#90 How Data Science is Transforming the Healthcare Industry (with Curren Katz)
Summary
TLDRThis episode of the "Data Framed" podcast features Karen Katzenbach from Johnson & Johnson discussing the transformative potential of data science and machine learning in healthcare, especially post-pandemic. She emphasizes the scalability of data solutions, ethical AI use, and handling the industry's operational challenges. Karen highlights the importance of ethical considerations in AI applications and interdisciplinary collaboration to ensure efficacy and fairness in outcomes. Successful use cases include optimizing patient care through better scheduling and enhancing diagnostics for early disease detection. She calls for overcoming data challenges like system interoperability and emphasizes the vital role of data literacy and transparency in using data responsibly within healthcare environments. She also talks about her experiences at J&J, sharing insights into executing significant projects like COVID-19 vaccine research efficiently using predictive modeling to target clinical trials effectively. Karen underlines the importance of understanding user needs in developing solutions and advocates for innovative approaches that include partnerships across different sectors.
Takeaways
- 🎯 Applying data science can solve crucial healthcare challenges effectively.
- 🤝 Interdisciplinary collaboration is key to healthcare solutions.
- 💡 Ethical AI considerations are necessary for effective healthcare applications.
- 🔍 Diagnostic improvements and operations can benefit from AI.
- 📈 Scalability in data solutions enhances healthcare impact.
- ⚙️ Data interoperability poses challenges but can be overcome with innovative solutions.
- ☑️ Data literacy is essential for successful data science use.
- 🔗 Cross-industry learning boosts healthcare applications.
- 📊 Predictive modeling can optimize healthcare trials.
- 🌟 Focus on patient-centric outcomes for impactful health solutions.
Timeline
- 00:00:00 - 00:05:00
Introduction to the podcast, highlighting its focus on data science trends and insights within healthcare, hosted by a data science educator and evangelist. The episode guest, Curran Katz from Johnson & Johnson, has extensive experience in healthcare data science.
- 00:05:00 - 00:10:00
Curran shares her background, career journey, and how she became engaged in the data science field. Her journey moved from cognitive neuroscience to roles in healthcare companies, eventually becoming a leader in data science innovation within large organizations.
- 00:10:00 - 00:15:00
Discussion on the current landscape of data science in healthcare, noting its slower evolution compared to other industries due to unique challenges and risks. However, there's a growing acceptance and deployment of data science models to drive business decisions.
- 00:15:00 - 00:20:00
Curran emphasizes the operational impact of data science in healthcare, such as improving patient scheduling for chemotherapy treatments, and the importance of operational efficiency in creating immediate impacts on patient experiences.
- 00:20:00 - 00:25:00
Exploration of main challenges in healthcare data science, such as data access, ethics, and sensitivity. Curran explains the importance of addressing ethical concerns and biases when using AI and machine learning, emphasizing collaborative, cross-functional approaches.
- 00:25:00 - 00:30:00
Curran discusses the importance of empathy and human-centered design in developing AI healthcare solutions. She highlights the need to focus on use cases that minimize harm and improve clinician decision-making rather than replace it.
- 00:30:00 - 00:35:39
Curran shares insights into managing data science teams within Johnson & Johnson, focusing on aligning with long-term goals and patient impact. She talks about leveraging resources within large organizations for developing data science solutions that enhance healthcare processes.
Mind Map
Video Q&A
What are operational challenges in healthcare?
It refers to using data science to optimize backend processes, such as scheduling, to enhance healthcare efficiency and patient experience.
How has the pandemic affected data science in healthcare?
The pandemic has highlighted opportunities for using data science in drug discovery, operational innovation, disease prevention, and more.
What trends has Karen noticed in data science and healthcare lately?
She notes scalable implementation, ethical AI use, and expanding capabilities in diagnostics and operations.
How was data science used in J&J's COVID-19 vaccine trials?
Using data science helps predict clinical trial locations where COVID-19 cases are rising, speeding up research processes.
How can healthcare overcome data challenges?
Multiple departments must work together, understand patient data interactions, and embrace innovative methods from other industries.
What is necessary for a successful data science project in healthcare?
A well-understood problem, collaborative implementation, and engagement from solution users are essential for success.
How does J&J ensure effective use of data science in their R&D department?
Interdisciplinary collaboration and focusing on unmet needs were key approaches.
How does Karen manage short-term wins and long-term projects in data science?
Finding clear problems that data science can address and creating timelines that align with business priorities.
What excites Karen about fairness in data science in healthcare?
It's about understanding and mitigating model biases while making real-time adjustments to ensure fair healthcare outcomes.
What main advice does Karen give to data science professionals?
She loves using data science to solve impactful healthcare issues and emphasizes the need for ethical consideration.
View more video summaries
SolarWinds: A Path to Excellence in Software Supply Chain Security
iPhone 16/16 Pro Review: Times Have Changed!
How Rwanda is Becoming the Singapore of Africa
Venus in 12 different houses and nature of spouse in detail||venus and past life karma in marriage||
I Investigated the City Where Every Drug is Legal...
Fluid Mechanics Lecture
- 00:00:01[Music]
- 00:00:08you're listening to data framed a
- 00:00:10podcast by data camp in this show you'll
- 00:00:13hear all the latest trends and insights
- 00:00:15in data science whether you're just
- 00:00:17getting started in your data career or
- 00:00:20you're a data leader looking to scale
- 00:00:22data-driven decisions in your
- 00:00:23organization join us for in-depth
- 00:00:26discussions with data and analytics
- 00:00:28leaders at the forefront of the data
- 00:00:29revolution let's dive right in
- 00:00:35hello everyone this is adele data
- 00:00:37science educator and evangelist at
- 00:00:38datacamp
- 00:00:40two years into the pandemic the
- 00:00:41potential for data science and machine
- 00:00:43learning in health care has never been
- 00:00:45more apparent
- 00:00:46whether it's drug discovery acceleration
- 00:00:48operational innovation virtual
- 00:00:50assistance and disease prevention the
- 00:00:52margin of opportunity for data science
- 00:00:54and health care is massive
- 00:00:56however that doesn't come without its
- 00:00:58own set of unique challenges and risks
- 00:01:00that require unique solutions
- 00:01:02this is why i'm excited to have current
- 00:01:04cats on today's episode of data framed
- 00:01:06curran is a senior director for data
- 00:01:08science portfolio management at johnson
- 00:01:10johnson she has decades of experience at
- 00:01:12the intersection of healthcare and data
- 00:01:14science and is deeply attuned to the
- 00:01:16state of data science and healthcare
- 00:01:17today
- 00:01:18throughout our conversation we discuss
- 00:01:20where the landscape of data science and
- 00:01:22healthcare is today the unique
- 00:01:23challenges of applying data science and
- 00:01:25healthcare the importance of ethical ai
- 00:01:28when working on healthcare use cases how
- 00:01:30to solve some of the data challenges of
- 00:01:32the healthcare industry use cases she's
- 00:01:34been excited about how data science was
- 00:01:36used to tackle covet 19 and much more if
- 00:01:40you enjoyed this podcast make sure to
- 00:01:42rate us and subscribe and add a comment
- 00:01:43but only if you enjoyed it now let's
- 00:01:46dive right in
- 00:01:47karen it's great to have you on the show
- 00:01:49yeah great to be here thank you for
- 00:01:51having me i'm excited to talk to you
- 00:01:53about data science and machine learning
- 00:01:54and healthcare your experience leading
- 00:01:56data teams and complex organizations and
- 00:01:58how you've led r d at johnson johnson
- 00:02:01but before i'd love to learn more about
- 00:02:02your background and what got you into
- 00:02:04the data space yeah absolutely
- 00:02:07so
- 00:02:08i guess like most people i've always
- 00:02:10loved data and my first statistics
- 00:02:13courses i started to think oh this could
- 00:02:15be really really fun and especially when
- 00:02:17i started applying it to data i had
- 00:02:19collected as a research assistant it was
- 00:02:22pretty addictive
- 00:02:23and then as i moved along in my career
- 00:02:26i'm a cognitive neuroscientist by
- 00:02:27training but did fmri research as well
- 00:02:30as looking at some like large
- 00:02:33epidemiology data sets and
- 00:02:3520 years ago wrote a paper on predictors
- 00:02:37of suicide attempts not exactly an ai ml
- 00:02:40approach to it but that interest in like
- 00:02:43how can we predict some event and then i
- 00:02:46had been in neuroscience studying neural
- 00:02:49networks all of these things and
- 00:02:50applying
- 00:02:51actually machine learning techniques to
- 00:02:55fmri images which are images while
- 00:02:57someone's doing something so it's a
- 00:02:58fairly complex although clean data set
- 00:03:02got me really excited and then i've
- 00:03:03always been passionate about healthcare
- 00:03:05and solving problems in healthcare
- 00:03:07and my first corporate data science job
- 00:03:11was at highmark health so i started on
- 00:03:13the payer side building a bunch of
- 00:03:16models and seeing how those models
- 00:03:18impacted care and was hooked and then
- 00:03:20moved to the parent company it's an
- 00:03:22integrated health care system second
- 00:03:24largest integrated payer provider system
- 00:03:26in the u.s
- 00:03:27and started a data science department at
- 00:03:30that parent company looking at the payer
- 00:03:32the insurance side the provider side and
- 00:03:34a few other diversified healthcare
- 00:03:35businesses and then came to johnson
- 00:03:38johnson where i am now and it's been a
- 00:03:40really exciting career where i get to
- 00:03:42see a lot of impact from data science to
- 00:03:44start off our conversation i'd love to
- 00:03:46understand the current state of data
- 00:03:48science machine learning in healthcare
- 00:03:50early in my career about five years ago
- 00:03:52and that's not too long ago healthcare
- 00:03:54was often and still is talked about as
- 00:03:57an industry with a large margin of
- 00:03:59opportunity for data science but it
- 00:04:00comes with its own unique sets of
- 00:04:02challenges which makes it slower in
- 00:04:04comparison to other industries given
- 00:04:06your experience as a data leader in
- 00:04:07healthcare i'd love to first start off
- 00:04:09our conversation by understanding how
- 00:04:11you would describe what the current
- 00:04:12landscape of data science and healthcare
- 00:04:14looks like today and how has it evolved
- 00:04:16in the past few years oh yeah that's an
- 00:04:19exciting question and it's it has
- 00:04:21evolved and different parts of
- 00:04:22healthcare i'll say are probably
- 00:04:25at different places and evolving and at
- 00:04:28different paces out of sometimes
- 00:04:31necessity and you say there's a lot of
- 00:04:33opportunity in healthcare there is
- 00:04:35and i think it's one of those industries
- 00:04:38where
- 00:04:38you have to take a bit of a careful
- 00:04:40approach to anything new they're
- 00:04:42practically their regulations and
- 00:04:44there's a lot of risk for something
- 00:04:46going wrong but huge benefits but what
- 00:04:48i've seen over
- 00:04:50uh the last
- 00:04:52few years
- 00:04:53is really a couple things that we're
- 00:04:55seeing in a lot of industries
- 00:04:57but in healthcare as well scale
- 00:05:00as we're moving into hey data science
- 00:05:03can be very very useful for solving real
- 00:05:06problems in healthcare there's a focus
- 00:05:08on
- 00:05:09deploying these models and not just
- 00:05:11having perfect
- 00:05:12concepts but really using them to drive
- 00:05:16core business decisions and core
- 00:05:18insights and and that requires data
- 00:05:21science at scale where at first
- 00:05:23it was a little more experimental a
- 00:05:25little more well let's just see
- 00:05:27how this goes alongside what we do today
- 00:05:29but we're not going to go all in and
- 00:05:31really use this to drive our business
- 00:05:33but we're moving towards that
- 00:05:35the other
- 00:05:37change i i guess are the problems that
- 00:05:40that we can
- 00:05:41solve or just we're realizing them right
- 00:05:43we're expanding the scope of what data
- 00:05:46science can do in healthcare and
- 00:05:49of course there's diagnostics there's
- 00:05:50also operations there's clinical trials
- 00:05:53and how those are run how patients are
- 00:05:55found there's so many things we can do
- 00:05:58and then a third i
- 00:05:59really important i wouldn't say change
- 00:06:02but something that's just continues to
- 00:06:04mature and we think about and i think
- 00:06:05it's helped accelerate data science and
- 00:06:08healthcare it's just thinking about the
- 00:06:10ethics of what we're doing considering
- 00:06:12it's impacting people and the care they
- 00:06:14receive and it can be
- 00:06:18life or death or it can either help or
- 00:06:20hurt the disparities we're seeing in
- 00:06:22care so really have thinking about
- 00:06:25ethics which is important in healthcare
- 00:06:27and then having tools and ways to
- 00:06:29address that at scale
- 00:06:31has really evolved over the past few
- 00:06:33years
- 00:06:34that's really great and i'm excited to
- 00:06:36unpack these with you even more so you
- 00:06:38mentioned at the beginning some of the
- 00:06:40areas of impact that data science and
- 00:06:42machine learning have in healthcare do
- 00:06:44you mind expanding on these main areas
- 00:06:45of value where you've seen data science
- 00:06:47and machine learning push the envelope
- 00:06:49forward within the healthcare space
- 00:06:51it's hard to
- 00:06:52pick a few but one i love to talk about
- 00:06:55and this is something my former team did
- 00:06:57and i really i loved the way they
- 00:07:00approached this and i saw to impact
- 00:07:02patients was looking at operations
- 00:07:05so sometimes in healthcare we go at the
- 00:07:09we're going to cure this disease we're
- 00:07:10going to diagnose this disease and of
- 00:07:12course how do we not say we're gonna put
- 00:07:16every data science tool we have towards
- 00:07:18cancer and we should
- 00:07:20but a safer way in in a way in that
- 00:07:23makes a huge impact can be the
- 00:07:24operations of healthcare itself or the
- 00:07:27operations of a clinical trial so i'll
- 00:07:28give you an example when i was at
- 00:07:30highmark health we built a tool
- 00:07:33to help schedule patients receiving
- 00:07:35chemotherapy and a big thing for me to
- 00:07:38start with the problem we heard about
- 00:07:40hey we're scheduling patients for
- 00:07:42chemotherapy they have long wait times
- 00:07:44which seemed not great
- 00:07:46we notice we're really busy in the
- 00:07:48mornings and then
- 00:07:50things are empty in in the afternoon so
- 00:07:52our clinicians are either overwhelmed or
- 00:07:55don't have a lot of patience and we dug
- 00:07:57in that was
- 00:07:58two things they didn't know how long a
- 00:08:00treatment could take and there could be
- 00:08:01side effects and clinicians want to care
- 00:08:03for their patients and make sure they
- 00:08:05have plenty of time so they're blind to
- 00:08:07how how long each patient might need
- 00:08:10staying there in that location so if
- 00:08:12we're able to predict that we can
- 00:08:14start efficiently scheduling and then
- 00:08:16just optimizing the scheduling
- 00:08:18optimizing the operations where in the
- 00:08:20calendar can this go where location wise
- 00:08:22can this go and we had this tool ready
- 00:08:25when the pandemic started and it became
- 00:08:26even more important to space vulnerable
- 00:08:28patients out it started with an
- 00:08:30operational challenge though scheduling
- 00:08:32very practical thing to solve and it
- 00:08:34made a huge difference i i've heard and
- 00:08:37stories from patients and saying hey i
- 00:08:39can get on and back to my life and not
- 00:08:41wait i can come at times convenient to
- 00:08:43me
- 00:08:44another area that i've seen
- 00:08:46an impact and a lot of promises
- 00:08:48diagnosis or detection early diagnosis
- 00:08:50early detection to give clinicians some
- 00:08:54some time to intervene we've heard about
- 00:08:56this in things like sepsis or acute
- 00:08:58diseases we're talking about early
- 00:09:02detection of things like pulmonary
- 00:09:03hypertension which is frequently
- 00:09:04diagnosed late and i know that's
- 00:09:06something where we're doing now these
- 00:09:08are big big areas of opportunity where
- 00:09:10we can treat patients
- 00:09:12because we can detect these diseases and
- 00:09:14diagnose them
- 00:09:15and then the third is patient's own
- 00:09:18experience like with the operational
- 00:09:21component of course that had a patient
- 00:09:23experience
- 00:09:24piece but just understanding patients
- 00:09:28their journeys where they're facing
- 00:09:30challenges how they're experiencing the
- 00:09:32healthcare system and where we're not
- 00:09:34maybe delivering care in the way we
- 00:09:35should
- 00:09:36data can help us see that
- 00:09:38and help us deliver a better experience
- 00:09:40deliver more personalized tailored
- 00:09:42experience on a biological level as well
- 00:09:46as
- 00:09:47just
- 00:09:48an individual level preferences ways of
- 00:09:50interacting and ways of receiving care i
- 00:09:53love how you frame the operations
- 00:09:54component here because whenever we talk
- 00:09:56about data science and machine learning
- 00:09:57in healthcare we always talk about
- 00:09:59aspirational use cases that i think
- 00:10:01we're all in agreement are extremely
- 00:10:02important for example i'm very excited
- 00:10:04to see the impacts of deep minds alpha
- 00:10:06fault and direct discovery but that
- 00:10:08doesn't mean we cannot create impact on
- 00:10:10people's lives right now with data
- 00:10:12science just by solving operational
- 00:10:14challenges when talking about data
- 00:10:16science and healthcare we often talk
- 00:10:18about challenges unique to the
- 00:10:19healthcare space such as access to
- 00:10:21relevant interoperable data ethics of ai
- 00:10:24and a host of other challenges i'd love
- 00:10:26it if you can break down what are the
- 00:10:28main data challenges you think that the
- 00:10:30healthcare industry is facing today
- 00:10:32i talk to my colleagues across
- 00:10:34industries everything manufacturing
- 00:10:37automotive
- 00:10:38just
- 00:10:39very different industries and no one
- 00:10:41tells me
- 00:10:42our data is perfect clean haven't really
- 00:10:45had a problem there or thought about it
- 00:10:47of course you're not surprised to hear
- 00:10:48this and in healthcare we base that as
- 00:10:51well and interoperability and different
- 00:10:53formats of data we're facing the same
- 00:10:55things but i think we're realizing that
- 00:10:58a other industries that face this and be
- 00:10:59you know there are solutions that will
- 00:11:01work here as well it's the whole topic
- 00:11:04the ethics of ai is is huge a huge one
- 00:11:08here
- 00:11:08and really really important so
- 00:11:11this becomes crucial in in healthcare
- 00:11:14i'm not saying if if you're selling a
- 00:11:16consumer good of course you don't want
- 00:11:18to make a mistake but if i get a
- 00:11:19recommendation to buy
- 00:11:21a toaster oven and i just bought a
- 00:11:23toaster oven so i'm probably not going
- 00:11:25to buy a second one and this just
- 00:11:26happened to me it's not a big deal it
- 00:11:28didn't really affect my life you can
- 00:11:30experiment with those algorithms get
- 00:11:31them out there and get them out there
- 00:11:34quickly and in healthcare we've
- 00:11:36obviously had to
- 00:11:37think and other industries face this as
- 00:11:39well
- 00:11:40there's risk so you have to really think
- 00:11:43through
- 00:11:44what you're doing and what could happen
- 00:11:47and how this algorithm is going to work
- 00:11:48what how you're going to build this
- 00:11:50process
- 00:11:51and get it right that's not to say there
- 00:11:53aren't things we can do there's a lot
- 00:11:55because there are a lot of problems and
- 00:11:56things we're not doing
- 00:11:58really well today so as long as we're
- 00:12:00not making it worse we should try some
- 00:12:01things but that's always going to be
- 00:12:05a pretty big challenge and an important
- 00:12:07challenge that we should take on
- 00:12:08relative to
- 00:12:09other industries
- 00:12:11it's just talking about the data
- 00:12:12obviously the sensitivity of the data
- 00:12:14itself
- 00:12:15makes it maybe a little harder to get
- 00:12:17access to data or think about how to use
- 00:12:20it share it what kinds of environments
- 00:12:22that data can be in
- 00:12:24and it should be i mean that's a
- 00:12:26challenge we should take on as a good
- 00:12:28challenge and the one we say we were
- 00:12:30never good enough because this is the
- 00:12:32most sensitive data in people's lives so
- 00:12:35that we should be continuously improving
- 00:12:38and thinking about how we protect this
- 00:12:40data how we use it how we make sure
- 00:12:43we're using it in a way
- 00:12:45that decreases inequalities in how we
- 00:12:48deliver care which i think it can but we
- 00:12:50have to use the data responsibly and
- 00:12:52consider it is very very sensitive data
- 00:12:55maybe more so than if there's a
- 00:12:57a leak of that i bought a toaster oven
- 00:13:00not that exciting
- 00:13:02i bought a coffee baker
- 00:13:04not that
- 00:13:05not that exciting but this this is a
- 00:13:07pretty big one
- 00:13:08i completely agree here and let's spark
- 00:13:10the chat a bit and talk about the ethics
- 00:13:12of ai in healthcare when we talk about
- 00:13:14using machine learning and ai in
- 00:13:16healthcare there's this aversion that
- 00:13:18whatever we develop will end up creating
- 00:13:20harmful outcomes or that it could be
- 00:13:22used irresponsibly and oftentimes the
- 00:13:24response is not to leverage machine
- 00:13:26learning in ai so i'd love to understand
- 00:13:28how you evaluate the risk of harmful
- 00:13:31outcomes of machine learning and ai in
- 00:13:32healthcare and how do you go about
- 00:13:33minimizing it
- 00:13:35well a great question one big thing to
- 00:13:38understand the potential harmful
- 00:13:41outcomes you have to understand the
- 00:13:43problem that you're solving
- 00:13:45be working collaboratively with a
- 00:13:48cross-functional team with clinicians
- 00:13:50with whoever is using and implementing
- 00:13:53and acting on your model with patients
- 00:13:55you have to have everyone in the room
- 00:13:56and involved in this process
- 00:13:59and understand that end-to-end because
- 00:14:01that's the only way you're going to find
- 00:14:03where the risks might lie you have to
- 00:14:05understand how how they're going to use
- 00:14:08this information and make a decision
- 00:14:10what mitigations can you build in where
- 00:14:12are the risks at every point in this
- 00:14:14system in that is sometimes something
- 00:14:17data scientists especially when they get
- 00:14:19started they're excited to build models
- 00:14:22and they skip over this piece of it
- 00:14:24unintentionally and when i read about
- 00:14:26you know resumes from the hr world like
- 00:14:29the algorithm is going to learn
- 00:14:31what you feed it and historically data
- 00:14:34reflects our human biases so the
- 00:14:37algorithm if you don't
- 00:14:39think about it and you don't account for
- 00:14:40that is going to learn to do exactly
- 00:14:42what people have done which is not
- 00:14:44uh
- 00:14:45really necessarily ethical but
- 00:14:49when with data and with an algorithm we
- 00:14:51have an ability to fix that and to
- 00:14:53control that a bit more than than we do
- 00:14:55in people but i always think about the
- 00:14:57end end how the decisions being made it
- 00:14:59can't just be about the algorithm and
- 00:15:01another part is it sounds kind of simple
- 00:15:04but empathy and the human centered
- 00:15:06design thinking approach is very
- 00:15:07valuable for data science because you
- 00:15:09start
- 00:15:10putting yourself in the shoes of
- 00:15:13the the person who's affected by this
- 00:15:15the patient all of the things they may
- 00:15:17be facing and all of the things that may
- 00:15:19happen based on the
- 00:15:21algorithm so you've got to really think
- 00:15:23about it from that angle and
- 00:15:26then it's of course the technology the
- 00:15:28data itself
- 00:15:30what biases are there the algorithms
- 00:15:32you're choosing the ways you can
- 00:15:34mitigate and correct it can you and
- 00:15:37that's job a technical expertise a data
- 00:15:39scientist has to have and it's essential
- 00:15:42now especially in in healthcare but
- 00:15:44everywhere we want to think about that
- 00:15:46the other obvious one is really going
- 00:15:49way back and saying did we pick the
- 00:15:50right use case and like the operations
- 00:15:52example there's a lot of problems to
- 00:15:54solve in healthcare we should be
- 00:15:56thinking about all of them but maybe
- 00:15:58the easier quick wins are ones where
- 00:16:01there's a little
- 00:16:03less
- 00:16:04opportunity for harm if it's maybe we're
- 00:16:08just randomly we're communicating with
- 00:16:10everyone in the same way today and maybe
- 00:16:12if we try to figure out some preferences
- 00:16:15and try to customize a bit and learn
- 00:16:17from there that may be lower risk than
- 00:16:19detecting a disease or changing the
- 00:16:22course of care and in medicine and
- 00:16:24healthcare this doesn't replace a
- 00:16:25clinician we want this to enhance the
- 00:16:27clinicians decision making that's
- 00:16:29awesome and i love how you draw
- 00:16:30inspiration from other fields like human
- 00:16:32centered design given that do you think
- 00:16:34also healthcare can draw from risk
- 00:16:36management risk analysis to create ai
- 00:16:38governance frameworks i think that is a
- 00:16:41great question and absolutely there is
- 00:16:44no industry
- 00:16:46we can't learn from we have to be
- 00:16:48looking outside of healthcare
- 00:16:50all the time and looking across
- 00:16:52healthcare to different parts of
- 00:16:53healthcare but definitely looking
- 00:16:55outside that's why i've
- 00:16:57very intentionally hired people from
- 00:16:58other industries
- 00:17:00on my teams i've wanted people from
- 00:17:02manufacturing and and it has worked
- 00:17:05they've come in and looked at things and
- 00:17:06said this is not an easy but a pretty
- 00:17:08easy problem to solve we deal with this
- 00:17:10all the time and
- 00:17:12something that someone my background is
- 00:17:14mainly in healthcare i would think
- 00:17:17certainly movement of chemotherapy drugs
- 00:17:19around to different locations that i i
- 00:17:21thought as though that's a pretty big
- 00:17:22challenge but i knew that other
- 00:17:24industries had solved it and so i looked
- 00:17:27to people from those industries to come
- 00:17:29in and bring some of that thinking to
- 00:17:30healthcare risk management of course
- 00:17:33that is something we do we have uh risk
- 00:17:37mitigation plans for everything we do
- 00:17:39think through everything early
- 00:17:41the every industry we need to be looking
- 00:17:44outside all the time in healthcare when
- 00:17:46thinking about some of the other
- 00:17:47obstacles that are unique to healthcare
- 00:17:49such as data access interoperability and
- 00:17:51collection what needs to change so that
- 00:17:54data science healthcare innovation
- 00:17:55accelerates here is it regulatory
- 00:17:57innovation industry standards that need
- 00:17:59to evolve
- 00:18:00the regulatory component is there it's
- 00:18:03important there's collaborative work and
- 00:18:05discussions going on across healthcare
- 00:18:08to make sure the the regulatory
- 00:18:11environment meets the needs of data
- 00:18:14science that's an ongoing process
- 00:18:16another one though that
- 00:18:18maybe is every industry but i see it a
- 00:18:21lot in healthcare the systems are very
- 00:18:23complex
- 00:18:24we have different
- 00:18:26emr systems those have a lot of steps
- 00:18:28and pieces data scientists don't always
- 00:18:31understand
- 00:18:32how a clinician interacts with that
- 00:18:34system yet that's that may be the place
- 00:18:36where their
- 00:18:37solution is delivered and acted on where
- 00:18:39the value is realized but they're very
- 00:18:42complicated systems and to get them all
- 00:18:44to connect maybe we want to use
- 00:18:46multimodal data from multiple sources
- 00:18:49imaging
- 00:18:51devices everything to really get a full
- 00:18:53picture of the patient at different time
- 00:18:55scales
- 00:18:56to really scale that solution and
- 00:18:58implement it we need those systems
- 00:19:00connected you can do it once grab all
- 00:19:02the data put it together build a model
- 00:19:04but how do you then deploy that model
- 00:19:06seeing some simplification of these
- 00:19:08systems and some consideration of hey
- 00:19:11it's very important to use this data to
- 00:19:13deploy solutions and to seamlessly
- 00:19:15connect and simplify things
- 00:19:17would be great to see and i think we're
- 00:19:20probably going to see that and i
- 00:19:21as i said it probably exists in in other
- 00:19:24industries as well
- 00:19:26um the other one is experience with data
- 00:19:28science data literacy or ai literacy
- 00:19:32we don't need
- 00:19:34clinicians and hospital administers they
- 00:19:37don't need to be experts in data science
- 00:19:39but
- 00:19:40i think as we all bring up that level of
- 00:19:43understanding and understanding
- 00:19:46how data science works how some of this
- 00:19:49stuff can be used and be able to speak a
- 00:19:51bit of the same language that would help
- 00:19:53and we're seeing that again in every
- 00:19:55industry but one i think we have a good
- 00:19:57chance of solving in in medicine a lot
- 00:20:00of people have a scientific background
- 00:20:02and
- 00:20:03it's data science has the science so
- 00:20:06it should be a good place and i've seen
- 00:20:09a lot of engaged clinicians and a lot
- 00:20:11coming in with a lot of knowledge
- 00:20:13experimental design and that's moving
- 00:20:15along but we could be better there and
- 00:20:18we need to keep pushing and that data
- 00:20:20literacy component is huge from a data
- 00:20:22quality perspective because a lot of
- 00:20:24healthcare professionals are the ones
- 00:20:25who are inputting this data into these
- 00:20:27systems and if they do not recognize the
- 00:20:29role the data plays in the value chain
- 00:20:31of data science then that value chain
- 00:20:33will end up breaking because no one is
- 00:20:35paying attention to the data quality
- 00:20:37right that's a great point and it
- 00:20:39actually that data literacy then it's
- 00:20:41going both ways it's a business literacy
- 00:20:44on the data scientist part of
- 00:20:45understanding
- 00:20:47how a clinician is inputting data and
- 00:20:50how they're interacting with an emr
- 00:20:52system or how on you know the insurance
- 00:20:54side maybe a care manager
- 00:20:56is identifying and reaching out to
- 00:20:59members of an insurance plan to help
- 00:21:01them coordinate their care and manage a
- 00:21:04chronic disease but
- 00:21:05we we have to understand how that data
- 00:21:07comes in
- 00:21:08and
- 00:21:09conversely
- 00:21:10if we show the value of data science the
- 00:21:13the people delivering care and part of
- 00:21:15that healthcare ecosystem
- 00:21:17are going to be able to work with us and
- 00:21:19say okay like i can i can
- 00:21:21see the value of uh this distinction as
- 00:21:25long as we don't take time away from
- 00:21:26their interactions with patients and
- 00:21:28make it harder don't want to do that
- 00:21:30that's awesome and given we're
- 00:21:32discussing the value of data science and
- 00:21:33healthcare i'd like to pivot to discuss
- 00:21:35your experience as a data and ai leader
- 00:21:37at johnson and johnson i'd love to
- 00:21:39understand and dig through some of the
- 00:21:41most exciting use cases you've seen data
- 00:21:43teams working on especially in
- 00:21:44healthcare at johnson and johnson
- 00:21:46especially given what must have been a
- 00:21:48very interesting time for r d teams with
- 00:21:51the release of the j j kovit 19 vaccine
- 00:21:53yeah there there are three that really
- 00:21:55come to mind and one we all are
- 00:21:58so deep in it it's always a great
- 00:22:00example so this is this is something i
- 00:22:02think is an excellent example of using
- 00:22:05data science to solve a real problem and
- 00:22:06make an impact
- 00:22:08when
- 00:22:09clinical trials are planned as you can
- 00:22:11imagine they're complex there's a lot of
- 00:22:12planning and you need to decide where to
- 00:22:14have those trials
- 00:22:16in the case of the vaccine
- 00:22:18we needed to find places where
- 00:22:21kovid was spreading so that we could see
- 00:22:23whether this worked quickly and get it
- 00:22:25out to people
- 00:22:26and what the teams were able to do using
- 00:22:28data science was predict where these
- 00:22:31future hotspots would be and plan the
- 00:22:34clinical trials in those places and it
- 00:22:36was effective and it allowed us to
- 00:22:38accelerate that and be really targeted
- 00:22:40and where we were doing clinical trials
- 00:22:42and where we're seeing high levels of
- 00:22:44covet so i think that's just a very
- 00:22:47great example and it shows data science
- 00:22:49can
- 00:22:50rise to the challenge and really solve
- 00:22:53big problems under pressure when it
- 00:22:55counts with there is no bigger really
- 00:22:57pressure in recent times than the whole
- 00:23:00world's in this pandemic and we need to
- 00:23:03do something about it with data science
- 00:23:04i'm really proud of that
- 00:23:06the other i think i mentioned the
- 00:23:07pulmonary hypertension example but just
- 00:23:09one example of how we can bring
- 00:23:12data together and use ai to diagnose a
- 00:23:15condition earlier and that and that's
- 00:23:17something we're doing and working on
- 00:23:18that's very very exciting this is an
- 00:23:21under diagnosed disease or it's not
- 00:23:23diagnosed early when when we could treat
- 00:23:25it and make an impact so if we can bring
- 00:23:28together diverse data sources and
- 00:23:30predict that diagnosis we can really
- 00:23:32make a difference in people's lives and
- 00:23:34then the third is just generally using
- 00:23:37data to accelerate what we're doing and
- 00:23:39how we're doing it at every part of the
- 00:23:41process we could talk about that all day
- 00:23:43but using digital data and digital
- 00:23:46endpoints to better measure outcomes
- 00:23:48using real world data claims data ehr
- 00:23:51data to really make sure we understand
- 00:23:53the patients we understand their needs
- 00:23:55we're developing drugs that are going to
- 00:23:59to make a difference and we're doing it
- 00:24:00efficiently and quickly because it
- 00:24:02always strikes me that every day that
- 00:24:04this is not out there a patient's not
- 00:24:06getting this treatment so i love that we
- 00:24:08are always focused on how do we get
- 00:24:10medicines to patients faster because
- 00:24:14this matters and we all either have been
- 00:24:17know someone or will be affected by this
- 00:24:20i absolutely love the kovit 19 use case
- 00:24:22here and it's really exemplary of a data
- 00:24:24science use case that requires
- 00:24:26relatively simple data science that can
- 00:24:28provide value now for patients and
- 00:24:30healthcare providers so i'd love it if
- 00:24:31you can impact that use case even more
- 00:24:33and maybe discuss the methodology used
- 00:24:35here
- 00:24:36i think it's a general process that
- 00:24:38really is important for solving any data
- 00:24:41science
- 00:24:42problem and at a high level
- 00:24:45and i've done this set up very multiple
- 00:24:47companies
- 00:24:48it starts with identifying a clear
- 00:24:50problem in this case right it was
- 00:24:52clearly we don't know where to
- 00:24:55plan to have these clinical trials and
- 00:24:57it's not something we can spin up in a
- 00:25:00day it takes some time so how could we
- 00:25:02know earlier it's finding that problem
- 00:25:04that can be solved with data science
- 00:25:07that's one piece that was crucial here
- 00:25:10and then it's collaborating working
- 00:25:13together
- 00:25:15with the
- 00:25:16business clinical areas
- 00:25:18to
- 00:25:19design and implement that solution in
- 00:25:21time sometimes data science if it gets
- 00:25:24too exploratory or
- 00:25:27just experimental
- 00:25:28we're not thinking about the urgency in
- 00:25:30the timelines where we need to deliver
- 00:25:32and working closely as a core member
- 00:25:34across the team and to to make something
- 00:25:37like this happen you have to do that
- 00:25:39those are just two key things that have
- 00:25:42to happen in any high impact
- 00:25:44data science use case and i think ones
- 00:25:46that have served well and then the third
- 00:25:49a piece of advice i got very early and
- 00:25:51i've always used and i've seen
- 00:25:54as a component of successful projects is
- 00:25:57really understanding how
- 00:26:00the solution you're building is going to
- 00:26:01be used and making sure the people who
- 00:26:03are going to use it are involved in the
- 00:26:05planning and have bought into this
- 00:26:06because you
- 00:26:08if you don't have adoption you're you're
- 00:26:10not going to solve the problem that that
- 00:26:12you wanted to solve so i think one thing
- 00:26:14that's evident is that there's a lot of
- 00:26:15different data teams at j doing
- 00:26:17different work it's one challenge to do
- 00:26:19this data science and health care but
- 00:26:21it's another challenge to work in a
- 00:26:22large matrix organizations where there
- 00:26:24are tons of stakeholders and a lot of
- 00:26:26different teams working on different
- 00:26:27problems i'd love to know how you ensure
- 00:26:30that you're staying effective despite
- 00:26:31this complexity and some of the best
- 00:26:33practices you can share
- 00:26:35in managing and working with data teams
- 00:26:38in large matrix organizations with other
- 00:26:40data leaders i think a big one is coming
- 00:26:43back to the shared mission vision what
- 00:26:46you're trying to do because in a
- 00:26:48healthcare organization or any
- 00:26:50organization but definitely in
- 00:26:51healthcare and at johnson and johnson it
- 00:26:53is
- 00:26:54very clear we are getting medicines to
- 00:26:57patients we're saving people's lives at
- 00:26:59the end of the day that
- 00:27:01cuts the the matrix the complexity of a
- 00:27:03large company sure it's there but the
- 00:27:06culture and the focus on the patient and
- 00:27:08what we're doing unifies and brings us
- 00:27:10all together and breaks down those silos
- 00:27:12and i think if at any company if you
- 00:27:15find and focus on that the problem
- 00:27:18and what you all care about how
- 00:27:20everyone's benefiting
- 00:27:21it it really helps the other is
- 00:27:23something i i think is just crucial
- 00:27:25bring people in early from across your
- 00:27:27company it becomes more complex when
- 00:27:29data science happens in the silo and
- 00:27:31then you show up with a solution
- 00:27:34and different parts of the business are
- 00:27:36thinking oh no we needed to be involved
- 00:27:38earlier this is slightly off here and it
- 00:27:41it can be harder than it needs to be
- 00:27:43which is
- 00:27:45brings me to the good part of a large
- 00:27:47matrix organization and why i keep
- 00:27:50working for them and i love to be at one
- 00:27:52i love to be the leader in a large
- 00:27:54matrix organization
- 00:27:56you have incredible resources
- 00:27:59you have
- 00:28:00experts you have legal teams you have
- 00:28:03supply chain there's there's so many
- 00:28:06experts in
- 00:28:08the area where you're developing
- 00:28:09solutions that's it is a luxury to have
- 00:28:12when you're a startup i talk to
- 00:28:13companies people that have great ideas
- 00:28:16and they have to work so hard to just
- 00:28:17get access to hey can you just tell me
- 00:28:19about
- 00:28:20some of the problems you have or how
- 00:28:22this works and they don't have all of
- 00:28:24these resources surrounding them at a
- 00:28:26large company
- 00:28:28you have so much support and you can
- 00:28:31never reach out too much or
- 00:28:33too early and think about hey you know
- 00:28:35what i'm struggling a bit with maybe
- 00:28:38how do you think about marketing oh we
- 00:28:40have a marketing team they everybody
- 00:28:42loves to get involved and they love
- 00:28:44to help and most companies i think
- 00:28:46you'll find this so reach out and use
- 00:28:48those resources that make a large
- 00:28:50company great because otherwise you're
- 00:28:52going to have
- 00:28:54all the bad parts of a big company and
- 00:28:55none of the good parts and that why do
- 00:28:57that that's great and it must be
- 00:28:59especially rewarding to have access to
- 00:29:01healthcare subject matter experts across
- 00:29:03the value chain because this will help
- 00:29:05you develop this empathy to create human
- 00:29:08centered data science solutions
- 00:29:10exactly no absolutely and we have that
- 00:29:13easily just phone call or
- 00:29:16quick message away like
- 00:29:18we're
- 00:29:19people are happy to talk and using that
- 00:29:22is key yes wonderful to have you
- 00:29:25great to use awesome so i'm sure these
- 00:29:27conversations with subject matter
- 00:29:29experts also influence your roadmap
- 00:29:31given the importance of r d in the
- 00:29:33healthcare space how do you ensure an
- 00:29:35adequate split between long-term
- 00:29:36research and short-term wins that can
- 00:29:39help you move the needle yeah absolutely
- 00:29:41and right now i'm in this r d
- 00:29:43environment developing medicines and
- 00:29:46it's a long-term view which is really
- 00:29:49interesting to see and to have
- 00:29:51that said there's a lot of short pieces
- 00:29:55and wins along the way to get to that
- 00:29:57end goal
- 00:29:58so if you're working with the
- 00:30:02clinical teams and as we do we really
- 00:30:05work together or in any company you're
- 00:30:06working with the business area and
- 00:30:09talking about
- 00:30:10what is that end to end what's the
- 00:30:12ultimate kind of long-term outcomes and
- 00:30:15then work backwards what are the short
- 00:30:17pieces and those quick wins as you say a
- 00:30:19lot to get you there
- 00:30:20you get that mix and then i think it's
- 00:30:23important to look at at the portfolio
- 00:30:26you have for data science and go through
- 00:30:28and see
- 00:30:30how many of these are really
- 00:30:32it's going to be years before we see the
- 00:30:34value and that's something in data
- 00:30:36science you need to know because you
- 00:30:37have to be careful not to let that
- 00:30:40timeline
- 00:30:41and that
- 00:30:42pace of technology and changes conflicts
- 00:30:45you've got to think about it early but
- 00:30:47yeah looking at how many long-term
- 00:30:49projects we have how many short quick
- 00:30:51wins do i have and then also
- 00:30:54it's okay to have purely exploratory i'm
- 00:30:56gonna play around with this data see if
- 00:30:58i can develop this model that's great to
- 00:31:01have too it's just looking across the
- 00:31:03portfolio and making sure
- 00:31:05that the
- 00:31:06percentage of work that's in all of
- 00:31:08these buckets is where you want it to be
- 00:31:10and need it to be and how do you
- 00:31:11determine which areas to research in
- 00:31:13your r d agenda
- 00:31:15the good thing is in an r d organization
- 00:31:18that happens at such a high high level
- 00:31:21but
- 00:31:22to bring it back to one simple concept
- 00:31:24it's unmet need and what do patients
- 00:31:27need and i think it's something
- 00:31:29that applies everywhere that where is
- 00:31:31there an unmet need where we can bring
- 00:31:33data science in but of course that's
- 00:31:35goes into the planning of what do we
- 00:31:37develop and it's a pharmaceutical r d
- 00:31:39organization it's a big process it's the
- 00:31:42core of of the business
- 00:31:44and then there's the data science
- 00:31:46component how does data science
- 00:31:48support and accelerate and enhance
- 00:31:52that that portfolio and that that r d
- 00:31:55process
- 00:31:56and as we mature and talk to each other
- 00:31:59and data science grows which we're doing
- 00:32:02at johnson johnson janssen r d which is
- 00:32:06pharmaceutical companies johnson johnson
- 00:32:08the data science team and capabilities
- 00:32:11are just exceptional
- 00:32:13jacqueline is our chief data science
- 00:32:15officer has built
- 00:32:17just a really incredibly advanced
- 00:32:20capability and and the company is
- 00:32:22putting a lot of investment
- 00:32:24into data science in r d and commercial
- 00:32:27and across the company
- 00:32:28it's great to see and that shows me that
- 00:32:31there is it right we've had the
- 00:32:32discussion about this can impact
- 00:32:35the r d portfolio this can um
- 00:32:38help you meet your goals and we've had
- 00:32:40that conversation conversations been
- 00:32:41successful and that's why we're able to
- 00:32:43to grow and really use data science
- 00:32:45now karen as we close out i'd love to
- 00:32:47have a look into the future and what you
- 00:32:49think are the data trends and
- 00:32:50innovations that you're particularly
- 00:32:52looking forward to see within healthcare
- 00:32:54one that is very important and i'm very
- 00:32:57excited about is the concept of fairness
- 00:33:00so we talked about the risks and reasons
- 00:33:03people don't want to use ai in
- 00:33:05healthcare and and this one comes up a
- 00:33:07lot and it really it any kind of high
- 00:33:11stakes industry it affects that industry
- 00:33:13but i'm really excited about the
- 00:33:16capabilities and the thinking that that
- 00:33:18is evolving around fairness both
- 00:33:20being able to detect bias and unfair
- 00:33:23pieces of the algorithm and then even
- 00:33:26fix that on the fly at scale make
- 00:33:29corrections i think that has the ability
- 00:33:31to allow us to really use data science
- 00:33:34ai and machine learning and healthcare
- 00:33:36but it really brings a ton of value to
- 00:33:39to people to patients and make sure
- 00:33:42they're getting
- 00:33:43care that is fair that we're considering
- 00:33:46things that maybe we haven't been great
- 00:33:48at in the past and maybe this can make
- 00:33:49medicine a bit better or any field a bit
- 00:33:51better so fairness is a huge one for me
- 00:33:54future trends of course i think we're
- 00:33:56going to continue to see
- 00:33:58scale we're going to continue to see a
- 00:34:00bit of a
- 00:34:02i don't want to say a ketchup but we're
- 00:34:03in a nice position to leapfrog other
- 00:34:05industries that have
- 00:34:07really perfected or made a huge a lot of
- 00:34:10the advancement in
- 00:34:12embedding ai into
- 00:34:14every part of their business we can take
- 00:34:16the
- 00:34:17technical learnings and platforms and
- 00:34:18pieces and start from there in
- 00:34:20healthcare and i think we're going to
- 00:34:22see that continue to grow because as we
- 00:34:24start making an impact we're going to
- 00:34:25need to consider how this becomes a core
- 00:34:28part of healthcare
- 00:34:29karen it was great to have you on the
- 00:34:31show do you have any final call to
- 00:34:33action before we wrap up
- 00:34:35you know it is to focus on the impact
- 00:34:38like i just
- 00:34:39always encourage data science and data
- 00:34:42science leaders to think through
- 00:34:44how is this data science solution
- 00:34:47solving a business problem how is it
- 00:34:48making an impact and how is it doing so
- 00:34:51the right way so focus on impact
- 00:34:54understand the context be fair but
- 00:34:56really go all and make a difference
- 00:34:58because data science we're ready for
- 00:35:00that
- 00:35:01thanks for being on data framed no thank
- 00:35:03you thanks for having me
- 00:35:11you've been listening to data framed a
- 00:35:13podcast by data camp keep connected with
- 00:35:16us by subscribing to the show in your
- 00:35:18favorite podcast player please give us a
- 00:35:20rating leave a comment and share
- 00:35:22episodes you love that helps us keep
- 00:35:25delivering insights into all things data
- 00:35:28thanks for listening until next time
- Data Science
- Healthcare
- Machine Learning
- AI Ethics
- Operational Challenges
- COVID-19
- Clinical Trials
- Data Integration
- Interdisciplinary Collaboration
- Data Literacy