Latest / Becker’s Healthcare Digital Health + Health IT / Using AI to Move Healthcare From Reactive to Proactive with Dr. Penelope Fu

Transcript

Using AI to Move Healthcare From Reactive to Proactive with Dr. Penelope Fu · Becker’s Healthcare Digital Health + Health IT

• 0:00
  - What we wanna do with AI is help
• 0:01
  - clinicians go from being reactive
• 0:04
  - to proactive.
• 0:05
  - At Philips, we can recognize patterns in monitoring
• 0:08
  - data across populations
• 0:10
  - and apply those insights to specific patients.
• 0:14
  - Safe and responsible AI can support sharper insights,
• 0:18
  - pave the way for faster diagnosis,
• 0:21
  - and help streamline care.
• 0:23
  - That can benefit clinicians, it can benefit health
• 0:26
  - systems, and most importantly, it can benefit patients
• 0:29
  - and their families.
• 0:33
  - Phillips.
• 0:34
  - Hi, everyone. This is Gracelyn Keller with Becker's
• 0:36
  - Healthcare. Thank you so much for tuning in
• 0:38
  - to the Becker's Healthcare podcast series. We are
• 0:40
  - recording live at the eleventh annual IT revenue
• 0:42
  - cycle conference, the future of AI and digital
• 0:45
  - health.
• 0:46
  - I'm joined for today's discussion by doctor Penny
• 0:49
  - Fu, who is the chief health informatics officer
• 0:51
  - at Michigan Medicine. So, doctor Fu, thanks for
• 0:53
  - being here. Thank you for inviting me. Yeah.
• 0:56
  - Absolutely. And I'm excited to jump right into
• 0:58
  - our conversation.
• 0:59
  - So health systems are receiving a growing stream
• 1:02
  - of patient data from bedside monitoring, ambulatory devices,
• 1:05
  - and care delivered at home. Where is that
• 1:07
  - data creating actionable insights for clinicians, and where
• 1:10
  - does it still create more work than insight?
• 1:12
  - That's such a great question. I think one
• 1:14
  - of the challenges that we have is around
• 1:17
  - what criticality of the data is, Because sometimes
• 1:21
  - you have normal signals and you don't need
• 1:22
  - to do anything, but sometimes they're out of
• 1:24
  - bounds. And that's really when you have to
• 1:26
  - have the tools to support because you can't
• 1:28
  - really have people looking at the data stream
• 1:31
  - twenty four seven.
• 1:32
  - We really need to make sure that we
• 1:33
  - have an ecosystem around the use cases where
• 1:35
  - we are collecting that data so that we
• 1:37
  - are appropriately signaling where we need to have
• 1:39
  - intervention.
• 1:41
  - And when clinicians are trying to reach a
• 1:43
  - diagnosis, they often must piece together information from
• 1:45
  - radiology, cardiology, pathology, urology, all of these other
• 1:49
  - systems.
• 1:50
  - So where does that fragmentation
• 1:52
  - cause the most friction in your organization, and
• 1:54
  - how do you bring together data sources and
• 1:56
  - address this fragmentation or close the gaps? You're
• 1:59
  - really picking some hard questions here. I think
• 2:01
  - that, you know, EHR systems like Epic and
• 2:04
  - Oracle really have helped to close the gap
• 2:06
  - in terms of aggregating all the information to
• 2:08
  - a certain place.
• 2:09
  - I think that accessing them is less of
• 2:11
  - a problem, and today more the burdens around
• 2:13
  - synthesizing because we have so much data available
• 2:16
  - to us as clinicians
• 2:17
  - that we need to be able to go
• 2:19
  - through in order to identify. And of course,
• 2:21
  - disinformation burden is also a challenge because we
• 2:24
  - have different various levels of learners who are
• 2:26
  - working with the same amount of data. So
• 2:28
  - somebody who's like a thirty year medical student
• 2:30
  - is going to have a very different approach
• 2:32
  - to how they evaluate the data than somebody
• 2:34
  - who's been in practice for twenty years.
• 2:36
  - I think that this kind of era of
• 2:39
  - artificial intelligence really gives us an opportunity to
• 2:42
  - help close those gaps.
• 2:44
  - Helping us to abstract the key findings that
• 2:46
  - are available in all these different note types
• 2:48
  - and all these template images and present them
• 2:50
  - in workflow to the provider.
• 2:53
  - I think that we have an unintended consequence
• 2:54
  - though,
• 2:56
  - around medical and nursing education because so much
• 2:58
  - of our existing system is built upon
• 3:01
  - individuals, physicians, nurses who can see the data
• 3:04
  - and interpret it appropriately, kinda doing critical analysis
• 3:07
  - in order to take the right action. I
• 3:09
  - think that in an era where getting to
• 3:12
  - the right answer is no longer a question,
• 3:13
  - like a medical student can just basically type
• 3:15
  - into open evidence or up to date AI
• 3:18
  - or or Doximity GPT and ask it a
• 3:20
  - question and get a very good summary of
• 3:22
  - what the right course of action to do,
• 3:25
  - knowing
• 3:25
  - to question that answer and say, it may
• 3:27
  - be right, but is it right for this
• 3:29
  - patient? Is it right for this population?
• 3:31
  - And I think that coming up with the
• 3:32
  - right education approaches is gonna be critical for
• 3:34
  - the next decade.
• 3:36
  - Absolutely.
• 3:37
  - And patients are increasingly moving among the hospital,
• 3:39
  - out patient settings in the home.
• 3:41
  - So where does clinical contacts most often get
• 3:44
  - lost during these transitions, and what would it
• 3:46
  - take to give care teams a more continuous
• 3:48
  - view of the patient?
• 3:49
  - I think that one of the challenges is
• 3:51
  - that when you move across multiple settings, you
• 3:53
  - typically have different care providers providing that.
• 3:57
  - Some larger integrated,
• 3:59
  - delivery systems really have all of that in
• 4:01
  - house. They're using the same systems and the
• 4:03
  - same teams to collect the data. And it
• 4:05
  - really gives a much better perspective
• 4:07
  - on what the kinda like patient's trajectory is.
• 4:11
  - Unfortunately,
• 4:12
  - the reality of our payer system is that
• 4:14
  - most of our patients are not receiving consistent
• 4:16
  - care across all those different settings. And so
• 4:19
  - I think that
• 4:20
  - one of the challenges besides interoperability is not
• 4:23
  - only accessing the data, but also, again, going
• 4:25
  - to the Us question, understanding the context and
• 4:28
  - the knowledge behind it.
• 4:29
  - It's just important for us to be able
• 4:31
  - to understand
• 4:33
  - where the things are that are important.
• 4:35
  - And with so many AI use cases competing
• 4:37
  - for attention, how do you distinguish a compelling
• 4:40
  - pilot from something ready to become part of
• 4:42
  - everyday care? And what must be in place
• 4:44
  - to scale this properly?
• 4:46
  - Oh, jeez. That is a challenge that we're
• 4:48
  - dealing with on an everyday basis because we
• 4:50
  - have so many different tools, some homegrown, some
• 4:53
  - acquired, some coming with our EHR. And the
• 4:55
  - model that we have to figure out is
• 4:57
  - what's the right value equation that we need
• 4:59
  - to be able to calculate for every one
• 5:01
  - of these tools. How do we create the
• 5:03
  - systems that look at the output
• 5:05
  - and are able to kind of critically assess,
• 5:08
  - are they effective? Are they continuing to be
• 5:10
  - effective? And because none of these things are
• 5:12
  - free, looking at the cost to actually run
• 5:14
  - and maintain and support these models over time.
• 5:17
  - We
• 5:18
  - need to come up with a method because
• 5:20
  - AI is not going away. As we can
• 5:22
  - see from being on the exhibit floor today,
• 5:23
  - many solutions now are embedding AI as well
• 5:26
  - as EHRs directly.
• 5:28
  - And so
• 5:29
  - we are looking for good measures. Like, I'm
• 5:32
  - a Dona BDM quality services... Quality health systems
• 5:32
  - researcher... Health services researcher, excuse
• 5:35
  - systems researcher... Health services researcher. Excuse me. And
• 5:39
  - so I always think of things in terms
• 5:40
  - of, like, structure and process and outcomes measures.
• 5:43
  - And so being able to
• 5:44
  - have good measures
• 5:46
  - on all three angles
• 5:48
  - to assess the efficacy and sustainability of any
• 5:51
  - AI tool is gonna be critically important going
• 5:52
  - forward.
• 5:54
  - And as technology relationships move beyond one time
• 5:56
  - purchases, what should a partner remain accountable for
• 5:59
  - after go live to make sure that the
• 6:01
  - relationship genuinely is valuable over time?
• 6:04
  - That is going to be so critical to,
• 6:06
  - I think, any vendor client partnership going forward.
• 6:10
  - I think we are no longer in an
• 6:12
  - era that doctor Bart called earlier today in
• 6:15
  - his keynote,
• 6:16
  - panel discussion,
• 6:17
  - era of best of breed, really looking for
• 6:18
  - more of a platform model approach. And we
• 6:21
  - have to make sure that we are working
• 6:23
  - with our vendor
• 6:24
  - partners to create those platforms where we can
• 6:26
  - build
• 6:27
  - collaboratively.
• 6:29
  - And that's really gonna be the critical thing,
• 6:31
  - especially for larger systems in order to deploy
• 6:33
  - and use these tools at scale.
• 6:36
  - And finally, as care increasingly moves beyond hospital
• 6:39
  - walls, how are you using connected technologies and
• 6:42
  - AI to identify patients who need earlier follow-up,
• 6:44
  - support care coordination teams, and close gaps in
• 6:47
  - post discharge care? It's a combination of the
• 6:49
  - things I mentioned previously.
• 6:51
  - There's the information dissemination challenge that we are
• 6:54
  - able to do because of existing tools like
• 6:57
  - like the EHR, but also when we start
• 6:59
  - to use AI tools to start doing augmentation
• 7:02
  - such as chart summarization or highlighting key factors
• 7:05
  - for the individual, the physician, the nurse in
• 7:07
  - the workflow.
• 7:08
  - The other challenge that we have is around
• 7:11
  - making sure that we have these platforms,
• 7:14
  - so that people have a consistent approach across
• 7:16
  - the entire system.
• 7:17
  - And so those are the two approaches that
• 7:18
  - we're looking at right now at Michigan Medicine.
• 7:21
  - Wonderful. Well, thank you, doctor Fu, for your
• 7:23
  - time and your insights today, and I also
• 7:25
  - wanna thank our podcast sponsor, Philips. You can
• 7:27
  - tune in to more podcasts from Becker's Healthcare
• 7:29
  - by visiting our podcast page at beckershospitalreview.com.