Latest / Becker’s Payer Issues Podcast / Leveraging Data and Human Centered Care in Medicare Advantage with Dr. Krystal Revai

Transcript

Leveraging Data and Human Centered Care in Medicare Advantage with Dr. Krystal Revai · Becker’s Payer Issues Podcast

• 0:02
  Generative AI is driving the era of infinite
• 0:05
  healthcare abundance.
• 0:06
  At Hippocratic AI, we partner with more than
• 0:09
  60 innovative healthcare organizations
• 0:11
  across payers, providers, and life sciences
• 0:14
  to drive patient and member outcomes.
• 0:16
  Offering the safest clinical AI voice agents, Hippocratic
• 0:20
  AI is the first company to define comprehensive
• 0:23
  voice AI benchmarks for healthcare.
• 0:25
  Safety, clinical accuracy,
• 0:27
  empathy, consistency.
• 0:29
  Learn more at hippocraticai.com.
• 0:32
  This is Gracelyn Keller with the Becker's Healthcare
• 0:34
  podcast, and we are recording live at the
• 0:37
  fourth annual spring pair issues roundtable.
• 0:40
  I'm currently joined by Crystal Revai, who is
• 0:42
  the chief medical officer at Atrio Health. Crystal,
• 0:46
  thanks for being here. I'm gonna have you
• 0:47
  start off by introducing yourself and sharing a
• 0:49
  little bit more about your role and organization.
• 0:51
  Great. Thank you so much. It's great to
• 0:54
  be here. Again, my name is Crystal Revai.
• 0:56
  I am the chief medical officer
• 0:59
  for Atrio Health.
• 1:01
  I have been in my role for just
• 1:03
  a short time at the time of this
• 1:04
  recording, just about two months.
• 1:07
  I have worked in managed care in the
• 1:09
  on the the commercial side and the Medicare
• 1:11
  Advantage side for about thirteen
• 1:14
  years.
• 1:15
  ATREO
• 1:16
  is a Medicare Advantage plan that is based
• 1:19
  out of the state of Oregon, and we
• 1:21
  really,
• 1:22
  partner
• 1:23
  with, small community
• 1:25
  organizations in order to,
• 1:28
  create the best health experience for our members.
• 1:30
  Wonderful. Well, thank you for taking the time.
• 1:33
  And let's start our conversation with from your
• 1:35
  perspective,
• 1:36
  what is the most pressing operational challenge facing
• 1:39
  payers right now, and how are you addressing
• 1:40
  it? I think one of the most pressing
• 1:43
  operational challenges
• 1:45
  for payers
• 1:46
  is around data.
• 1:48
  Who has the data,
• 1:50
  and how are we best able to utilize
• 1:53
  it to answer the questions that we're looking
• 1:56
  for? We've come a long way in terms
• 1:59
  of electronic medical record interoperability,
• 2:03
  and a lot of that came because of
• 2:05
  federal mandates.
• 2:06
  And hospitals and health care providers are getting
• 2:10
  more comfortable with payers
• 2:12
  having access to their EMR.
• 2:15
  But when it comes to data
• 2:17
  and actual, you know, claims data,
• 2:20
  diagnostic data,
• 2:22
  we are still a little bit behind
• 2:24
  the times in terms of being able to
• 2:26
  freely
• 2:27
  share data with each other. For some of
• 2:30
  the national plans, you know, they have the
• 2:32
  ability and they have the capability
• 2:34
  to harness large
• 2:36
  data banks and be able to and have
• 2:39
  teams that can readily,
• 2:41
  analyze that data and ask a myriad of
• 2:44
  different questions.
• 2:46
  For small plants like us, we are often
• 2:48
  relying on vendors
• 2:50
  who might be collecting the data for us.
• 2:53
  And if there are a number of different
• 2:55
  vendors
• 2:56
  and we
• 2:57
  need their two data systems to speak to
• 2:59
  each other, that can be operationally very challenging.
• 3:03
  Absolutely. And there is a constant tension between
• 3:06
  hooping costs down and actually delivering a good
• 3:08
  member experience, especially as it comes to care
• 3:10
  coordination.
• 3:12
  So how is your organization balancing cost containment
• 3:15
  with improving member experience and quality outcomes, and
• 3:18
  where are you seeing the greatest opportunity for
• 3:20
  impact?
• 3:21
  Yeah. So what we're targeting
• 3:23
  right now from a cost containment standpoint is
• 3:26
  really
• 3:27
  focused on our
• 3:29
  most critically
• 3:30
  ill members.
• 3:31
  So really using targeted
• 3:35
  care management
• 3:36
  to address the members who not just have
• 3:40
  the highest utilization,
• 3:42
  but also are at highest risk because of
• 3:45
  the number of comorbidities
• 3:47
  that they have, perhaps the number of medications
• 3:50
  that they're taking, and bring them into a
• 3:53
  chronic care management system
• 3:56
  so that
• 3:57
  we can help to
• 3:59
  even
• 4:00
  anticipate
• 4:01
  where some of their costs may come in
• 4:04
  the future
• 4:05
  and address them before it becomes an issue
• 4:07
  for the member. So we're actually going a
• 4:10
  little bit old school in terms of we
• 4:12
  are
• 4:13
  doing person to person
• 4:15
  interactions.
• 4:16
  We're not doing it with an AI bot
• 4:18
  or automated calls,
• 4:20
  but we're targeting
• 4:22
  we're starting with a target of high impact
• 4:25
  members to be able to bring
• 4:28
  their costs down. And we're also not only
• 4:30
  targeting their,
• 4:32
  barriers in terms
• 4:34
  of social determinants of health,
• 4:36
  but also where are their quality gaps? Where
• 4:39
  where are they needing to fulfill some of
• 4:41
  those,
• 4:42
  requirements in terms of their hemoglobin a one
• 4:44
  c? How's their blood pressure coming under control?
• 4:47
  It's it's both. Can you get to the
• 4:49
  doctor?
• 4:50
  Do you understand your medicine? And
• 4:53
  I understand that you don't really like the
• 4:54
  idea of doing a colonoscopy.
• 4:57
  There are some other options to
• 4:59
  get screened for colon cancer.
• 5:02
  And how do you see AI supporting member
• 5:04
  facing interactions while maintaining that trust and quality?
• 5:08
  I'll answer that with a negative first. And
• 5:11
  the first is that I don't
• 5:14
  see the AI chatbots
• 5:17
  as
• 5:18
  the way to do it, especially for some
• 5:21
  of these higher risk members.
• 5:24
  I think that you can have AI
• 5:27
  in the background
• 5:28
  assisting
• 5:29
  our
• 5:30
  customer care
• 5:32
  providers
• 5:32
  in
• 5:33
  doing real time evaluation, real time coaching
• 5:37
  to help them interact best with the members
• 5:39
  that they are speaking to face to face.
• 5:42
  For me, I think that
• 5:44
  members and the public in general is getting
• 5:47
  much more sophisticated
• 5:50
  at being able to tease out when something
• 5:52
  is AI and when it's a real human.
• 5:55
  And so having
• 5:57
  support systems in place where AI
• 6:01
  helps the real person in that interaction
• 6:04
  in interaction is, for me, the way we
• 6:07
  maintain trust and quality
• 6:10
  because we all know the AI system is
• 6:12
  only as good as the information it's fed.
• 6:15
  And so if somewhere down the line, it's
• 6:17
  fed or they have I think it's called
• 6:21
  there's a term for it when
• 6:23
  the AI has one piece of misinformation, and
• 6:25
  it just keeps excel it just it just
• 6:27
  keeps perpetuating
• 6:28
  that piece of misinformation.
• 6:30
  And so we know that that happens, and
• 6:32
  then the quality just degrades. Mhmm.
• 6:35
  And health care workforce shortages continue to impact
• 6:38
  payers and providers alike. So how are you
• 6:40
  thinking about AI as a way to augment,
• 6:42
  not replace care teams, especially in areas like
• 6:46
  call centers, care management, and population health?
• 6:49
  Well, I think from the call centers, the
• 6:51
  a the AI coach, we can call it
• 6:54
  Mhmm.
• 6:55
  Is really essential, especially since many of our
• 6:58
  call center employees
• 6:59
  are not clinically educated.
• 7:02
  So these aren't nurses
• 7:04
  or pharmacists
• 7:04
  or or occupational therapists,
• 7:07
  but when the members call in, they often
• 7:09
  have a clinical question. And so
• 7:12
  training the AI
• 7:14
  to help the call center
• 7:17
  employee
• 7:18
  not give information and not to give education,
• 7:21
  but how best to word the response
• 7:24
  that
• 7:25
  engages the member in a way that's helpful
• 7:27
  without
• 7:28
  crossing the boundary into clinical care.
• 7:31
  From in care in terms of care management
• 7:33
  and population health, I mean, I think AI,
• 7:36
  one thing that we know that it's very
• 7:37
  good at is analyzing data,
• 7:40
  giving it a fixed dataset
• 7:43
  to give. So things like when you have
• 7:45
  the appropriate,
• 7:47
  safety measures in place, uploading clinical records,
• 7:50
  analyzing even conversations. Right? Approved you had a
• 7:54
  conversation with a care management has a conversation
• 7:57
  with a member.
• 7:59
  Two human beings, it's difficult to remember. As
• 8:01
  good of a note taker as you are,
• 8:03
  it's good to remember it's hard to remember
• 8:05
  every single thing that said. So using
• 8:08
  AI potentially
• 8:09
  to analyze
• 8:10
  that
• 8:11
  case that that
• 8:13
  that,
• 8:14
  care coordinator's conversation
• 8:16
  with the member and say, hey. Did you
• 8:19
  notice
• 8:20
  they paused when you asked about
• 8:23
  if they had enough food in the house?
• 8:24
  That could indicate that there is some food
• 8:26
  insecurity.
• 8:27
  You might wanna dig into that
• 8:30
  at a at a future date.
• 8:31
  Absolutely.
• 8:33
  And wrapping up our conversation,
• 8:34
  AI voice technology
• 8:36
  has changed pretty dramatically,
• 8:38
  and we're not talking about robo calls or
• 8:40
  IVR menus anymore. There's now AI that can
• 8:43
  hold a real conversation, show empathy, ask follow-up
• 8:45
  questions,
• 8:46
  and know when to bring in that human.
• 8:49
  So you do you think that we're at
• 8:50
  the point where a patient or member can
• 8:52
  have that kind of interaction with an AI
• 8:55
  and actually trust it, or what has to
• 8:57
  be true in your opinion for that to
• 8:59
  work?
• 9:01
  I, honestly, I don't think that we are
• 9:04
  there.
• 9:06
  I think that, again, because
• 9:09
  AI has
• 9:10
  it it doesn't really have the ability to
• 9:12
  judge. Mhmm. It's just giving information.
• 9:15
  You know, what has to be true for
• 9:17
  that to work? They have to be able
• 9:18
  to make, you know, human judgment calls, which
• 9:21
  is just not feasible.
• 9:23
  For me, I think that
• 9:25
  this borders on a little bit of,
• 9:28
  you know, just fantasy, sort of futuristic
• 9:33
  thinking in that at some point, we might
• 9:36
  have, you know, true
• 9:38
  AI counselors that are able to be
• 9:40
  empathetic and truthful and have human level judgment.
• 9:44
  Yeah. At that point, the machines might really
• 9:46
  take over, and that's frightening future for me.
• 9:49
  Well, Crystal, thank you so much for taking
• 9:51
  the time to join me today on the
• 9:53
  Becker's healthcare podcast,
• 9:54
  share these thoughts and insights. Again, we are
• 9:57
  recording live at the fourth annual spring payer
• 9:59
  issues roundtable.
• 10:01
  Thank you.