Latest / Becker’s Healthcare Podcast / Nabile M. Safdar, Chief AI Officer, Emory Healthcare

Transcript

Nabile M. Safdar, Chief AI Officer, Emory Healthcare · Becker’s Healthcare Podcast

• 0:00
  Philips is a health tech leader focused on
• 0:03
  innovation that improves the health and well-being of
• 0:06
  people. Our health care technology and informatics solutions
• 0:09
  help care teams diagnose, treat, and manage more
• 0:12
  patients with greater precision, speed, and confidence across
• 0:16
  the care journey. With Philips, clinicians are empowered
• 0:19
  with streamlined insights in the moments that matter
• 0:21
  for every patient. Better care for more people.
• 0:25
  Philips.
• 0:28
  This is Gracelyn Keller with the Becker's Healthcare
• 0:30
  Podcast, and we are live at the tenth
• 0:33
  annual health IT digital health and RCM meeting.
• 0:36
  I'm currently joined by Nabil Sabdar who is
• 0:39
  the chief AI officer at Emory Healthcare.
• 0:42
  Nabil, thanks for being here. We'd love to
• 0:44
  have us start off by having you share
• 0:45
  a little bit more about your work in
• 0:47
  health care and your background.
• 0:49
  Yeah. Sure. Thank you, Grace.
• 0:51
  I am a radiologist with a background in
• 0:54
  imaging informatics and clinical informatics,
• 0:58
  and that has led me to my current
• 1:00
  role as the chief AI officer for Emory
• 1:03
  Healthcare and Emory University.
• 1:06
  You may be wondering how do you get
• 1:08
  from point a to point b. And if
• 1:10
  you recall,
• 1:11
  radiology
• 1:12
  is, one of the most digital specialties and
• 1:15
  one of the areas
• 1:17
  which has been,
• 1:18
  early adoption
• 1:19
  where early adoption of AI has been really
• 1:22
  prominent,
• 1:23
  to the point where many people have speculated
• 1:25
  that,
• 1:26
  we could go away as a profession. Well,
• 1:28
  I was interested in that, out of self
• 1:30
  preservation,
• 1:32
  and, I thought I'd learn more. And in
• 1:33
  fact, I've been involved with some research,
• 1:36
  implementation,
• 1:37
  deployment,
• 1:38
  governance
• 1:39
  in AI for quite some time,
• 1:41
  have been working with Emery Digital.
• 1:43
  And so now I have the privilege of
• 1:45
  serving on a team that
• 1:47
  helps design governance
• 1:50
  for the system, helps
• 1:52
  to manage the portfolio
• 1:54
  of AI efforts
• 1:55
  that we are doing across
• 1:58
  different types of,
• 2:00
  applications, whether they're clinical
• 2:02
  or operational
• 2:04
  or marketing,
• 2:05
  patient facing or internally facing,
• 2:08
  and really helps to evaluate,
• 2:11
  performance, evaluation, determine ROI, all those great things.
• 2:15
  Wonderful.
• 2:15
  Well, let's start our conversation talking about AI.
• 2:18
  Nearly half of medical practices reported using AI
• 2:21
  in some capacity in the past year Mhmm.
• 2:23
  And it remains a key topic for health
• 2:25
  IT leaders.
• 2:26
  So from your perspective, what are the use
• 2:28
  cases that are making a difference right now,
• 2:30
  and how are you leveraging them in your
• 2:31
  organization?
• 2:33
  Yeah. I think, you know, many,
• 2:36
  many of us who are at a conference
• 2:37
  like this or are discussing this with each
• 2:40
  other at health care systems across the nation
• 2:43
  are familiar with the the top three or
• 2:45
  three or four use cases. So,
• 2:48
  we've had a lot of success with Ambien
• 2:50
  AI.
• 2:51
  We've been using,
• 2:53
  FDA cleared
• 2:54
  AI in areas like cardiology
• 2:57
  and radiology
• 2:58
  and, hopefully, soon pathology as well as we
• 3:00
  do our digital pathology,
• 3:02
  transition.
• 3:04
  We,
• 3:05
  see
• 3:06
  early success
• 3:07
  with
• 3:08
  agentic chatbots
• 3:10
  who are reaching out to patients to help
• 3:12
  remind them to make sure they fill their
• 3:15
  prescriptions,
• 3:16
  their get their medications
• 3:18
  after discharge.
• 3:20
  We've also had some good early success,
• 3:23
  with those same types of agentic
• 3:26
  agents, chat agents reaching out to patients to,
• 3:30
  help them help walk them through taking their
• 3:32
  blood pressure at home
• 3:34
  and then making sure that they're in a
• 3:36
  range which is safe for them and if
• 3:38
  not quickly escalating,
• 3:40
  to a human,
• 3:41
  who can help them navigate to the next
• 3:43
  appropriate steps.
• 3:45
  And and what we've realized is there is
• 3:47
  an incredible
• 3:49
  demand
• 3:50
  for
• 3:51
  the technology
• 3:52
  that is powering our daily lives now. So
• 3:55
  many
• 3:57
  people,
• 3:58
  physicians,
• 3:59
  nurses,
• 4:00
  technologists,
• 4:00
  administrators,
• 4:01
  managers
• 4:03
  in health care are using something like CHAT
• 4:05
  g p t,
• 4:06
  something like Claude, something like Gemini at home
• 4:10
  in their everyday lives, they wanna have that
• 4:12
  same efficiency of work
• 4:15
  in their,
• 4:16
  same efficiency in their work as well.
• 4:19
  And so we are also bringing them safe
• 4:22
  kind of HIPAA,
• 4:24
  safe harbor,
• 4:25
  you know, tools to be able to use
• 4:27
  that, those types of frontier models, those LLMs,
• 4:30
  those chatbots in their daily work as well.
• 4:32
  So a lot of opportunity,
• 4:34
  a lot going on,
• 4:36
  a a lot more opportunity in areas like
• 4:38
  rev cycle,
• 4:40
  marketing,
• 4:41
  you know, access, etcetera,
• 4:43
  where we're we're dipping our toes in the
• 4:45
  water there as well.
• 4:46
  Sounds like it. And as virtual care expands
• 4:50
  from AI enabled tools and remote monitoring
• 4:52
  to broader digital health platforms,
• 4:54
  introducing new technology brings challenges. So what advice
• 4:57
  do you have for leaders navigating everything from
• 5:00
  governance to patient engagement? And can you share
• 5:02
  an example of how your organization has balanced
• 5:04
  innovation with operational constraint?
• 5:06
  Yeah. This is, I think there are two
• 5:08
  extremes in governance which are to be avoided.
• 5:10
  One is the absence of AI governance that
• 5:13
  will get your institution, your organization
• 5:16
  into trouble.
• 5:17
  You know, that see looks like basically everything
• 5:20
  goes, you know, data deals,
• 5:22
  at will,
• 5:24
  AI deployments at will. If somebody has the
• 5:26
  budget, they can move forward and do what
• 5:28
  they want, and there's no systematic
• 5:31
  oversight. There's no risk posture.
• 5:34
  There's no statement of principles from top leadership
• 5:37
  about appropriate use of AI.
• 5:39
  On the other
• 5:40
  hand, governance can be so restrictive
• 5:43
  that there is only a single committee which
• 5:45
  becomes
• 5:47
  a choke point in the system for the
• 5:49
  deployment
• 5:50
  of AI of novel AI solutions, and,
• 5:54
  any large system needs to be more nimble
• 5:56
  than that. So,
• 5:58
  what we've had success with is
• 6:00
  helping
• 6:01
  specific functional areas, departments,
• 6:04
  clinical specialties, etcetera, to develop their own AI
• 6:07
  governance, whether that's in imaging, whether that is
• 6:11
  in access and rep cycle, whether that is,
• 6:14
  in pathology,
• 6:16
  or clinical
• 6:17
  informatics.
• 6:18
  You know, the they're the experts that know
• 6:21
  what they need. They can, be helped to
• 6:23
  understand what the priorities of the organization are,
• 6:26
  what that alignment looks like, and, really, we
• 6:28
  should trust them to make appropriate decisions,
• 6:31
  especially if we provide them with the triggers,
• 6:35
  that,
• 6:36
  would
• 6:37
  indicate a need for a higher level of
• 6:39
  scrutiny or review, at which point it can
• 6:42
  be a decision around AI can be kicked
• 6:44
  up to,
• 6:45
  another governance group. So that's been our our
• 6:47
  strategy is let people make decisions about AI,
• 6:50
  but give them clear guidelines about what is
• 6:53
  and what is not inbounds so they can
• 6:56
  kick out the, decisions that are a little
• 6:58
  bit more risky, maybe something that's more patient
• 7:01
  facing, more autonomous,
• 7:02
  more,
• 7:03
  agentic,
• 7:05
  maybe if there are conflicts of interest or
• 7:07
  if we suspect that there's unauthorized use of
• 7:09
  AI,
• 7:10
  for for further scrutiny.
• 7:13
  And shifting gears slightly, how are you seeing
• 7:15
  recent legislation,
• 7:17
  both state and federal, affect health care organizations
• 7:20
  and health IT specifically,
• 7:22
  and have you adjusted strategies in response?
• 7:25
  Yeah. Great question. So, you know, keeping an
• 7:27
  eye on regulations
• 7:29
  is challenging.
• 7:30
  First of all, you have to know the
• 7:32
  players.
• 7:33
  You know, there's congress.
• 7:35
  There's state legislatures,
• 7:37
  there's CMS, HHS,
• 7:39
  ONC,
• 7:41
  even OCR, the Office of Civil Rights. All
• 7:43
  of these are relevant agencies,
• 7:46
  for health care and AI and health care.
• 7:48
  So
• 7:49
  it can be a full time job to
• 7:51
  keep an eye on the regulations affecting
• 7:54
  digital and AI
• 7:56
  in a health care environment.
• 7:58
  We are lucky to have an office that
• 8:00
  looks at these things, and so we've been
• 8:02
  partnering with them.
• 8:03
  But, also,
• 8:05
  our AI team has been looking at these
• 8:07
  regulations very carefully.
• 8:09
  What we have seen,
• 8:10
  a couple things we're keeping track of is
• 8:13
  earlier in the year,
• 8:14
  the federal government at some point,
• 8:17
  said that they would try to bar states
• 8:20
  from having AI legislation,
• 8:22
  that,
• 8:24
  seemed to go away in July.
• 8:26
  There are over well over 200
• 8:30
  pieces of potential legislation or legislation in state,
• 8:34
  legislator legislatures
• 8:36
  across the country.
• 8:38
  And, we're watching those carefully, especially in the
• 8:40
  big influential states.
• 8:42
  California recently passed SB 53,
• 8:46
  which would require some of the large AI,
• 8:49
  foundation
• 8:50
  labs, the frontier that create frontier models, the
• 8:53
  LLMs,
• 8:54
  to,
• 8:55
  publish their, safety protocols, their their safety profiles.
• 8:59
  We're waiting to see what kind of impact
• 9:01
  that's gonna have on us in health care
• 9:03
  and to those that use their products as
• 9:06
  derivatives through APIs,
• 9:08
  or through other means, and what it means
• 9:10
  when you're trying to fine tune a model
• 9:11
  for health care. A lot to keep track
• 9:13
  of. It's an active process.
• 9:16
  Right now, I think we're able to proceed
• 9:19
  because we do have safe areas, like the
• 9:21
  areas that are, FDA cleared,
• 9:24
  the FDA cleared devices. It's it's also an
• 9:26
  evolving area. The FDA has been think changing
• 9:28
  its thinking as time goes on, so we're
• 9:30
  keeping an eye on that as well.
• 9:32
  Really, what this you know, our my advice
• 9:35
  is to keep an eye on it, to
• 9:37
  talk to your vendor partners,
• 9:39
  to talk if you have an office
• 9:41
  of regulatory affairs, to have close,
• 9:44
  coordination with them,
• 9:46
  and and potentially even impact,
• 9:48
  your own state's legislation if appropriate.
• 9:52
  And final question as we wrap our conversation
• 9:54
  up. What is your top piece of advice
• 9:56
  for health care leaders as they prepare for
• 9:58
  further advancements in technology and rising demands for
• 10:00
  care?
• 10:01
  Yeah. I mean, this is a,
• 10:03
  this is a journey in which it is
• 10:05
  harder and harder to look,
• 10:07
  out the front windshield and see far down
• 10:10
  the road.
• 10:11
  Things are changing fast, so it's a curvy
• 10:13
  road, and you need to be very agile
• 10:15
  and nimble.
• 10:16
  I think it is important to get, alignment
• 10:20
  with, the leadership of your organization. We often
• 10:23
  focus on digital governance, IT governance, or maybe
• 10:26
  even data and AI governance.
• 10:28
  However,
• 10:29
  corporate governance,
• 10:31
  is, important more important than ever. And aligning
• 10:34
  AI strategy with corporate governance, I think, is
• 10:37
  critical. And then also having
• 10:40
  lots of connections
• 10:41
  to those on the front lines,
• 10:43
  nurses, physicians, providers,
• 10:46
  administrators who are doing the work,
• 10:48
  to make sure that we are not imposing,
• 10:51
  AI,
• 10:53
  or other solutions on them, that really they're
• 10:55
  a part of the process. They're engaged early
• 10:58
  on. They're co developing solutions, co creating solutions.
• 11:02
  Those are integrated well into their workflows, and
• 11:04
  then that we celebrate the wins that they
• 11:06
  have as they use these tools in their,
• 11:08
  everyday lives.
• 11:09
  Wonderful. Well, thank you so much for joining
• 11:11
  me today on the Becker's Healthcare podcast, sharing
• 11:14
  these thoughts. Again, we are recording live at
• 11:15
  the tenth annual health IT, digital health, and
• 11:17
  RCM meeting. Thank you.