Latest / Becker’s Healthcare Digital Health + Health IT / Yasir Tarabichi, Chief Medical Informatics Officer & Interim Medical Director, Ovatient, & Director of Clinical Research Informatics, The MetroHealth System

Transcript

Yasir Tarabichi, Chief Medical Informatics Officer & Interim Medical Director, Ovatient, & Director of Clinical Research Informatics, The MetroHealth System · Becker’s Healthcare Digital Health + Health IT

• 0:00
  This is Laura D with the Becker Health
• 0:02
  care podcast.
• 0:03
  I'm thrilled today to be joined by Doctor.
• 0:05
  Y T mitsubishi, Chief Medical Informatics officer an
• 0:08
  interim medical director for Ovation and Director of
• 0:11
  clinical research informatics at the Metro health system.
• 0:14
  Doctor. T she she's a pleasure to have
• 0:15
  you on the podcast today.
• 0:17
  Thank you for having me. I'm really looking
• 0:19
  forward to our discussion and diving a little
• 0:21
  bit deeper to what you're doing at both
• 0:23
  ovation as well as the Metro health system,
• 0:25
  but before we dive in, I was wondering
• 0:27
  if you could tell us a little bit
• 0:28
  more about yourself in your background?
• 0:30
  Sure. So I am a pulmonary and critical
• 0:32
  care physician and position in
• 0:35
  Who's been doing a lot more of the
• 0:37
  latter in my career.
• 0:40
  I've been at Mental Health now for almost
• 0:42
  9 years.
• 0:43
  And
• 0:44
  increasingly
• 0:46
  doing
• 0:47
  things in the data science and clinical decision
• 0:50
  support space as well as more recently in
• 0:52
  the virtual care space
• 0:54
  through this ovation venture, which is a metro
• 0:56
  health and
• 0:58
  Collaboration. So multi health care system
• 1:02
  virtual care platform, which is built frankly on
• 1:04
  interoperability
• 1:05
  and system integration as a differentiating factor.
• 1:12
  That's really cool to hear. And and definitely
• 1:14
  exciting to see how the technology can really
• 1:17
  inform and influence the things you're doing at
• 1:18
  the health system level as well. And and
• 1:20
  given your perspectives in both fronts, what are
• 1:22
  some of the biggest issues that you're following
• 1:24
  in healthcare today?
• 1:27
  Sure. So I think
• 1:28
  artificial intelligence is
• 1:30
  obviously on everybody's in mind.
• 1:33
  When I think about
• 1:35
  healthcare systems through the
• 1:37
  clinical informatics lens, a lot of it
• 1:40
  least from our perspective right, sits on this
• 1:43
  cushion of data
• 1:44
  and how we can turn data,
• 1:47
  data input into meaningful outputs to help drive
• 1:50
  patient care.
• 1:52
  And so as you think about kind of
• 1:53
  the evolution of where we've been and what
• 1:55
  we've been able to achieve in the health
• 1:57
  informatics space
• 1:58
  around
• 1:59
  really getting everybody on an electronic health records,
• 2:02
  standardizing
• 2:03
  data to the degree that we've been able
• 2:04
  to do so and increasingly more looking at
• 2:07
  interoperability
• 2:08
  I think that's created an incredible
• 2:10
  really charged atmosphere or
• 2:14
  bigger and better things that we can do
• 2:16
  with this data
• 2:17
  including obviously implementing Ai intervention.
• 2:21
  And so
• 2:22
  I think that that
• 2:24
  really becomes the key driver
• 2:26
  of where we go next? We can talk
• 2:28
  a lot about some of those aspects. So
• 2:31
  I think about things like
• 2:33
  how does data standards and interoperability
• 2:37
  drive
• 2:38
  evolution in the Ai
• 2:40
  clinical and support space,
• 2:42
  and that comes down to things like being
• 2:44
  able to build and scale and fed
• 2:47
  learning and models
• 2:49
  in a way that really is meaningful
• 2:51
  and externally valid.
• 2:53
  And also how do we bring those insights
• 2:55
  and models
• 2:56
  and knowledge back into our healthcare system. And
• 2:59
  so talking about things like
• 3:01
  Ai assurance labs and how do we scale
• 3:03
  those? Like how do we know that models
• 3:07
  approved or sanctioned by other organizations or entities?
• 3:10
  Really come back into our system and make
• 3:12
  a difference. So a lot of what I
• 3:14
  focused this on is
• 3:15
  thinking about Ai
• 3:17
  implementation
• 3:18
  at the local level
• 3:20
  and us kind of reached this conclusion that
• 3:23
  really Ai is a implementation, frankly is a
• 3:27
  local
• 3:28
  concept
• 3:29
  and the value of a model
• 3:31
  is only really
• 3:34
  made
• 3:35
  obvious when it's implemented and valid at the
• 3:37
  local setting.
• 3:39
  And so I like to think about how
• 3:41
  all of these
• 3:42
  tools in these ecosystems
• 3:44
  evolve to
• 3:46
  support that mission.
• 3:48
  And I get we can get into some
• 3:49
  of the details, but I think that's both
• 3:51
  the opportunity and I think the excitement in
• 3:53
  this space.
• 3:55
  I think about the challenges it's
• 3:57
  really the hype
• 3:59
  and
• 4:00
  making sure as an informatics as a physician
• 4:02
  executive, you're not making the wrong
• 4:05
  decisions and choices
• 4:06
  when it comes to which solutions to implement,
• 4:09
  right? So the obvious 1 being saying away
• 4:11
  from point solutions and making good solid
• 4:14
  decisions that enhance your digital ecosystem and infrastructure
• 4:18
  and that can
• 4:19
  allow you to build on top of that
• 4:21
  and develop
• 4:22
  the Ai driven analytics that are going to
• 4:24
  propel your organization and healthcare forward.
• 4:28
  So very high level generic
• 4:30
  take on the Ai space.
• 4:33
  But I'm happy to get into the details
• 4:34
  if you're interested.
• 4:36
  Yeah. Absolutely. And I I really appreciate your
• 4:38
  analysis here because you know, I I know
• 4:41
  some money leaders and organizations are trying to
• 4:43
  figure out the best way to move forward
• 4:45
  with Ai and technology and the different types
• 4:47
  of investments that they're trying to make. Now
• 4:50
  when you talk about artificial intelligence especially for
• 4:53
  organizations that, you know, may have started to
• 4:55
  work through some of that and and figure
• 4:57
  out what
• 4:58
  their their mission is and how Ai can
• 5:01
  support their mission. How do you then grow
• 5:03
  into more sophisticated space? What you need from
• 5:05
  a team level as well as a technology
• 5:08
  level that, will really make a difference and
• 5:10
  propel you quickly into the future.
• 5:12
  Yeah.
• 5:13
  So
• 5:14
  I'll say that a lot of
• 5:16
  the infrastructure that you need to embrace this
• 5:20
  evolving technology.
• 5:22
  Now at the end of the day, it's
• 5:23
  a tool. Obviously, it's a different tool
• 5:26
  with
• 5:27
  unique implications
• 5:29
  and considerations from
• 5:31
  several perspectives, right? So like I said, this
• 5:34
  notion of truly validating that a model is
• 5:36
  going to work for your patient's place in
• 5:38
  time that it's going to be fair and
• 5:41
  unbiased and even beyond that actually being able
• 5:45
  to implement it in a way that actually
• 5:47
  does not
• 5:49
  exacerbate disparities, which is especially important for a
• 5:52
  place like metro health where we're safety healthcare
• 5:54
  care systems that we see patients that are
• 5:56
  at risk
• 5:58
  minorities and soc economic
• 6:00
  deprivation being a common theme here. So
• 6:03
  when you think about
• 6:05
  what is the skills set required?
• 6:08
  Right? To bring that technology in.
• 6:10
  I think the foundation
• 6:12
  is fairly similar to really everything we do
• 6:15
  in the informatics space.
• 6:17
  Those understanding systems,
• 6:19
  understanding that any technology you bring in enters
• 6:22
  a
• 6:23
  person process
• 6:25
  a complex person process associated with technical space,
• 6:29
  right that you need to
• 6:31
  appreciate and understand
• 6:34
  and
• 6:34
  the implications too of change,
• 6:38
  innovation,
• 6:39
  implementation science,
• 6:41
  quality improvement. So all of these things that
• 6:43
  we we've been doing, right, from the outset,
• 6:46
  implementing all these technologies in a thoughtful and
• 6:48
  meaningful way.
• 6:49
  Do you have to take all of that
• 6:50
  solid base that you developed
• 6:52
  and add this new tooling to it
• 6:54
  and the stretch that you have to make
• 6:57
  when you're getting to that point.
• 6:59
  Is you need somebody who understands enough about
• 7:02
  these models
• 7:03
  and the implication,
• 7:05
  right to bring it into the space.
• 7:07
  Because the converse
• 7:08
  unfortunately is it's not true, right? So you
• 7:10
  can have amazing Ai and amazing technology,
• 7:14
  but shoving it into a healthcare system.
• 7:17
  Is not always the best way, right? That's
• 7:19
  just a clear path of failure.
• 7:22
  And so
• 7:23
  I think that the model that I
• 7:26
  I think speaks to this
• 7:28
  is this concept of a like quan
• 7:31
  and that somebody who
• 7:33
  understands enough about the technology
• 7:36
  implications and potential, right, to bring it into
• 7:40
  their environment
• 7:41
  and then bring it into their environment into
• 7:43
  this complex healthcare system
• 7:46
  in the way that it actually
• 7:48
  feeds and enables people to do the things
• 7:51
  that they're setting out to do, which obviously
• 7:53
  in this space primarily is around
• 7:55
  providing excellent care.
• 7:58
  Understand that the implications understands the risk and
• 8:00
  biases that they have to address.
• 8:02
  And so you obviously need some
• 8:06
  that informatics type individual coming out from healthcare.
• 8:09
  With enzyme understanding into the data science and
• 8:11
  the artificial intelligence,
• 8:13
  also obviously paired with somebody who can do
• 8:15
  to data science.
• 8:17
  I think we're slowly moving away from the
• 8:19
  need to have
• 8:21
  hardcore data scientists.
• 8:23
  I'm also optimistic that
• 8:26
  as the tools evolve
• 8:28
  that these vendors that are producing these solutions,
• 8:32
  package
• 8:33
  their solutions with ideally open source tools that
• 8:36
  have been
• 8:37
  really kind of taken through their paces,
• 8:40
  standardized
• 8:42
  vetted,
• 8:44
  allowing these organizations then to kind of flex
• 8:46
  up with whatever resources they have to reach
• 8:50
  into that space. And bring that solution in.
• 8:53
  And so that way you can start to
• 8:55
  think about how you're democrat
• 8:57
  the Ai to the smaller healthcare organization. We
• 9:00
  don't have the money and the resources and
• 9:02
  the personnel
• 9:04
  to individually develop their own models
• 9:08
  and their own Ai expertise. Now those places
• 9:10
  exist, obviously
• 9:12
  academic centers
• 9:14
  obviously, I could name some organizations,
• 9:17
  we all know which ones they are. It's
• 9:19
  the same organizations who are leading a lot
• 9:21
  of The charge and this color is the
• 9:24
  rest of the discussion that I I kind
• 9:26
  of brought in earlier.
• 9:28
  I think
• 9:29
  this notion of Ai assurance,
• 9:32
  right, that an organization can tell this model
• 9:34
  looks good because we tested in our data
• 9:37
  as someone who's independently implemented and validated Ai
• 9:41
  solutions at the point of care.
• 9:43
  Models are messy inherently biased and they don't
• 9:46
  always work well for your patient population. And
• 9:49
  this has been true throughout time. It doesn't
• 9:51
  not be a sophisticated model to do that,
• 9:52
  even models at pretty risk
• 9:54
  like heart disease for instance. When we bring
• 9:56
  them into a system like Metro health
• 9:59
  and more diverse and our risk coa population,
• 10:02
  we can see that the model actually tends
• 10:05
  to under predict risk of cardiac, disease and
• 10:08
  poor outcomes
• 10:09
  which case, we actually could be doing our
• 10:11
  patients on disservice,
• 10:13
  a model like that is. So this notion
• 10:15
  of rules and models being biased
• 10:17
  just the world we live in and it's
• 10:19
  not new.
• 10:20
  And I think
• 10:21
  when you take these models out from systems
• 10:24
  that they've been developed, we bring them into
• 10:26
  systems
• 10:27
  like metro health or
• 10:29
  F
• 10:31
  or other similar systems.
• 10:34
  There's always that fair and concern that you're
• 10:36
  actually going to be doing and disadvantage. And
• 10:38
  so some of the work that we do
• 10:39
  for instance,
• 10:42
  we've done
• 10:43
  implementation of
• 10:45
  potentially challenging, predictive models like we predicted the
• 10:48
  risk that a patient when to show up
• 10:49
  to a clinic.
• 10:50
  And what that brings with it is obviously
• 10:54
  more often than not more likely, you're actually
• 10:56
  going to be targeting. Whatever intervention you build
• 10:58
  on that, you're going to be targeting
• 11:00
  patients that are at risk, right? So usually
• 11:02
  minorities.
• 11:03
  And actually did the implementation through a study
• 11:06
  where we decided we're not going to over
• 11:08
  book because we think that that actually worsen
• 11:10
  disparity in this but we're actually going to
• 11:12
  provide more resources for outreach to bring patients
• 11:14
  that risk not showing up into the system.
• 11:17
  And then doing so actually reduced d show
• 11:19
  rate in our Black patients by 15 percent.
• 11:22
  And so we kind of hold that up
• 11:23
  as an example of how an
• 11:26
  organization
• 11:27
  can responsibly
• 11:28
  validate implement
• 11:31
  a model that could otherwise worsen disparities in
• 11:34
  a way where not only do all new
• 11:36
  tied with all boats, but
• 11:38
  really you were able to generate a benefit
• 11:41
  to your patient population
• 11:43
  in a very equitable way. We're actually the
• 11:46
  outside effect help build that were less fortunate.
• 11:49
  So that's
• 11:50
  kind of the work that I do that's
• 11:52
  where I find myself
• 11:54
  spending a lot of time thinking
• 11:56
  about these problems and thinking about how to
• 11:57
  bring these
• 11:59
  solutions in. And I think the opportunity like
• 12:01
  I said earlier
• 12:02
  is the tooling needs a lot of work,
• 12:05
  but I'm excited that
• 12:06
  with better standards, data standardization and interoperability,
• 12:10
  getting people on the same
• 12:12
  kind of playing field where I can share
• 12:15
  my model and my experience
• 12:17
  and data
• 12:19
  in a way where I can validate your
• 12:20
  model
• 12:21
  right? On my data for instance, Having that
• 12:24
  interchange happen
• 12:25
  in a way where I get it's
• 12:27
  protected,
• 12:28
  no particular information is passed,
• 12:30
  I think that's incredibly powerful. So I see
• 12:33
  this as a continuum of
• 12:36
  everything that we've been doing in informatics
• 12:38
  making that world just so much much more
• 12:41
  potentially powerful and empowering
• 12:43
  some of the organizations with western resources is
• 12:45
  to join the Ai revolution that
• 12:48
  otherwise, may just be relegated to, you know,
• 12:50
  the the large academic centers.
• 12:54
  That's fascinating to hear. Thank you so much
• 12:55
  for going through that with us dig you
• 12:57
  1 bit deeper, especially on Ai. Now what
• 13:00
  are you most excited about for the future?
• 13:01
  And what makes you nervous?
• 13:05
  So
• 13:06
  I'm excited about
• 13:08
  the opportunity
• 13:09
  that we have,
• 13:11
  I think
• 13:12
  maybe even sufficiently excited about
• 13:15
  the dying down of the Ai hike.
• 13:17
  So the more pragmatic
• 13:20
  folks rate that weather the storms of new
• 13:23
  technologies
• 13:23
  and the hype cycles, right? We'll tell you
• 13:26
  that we really just want to know what
• 13:28
  works, right and what works for our patients
• 13:30
  and what enable us to provide the care
• 13:32
  that we'd like.
• 13:33
  So
• 13:34
  I think that the current energy
• 13:37
  is around
• 13:39
  trying to validate an Fda regulate and even
• 13:42
  create
• 13:43
  assurance
• 13:44
  bodies to basically sanction
• 13:47
  certain models as
• 13:48
  usable
• 13:49
  or safe.
• 13:51
  I would say that really we should
• 13:54
  I worry a
• 13:56
  about the amount of money that we're spending
• 13:58
  as the healthcare
• 13:59
  organizations following that high cycle.
• 14:01
  So I am looking forward to that dying
• 14:04
  down, right? And for us to really to
• 14:06
  see these the real solutions come out also
• 14:09
  to
• 14:10
  likely actually lead to more
• 14:13
  or should I should say surface the more
• 14:16
  obvious integrated solutions rather than the point solutions
• 14:19
  that
• 14:20
  are more hype than actual value.
• 14:24
  So that's the concern right. We spend more
• 14:25
  money there that concerns me. But the opportunity
• 14:28
  again, is once the field starts to kind
• 14:30
  of
• 14:32
  or I should say once the weaker options
• 14:35
  shake out, I think the path forward for
• 14:37
  driving true value from this
• 14:40
  Ai clinical
• 14:41
  revolution, I think it becomes more obvious. And
• 14:45
  that's the kind of work that many us
• 14:47
  in this area are excited about.
• 14:49
  Really
• 14:50
  getting beyond the hype cycle goal and actually
• 14:53
  starting to integrate
• 14:55
  these
• 14:57
  solutions into our process and the workflow
• 14:59
  in a meaningful way.
• 15:01
  The other parallel to that for me in
• 15:03
  the virtual care space. So
• 15:06
  mentioned that we're building this cross
• 15:08
  organization
• 15:09
  so
• 15:11
  Metro health and Have partnered together
• 15:14
  to build a
• 15:15
  common virtual care platform, which is meant to
• 15:18
  be
• 15:20
  inter and integrated with healthcare care systems that
• 15:22
  really want to
• 15:24
  extend the care that they do.
• 15:27
  And the differentiation for us there is
• 15:30
  being able to provide the additional service in
• 15:32
  a way where really from the patient perspective
• 15:34
  it's a seamless which part of their journey.
• 15:36
  That when you have reasons to come back
• 15:38
  into and organization's brick and mortar services,
• 15:42
  that happens
• 15:43
  seamlessly. And so that also
• 15:46
  common theme here being built on
• 15:49
  great data standards
• 15:51
  and
• 15:52
  interoperability.
• 15:53
  And so these are the kinds of things
• 15:54
  that you can unlock
• 15:56
  when we all start speaking that same language.
• 15:58
  And that for me is very exciting. And
• 16:01
  so to kind of bring those 2 examples
• 16:03
  for circle,
• 16:05
  we need to be continuing to spend more
• 16:07
  money
• 16:08
  on the right things
• 16:10
  and the right things are
• 16:12
  continuing to integrate our systems
• 16:14
  to create an opportunity across the board for
• 16:16
  a lot of these
• 16:18
  initiatives and collaborations to really
• 16:22
  to break through and really deliver on that
• 16:24
  promise,
• 16:25
  rather than spending money
• 16:27
  on
• 16:28
  shiny objects
• 16:30
  The other thing in terms of kind of
• 16:32
  spending money is new regulation, so that concerns
• 16:35
  be coming around the corner. So
• 16:38
  how is the Fda planning to regulate a
• 16:40
  lot of this technology?
• 16:42
  And what is that do to the innovation
• 16:45
  space, right? So does that increase the barrier
• 16:48
  of entry? Does it take away the ability
• 16:51
  to democrat
• 16:52
  the both
• 16:55
  and
• 16:56
  of of some of these Ai
• 16:59
  initiatives.
• 17:01
  It's fascinating in here. You know, and really,
• 17:04
  cool to to think through
• 17:06
  all the possibilities that are ahead. Before we
• 17:08
  wrap up here, I'm wondering, could you talk
• 17:10
  about what the most effective healthcare leaders will
• 17:12
  need in order to be successful over the
• 17:14
  next 2 to 3 years, especially plans as
• 17:16
  Ai continues to evolve and in grow this
• 17:19
  space and technology becomes an even bigger part
• 17:21
  of health delivery and in systems
• 17:24
  operations.
• 17:26
  Yes. So
• 17:28
  I'll recap that
• 17:30
  to create change
• 17:32
  in a healthcare environment
• 17:34
  I think we have
• 17:36
  already
• 17:37
  a slew of amazing models to build off.
• 17:41
  And so
• 17:42
  modeling around this notion of learning health systems,
• 17:46
  right, building
• 17:48
  organizations and
• 17:49
  that attempt to be agile and learn from
• 17:52
  the data they generate to create this loop
• 17:55
  of continuous quality improvement.
• 17:57
  How do you
• 17:59
  how do you change, right? And how do
• 18:01
  you manage change And how do you implement
• 18:04
  things that you know should be implemented and
• 18:07
  how do you
• 18:08
  monitor the effects from a quality improvement perspective.
• 18:11
  I think these are the many
• 18:14
  very
• 18:15
  specific
• 18:16
  areas that somebody who's in clinical informatics
• 18:20
  is a clinician leader even needs to think
• 18:22
  about right from an
• 18:23
  organizational perspective.
• 18:25
  If you want to add Ai to that,
• 18:28
  I'll recap this notion of the Light Quan,
• 18:30
  which is start understanding enough about the technology
• 18:34
  to understand the implication, the limitations and the
• 18:38
  opportunities
• 18:39
  that allow to bring it back
• 18:41
  into the system.
• 18:43
  And so
• 18:45
  what I guess I'm trying to say in
• 18:46
  the nutshell is, we have these really great
• 18:48
  tools, right in the informatics universe that
• 18:51
  we leverage to do what we do
• 18:54
  And so building on that, I think
• 18:57
  where an informatics leader specifically
• 19:01
  used to sit, right, 10 to 20 years
• 19:03
  ago they were the physician champion of the
• 19:06
  electronic health record,
• 19:07
  right but that digital ecosystem now has expanded
• 19:11
  And so now they need to be thinking
• 19:13
  much more broadly
• 19:15
  about
• 19:16
  where we meet patients, how we meet patients
• 19:19
  and that's exciting, right, because now we're getting
• 19:22
  into the act I think the point of
• 19:23
  contact
• 19:25
  and the doors if you will have expanded
• 19:26
  in terms of opportunity.
• 19:28
  And so they need to be thinking about
• 19:31
  digital care. They need to be thinking about
• 19:33
  patients as consumers with choice.
• 19:35
  And beyond that, they need to be thinking
• 19:37
  less
• 19:38
  tactically, like how do we survive the day
• 19:41
  and more strategically,
• 19:43
  why should we acquire and implement this platform
• 19:46
  or technology?
• 19:47
  How can we carry that forward? Great? How
• 19:49
  can we how can we turn this into
• 19:50
  a successful return our investment
• 19:53
  for our organization? And so I really do
• 19:55
  just feel like this
• 19:56
  a more initially limited role of a clinical
• 19:59
  and informatics leader
• 20:01
  has just kind of blown up. And I
• 20:02
  think you've seen
• 20:04
  pieces of that right
• 20:05
  specializations now and there's your Chief Digital Officer
• 20:08
  and
• 20:10
  your Chief Ai
• 20:11
  sir and it's really just shown the need
• 20:14
  for people to start exploring some of the
• 20:17
  expanding tentacles of the health technology states.
• 20:21
  I guess like conclusion really, there's
• 20:23
  a lot that we already know about how
• 20:26
  to create effective change and lead within healthcare
• 20:29
  and from within healthcare, which is I think
• 20:31
  where most of the gist are going to
• 20:32
  happen.
• 20:33
  It's just a matter of where you think
• 20:35
  you need to sit in terms of driving
• 20:37
  that change. So Ai is not for everybody.
• 20:40
  It's going to be ubiquitous, but it's
• 20:42
  definitely going to come to a point where
• 20:45
  there's going some people don't understand enough about
• 20:47
  a drive forward and the rest of us
• 20:48
  are just basically going to have to
• 20:51
  so that's just going to be a part
• 20:52
  of life.
• 20:53
  Doctor Terry. Thank you so much for joining
• 20:55
  us on the podcast today. This has been
• 20:56
  a fascinating discussion, and I look forward to
• 20:58
  connecting with you again soon.
• 21:01
  Yeah. Thank you for having. It's it's been
• 21:02
  fun.