Latest / Becker’s Healthcare Digital Health + Health IT / Turning Healthcare Data Into Actionable Clinical Insights with Dr. Tinu Tadese

Transcript

Turning Healthcare Data Into Actionable Clinical Insights with Dr. Tinu Tadese · 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
  - Hello, everyone. This is Jeremy with the Becker's
• 0:36
  - Healthcare Podcast.
• 0:37
  - Thank you so much for tuning in to
• 0:39
  - the Becker's Healthcare Podcast series recorded live at
• 0:41
  - the eleventh annual IT and revenue cycle conference,
• 0:44
  - the future of AI and digital health. I'm
• 0:45
  - joined today by doctor Tinu Tadesha,
• 0:48
  - system chief medical informatics officer at Boston Medical
• 0:51
  - Center.
• 0:52
  - Tinu, how are you doing today? I am
• 0:54
  - doing well, Jeremy. Thank you very much for
• 0:56
  - having me on this podcast. Can Can you
• 0:58
  - take a brief moment to introduce yourself and
• 1:00
  - tell a little bit about your, career journey
• 1:01
  - so far?
• 1:02
  - Thank you, Jeremy. My name again is Tinu,
• 1:05
  - and,
• 1:06
  - I have been,
• 1:07
  - in informatics for quite a bit now. I
• 1:10
  - am a pediatrician
• 1:11
  - by training.
• 1:13
  - And right now, I am vice president and
• 1:15
  - system chief medical information officer at Boston Medical
• 1:18
  - Center.
• 1:19
  - Over the... My career, I have been on
• 1:22
  - the vendor side of things being,
• 1:25
  - executive physician at Sona, which is now Oracle
• 1:28
  - Health.
• 1:29
  - And then I've then gone into roles such
• 1:31
  - as VP,
• 1:33
  - chief medical information officer
• 1:35
  - in three or four health systems.
• 1:38
  - My bent is really operational informatics,
• 1:41
  - how we make technology
• 1:43
  - work for us. Mhmm.
• 1:45
  - As clinicians,
• 1:47
  - I sit in that space between IT
• 1:50
  - and the clinical space and operations.
• 1:52
  - So that's what I do today. I'm excited
• 1:54
  - to be here.
• 1:56
  - Awesome. Well, health systems are receiving a growing
• 1:59
  - stream of patient data from bedside monitoring,
• 2:02
  - ambulatory devices, and care delivered at home.
• 2:06
  - Where is that data creating actionable insights for
• 2:08
  - clinicians,
• 2:09
  - and where does that still create more work
• 2:11
  - than insight?
• 2:12
  - That's a great question. We have data streaming
• 2:15
  - in into the electronic health record or different
• 2:18
  - applications in our health systems.
• 2:21
  - Right now, we have actionable insights in areas
• 2:24
  - such as,
• 2:25
  - continuous glucose monitoring.
• 2:27
  - That's an example where a patient
• 2:30
  - wears what we call the CGM
• 2:32
  - patches, and we start to find out what
• 2:35
  - their blood glucose level is like,
• 2:38
  - either at home, at night,
• 2:40
  - or after a meal. We're able to act
• 2:43
  - on that when that patient comes in because
• 2:45
  - we have the data in front of us.
• 2:47
  - Or take take,
• 2:49
  - bedside monitoring,
• 2:50
  - a patient in the ICU.
• 2:52
  - This is how we are able to track
• 2:54
  - their data. Are they doing well? Are they,
• 2:57
  - decompensating?
• 2:59
  - So right now, we have actionable data from
• 3:01
  - different data sources, and,
• 3:04
  - we're able to use them.
• 3:06
  - We're starting to find areas where we just
• 3:09
  - have data. Having data available does not mean
• 3:11
  - they're actionable. It just means more work Mhmm.
• 3:14
  - For the clinician.
• 3:16
  - So when clinicians are trying to reach a
• 3:17
  - diagnosis, they often must piece together information from
• 3:20
  - radiology, cardiology, pathology, urology, and other systems. Where
• 3:24
  - does that fragmentation create the most friction in
• 3:27
  - your organization,
• 3:28
  - and how do you bring together data sources
• 3:30
  - and address that fragmentation or close these gaps?
• 3:33
  - Yeah. That's a great question.
• 3:35
  - Again,
• 3:36
  - the... Right now, take thirty years ago or
• 3:39
  - maybe twenty years ago when almost all health
• 3:42
  - systems were
• 3:44
  - on,
• 3:45
  - paper.
• 3:46
  - You didn't have that problem. But what you
• 3:48
  - also did not have was ability to synthesize
• 3:51
  - data. It took you a long time as
• 3:53
  - a physician
• 3:57
  - patient.
• 3:58
  - Now we have data streaming in from everywhere.
• 4:01
  - We have imaging. We have cardiac data. We
• 4:04
  - have just regular wearables where you're monitoring
• 4:07
  - your,
• 4:08
  - your steps,
• 4:10
  - blood blood pressure,
• 4:13
  - monitors, or even your bathroom scale is speaking
• 4:16
  - to you. Now the fragmentation
• 4:19
  - is there, and then the question is, what
• 4:20
  - is the clinician supposed to do with this?
• 4:23
  - How do they walk with this to be
• 4:25
  - able to provide
• 4:27
  - safe patient care? The patient sitting in front
• 4:29
  - of you, what do I do with all
• 4:31
  - this data? So that fragmentation
• 4:33
  - is certainly there, and every day we have
• 4:36
  - more data being produced. I'll take for example,
• 4:39
  - one of the things we went through at
• 4:41
  - our health system.
• 4:42
  - We acquired two community hospitals, and we are
• 4:46
  - mainly an academic
• 4:47
  - health center. We had to,
• 4:50
  - onboard these community hospitals at a very compressed
• 4:52
  - timeline,
• 4:53
  - something that should have taken eighteen months, we
• 4:55
  - did in six months. The question then is
• 4:58
  - if a patient walks into
• 4:59
  - a ER day one after go live, all
• 5:03
  - this data
• 5:04
  - from three different hospitals,
• 5:06
  - what do we need and what can we
• 5:08
  - do?
• 5:09
  - Do we have enough data?
• 5:11
  - We had that challenge because data takes time
• 5:14
  - to ingest,
• 5:15
  - and so we have to determine
• 5:17
  - what is necessary by each service line. So
• 5:20
  - if I'm cardiology and a patient comes into
• 5:23
  - the ER, what must I do? So what
• 5:24
  - can I migrate
• 5:26
  - deliberately into the EHR
• 5:28
  - from the from the current EHR to our
• 5:30
  - own new?
• 5:32
  - Which is new to the community hospitals?
• 5:35
  - Imaging,
• 5:36
  - radiologist has seen that patient, is trying to
• 5:38
  - figure out
• 5:40
  - what was the prior image
• 5:42
  - oncology
• 5:43
  - cancer, what was the previous pathology test. All
• 5:46
  - of these are fragmented data, and we had
• 5:49
  - to deliberately, intentionally
• 5:51
  - determine if I were the patient walking into
• 5:54
  - the ER, what would my doctor need to
• 5:56
  - see me? So we had to deliberately pull
• 5:58
  - all of that together, what stays in archival
• 6:01
  - status, what's in an interim status, and what
• 6:04
  - must be in the EHR on day one.
• 6:06
  - So fragmentation is a big problem. Again, having,
• 6:10
  - data available
• 6:12
  - does not mean that the patient is safe
• 6:14
  - or that you can deliver safe care. You
• 6:17
  - actually literally
• 6:18
  - have to deliberately determine the patient journey and
• 6:21
  - what will be needed Mhmm. At that time.
• 6:24
  - Mhmm.
• 6:25
  - And so patients are, they increasingly move among
• 6:28
  - the hospital, outpatient settings, and home. Where does
• 6:31
  - clinical context most often get lost during those
• 6:34
  - transitions, and what would it take to give
• 6:36
  - care teams a more continuous view of the
• 6:38
  - patient?
• 6:39
  - That's a great question.
• 6:41
  - I refer to that project that we had,
• 6:44
  - at our health system where we had to
• 6:47
  - ingest data. Now think about this patient who
• 6:50
  - has been hospitalized
• 6:51
  - in one of our three hospitals.
• 6:53
  - This patient then goes back home, and the
• 6:56
  - admitting physician or discharging physician say see your
• 7:00
  - primary care physician at such and such a
• 7:02
  - date.
• 7:03
  - The challenge
• 7:04
  - is what's the context? Mhmm. Did that patient
• 7:07
  - actually follow-up? These are the things that are
• 7:09
  - lost in translation.
• 7:11
  - A test was ordered. Who actually follows up
• 7:15
  - to make sure that that test was read?
• 7:18
  - Oh, a medication that the patient was on
• 7:20
  - previously was stopped.
• 7:22
  - Well, is that patient supposed to start taking
• 7:25
  - it after,
• 7:27
  - a period of time? It's context that is
• 7:29
  - lost Mhmm. In transitions of care. Not necessarily
• 7:33
  - the data, but why do we have this
• 7:35
  - data? And that's what we have to pay
• 7:38
  - rigorous attention to. Otherwise, then we cannot deliver
• 7:41
  - safe patient care.
• 7:43
  - Patient are left holding
• 7:45
  - the bag, really. They... And they... The physician
• 7:48
  - is asking, why did your doctor stop this?
• 7:51
  - They don't know. And so this is why
• 7:53
  - that's... We need to pay attention to clinical
• 7:55
  - context and continuity,
• 7:57
  - when we're changing care settings and care teams.
• 8:00
  - Mhmm. Mhmm.
• 8:02
  - And with so many AI use cases
• 8:04
  - competing for attention, how do you distinguish a
• 8:07
  - compelling pilot from something ready to become part
• 8:10
  - of everyday care, and what must be in
• 8:12
  - place to scale it responsibly?
• 8:14
  - That's a great question. Slightly different from what
• 8:16
  - we have been, talking about. But, of course,
• 8:20
  - in the... In this,
• 8:22
  - time of AI,
• 8:24
  - we have every pilot you can think of
• 8:26
  - coming through... Well,
• 8:28
  - proposals for pilots coming through
• 8:30
  - through our doors and, AI committee
• 8:33
  - having to wait through. So
• 8:35
  - the question is,
• 8:37
  - was that pilot, the one we determined to
• 8:41
  - start, was it successful? So you have to
• 8:43
  - now
• 8:44
  - look back to what was the original need
• 8:47
  - or
• 8:49
  - problem you were trying to solve? Did it
• 8:51
  - solve it? What were the KPIs you were
• 8:53
  - looking? Was it a financial,
• 8:55
  - solution you were looking for? Was it clinical?
• 8:59
  - Was it operational?
• 9:00
  - If the pilot is able
• 9:02
  - to actually solve that, then we say, yes.
• 9:04
  - This pilot is successful. However, that does not
• 9:07
  - mean that it's scalable. Mhmm. Being... Having a
• 9:10
  - successful pilot does not mean that operations
• 9:14
  - are ready
• 9:16
  - for what it needs to take. So you...
• 9:18
  - Maybe you you piloted with 30 clinicians.
• 9:21
  - 30 physicians use this pilot. Now you wanna
• 9:24
  - scale to 2,000 physicians. Mhmm. Do we have
• 9:27
  - all the support staff in place? Do we
• 9:30
  - have operational leaders in place?
• 9:32
  - Do we have the KPIs we're looking for?
• 9:35
  - Who is gonna monitor? Who is going to
• 9:37
  - determine that this is still safe there? AI,
• 9:40
  - for example,
• 9:42
  - these are models that change on the server
• 9:44
  - side that you have no control over Mhmm.
• 9:46
  - In your health system. Who's going to continuously
• 9:49
  - monitor this? So when you think about all
• 9:52
  - of this,
• 9:53
  - a successful pilot does not mean it's ready
• 9:56
  - for scale. But if you can guarantee that
• 9:59
  - we have the operational readiness for it, then
• 10:02
  - it's time to scale. Okay.
• 10:04
  - And as technology,
• 10:06
  - relationships
• 10:07
  - move beyond one time purchases,
• 10:09
  - what should a partner remain accountable for after
• 10:12
  - go live to make the relationship genuinely,
• 10:15
  - valuable over time?
• 10:16
  - That's a great question. Remember, part of my
• 10:18
  - history or my journey in informatics was I
• 10:21
  - did work on the vendor side, which where,
• 10:24
  - what is now Oracle Health. And and so
• 10:27
  - I kinda,
• 10:28
  - have an idea of the things that you
• 10:30
  - really should bring to my table as a
• 10:32
  - system CMIO.
• 10:34
  - I wanna see you walk with me, stay
• 10:37
  - with me to make sure that the value
• 10:39
  - we determined we were able to actually actually
• 10:42
  - happens. Mhmm. Don't leave me in the lodge.
• 10:44
  - Don't tell me that the technology works. Go
• 10:47
  - figure it out. Mhmm. And on the flip
• 10:49
  - side of that, the vendor does not own
• 10:51
  - your operational readiness. The vendor does not own
• 10:54
  - your change management.
• 10:55
  - The vendor can advise and say to you,
• 10:58
  - in other... With our other customers, this is
• 11:01
  - what we've seen that was successful. So bring
• 11:03
  - me that intelligence, bring me that insight so
• 11:06
  - that I can continue to succeed. So it...
• 11:08
  - It's a partnership like you mentioned. They are
• 11:11
  - partners,
• 11:12
  - and we need to do our part. They
• 11:14
  - need to do their part, but they need
• 11:16
  - to make sure that what we said was
• 11:18
  - gonna happen actually happens. At least work with
• 11:20
  - us to make sure it happens. Right. Right.
• 11:22
  - Right. And as care increasingly moves beyond hospital
• 11:25
  - walls, how are you using connected technologies and
• 11:27
  - AI to identify patients who need earlier follow-up,
• 11:30
  - support care coordination teams, and close gaps in
• 11:33
  - post discharge care? That is great. If if
• 11:36
  - we... There's a lot of this...
• 11:39
  - What we talk about technology, but what we
• 11:41
  - really wanna talk about really is, do you
• 11:44
  - have the workflows in place? Mhmm. Because it's
• 11:46
  - very easy to build,
• 11:49
  - the
• 11:50
  - technology that would alert you that a patient
• 11:53
  - is deteriorating.
• 11:54
  - Mhmm. The question then is,
• 11:56
  - who
• 11:57
  - picks up that alert? Right. Who does
• 12:01
  - what needs to be done when that patient
• 12:03
  - is deteriorating?
• 12:04
  - Who is responsible?
• 12:06
  - And so, yes, we have technologies. For example,
• 12:09
  - we admission.
• 12:10
  - Do we have the risk,
• 12:12
  - algorithms in place so that we can detect
• 12:15
  - even before the patient
• 12:17
  - leaves the hospital, this patient is probably gonna
• 12:19
  - be readmitted? And if that's the case, what
• 12:22
  - is our follow-up to make sure the patient
• 12:24
  - is not readmitted?
• 12:25
  - Is it a social determinant of health issue,
• 12:28
  - or is it the fact that this patient
• 12:30
  - just is noncompliant?
• 12:31
  - Mhmm. Since we already know that technology alerts
• 12:35
  - us based on algorithms we have developed, What
• 12:37
  - then do we do with it? And so
• 12:39
  - it's not just technologies, both technology
• 12:42
  - and operational workflows and responsibility on the care
• 12:45
  - team. Right. Right.
• 12:47
  - Well, I wanna thank you so much again
• 12:49
  - for joining us today and sharing your insights.
• 12:52
  - We also wanna thank our podcast sponsor, Philips.
• 12:54
  - You can turn into more podcasts from Becker's
• 12:56
  - Healthcare by visiting our podcast page at beckershospitalreview.com.
• 13:00
  - Thank you so much, Jeremy. I'm glad to
• 13:01
  - have come on your podcast. Thank you so
• 13:03
  - much. It was an absolute pleasure.