Latest / Becker’s Healthcare Digital Health + Health IT / Using AI to Move Healthcare From Reactive to Proactive Care with Romila Aloysius

Transcript

Using AI to Move Healthcare From Reactive to Proactive Care with Romila Aloysius · 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,
• 0:25
  - it can benefit health systems, and most importantly,
• 0:28
  - it can benefit patients and their families.
• 0:33
  - Phillips.
• 0:34
  - Hi, everyone. This is Gracelyn Keller with Becker's
• 0:36
  - Healthcare, and we are recording live at the
• 0:39
  - eleventh annual IT and revenue cycle conference, the
• 0:41
  - future of AI and digital health. Thank you
• 0:43
  - so much for tuning in to the Becker's
• 0:45
  - Healthcare Podcast series.
• 0:47
  - Joining me for today's discussion is Romila Alosius,
• 0:50
  - who is the executive director of Heart, Lung,
• 0:52
  - and Vascular Institute at AdventHealth.
• 0:55
  - Rameela, thanks for being here. And I'd like
• 0:57
  - to just jump into our conversation, if that's
• 0:59
  - alright with you. Thank you for having me.
• 1:01
  - Let's get started.
• 1:03
  - So health systems are receiving a growing stream
• 1:05
  - of patient data from bedside monitoring, ambulatory devices,
• 1:09
  - and care delivered at home. Where is that
• 1:11
  - data creating actionable insights for clinicians, and where
• 1:14
  - does it still create more work than insight?
• 1:17
  - Yeah. I think the strongest business case for,
• 1:20
  - data is really when data is affecting a
• 1:23
  - decision that is made.
• 1:25
  - That could be anything from identifying deterioration
• 1:28
  - earlier,
• 1:29
  - prioritizing patients by risk, or determining the right
• 1:32
  - next step in the, care journey.
• 1:35
  - At the bedside,
• 1:36
  - continuous,
• 1:38
  - vital sign reading can be useful.
• 1:40
  - In RPM models, continuous monitoring of heart failure
• 1:44
  - or hypertension
• 1:45
  - can be useful.
• 1:46
  - But the problem is health systems get a
• 1:48
  - lot of data, and there's usually data fatigue.
• 1:52
  - And and that's where technology comes in. Technology
• 1:55
  - comes in to filter the right datasets.
• 1:57
  - Everything that is built has to be built
• 1:59
  - in partnership with clinical teams to identify
• 2:02
  - what is the right metric or the,
• 2:05
  - the data point that would influence the decision.
• 2:08
  - And that's why I always tie back data
• 2:10
  - to,
• 2:11
  - the decision that needs to be made, and
• 2:13
  - that's what improves the overall clinical workflows.
• 2:17
  - And when clinicians are trying to reach a
• 2:19
  - diagnosis, they often must piece together information from
• 2:22
  - radiology, cardiology, pathology,
• 2:24
  - and many other systems.
• 2:26
  - So where does that fragmentation create the most
• 2:28
  - friction in your organization, and how do you
• 2:30
  - bring together data sources and address this fragmentation
• 2:33
  - or close the gaps?
• 2:35
  - Yeah. So fragmentation
• 2:36
  - is is a common problem. We often see
• 2:38
  - this in read radiology or imaging where we
• 2:41
  - we need data scans or, the notes really
• 2:45
  - in
• 2:45
  - from the radiology scans.
• 2:47
  - It's it lies in where the data is
• 2:50
  - stored, and translating that sometimes into epic discrete
• 2:53
  - fields is something that, you know, most health
• 2:55
  - systems are still dealing with.
• 2:57
  - First, I think, is recognizing
• 2:59
  - where that fragmentation is happening. So as you
• 3:01
  - think about
• 3:03
  - air delivery,
• 3:04
  - I don't think of data as a separately
• 3:06
  - from the overall workflow and the overall care
• 3:09
  - model. And I think that's where we need
• 3:11
  - to recognize that
• 3:13
  - designing the care model must be designed in
• 3:15
  - a way that the data is connected from
• 3:17
  - one point to another, and there there is
• 3:20
  - the next step related to the data that
• 3:22
  - is picked up by the next person and
• 3:24
  - there's accountability there. So I think that's where,
• 3:27
  - fragmentation
• 3:28
  - can be solved for is when we are
• 3:30
  - looking at data in the context of care
• 3:33
  - models and in the context of care design.
• 3:36
  - And patients increasingly move among the hospital, outpatient
• 3:39
  - settings, and the home.
• 3:41
  - So where does clinical
• 3:43
  - context most often get lost during these transitions,
• 3:45
  - and what would it take to give care
• 3:47
  - teams a more continuous view of the patient?
• 3:50
  - Yeah. I think that first of all, health
• 3:52
  - systems have come a long way in closing
• 3:54
  - transitions closing care gaps.
• 3:56
  - Clinical context is usually lost when this transition
• 3:59
  - happen happens.
• 4:01
  - Some of the, when responsibility changes hands.
• 4:05
  - I think what's what's helped in this arena
• 4:08
  - is when there is human intervention.
• 4:10
  - The need for navigation teams and care coordinators
• 4:14
  - continues to be,
• 4:16
  - relevant in this context despite having sophisticated data
• 4:20
  - systems
• 4:21
  - that,
• 4:22
  - human intervention
• 4:23
  - continues to play a big role
• 4:25
  - in,
• 4:26
  - in the way that we use data and
• 4:27
  - pass it from one person to another. So
• 4:29
  - I would say that, technology can certainly help
• 4:32
  - in this case. Data organization can help in
• 4:34
  - this case, But I don't feel like,
• 4:37
  - it prevents the need of human intervention.
• 4:41
  - Absolutely.
• 4:42
  - And and with so many AI use cases
• 4:44
  - competing for attention, how do you distinguish a
• 4:47
  - compelling pilot from something ready to become part
• 4:49
  - of everyday care, and what must be in
• 4:51
  - place to scale that properly?
• 4:54
  - Yeah. A compelling pilot, it becomes it it's
• 4:57
  - begins with a,
• 4:59
  - relevant operational problem.
• 5:01
  - So the AI capability
• 5:03
  - should be used to solve the operational problem.
• 5:06
  - It isn't so much to
• 5:08
  - create an AI. It it create a use
• 5:10
  - case for an AI capability. It is to
• 5:12
  - identify what the operational problem is and then
• 5:15
  - identify the AI capability that can solve that
• 5:18
  - problem.
• 5:19
  - So in order to see early results, we
• 5:21
  - really have to be able to take the
• 5:23
  - tool that can,
• 5:25
  - improve the decision making, reduce administrative burdens, improve
• 5:29
  - our capacity, improve quality,
• 5:31
  - and,
• 5:32
  - produce meaningful,
• 5:33
  - difference in cost
• 5:35
  - and, risk.
• 5:37
  - Before scaling, we've gotta ask the question, is
• 5:39
  - this repeatable? Is it gonna
• 5:41
  - deliver consistent results?
• 5:43
  - And,
• 5:44
  - it has to be embedded within the workflows,
• 5:46
  - which I've consistently mentioned through this dialogue.
• 5:49
  - Human accountability is a is still a must,
• 5:52
  - as I've mentioned before,
• 5:53
  - and,
• 5:55
  - responsible
• 5:55
  - scaling requires proper governance structures and data quality
• 6:00
  - teams as we think about scaling.
• 6:02
  - The last thing I'll say is performance must
• 6:04
  - be, meet measured and,
• 6:07
  - scaled,
• 6:08
  - and the system should be scaled only when
• 6:11
  - through a stage gated process, so only when
• 6:14
  - results are achieved.
• 6:15
  - And as technology relationships move beyond onetime purchases,
• 6:19
  - what should a partner remain accountable for after
• 6:22
  - go live to make the relationship genuinely valuable
• 6:24
  - over time?
• 6:27
  - Yeah. So I think this starts off before
• 6:29
  - the contract happens. So I think in in
• 6:31
  - the early contract stage, it's really defining what
• 6:34
  - those measures of success are gonna be for
• 6:37
  - a workflow to be successful. The partner should
• 6:39
  - be engaged post,
• 6:41
  - post postoperationalization.
• 6:44
  - Nothing ever happens as planned in health care
• 6:47
  - because it's so complex.
• 6:49
  - And so,
• 6:50
  - it is,
• 6:51
  - necessary for the data partner
• 6:54
  - to be involved for continuous optimization of workflows
• 6:57
  - post,
• 6:58
  - post launch, post implementation.
• 7:01
  - Reporting
• 7:02
  - a clear definition of what success looks like
• 7:04
  - for both parties has to be agreed upon
• 7:07
  - prior
• 7:08
  - to to the contracting phase, and reporting has
• 7:10
  - to be done on a consistent basis.
• 7:13
  - And,
• 7:14
  - the relationship, as I mentioned before, should have
• 7:16
  - clear governance and the contract should be clear
• 7:18
  - on renewals, expansion,
• 7:21
  - and, exit clauses.
• 7:23
  - So,
• 7:24
  - the partner needs to be engaged and aligned
• 7:27
  - through several,
• 7:28
  - I would say, even several years of the
• 7:30
  - engagement
• 7:31
  - just because you're continuously optimizing, you're continuously expanding.
• 7:35
  - Priorities change within health care, and, therefore,
• 7:39
  - any sort of relationship should be a long
• 7:41
  - long term and ongoing one.
• 7:43
  - And finally, as care increasingly moves beyond hospital
• 7:46
  - walls, how are you using connected technologies and
• 7:49
  - AI to identify patients who need earlier follow-up,
• 7:52
  - support care coordination teams, and close gaps in
• 7:54
  - post discharge care?
• 7:57
  - Yeah. So,
• 7:59
  - with AI, we're using it in several places
• 8:01
  - within our teams. We use it for early
• 8:04
  - identification.
• 8:05
  - So for example, in lung cancer,
• 8:07
  - AI is used to identify incidental nodules on
• 8:11
  - my team. And, the idea is if we
• 8:13
  - can find the nodule early on, we can
• 8:15
  - prevent an acute or malignant cancer. The AI
• 8:19
  - screens every single person person who interacts with
• 8:21
  - our system and identifies those who have a,
• 8:25
  - nodules that could potentially turn into something cancerous.
• 8:28
  - So in that way, we're using AI to
• 8:30
  - shift the the point of intervention and therefore
• 8:32
  - catching diagnosis early on, and this
• 8:36
  - overall increases our ability to better manage patients.
• 8:39
  - In other cases, we also use it in
• 8:41
  - in, I think, in health care in general.
• 8:43
  - So far, it's really being used to,
• 8:46
  - eliminate administrative burden. So it's used to translate
• 8:49
  - notes into
• 8:50
  - more clarified into discrete data elements.
• 8:54
  - Some systems have started to use it in
• 8:56
  - support of clinical decisions.
• 8:58
  - And this is obviously done with, significant human
• 9:01
  - intervention and monitoring.
• 9:03
  - But I think that is progressive is when
• 9:05
  - the when AI can be used to support
• 9:07
  - clinical decisions.
• 9:09
  - Absolutely. Well, Mila, thanks so much for your
• 9:12
  - time and insights today. And we also want
• 9:14
  - to thank our podcast sponsor, Philips. You can
• 9:16
  - tune in to more podcasts from Becker's Healthcare
• 9:18
  - by visiting our podcast page at beckershospitalreview.com.