity
• 0:55
  that exists at the intersection of both
• 0:57
  the Washington University School of Medicine and then
• 1:00
  our partners at BJC.
• 1:02
  And when you think about it, really what
• 1:03
  we're trying to do is advance the design
• 1:06
  and adoption and use of AI
• 1:08
  in a way that really drives excellence in
• 1:11
  both operations and ultimately care delivery.
• 1:14
  And when you think about the combination of
• 1:15
  our 2 organizations, we have a wealth of
• 1:17
  expertise in that medical school from a biomedical
• 1:20
  research and computational science and AI perspective. And
• 1:24
  that really provides us with this sort of
• 1:26
  academic foundation for our work,
• 1:28
  and that leads to activities like cutting edge
• 1:31
  algorithm design or training new models or validating
• 1:34
  systems.
• 1:35
  And then at the same time, BJC
• 1:38
  Health System is
• 1:42
  and then at the same time, BJC Health
• 1:43
  System is a leading not for profit delivery
• 1:43
  system and really provides the ideal real world
• 1:44
  environment in which we can implement and scale
• 1:46
  these AI solutions.
• 1:48
  If you think about it, it's really sort
• 1:49
  of, as I just mentioned, a living laboratory
• 1:52
  in which we can think about how these
• 1:53
  technologies scale, how they have impact,
• 1:56
  and really how they provide value in real
• 1:59
  world complex workflows.
• 2:01
  And that means the innovations that we'll be
• 2:02
  developing won't just work in the laboratory, but
• 2:04
  they'll really have a direct and practical impact
• 2:06
  on daily patient care. And this new center
• 2:09
  is really all about bridging that gap and
• 2:11
  connecting sort of those 2 sets of capabilities
• 2:14
  so we can really,
• 2:15
  navigate that
• 2:17
  journey from identifying opportunities where AI can provide
• 2:20
  value to deploying technologies, to evaluating them and
• 2:23
  then scaling them.
• 2:24
  And I also think this is emblematic of
• 2:26
  just the collaborative nature of the relationship between
• 2:29
  our 2 organizations,
• 2:31
  especially as we work, even more closely together
• 2:33
  to drive, care delivery for a substantial population
• 2:37
  across
• 2:38
  not only Missouri, but a number of surrounding
• 2:40
  states.
• 2:42
  That's amazing to hear. You know? It really,
• 2:45
  very much, it penalizes
• 2:46
  how these collaborations
• 2:48
  between academic side and health system side can
• 2:50
  really make a big difference. Deborah, is there
• 2:52
  anything you wanted to add here?
• 2:55
  I just echo what Philip said. I think
• 2:56
  this is,
• 2:58
  for us, the opportunity for us to bring
• 3:00
  together both the benefits of the what the
• 3:03
  university brings in terms
• 3:05
  of their,
• 3:06
  advanced research and what they bring from biomedical
• 3:09
  research and AI development. But BJC brings the
• 3:12
  place where we can operationalize that and scale
• 3:15
  that. So it's a really effective partnership that
• 3:17
  we have developed and, looking forward to continuing
• 3:20
  to advance AI capabilities through that.
• 3:24
  That makes a lot of sense. Now I
• 3:25
  know AI has the potential to significantly enhance
• 3:28
  patient care.
• 3:29
  Deborah, starting with you, can you share some
• 3:31
  of the specific ways that technology is being
• 3:34
  applied right now to improve patient outcomes and
• 3:36
  overall patient experience?
• 3:40
  So,
• 3:40
  you know, AI is already transforming health care
• 3:43
  operations and clinical outcomes both, through our university,
• 3:47
  Washington University School of Medicine and BJC Health
• 3:49
  Care. So we have, for example, predictive algorithms
• 3:53
  are optimizing patient flow,
• 3:55
  looking at real time analysis,
• 3:58
  historical data to help us identify
• 4:00
  patient trajectories and optimize transitions in care.
• 4:04
  AI tools are are enhancing early detection and
• 4:07
  diagnosis through medical imaging data and identifying conditions
• 4:11
  like cancer or cardiovascular
• 4:13
  degree,
• 4:14
  disease. So we are, really helping
• 4:17
  to support clinicians in making timely and precise
• 4:21
  decisions.
• 4:22
  AI is also being used
• 4:24
  to personalize treatment plans by analyzing electronic health
• 4:28
  records and genomic data,
• 4:30
  to really recommend
• 4:32
  therapies that are tailored and particular in fields
• 4:34
  like oncology.
• 4:36
  So predictive analytics are driving,
• 4:39
  population health, identifying high risk patients. So there's
• 4:42
  so many opportunities where we're seeing,
• 4:44
  the demonstration of how AI can,
• 4:47
  enhance both the efficacy and the the operations
• 4:50
  so that we can provide superior quality of
• 4:53
  care.
• 4:54
  Philip, what would you add?
• 4:56
  No. I mean, I think that was an
• 4:57
  excellent summary. I mean, I think a lot
• 4:59
  of this is anchored on, I think,
• 5:02
  a driving philosophy here, which is that
• 5:05
  our goal is to use AI to complement
• 5:08
  the unique strengths of our providers and our
• 5:10
  clinical workforce so that it's really this combination
• 5:13
  of humans and computers working together
• 5:16
  that is superior to either working alone. And
• 5:18
  I think that's borne out in a lot
• 5:20
  of the examples that Deborah gave,
• 5:22
  which really come down to how do we
• 5:24
  sort through and understand these massive quantities of
• 5:27
  data and evidence and knowledge we have and
• 5:28
  make better,
• 5:30
  safer,
• 5:32
  care related decisions for our patients and improve
• 5:35
  the the value of that care along the
• 5:36
  way.
• 5:38
  I love that. It it seems like such
• 5:40
  a meaningful way to apply technology into, the
• 5:43
  health system and health system and health care
• 5:44
  organizations
• 5:45
  for patients as well as, the team members?
• 5:47
  And and speaking of your team, how do
• 5:49
  you envision AI tools improving health care employee
• 5:52
  experience?
• 5:53
  Are there any early signs of success or
• 5:55
  feedback that you've received from those on the
• 5:57
  front lines?
• 5:59
  Yeah. I mean, I think you're sort of
• 6:01
  touching on probably one of the most important
• 6:03
  areas
• 6:04
  that we have been looking at,
• 6:06
  which is how do we remove sort of
• 6:08
  these high friction, low value tasks,
• 6:13
  that
• 6:14
  are taking our providers away from focusing on
• 6:16
  our patients?
• 6:18
  And in a similar way, how do we
• 6:19
  get sort of the computer out of the
• 6:21
  middle of those interactions, again, between our
• 6:24
  providers and patients so that we restore a
• 6:27
  certain degree of humanism to how we practice
• 6:29
  medicine.
• 6:30
  So, of course, like many large delivery systems,
• 6:34
  we've been
• 6:35
  leveraging
• 6:36
  ambient AI technology to help automate documentation during
• 6:40
  patient visits and we've had some very promising
• 6:42
  results not only in primary care settings but
• 6:45
  now in procedural disciplines and we're looking at
• 6:47
  other domains.
• 6:48
  And this has been described by some of
• 6:51
  our providers as transformative.
• 6:53
  We have individuals who used to spend several
• 6:55
  hours a day after normal business hours documenting
• 6:59
  that are no longer doing that because they're
• 7:00
  able to be more efficient and timely with
• 7:02
  their documentation
• 7:03
  during the course of their,
• 7:05
  clinical work.
• 7:07
  We're also
• 7:08
  using those same types of technologies to look
• 7:10
  at how can we do a better job
• 7:11
  of summarizing the content of electronic health records
• 7:14
  so that we can get the right information
• 7:15
  to our providers more quickly, whether they're
• 7:18
  getting ready to admit a patient
• 7:20
  or they are seeing a new patient for
• 7:22
  the first time,
• 7:24
  or any number of other use cases where
• 7:26
  prior workflows would involve them spending
• 7:29
  potentially up to several hours trying to prepare
• 7:31
  for these types of clinical encounters. And now
• 7:34
  we've turned that summarization activity into sort of
• 7:36
  a matter of minutes and mouse clicks. So
• 7:39
  for us, it's all about how do we,
• 7:41
  again, restore that humanism in medicine and really
• 7:43
  let our providers focus on our patients by
• 7:46
  taking those other tasks out of the mix
• 7:48
  by automating them. So we're really excited and
• 7:51
  I think our providers are really excited about
• 7:53
  this.
• 7:55
  That's amazing to hear. And, certainly, you know,
• 7:57
  what what a great,
• 7:58
  an interesting way to think about technology and
• 8:00
  using that to then give the providers back
• 8:03
  time to be more human with their patients
• 8:06
  and have that really human to human connection.
• 8:08
  Deborah, is there anything else you wanted to
• 8:10
  speak to when you think about the employee
• 8:12
  experience and those that you're working with, to
• 8:14
  really truly elevate the way things are at
• 8:17
  BJC Health Care?
• 8:19
  The only other thing that I'll add is
• 8:20
  that, when we think about what the new
• 8:23
  center can can do and how we can
• 8:25
  leverage technology, it's it's also our employees and
• 8:29
  staff
• 8:30
  that might not provide direct patient care. So
• 8:32
  where can we look at opportunities to help
• 8:35
  our,
• 8:36
  our other capabilities, whether they be in our
• 8:38
  finance team or supply chain to help with
• 8:42
  tools that make day to day life more
• 8:44
  efficient, more effective?
• 8:46
  So I think the center is is here
• 8:49
  primarily to help that clinical that clinician and
• 8:52
  that patient experience, but we do see opportunities
• 8:55
  that can help us as a health system
• 8:56
  to also be
• 8:57
  more effective, which really does,
• 8:59
  ultimately have a positive impact on the patients
• 9:02
  as well.
• 9:04
  Absolutely. That's such a great point and and
• 9:06
  such an important,
• 9:07
  thing to remember when you're looking at the
• 9:09
  system as a whole. Now I know the
• 9:10
  Center For Health AI will offer opportunities for
• 9:13
  medical students and residents to train on AI
• 9:15
  driven care. How do you see this influencing
• 9:18
  the next generation of health care leaders, and
• 9:20
  what skills will they need in order to
• 9:21
  succeed,
• 9:23
  in an AI enhanced medical landscape? What are
• 9:26
  your thoughts on this, doctor Payne?
• 9:28
  Yeah. I mean, I think that's a really
• 9:30
  important question because the fundamental
• 9:32
  environment in which medicine is being practiced is
• 9:35
  changing as a function of AI. And so
• 9:37
  the question is, really, how do we prepare
• 9:39
  our future practitioners,
• 9:41
  researchers, and leaders to thrive in that
• 9:48
  fortuitous timing for us that we've just gone
• 9:50
  through a
• 9:52
  refresh of our medical curriculum here at Washington
• 9:54
  University.
• 9:55
  And as part of that, we thought about
• 9:56
  what are really the core areas of expertise
• 9:58
  that the physician leaders of the future need
• 10:01
  to have. And one of them was clearly,
• 10:03
  informatics,
• 10:04
  AI and data science.
• 10:06
  And that means we're now integrating that into
• 10:09
  our medical curriculum. We're teaching our medical students
• 10:12
  to understand
• 10:13
  how the use of technology impacts
• 10:16
  their decision making and how they need to
• 10:18
  sort of consider that whether that be for
• 10:20
  diagnostic purposes or therapy planning or any number
• 10:23
  of other scenarios.
• 10:25
  We're also using AI powered simulation tools now
• 10:28
  to create really realistic scenarios so students can
• 10:31
  practice clinical decision making in sort of a
• 10:33
  safe space before they end up in the
• 10:35
  clinical environment. And that provides instant feedback and
• 10:38
  insights into their performance that are highly tailored.
• 10:41
  So much like we talk about precision medicine
• 10:43
  for patients, this is really precision curricula.
• 10:46
  And then I think much like it has
• 10:49
  historically been the case that we teach sort
• 10:51
  of future,
• 10:52
  again, practitioners
• 10:54
  and researchers and leaders to critically evaluate,
• 10:57
  the literature. We're now teaching them how to
• 10:59
  critically evaluate the output of,
• 11:02
  AI tools like predictive algorithms
• 11:04
  so that they understand sort of both the
• 11:06
  strengths and weaknesses of these types
• 11:10
  of predictive tools and how they can and
• 11:12
  should influence clinical decision making in equal measure
• 11:15
  where
• 11:16
  sort of appropriate caution may be necessary.
• 11:20
  And the last area that I would emphasize
• 11:22
  is
• 11:23
  we're also giving our students the opportunity to
• 11:26
  really analyze extremely large datasets and gain insights
• 11:29
  around
• 11:30
  phenomena that might be of interest around
• 11:33
  policy
• 11:34
  or population health or other sort of
• 11:38
  interventions that happen beyond the individual patient in
• 11:40
  the exam room or hospital room. And we're
• 11:43
  doing that by really democratizing how people access
• 11:45
  data and analyze data, again, using AI tools
• 11:48
  that sort of guide the trainee through that
• 11:50
  process.
• 11:51
  So really, when you put that all together,
• 11:53
  we're not only
• 11:55
  improving the quality and sort of the completeness
• 11:57
  of the education we provide, but we're also
• 12:00
  ensuring that those students are ready to go
• 12:01
  out and practice in a world where
• 12:04
  AI is going to be ubiquitous and, you
• 12:06
  know, really a
• 12:08
  sort of, in some ways,
• 12:10
  sort of key dimension of the modern practice
• 12:13
  of medicine.
• 12:15
  Absolutely. Well, that's fascinating to hear, you know,
• 12:17
  and certainly, as you think about those next
• 12:20
  generations of medical students,
• 12:21
  what a a great tool to have in
• 12:23
  their back pocket with AI and understanding its
• 12:25
  limitations, but also opportunities.
• 12:27
  Deborah, is there anything you wanted to add
• 12:29
  in there?
• 12:30
  I'll just say I think, BJC wants to
• 12:33
  continue to be,
• 12:35
  a world class institution that brings world class
• 12:37
  providers here, and so we need to stay,
• 12:40
  at the forefront of this technology
• 12:42
  that,
• 12:44
  our providers are learning to use in school,
• 12:47
  and we want them to have access to
• 12:49
  that,
• 12:50
  at a scale of a system our size
• 12:52
  and breadth. So, this is why the partnership
• 12:55
  between BJC and WashU is so important that
• 12:58
  we can continue to do this in a
• 12:59
  very unique way, to the medical school and
• 13:02
  then the size and scale of the BJC
• 13:04
  system.
• 13:05
  That's amazing to hear, you know, and really
• 13:07
  great that on the health system level, you
• 13:09
  can have that foresight
• 13:10
  and and support what the medical school is
• 13:12
  doing. Now I have a few questions for
• 13:14
  you, doctor Payne, and then we'll follow you
• 13:16
  back to you, Deborah. But, doctor Payne, one
• 13:18
  example that you've shared, is using AI to
• 13:20
  predict excessive
• 13:22
  blood loss during surgery, which helps improve resource
• 13:25
  management. Can you expand how AI is improving
• 13:28
  surgical procedures, diagnostic accuracy, and precision medicine overall?
• 13:33
  Yeah. I mean, it's a it's a great
• 13:34
  question and it's a project I'm particularly excited
• 13:36
  about.
• 13:37
  There's actually a research team here that spans
• 13:40
  not only our Institute For Informatics and our,
• 13:43
  Department of Anesthesiology
• 13:45
  and Department of Surgery, but also,
• 13:47
  key partners at BJC.
• 13:49
  And that group has really sort of demonstrated
• 13:52
  in a very compelling way how you can
• 13:54
  use machine learning models to predict
• 13:57
  blood
• 13:58
  loss during surgical procedures. And if you're familiar
• 14:01
  with this domain, sort of the current state
• 14:03
  of the art is to use sort of
• 14:05
  historical and largely,
• 14:07
  I would say out of date data concerning
• 14:09
  predictive blood loss as a function of how
• 14:12
  the average patient may
• 14:14
  sort of
• 14:15
  experience blood loss during surgery. And the result
• 14:17
  was a lot of blood products being used
• 14:20
  in the OR
• 14:22
  that were necessary but an equal measure of
• 14:24
  blood products being sent to the OR that
• 14:26
  were never used and therefore wasted. And so
• 14:28
  this team actually used
• 14:30
  over 4,000,000 surgical records from the American College
• 14:32
  of Surgeons National Surgical Quality Improvement Program database
• 14:36
  and actually built a highly accurate model that
• 14:39
  recommended type and screen orders,
• 14:42
  for patients during the sort of perioperative management
• 14:46
  sort of period.
• 14:47
  And actually,
• 14:49
  this resulted in, if you sort of compare
• 14:51
  the traditional model versus current model, a difference
• 14:54
  of 96%
• 14:55
  sensitivity in anticipating
• 14:58
  perioperative blood loss using the machine learning model
• 15:01
  as compared to sort of around 50%
• 15:04
  sort
• 15:05
  of sensitivity in a baseline model. And that
• 15:07
  results in real savings of blood products. So,
• 15:10
  in a lot of ways, this is about
• 15:12
  how we can use personalized
• 15:13
  data driven interventions in order to enhance that
• 15:16
  perioperative care period and at the same time,
• 15:20
  really be much more judicious in how we
• 15:22
  manage
• 15:22
  a very finite resource that's critical to surgical
• 15:25
  safety. So, that's I think, a a great
• 15:27
  example of how all these pieces come together.
• 15:31
  That it really colors the work that you're
• 15:33
  doing,
• 15:34
  well and and spotlights the outcomes and impact
• 15:36
  it can have not only on patients but
• 15:38
  the organization too. You mentioned AI can help
• 15:41
  identify the most effective treatments for individual patients
• 15:44
  as well. How is this process being applied
• 15:47
  in practice at WashU and BJC,
• 15:49
  especially for the complex or rare diseases?
• 15:52
  Yeah. So,
• 15:55
  I think one of the things that we
• 15:56
  know is that especially in rare disease,
• 15:59
  the availability of therapeutic options is very limited
• 16:01
  because they are frankly not as appealing to
• 16:05
  pharmaceutical
• 16:05
  or bio technology companies. And so we often
• 16:08
  are in a position where we have to
• 16:10
  create tailored therapeutic
• 16:13
  approaches for those individuals.
• 16:15
  Great examples, we have a research
• 16:17
  program here working with our neurofibromatosis
• 16:21
  type 1 center or NF1 center. And this
• 16:23
  is,
• 16:23
  the most common rare genetic diseases in children.
• 16:26
  It's relatively easily diagnosed.
• 16:29
  But then there's a multitude of phenotypes that
• 16:31
  patients will present with. Some patients will be
• 16:33
  fine and will not experience any meaningful symptoms.
• 16:36
  Some will develop
• 16:37
  optic pathway gliomas or plexiform neurofibromas,
• 16:41
  which are painful nerve sheet tumors. Some will
• 16:43
  develop ADHD. Some will develop scoliosis. I could
• 16:46
  keep on going down a lengthy list. And
• 16:48
  the problem is we're not very good at
• 16:50
  anticipating
• 16:50
  which of those phenotypes patients will
• 16:53
  exhibit it. So we tend to manage those
• 16:55
  patients extremely aggressively trying to find early signs
• 16:58
  of those various sort of deleterious outcomes, but
• 17:01
  we're not always good at finding them.
• 17:03
  And we've actually been able to, again, using
• 17:06
  artificial intelligence methods and a combination of imaging
• 17:09
  data, data from the EHR and then various
• 17:12
  neurobehavioral
• 17:13
  phenotyping instruments applied to patients directly,
• 17:16
  we're able to generate highly accurate
• 17:19
  predictive models that can be used to forecast
• 17:21
  sort of the disease course for children with
• 17:23
  NF1. And so what that means now is
• 17:25
  instead of every child having to have a
• 17:28
  large amount of imaging or other potentially challenging
• 17:31
  diagnostic
• 17:32
  procedures, we can focus on those individuals who
• 17:35
  are most at risk and often detect the
• 17:37
  outcomes earlier, which leads to better therapeutic outcomes.
• 17:41
  And I often say to people, sometimes precision
• 17:43
  medicine is about finding the right treatment, but
• 17:45
  sometimes precision medicine is just about demystifying diseases.
• 17:48
  So we know what to expect and we
• 17:50
  can be smarter about how we manage those
• 17:52
  symptoms. And I think this is a great
• 17:53
  example where we're getting smarter using AI so
• 17:56
  that we can be very precise in how
• 17:58
  we manage
• 17:59
  the clinical course of kids with a very
• 18:01
  challenging genetic disease.
• 18:04
  And we need to do more of that.
• 18:05
  There's no doubt that we have a wealth
• 18:07
  of data to build these models. It's just
• 18:09
  that we need to start doing the hard
• 18:11
  work of bringing those data together,
• 18:13
  working with our clinical colleagues and frankly running
• 18:16
  large studies to validate that the models provide
• 18:19
  clinical value, which is another big part of
• 18:21
  the project I just described to you.
• 18:24
  Absolutely. Thank you so much for that example.
• 18:26
  And real quick, what steps are you taking
• 18:28
  to ensure that those AI technologies are used
• 18:30
  responsibly and safely in patient care?
• 18:33
  Yeah.
• 18:34
  That is a huge and really important question.
• 18:37
  Right?
• 18:38
  How do we make sure that when we
• 18:39
  deploy ai, it's safe, it's efficacious
• 18:42
  that we are considering the ethical
• 18:44
  and legal and social implications of these emerging
• 18:46
  technologies. And
• 18:47
  I would start by saying we're committed across
• 18:49
  Washington BJC to making sure that everything we
• 18:52
  do through this new Center For Health AI
• 18:54
  emphasizes
• 18:55
  delivering these tools in a secure, scalable and
• 18:58
  reliable manner, but also in a way that
• 19:00
  prioritizes patient safety and privacy
• 19:02
  and confidentiality.
• 19:04
  And that means we have to really sort
• 19:05
  of use what is often referred to as
• 19:06
  sort of a privacy by design approach, which
• 19:09
  means that data governance and the application of
• 19:12
  appropriate regulatory standards as well as more technical
• 19:16
  controls,
• 19:16
  and then engagement of
• 19:18
  key stakeholder groups to define early and often
• 19:21
  the ethical
• 19:22
  frameworks in which these tools exist, all of
• 19:24
  that has to happen at the beginning of
• 19:26
  projects, not at the end of projects. And
• 19:28
  I think one of our challenges often is
• 19:30
  that we treat these, again, ethical, legal, and
• 19:32
  social implications of technologies
• 19:35
  as something
• 19:36
  that is assessed after the fact.
• 19:38
  But the likelihood of success is greatly diminished
• 19:41
  when we do that. And I think the
• 19:43
  cornerstone of all of this is transparency, right?
• 19:45
  Making sure that we're empowering our patients,
• 19:47
  our providers and all the other stakeholders involved
• 19:50
  in our delivery system so they know when
• 19:52
  and how we're using AI.
• 19:55
  And I
• 19:57
  think ultimately patient safety and these ethical considerations,
• 20:00
  they're gonna be central to every decision we
• 20:02
  make about AI deployment.
• 20:04
  And that means that the teams that we
• 20:05
  put together will have to be interdisciplinary
• 20:08
  and include not only technologists and clinicians
• 20:10
  but ethicists and patient advocates and community representatives.
• 20:14
  We're gonna have to make sure that we
• 20:16
  focus on and commit to really rigorous
• 20:18
  validation of these tools that includes
• 20:20
  an assessment of sort of the ethical and
• 20:23
  social impact of their use. And then we
• 20:25
  have to create these continuous monitoring and feedback
• 20:28
  loops so that we understand unintended consequences or
• 20:31
  outcomes
• 20:32
  of the use of AI rather than treating
• 20:34
  the process as somewhat unidirectional where we declare
• 20:36
  success when the tool works, deploy it in
• 20:39
  the healthcare delivery system, and then move on
• 20:40
  to the next project.
• 20:42
  And I would say,
• 20:44
  much as people have talked about the need
• 20:47
  to be vigilant about adverse outcomes with new
• 20:49
  therapeutics as they come to market. I think
• 20:52
  the same will be very much the case
• 20:54
  with AI.
• 20:55
  And I'll just wrap up by saying, in
• 20:57
  my mind, it is our responsibility
• 20:59
  at Washington Medicine and BJC
• 21:01
  to set the benchmark
• 21:03
  for responsible, safe, and ethical AI innovation in
• 21:05
  health care. So I don't believe we're satisfied
• 21:08
  just with, adopting best practices from others, but
• 21:10
  we really wanna set the standard for what
• 21:12
  those best practices are.
• 21:15
  That's amazing to hear. A big job, but
• 21:17
  one certainly that you're set up well-to-do
• 21:20
  responsibly.
• 21:21
  Now, Deborah, from your perspective, one of the
• 21:23
  primary goals of the center is to improve
• 21:26
  operational efficiency, including streamlining scheduling and reducing administrative
• 21:29
  burdens. Can you walk us through some of
• 21:31
  the AI tools currently being piloted and how
• 21:34
  they're making a tangible difference for health care
• 21:36
  providers and patients?
• 21:38
  Yeah. We're actually,
• 21:39
  really looking at what our current vendor partners
• 21:43
  offer embedded in tools that we might already
• 21:45
  have. And I think this is very common
• 21:47
  across,
• 21:48
  the health care environment today that we're in
• 21:50
  any environment, actually, where they're seeing,
• 21:53
  existing
• 21:54
  software and tools that we're using and they're
• 21:57
  adding in components. So whether that be our
• 22:00
  EHR vendor or our ERP vendor,
• 22:03
  we are we are looking at those opportunities
• 22:05
  to see how we can implement them. I
• 22:07
  think
• 22:08
  as Philip and I,
• 22:09
  are standing up this new center,
• 22:12
  we really talk about how we can
• 22:14
  help move these projects along, how we can
• 22:16
  unstick,
• 22:18
  where things get get stuck with new technology
• 22:21
  and,
• 22:22
  implementation and change of processes. So we're looking
• 22:25
  at really how we can leverage what we
• 22:27
  already have today as well as,
• 22:29
  the experience and knowledge of all the examples
• 22:32
  that Philip just gave that come from our
• 22:34
  university research and provider side.
• 22:37
  So we have many different pilots going on
• 22:39
  in those spaces where it might feel,
• 22:42
  like, we don't see it every day, because
• 22:44
  we're
• 22:46
  processing
• 22:47
  payment or we're ordering supplies or we're working,
• 22:51
  through vendor
• 22:52
  agree legal agreement. But those are places where
• 22:55
  AI
• 22:55
  tools can really help us be more effective
• 22:57
  and efficient and move things through faster.
• 23:01
  That's great to hear. You mentioned that AI
• 23:03
  could help predict demand for equipment and staff
• 23:05
  resources.
• 23:06
  What are some of the real world examples
• 23:08
  of how AI is improving logistics and resource
• 23:10
  management in health systems, especially in large and
• 23:13
  complex organizations like BJC?
• 23:16
  I think today in health care, a lot
• 23:18
  of health systems are really struggling with similar
• 23:20
  things around capacity,
• 23:22
  and throughput. And so understanding how we expect
• 23:27
  people to come through, the needs and equipment
• 23:29
  to come through, whether it's shortages
• 23:31
  that we have in certain supplies,
• 23:33
  and being able to project and predict when
• 23:35
  those things are going to happen and what
• 23:37
  the impact is going to be.
• 23:39
  So I think
• 23:40
  most of the country is experiencing some of
• 23:43
  those at any given time, whether that be,
• 23:45
  we have backups in our emergency department or
• 23:48
  we have,
• 23:49
  IV shortages due to
• 23:51
  unforeseen,
• 23:52
  world situations,
• 23:54
  we need to,
• 23:55
  we need to be move quickly at helping
• 23:59
  improve how we can answer those problems and
• 24:01
  address those resource needs. And so, those are
• 24:05
  things that we do on a daily basis,
• 24:06
  and we're hoping the new center can really
• 24:08
  help us to figure out how to continue
• 24:10
  to drive those sorts of decision making.
• 24:13
  I love that. What a great use case
• 24:15
  and and really strong example. Before we wrap
• 24:17
  up here, I have one more quick question.
• 24:19
  How do you plan to scale successful AI
• 24:21
  tools and solutions from pilot projects to widespread
• 24:24
  adoption across the health system? What challenges do
• 24:27
  you anticipate in this process, and how can
• 24:29
  the center help overcome them?
• 24:32
  I think, the the challenges are gonna be
• 24:34
  challenges with any new technology, new change, new
• 24:37
  processes,
• 24:39
  learning what works, what doesn't, trying some things.
• 24:42
  Process change is always,
• 24:44
  getting some people will be on board. Some
• 24:47
  people will need more,
• 24:49
  support and coaching around how this can be
• 24:52
  different and how this can change their,
• 24:54
  their workflow and their day. So I think,
• 24:58
  luckily, I think Philip and I are privileged
• 25:00
  to work in an organization
• 25:02
  that,
• 25:03
  really wants to learn and grow, and we
• 25:06
  are often just trying to keep up with
• 25:08
  the needs and demand around how we can
• 25:10
  do things more efficiently, more effectively, and provide
• 25:12
  better care.
• 25:13
  So I think the center will we will,
• 25:17
  we're really here to help our leaders and
• 25:20
  our providers
• 25:21
  do the best that they can for our
• 25:22
  patients. So however that,
• 25:25
  however that manifests itself in what we can
• 25:27
  do every day to support them, I think,
• 25:30
  I'm excited about it. I'm excited to continue
• 25:33
  to work with Philip in in how we
• 25:34
  can scale this and really support what our
• 25:37
  providers do every day for our community.
• 25:41
  That's amazing to hear. Doctor Payne, Deborah, thank
• 25:43
  you so much for joining us on the
• 25:44
  podcast today. This has been such a full
• 25:46
  and rich conversation, and I look forward to
• 25:48
  connecting with you again soon and catching up
• 25:50
  on some of the great things that you
• 25:51
  continue to do with the center.
• 25:54
  Well, thank you for the opportunity.