Latest / Becker’s Healthcare Podcast / Beyond Point Solutions: Building the Clinical AI Layer

Transcript

Beyond Point Solutions: Building the Clinical AI Layer · Becker’s Healthcare Podcast

• 0:00
  - Hello, and welcome to the Becker's Healthcare podcast.
• 0:02
  - Today, we're gonna talk about
• 0:04
  - BeyondPoint Solutions, building the clinical AI layer. Joining
• 0:08
  - me for today's discussion is doctor Ed Lee.
• 0:11
  - Doctor Ed Lee is a practicing internal medicine
• 0:14
  - physician,
• 0:15
  - physician executive,
• 0:16
  - and nationally recognized leader in clinical informatics and
• 0:19
  - ambient AI. He currently serves as chief medical
• 0:22
  - officer at Nabla where he drives strategy and
• 0:25
  - clinical innovation for an AI powered ambient documentation
• 0:28
  - platform
• 0:29
  - that enhances clinician well-being and care delivery.
• 0:32
  - Previously, doctor Lee was chief information officer and
• 0:34
  - associate executive director at Kaiser Permanente
• 0:38
  - where he led groundbreaking digital transformation efforts across
• 0:41
  - one of the nation's largest integrated health care
• 0:44
  - systems.
• 0:44
  - His leadership advanced AI governance,
• 0:47
  - expanded access to care through telehealth,
• 0:50
  - and improved clinician experience through smarter EHR tools
• 0:54
  - and digital infrastructure.
• 0:56
  - In parallel with his executive roles, Doctor. Lee
• 0:59
  - recently served as Chair of Clinical Education
• 1:02
  - and Director of Clinical Informatics at California North
• 1:05
  - State University College of Medicine, where he oversaw
• 1:08
  - a technology enabled curriculum and clinical education strategy.
• 1:12
  - Doctor Lee, thank you so much for joining
• 1:14
  - me today. Thanks for having me. Perfect. Well,
• 1:16
  - to get us started out,
• 1:18
  - Nabla has undergone significant evolution over the past
• 1:21
  - year from leadership changes to an expanded product
• 1:25
  - portfolio.
• 1:26
  - How would you describe where the company stands
• 1:28
  - today, and what's been the biggest catalyst behind
• 1:30
  - that evolution?
• 1:32
  - Well, I think it's a really exciting time
• 1:34
  - for Nabla. We started a few years ago
• 1:36
  - by solving a very specific
• 1:38
  - and very painful problem for clinicians,
• 1:41
  - which is documentation.
• 1:43
  - And we've been fortunate to see AmbientDi grow
• 1:45
  - into something that's becoming increasingly
• 1:48
  - important part of standard clinical workflow.
• 1:51
  - But here's the thing. Once you're embedded in
• 1:53
  - the clinical workflow,
• 1:55
  - you can see opportunities beyond
• 1:57
  - generating a note. So documentation connects to coding,
• 2:01
  - orders,
• 2:02
  - clinical context,
• 2:04
  - decision support, quality, and and ultimately the entire
• 2:07
  - care journey.
• 2:08
  - The biggest reason for the evolution is undoubtedly
• 2:11
  - our clinical partners, though.
• 2:13
  - As we've scaled across health systems and supported
• 2:15
  - many clinicians and specialties,
• 2:18
  - they've pushed us to think more broadly about
• 2:21
  - what they need from AI.
• 2:23
  - It's not a collection of disconnected tools, but
• 2:26
  - more an integrated clinical AI layer that can
• 2:29
  - support clinicians throughout the work.
• 2:32
  - And, you know, the the leadership change reflects
• 2:34
  - that,
• 2:35
  - as we've brought in a new CEO,
• 2:38
  - Brian Manning.
• 2:39
  - He's an experienced health care technology executive who
• 2:42
  - has an impressive track record for elevating companies
• 2:45
  - to the next level, and I really expect
• 2:47
  - them to do the same with Nabla.
• 2:50
  - Got it. Thank you so much. And now
• 2:51
  - you talked about Nabla evolving into a clinical
• 2:54
  - AI layer rather than a collection of point
• 2:57
  - solutions. What does that vision look like in
• 2:59
  - practice, and why do you believe health systems
• 3:01
  - are increasingly looking for platform approaches instead of
• 3:04
  - standalone AI tools?
• 3:06
  - Well, the the idea is that AI can
• 3:08
  - sit across the clinical workflow,
• 3:11
  - understand the context of the patient and the
• 3:13
  - encounter,
• 3:14
  - and then help power a variety of experiences.
• 3:17
  - So things like documentation,
• 3:19
  - dictation,
• 3:20
  - coding,
• 3:21
  - clinical summaries, and and eventually more advanced forms
• 3:24
  - of clinical assistance.
• 3:26
  - Today,
• 3:27
  - health systems might have one AI vendor for
• 3:30
  - documentation,
• 3:31
  - another for coding,
• 3:32
  - another for patient summaries, and and another for
• 3:34
  - decision support.
• 3:36
  - And and so
• 3:37
  - each may work well independently,
• 3:40
  - but over time, that creates fragmentation.
• 3:42
  - And that's
• 3:43
  - the case not only for clinicians, but also
• 3:45
  - for IT teams,
• 3:46
  - for governments, and for underlying technology infrastructure.
• 3:50
  - The alternative
• 3:52
  - is that health systems are increasingly working towards
• 3:55
  - creating an AI architecture
• 3:57
  - that can scale across the entire enterprise.
• 4:00
  - So that takes a lot of the overhead
• 4:02
  - away from the teams that so so they
• 4:04
  - can focus their efforts on really providing the
• 4:07
  - best care possible to their patients.
• 4:09
  - Absolutely. That makes a lot of sense. And
• 4:11
  - now Nabla has expanded into
• 4:14
  - medical grade dictation and coding support. How do
• 4:17
  - these capabilities work together within a clinician's workflow,
• 4:20
  - and what operational
• 4:21
  - advantages
• 4:22
  - does that integrated approach create for health systems?
• 4:25
  - Yeah. So clinicians don't work in in one
• 4:28
  - mode all day.
• 4:29
  - Sometimes ambient documentation is ideal.
• 4:32
  - You have a conversation with the patient, and
• 4:34
  - the AI can help generate the documentation in
• 4:36
  - the background.
• 4:37
  - But there are also times when a clinician
• 4:39
  - simply wants to dictate something word for word,
• 4:42
  - like a procedure note, an addendum,
• 4:45
  - a patient message.
• 4:47
  - And
• 4:47
  - we think those experiences should coexist
• 4:50
  - rather than forcing clinicians into one workflow.
• 4:53
  - Coding
• 4:54
  - is a natural extension of that because
• 4:56
  - it comes from the same underlying clinical encounter.
• 5:00
  - And if the system already understands what's happening
• 5:02
  - during the visit and what was documented,
• 5:05
  - it can help identify appropriate coding
• 5:08
  - and documentation opportunities.
• 5:10
  - So for the health system,
• 5:12
  - you know, that integration integration is is important
• 5:15
  - because instead of deploying separate technologies with separate
• 5:18
  - interfaces,
• 5:19
  - integrations, contracts,
• 5:21
  - support models, you know, you're all building on
• 5:24
  - the same clinical context and infrastructure.
• 5:26
  - And then for clinicians,
• 5:28
  - the goal is pretty simple. I mean, it's
• 5:30
  - it's about fewer clicks, fewer systems,
• 5:33
  - and overall less cognitive overhead.
• 5:36
  - Got it. Got it. And now looking ahead
• 5:37
  - three to five years, where do you see
• 5:39
  - clinical AI
• 5:41
  - evolving?
• 5:41
  - You've spoken about emerging technologies like world models
• 5:45
  - and your partnership with AMI Labs. How do
• 5:47
  - those efforts fit into your long term vision?
• 5:51
  - Well, I I think it's still very early.
• 5:54
  - The early generation of clinical AI was
• 5:57
  - largely based on reducing administrative work,
• 6:01
  - and it's been great. It's created efficiency,
• 6:03
  - time savings, and reduction and burnout.
• 6:06
  - You know? But over the next three to
• 6:08
  - five years, I think we'll move from
• 6:10
  - AI that primarily
• 6:12
  - documents what happened
• 6:14
  - to AI that increasingly understands
• 6:16
  - what's actually happening.
• 6:19
  - Clinical care is longitudinal.
• 6:21
  - It involves understanding a patient's history,
• 6:23
  - which change over time,
• 6:25
  - the clinician's intent, and the consequences of different
• 6:28
  - decisions.
• 6:30
  - Today's
• 6:30
  - language models, they're incredibly powerful,
• 6:34
  - But there are still limitations in areas like
• 6:36
  - reasoning about the physical world,
• 6:39
  - causality, temporal relationships,
• 6:41
  - and complex sequences of events.
• 6:44
  - And and I think that's where the work
• 6:46
  - around war models is so interesting to us
• 6:49
  - and why our relationship with AMU Labs
• 6:52
  - is really so exciting.
• 6:53
  - You know, we're thinking about the next generation
• 6:55
  - of clinical intelligence
• 6:57
  - and
• 6:58
  - where those systems
• 7:00
  - will enrich the model of the patient, the
• 7:03
  - clinical environment,
• 7:05
  - and how things evolve over time.
• 7:08
  - There's still a bunch of research that needs
• 7:10
  - to be done, and health care appropriately
• 7:13
  - requires a high bar for evidence and safety.
• 7:17
  - But long term,
• 7:19
  - the opportunity is that AI moves from primarily
• 7:22
  - being an administrative assistant towards becoming a a
• 7:25
  - much more capable
• 7:26
  - clinical partner.
• 7:28
  - That's really helpful context and leads me to
• 7:30
  - my next question.
• 7:32
  - As CIOs, CMIOs, and chief AI officers make
• 7:35
  - strategic investments over the next several years, what
• 7:39
  - capabilities or architectural decisions should they prioritize today
• 7:42
  - to build a foundation for what's coming next?
• 7:45
  - Well, I I think the biggest mistake organizations
• 7:48
  - can make right now is building their AI
• 7:51
  - strategy on one tool at a time.
• 7:54
  - I mean, the technology is changing so quickly
• 7:56
  - that the c suite really needs to understand
• 7:59
  - architecture and capabilities that will last,
• 8:02
  - not just
• 8:03
  - individual use cases that are exciting and what
• 8:06
  - they're solving for today.
• 8:09
  - So a few things come to mind. You
• 8:11
  - know, the first is to make sure you
• 8:13
  - anchor your AI investments to
• 8:15
  - enterprise priorities, so things like clinician experience,
• 8:20
  - patient access, quality,
• 8:22
  - and revenue integrity.
• 8:24
  - You know? And and also to make sure
• 8:25
  - you're designing around clinical workflows.
• 8:28
  - Clinicians shouldn't have to navigate,
• 8:30
  - a growing collection of disconnected AI tools.
• 8:34
  - The goal should be to bring AI into
• 8:36
  - workflows that
• 8:38
  - they're already using,
• 8:40
  - while supporting the differences across
• 8:42
  - various specialties and care settings.
• 8:44
  - And then, you know, there's there's governments and
• 8:47
  - making sure you're building it in a way
• 8:49
  - that it can scale.
• 8:51
  - You need a repeatable way to evaluate safety,
• 8:54
  - performance,
• 8:55
  - transparency,
• 8:56
  - and and value,
• 8:58
  - and to have ongoing monitoring
• 9:00
  - rather than creating a new governance process for
• 9:02
  - every AI application.
• 9:04
  - And
• 9:05
  - then finally, I I think about
• 9:08
  - platform architecture.
• 9:09
  - Ask whether the technology you're investing in today
• 9:12
  - can support additional capabilities
• 9:15
  - for tomorrow, something that can scale across the
• 9:17
  - enterprise and and simplify rather than add complexity
• 9:20
  - to your your technology environment.
• 9:23
  - I love that. And it really does make
• 9:25
  - a lot of sense to have all of
• 9:26
  - it, you know, kind of integrated across the
• 9:28
  - entire system rather than fragmented. I love that.
• 9:31
  - Well, Well, Doctor. Li, I wanna thank you
• 9:32
  - for your time today. But before I let
• 9:33
  - you go, is there anything else that listeners
• 9:35
  - should know?
• 9:36
  - Well, I mean, I I think the biggest
• 9:38
  - takeaway is that we're moving into a a
• 9:40
  - different phase of clinical AI. You know, the
• 9:42
  - first phase was about proving that AI could
• 9:44
  - solve meaningful problems for commissions,
• 9:47
  - particularly around documentation.
• 9:48
  - But the next phase is about how we
• 9:50
  - build that foundation
• 9:52
  - thoughtfully
• 9:53
  - across workflows, across specialties, across the enterprise. And
• 9:56
  - and for health systems, that really means thinking
• 9:58
  - beyond individual tools,
• 10:00
  - asking how each investment fits into the broader
• 10:03
  - AI strategy.
• 10:04
  - And for all of us building these technologies,
• 10:06
  - the standard has to remain the same. Does
• 10:09
  - it make clinicians work easier? Does it improve
• 10:12
  - the way care is delivered?
• 10:14
  - And does it ultimately create value for patients
• 10:17
  - and health systems?
• 10:18
  - Well, that's a great takeaway to end on.
• 10:20
  - Doctor. Lee, I want to thank you once
• 10:22
  - again for your time and insights today. And
• 10:24
  - I also want to thank our podcast sponsor,
• 10:26
  - Nava.
• 10:26
  - You can tune into more podcasts from Becker's
• 10:28
  - Healthcare by visiting our podcast page at beckershospitalreview.com.
• 10:32
  - Thank you. Thank you.