Latest / Becker’s Healthcare Podcast / Building AI Solutions That Transform Cardiovascular Care with Dr. Faraz Ahmad

Transcript

Building AI Solutions That Transform Cardiovascular Care with Dr. Faraz Ahmad · Becker’s Healthcare Podcast

• 0:02
  - There's a version of AI in health care
• 0:04
  - that lives mostly in press releases,
• 0:06
  - pilot programs, partnership announcements,
• 0:09
  - theoretical
• 0:10
  - use cases.
• 0:11
  - And then there's the version that lives in
• 0:13
  - the work, the CIO trying to get a
• 0:15
  - model into production without breaking a dozen workflows,
• 0:19
  - the revenue cycle leader weighing automation against accuracy,
• 0:22
  - the CISO who knows that every new integration
• 0:25
  - is a new attack surface.
• 0:28
  - Becker's eleventh annual IT and revenue cycle conference
• 0:31
  - is built for the people doing that work.
• 0:33
  - More than 2,500
• 0:35
  - executive level attendees,
• 0:37
  - 600 plus speakers,
• 0:39
  - and four days of sessions designed around the
• 0:41
  - real operational questions behind
• 0:44
  - AI, cybersecurity,
• 0:45
  - revenue cycle performance, and digital health.
• 0:48
  - Join us September fourteenth through the seventeenth.
• 0:51
  - For the agenda and event details,
• 0:53
  - visit beckershospitalreview.com
• 0:55
  - and click on the events tab in the
• 0:57
  - upper right.
• 0:59
  - Welcome to the Becker's Healthcare Cardiology and Heart
• 1:02
  - Surgery podcast. I'm Mariah Taylor, assistant editor, and
• 1:05
  - I'm thrilled to be speaking with Faraz Ahmad,
• 1:08
  - medical director of AI strategy for Bloom Cardiovascular
• 1:11
  - Institute Center of AI and Northwestern Medicine's Ventures
• 1:14
  - and Innovation.
• 1:15
  - Thank you so much for joining me today.
• 1:17
  - I'm really excited to learn more about your
• 1:19
  - AI center and the strategy.
• 1:21
  - But before we dive into that, I'd love
• 1:23
  - for you to introduce yourself and tell us
• 1:25
  - a little bit about your background.
• 1:26
  - Thank you so much for having me. I'm
• 1:28
  - really excited to be here.
• 1:29
  - So I'm a heart cardiologist here at Northwestern.
• 1:31
  - I came here in 2018 for my fellowship,
• 1:34
  - my cardiology fellowship here, my informatics training, as
• 1:37
  - well as my advanced heart failure training. So
• 1:38
  - currently, I spend part of my day seeing
• 1:40
  - patients with heart failure across the spectrum of
• 1:42
  - heart failure.
• 1:43
  - I also run a research lab that's they're
• 1:45
  - really dedicated to developing and implementing AI tools
• 1:47
  - in our health system. And then I also
• 1:49
  - in our center for AI and also in
• 1:51
  - our and with our NM Ventures innovation group,
• 1:54
  - help think through our AI strategy.
• 1:56
  - Where should we build? Where should we buy?
• 1:58
  - Where should we partner? And where are the
• 1:59
  - tools that are really gonna help us transform
• 2:01
  - medicine
• 2:02
  - over the next few years?
• 2:04
  - Excellent.
• 2:05
  - Now Northwestern is a little bit unique because
• 2:08
  - it's one of the only centers in The
• 2:09
  - US that has a center dedicated to AI.
• 2:12
  - So can you tell us a little bit
• 2:13
  - more about how that center started and what
• 2:15
  - its focus is? Yeah. I'd love to. So
• 2:18
  - the center started in 2018. That was actually
• 2:20
  - the year I joined faculty here at Northwestern.
• 2:22
  - It was found as part of a really
• 2:24
  - generous flow and fabric gift by mister Neil
• 2:25
  - Blum. At the time, the executive director for
• 2:28
  - our institute, doctor Pat McCarthy,
• 2:31
  - really saw the future. You know, it was
• 2:32
  - pre chat CPT and realized that it's really
• 2:34
  - important to have a center in our health
• 2:36
  - system, in our Crested Institute really dedicated to
• 2:38
  - AI. And this is where we serve as
• 2:40
  - the hub for both our AI research and
• 2:42
  - our implementation efforts. And really at the core
• 2:44
  - of our center has also been education, and
• 2:46
  - we started a fellowship at the same time.
• 2:48
  - And so in terms of the areas we've
• 2:49
  - really been focused on, so we've really been
• 2:51
  - focused on thinking about how can we use
• 2:52
  - AI to empower care teams and patients, how
• 2:55
  - can we create really meaningful clinical and operation
• 2:57
  - impact. We really focus a lot on thinking
• 2:59
  - about our team and who do we need
• 3:01
  - at the table to really drive innovation in
• 3:03
  - the space. So that includes our clinical experts,
• 3:05
  - our AI experts, our operations experts, engineering, IT.
• 3:08
  - Really, everyone we need to really make impact
• 3:10
  - in the space.
• 3:12
  - We've also been recently we've launched what we
• 3:14
  - call a clinical implementation core in our South
• 3:17
  - Region. It's called AI click. And that's, again,
• 3:19
  - another emphasis in the space of implementation where
• 3:21
  - we see ourselves as continuing to want to
• 3:23
  - lead in.
• 3:24
  - And then lastly, we've also had a number
• 3:26
  - of projects in the space, that that leverage
• 3:29
  - various AI diagnostics. We built that into a
• 3:31
  - larger program that we call Predict, which I'll
• 3:33
  - talk a little bit more later on on.
• 3:35
  - Perfect.
• 3:36
  - Now I also hear that Northwestern launched an
• 3:39
  - AI fellowship for clinical trainees. Can you tell
• 3:41
  - me a little bit more about how that
• 3:42
  - works and what the results have been? Yeah.
• 3:44
  - This is, well, I think, one of the
• 3:45
  - most exciting aspects of the center when we
• 3:47
  - started back in 2018.
• 3:49
  - And so, really, the vision at that time
• 3:50
  - from leadership was to really make an impact
• 3:52
  - in the space, we need people who understand
• 3:54
  - both clinical medicine and engineering.
• 3:56
  - So fellows come to us. They spend entire
• 3:58
  - year with us in our center, and they
• 4:00
  - get actually a master's in engineering,
• 4:03
  - at the Northwestern University McCormick School of Engineering.
• 4:06
  - They They learn how to code. They learn
• 4:07
  - how to use LLMs. They learn the methods
• 4:09
  - of deep learning,
• 4:10
  - best practices for implementation evaluation.
• 4:12
  - They also work with us on research projects
• 4:14
  - with a dedicated mentor, and we also have
• 4:16
  - supplementary lectures as well.
• 4:18
  - And And it's been a pretty high impact
• 4:19
  - program for us. You know, we've had 15
• 4:21
  - fellows since 2018,
• 4:23
  - and these aren't just cardiology fellows. You know,
• 4:24
  - these are fellows from cardiac surgery, nephrology, other
• 4:27
  - specialties. We even had a medical student.
• 4:31
  - And, you know,
• 4:32
  - and through this and the fellowships had a
• 4:34
  - pretty strong publication record with dozens of publications.
• 4:37
  - And I think probably one of the most
• 4:39
  - exciting parts for me personally,
• 4:40
  - as we know, we've hired a bunch of
• 4:42
  - them. You know, six six in total, four
• 4:43
  - in cardiology across the across the health system.
• 4:46
  - This has really helped us build this community
• 4:47
  - of AI minded clinicians who's really helping us
• 4:50
  - now drive implementation across our system as part
• 4:52
  - of our AI click core lab.
• 4:55
  - Excellent.
• 4:56
  - And earlier, you mentioned the predict program. So
• 4:59
  - I'd love to hear a little bit more
• 5:01
  - about that and how it's helping turn helping
• 5:03
  - your system turn data into actionable decision making.
• 5:07
  - Yeah. So PREDICT has a pretty long acronym.
• 5:09
  - It stands for proactive risk evaluation detection using
• 5:12
  - an AI in and cardiovascular testing program.
• 5:15
  - And so I'd say risk modeling and AI
• 5:18
  - diagnostics is one of the more mature areas
• 5:20
  - of AI and cardiovascular medicine.
• 5:22
  - Our experience with this started in 2020 where
• 5:24
  - we launched an internally developed model to find
• 5:27
  - patients with advanced heart failure across our health
• 5:29
  - system. And we re engineered our workflow to
• 5:31
  - bring them into specialty to new access heart
• 5:33
  - failure clinics across the system
• 5:35
  - to make sure they're on the right medications,
• 5:37
  - see if they're candidates for therapies like transplant
• 5:39
  - or VAD,
• 5:41
  - and discuss their goals of care. And so
• 5:43
  - that was a pretty successful program for us.
• 5:44
  - You know, over our twenty month pilot, we
• 5:46
  - brought over 70 patients, led to us several
• 5:49
  - evaluations for transplant and VAD advanced therapies,
• 5:52
  - small number of surgeries.
• 5:53
  - But since then, the program has really grown
• 5:55
  - as this field of AI diagnostics and risk
• 5:57
  - prediction has matured. So now we have over
• 5:59
  - 12 algorithms in various stages of research and
• 6:02
  - development
• 6:03
  - in place. Six are live, actually.
• 6:05
  - And this is really a mix of commercial
• 6:08
  - models,
• 6:09
  - internal models, and research models too with our
• 6:11
  - some external academic partners.
• 6:14
  - And the approach we really take with this
• 6:15
  - program is that we, you know, we wanna
• 6:16
  - first figure out the problem we wanna solve,
• 6:19
  - and then we wanna figure out is it
• 6:20
  - something we wanna build ourselves, something we wanna
• 6:21
  - buy our partner.
• 6:23
  - But no matter what we end up doing,
• 6:24
  - at the core of what we do is
• 6:25
  - really clinical reengineer clinical workflow engine reengineering. That's
• 6:28
  - really sort of the secret sauce of all
• 6:30
  - these implementations.
• 6:31
  - And then really also making sure we have
• 6:32
  - the right systems in place for ongoing evaluation.
• 6:35
  - Interesting. Are you able to share what one
• 6:38
  - of those programs is and how it works
• 6:40
  - at your hospitals?
• 6:41
  - Yeah. So we,
• 6:43
  - so one example is right now, we recently
• 6:46
  - started using
• 6:47
  - two models for
• 6:49
  - AI ECG. So, you know, a a ACG
• 6:52
  - has been around ECG has been around forever
• 6:54
  - being read by humans and even machines using
• 6:56
  - rules. But since sort of, like, late twenty
• 6:59
  - tens, a number of algorithms have come out
• 7:01
  - that can see things that humans can't see,
• 7:03
  - things like risk of having
• 7:05
  - low ejection fraction or weak heart muscle, things
• 7:07
  - like having a high risk for AFib in
• 7:09
  - the next year.
• 7:11
  - And so working with one of our strategic
• 7:13
  - partners, we've deployed,
• 7:16
  - these algorithms across our cardiac clinics and now
• 7:18
  - our skill into primary care. And then we
• 7:20
  - spend a lot of time
• 7:22
  - learning what are the best ways to deploy
• 7:24
  - these models. You know, if you just send
• 7:25
  - a note to the clinician, they may not
• 7:27
  - act on it. So how do you really
• 7:28
  - build a safety net with nurses and
• 7:31
  - advanced practitioners and and physicians to really make
• 7:33
  - sure those results are acted upon in a
• 7:35
  - evidence based way.
• 7:37
  - And in this new space of AI diagnostics,
• 7:40
  - what we are really interested in here at
• 7:42
  - at the BCVI Center for AI is,
• 7:45
  - creating is generating the evidence
• 7:47
  - to scale these technologies. And so as we
• 7:50
  - built this program, we've really made an evaluation
• 7:52
  - at the heart of it, so we can
• 7:54
  - study the outcomes, hopefully join other health systems
• 7:56
  - and really understand, do these technologies
• 7:58
  - improve patient outcomes, and how do they bring
• 8:00
  - ROI to all the key stakeholders in the
• 8:02
  - health system?
• 8:03
  - Excellent.
• 8:04
  - Now what is a step other hospitals or
• 8:07
  - health systems could take this week to improve
• 8:09
  - their data to help with decision making with
• 8:12
  - AI?
• 8:13
  - Yeah. So
• 8:15
  - so I'd say two things. So I think,
• 8:16
  - one,
• 8:18
  - it's really important for health systems to figure
• 8:19
  - out how to empower
• 8:21
  - all your
• 8:22
  - workforce to use these tools. You know, these
• 8:25
  - tools began so easy to use with natural
• 8:27
  - language. People can build all sorts of things.
• 8:30
  - So I think that making sure people have
• 8:32
  - the the training, the secure environment, and access
• 8:35
  - to the data to take their ideas into
• 8:37
  - action is crucially important. Now when you do
• 8:39
  - this, you wanna make sure the right guardrails
• 8:41
  - for security. You wanna make sure you have
• 8:42
  - ways to monitor costs so the system isn't
• 8:45
  - getting a huge bill from all the token
• 8:47
  - credits people are using with these cloud providers.
• 8:49
  - But, really, I think health systems need to
• 8:51
  - figure out how to empower the frontline workers
• 8:53
  - to start solving problems or pair them with
• 8:55
  - engineers who can help them solve these problems
• 8:57
  - in an efficient way.
• 8:59
  - The second point, I think, I'd say is
• 9:01
  - that
• 9:02
  - maybe they don't need to take a a
• 9:04
  - step this week. You know, this is AI
• 9:06
  - is a really fast moving space. It feels
• 9:08
  - like it's everywhere. I think health and feel
• 9:09
  - a lot of pressure to adopt AI.
• 9:11
  - But I think it's probably more important to
• 9:13
  - think about what are your priorities,
• 9:15
  - what are your systems you have in place
• 9:16
  - for evaluation governments, and then what partners do
• 9:19
  - you really wanna go deep with. You know,
• 9:21
  - you no one wants to just have hundreds
• 9:23
  - of pilots or dozens of pilots that never
• 9:24
  - really scale. So what are the you really
• 9:26
  - your key priorities that you'll go deep with?
• 9:28
  - And then, hopefully, if they work, scale them.
• 9:30
  - And that's also the fun part, going deep
• 9:32
  - in these partnerships and and programs, and that's
• 9:35
  - where the work happens.
• 9:37
  - Great. Well, this has been such a great
• 9:40
  - discussion. I appreciate you taking the time to
• 9:42
  - join me on the podcast today. Thanks for
• 9:44
  - having me.