ACM ByteCast - Wendy Chapman - Episode 24

Episode Date: April 21, 2022

In this episode, the first of a special collaboration between ACM ByteCast and the American Medical Informatics Association (AMIA)’s For Your Informatics podcast, hosts Karmen Williams and Sabrina Hsueh welcome Wendy Chapman, Associate Dean of Digital Health and Informatics at the University of Melbourne and Director of the Centre for Digital Transformation of Health. Her research focuses on developing computer algorithms to understand information typed into electronic medical records and natural language processing of clinical texts. She is an elected fellow of the American College of Medical Informatics and the US National Academy of Medicine. Wendy discusses her journey from an undergraduate background in linguistics and Chinese literature to completing a PhD in Medical Informatics at the University of Utah and learning to program from scratch. She also describes moving to Australia when saw an opportunity to grow the field of digital health in Melbourne. She identifies the most pressing issues she is faced with in her new role and provides valuable advice based on her most impactful career moves. Wendy also shares with Karmen and Sabrina the development of the Digital Health Validitron at the University of Melbourne, which will guide innovators through questions in order to obtain funding and reimbursement. Finally, she identifies the areas in which ACM and AMIA can partner together in order to create a real impact in the field.

This episode is part of a special collaboration between ACM ByteCast and AMIA For Your Informatics Podcast, a joint podcast series for the Association of Computing Machinery, the world's largest educational and scientific computing society, and the American Medical Informatics Association, the world's largest medical informatics community. In this new series, we talk to women leaders, researchers, practitioners, and innovators who are at the intersection of computing research and practice to apply AI to healthcare and life science. They share their experiences in their interdisciplinary career paths, the lessons learned for health equity, and their own visions

for the future of computing. Hello and welcome to the ACM AMA joint podcast series. This joint podcast series aims to explore the interdisciplinary field of medical informatics, where both the practitioners of AI and ML solution builders and stakeholders in the healthcare ecosystem take interest. I am Dr. Sabrina Hsueh with the Association of Computing Machinery podcast series. And co-hosting with me today is Dr. Carmen Williams from 4-Year Informatics Podcast with the American Medical Informatics Association. We have the pleasure of speaking with our first guest of the series, Dr. Wendy Chapman. Thank you for joining us. Dr. Wendy

Chapman is the director of the Center for Digital Transformation of Health at University of Melbourne. Her research focuses on developing computer algorithms to understand information typed into electronic medical records, natural language processing of clinical texts. She sees incredible opportunities for improving healthcare delivery and individual health if we can harness information to make decision-making, improve processes, and behavioral change. She enjoys skiing, mountain biking, cooking, and visiting dog parks and beaches with her two huskies. She moved to Melbourne with her husband and youngest son and looks forward to the time when her older children can visit. Thank you so much again for joining us, Wendy.

Yeah, thank you for having me. Great to have you here, Wendy. As you know, in this new podcast series, we would like to focus a bit more on the career paths of people who have been working in the disciplinary areas. And you have been working in the interface between computer science and medicine for a long time. What made you decide to start your career on this path initially? And did anything particular motivate you in that beginning? Yeah, my story is like a lot of people's stories in this field, where there was a lot of wandering and a bit of serendipity. I started graduate school in 1994, and I had a bachelor's degree in linguistics with a minor in Chinese and had been admitted to a PhD program in Chinese literature.

My husband was doing electrical engineering, and he decided he wanted to do medical applications. So he applied to the informatics department at the University of Utah. And when he interviewed there, he met with Peter Haug, who does natural language processing. And he talked with one of the graduate students who was an electrical engineer. And he said, it's not rocket science, but the linguistics is killing me. And so Brian told me about this. And I thought that might be a really nice field to go into because I can apply my love of language and it's something practical and potentially helpful to the world. So we moved back to Utah. We were in Wisconsin at the time. School had already started, but my

husband's advisor encouraged me to apply. So I applied and they let me in on probation because I had never had biology or physiology. I didn't know how to program computers. And all I did have is I had good GRE scores and I'd had calculus and gotten good grades in calculus. So they let me in on probation. And after the first semester, I did well. And so they let me keep going. It took me six years to get my Ph.D. I had a lot of classes to take, a lot to learn. And so my motivation was really pretty shallow. I was just really looking for a career and this seemed like something interesting. Thank you. It's always fascinating to hear your story.

And then I'm wondering, you already mentioned a few of the interdisciplinary kind of areas you're into. And so were there any challenges that you confronted and then how did you overcome them? Yeah. Graduate school was really hard. The courses for me, because taking physiology and things that I hadn't had before, but the real challenge was learning how to program. And because I was in graduate school, but had never taken any programming classes, I had to take undergraduate programming classes, but they didn't get credit programming classes. I had to take undergraduate programming classes, but they didn't get credit for them. I learned to program in C++, and I had a one-year-old when I started graduate school. And this was back in the days of modems. I mean, we had just started using the internet, and you couldn't really do what you needed to at home. And so,

I would go to the computer lab at the university. And one time I took my one-year-old because he was, I don't remember why I took him, but it was like two in the morning and he stretched across my feet sleeping. And I was the only woman in this computer lab with 40 men sitting there working on their computer programs. So that was the hardest thing for me was learning how to program and there wasn't a clear pathway for people in informatics. And I think there still isn't in general. Yeah, I had a similar experience, Wendy. And it's always fascinating to hear from another woman who has overcome so many barriers to be where she is today. And you have taken up several leadership positions from there in your career path. You have been director at Utah, and most recently at the Center of Digital

Transformation of Health at the University of Melbourne. Just wondering, then how did you decide to pursue a leadership role? And what led you to your current role? Yeah, another serendipitous story. So in 2013, the position of chair of biomedical informatics opened up at the University of Utah. And that's where I graduated from. And I love that department. And so it took me several months to get up the courage to apply. And actually, a lot of discussions with people and arguments with my husband about, you know, am I really chair material? I was the one saying, I can't do this. I'm too young.

It's too early in my career. People will look at me and say, who does she think she is applying for this position? And he's like, you sound so stupid. Come on. You need to at least apply. So I finally got up the courage and applied and I ended up taking the job. And it was a real career pivot for me because I left my research. I didn't leave it, but I had to spend a lot of time on the leadership parts. And so my research time really decreased. But I took it because I wanted to have

more impact. And I felt like where I was at in natural language processing was too far away from impact on the patient. And I wasn't a good software engineer where I could really be applying the NLP in real settings and having impact that way. So I decided to go down the leadership role where I could have potentially influence over a broader stage. And so after six years there, we were looking at other opportunities. I was at AMIA in San Francisco and I went across the street and had lunch with a friend. And another colleague came and sat with us and she had just happened to be in the restaurant at the same time. And I told her I was looking around and she said, oh, we have this position in Melbourne that they never filled. And so ultimately I ended up interviewing and we made

the big move to come down under. And what appealed to me was it's a new investment here at the University of Melbourne. They haven't had very much strength in informatics or digital health. And Australia is really behind in a lot of ways. And there are some fabulous pockets of excellence here in informatics and digital health. But in general, the field, the healthcare field is behind. And in fact, the hospitals that we are affiliated with, most of them didn't even have electronic medical records when I moved here. And so the four of the main hospitals we work with just installed Epic a year ago. I thought there was a lot of opportunity to take the things that I had learned in the past and bring them to a new place where I could potentially have more impact. What a story. This is exactly why we have this joint series through looking to this

interdisciplinary area between ACM and AMIA. We want people to see how to make an impact in real world applications, and you are certainly making examples of that. Thank you. And as we also been a big learning curve to figure that out because as we all know, the funding model really drives the type of innovation that you can put into place or not. And so I think there's a ton of opportunity for better connection and our vision and our center is connected health, that the patient really has a more, when they interact with the healthcare system, they have a more connected journey. And there's huge opportunity to improve in that here. So that would be one of my main goals. Wonderful. You might have just attracted more people to following your footsteps. And switching gears a little bit here, I also want to talk to you more about the Women Leadership

Program you have been pushing for years. Yeah, so the Women in AMIA Committee decided, has a leadership subcommittee, which you know, because you lead that right now, or co-lead that. And we thought that there was need for a leadership program to help. And it really stemmed from a lot of our experience, where we weren't applying for awards, or positions, or promotions, because we weren't applying for awards or positions or promotions because we weren't confident enough that we could do it. You know, just like me with the job at Utah, it's part confidence, it's part skills. You know, there's things that you don't learn as a researcher to be a leader. You have to understand more about finances and there's,

more politics and change management and those types of things you don't learn. And so we wanted to make a program where we could increase the confidence of women. We could increase their confidence in those areas that they had less training in, and we could improve their connections with each other and helping encourage each other and give each other advice. So we launched the Women in AMIA Leadership Program, and it's now in its second year. It has its second cohort. They're about halfway through. And I think it's been really fabulous for a lot of people. We also launched it here in Australia. We had the kickoff two days ago, and we have 27 women starting the Women in Digital Help Leadership Program here.

That's wonderful. And as being a part of the second cohort of that leadership program, it's amazing. And I am already learning so many great things. And so thinking about that, what would, how to do that. And this happens with all of us in this field. It's very multidisciplinary. And we know that in order to get the outcomes that we want, we need people with computer science expertise, statistics, sociology, psychology, linguistics sometimes, informatics. And bringing those diverse people together to work together can be really challenging. So I think that's my biggest challenge is figuring out how to work as a team. Another challenge is I'd say the funding models are really getting a hold of them on them and

understanding them and then being able to innovate within the environment that you're in, because there's so many barriers. I would like to dive in a little bit more here. When you mentioned the challenge coming from the funding model, is it because it's more driven by the government? Or is it because you will have to go out there to do fundraising a little bit more than what you used to do? Can you elaborate a little bit more on the challenge on this front? Yep. So they have a mixed system here. There is government funded health care, but there's also private health care. But the government funded health care,

the big challenge is that hospitals and acute care are funded by states and primary care is funded by the federal government. And so they have very different models, and they don't connect with each other, and they're funded in different ways. So it's that fragmentation between primary care and acute care. And you see in the US and in other places starting to have integrated systems. We don't have that here. I see. see yeah so you also have to integrate these different data sources on your site and try to navigate the process of doing all this integration on your own just just like in when you're collaborating like between computer science and medicine and you see that the funding models for our jobs are so different as faculty. In computer

science, they have a nine-month contract. The way that they're paid and the way that they're incentivized is different than in medicine. And that makes it hard to work together. And it's similar when the federal government and the state government systems are incentivized differently, and it makes it hard to integrate them. Right. But to drive patient outcomes, you really need to have them integrated in order to understand the pictures end-to-end. That's right. Thanks a lot for sharing. That's certainly a very common problem in this field

that we all need to think more about how to tackle. Switching gear back to our career-related questions, for our audience here who are early in their career, they would like to know more about what were some of the earlier career moves that you have found useful in hindsight? And now you are here, can you recommend to newcomers in this area who might have backgrounds on one side but not on the other side what would be your advice for them to make impact in a multidisciplinary team yeah it would be the same for no matter what stage i think okay i would say first say yes and we're always hearing you know you need to learn to say no. And that's true. But most of my opportunities came from me saying yes to things that either I didn't really want to

do, or I thought I shouldn't be doing it that, you know, it was the wrong timing, or I thought I couldn't do. And people could see potential in me, leadership potential that I couldn't see myself. And so they would give me opportunities to lead in certain areas. For example, the AMIA Natural Language Processing Working Group, and somebody nominated me to be the chair of that. And I never would have thought of doing that. And I talked to him, he was in France, and I said, I don't want to do this. I have no idea how to do this. And he said, oh, it's just, it's so easy. You don't really have to do anything. So I'm like, okay. And then I made it, I got elected and I thought, well, I better do a good job. And so I made it my own

thing and went down that path and surprised myself in the things that we were able to accomplish. And I've had so many examples of that where someone kind of sponsors me and says, I want you to do this. And it's going to take more time. And it takes time away from the things that I thought were important and that it's too early in my career, but I did them and they opened up so many new doors. So that'd be my first piece of advice. And the second piece of advice would be, go ahead and pivot. You know, this is more like mid-career. Pivoting is scary and you lose a lot. And, you know, becoming the chair was really hard for me. And even through the whole six years,

I just felt like a novice. When I left UC San Diego as an NLP researcher, I was feeling like kind of at the peak of my research career and really confident in that space. And then when I became the chair, it's like I knew nothing about how to do my job. And I was thrown in with all these clinical chairs, chairs of surgery and pathology and medicine and the CEO of the hospital. And there were all these things I didn't know about how healthcare worked and the things that they cared about. And I made a fool of myself a lot. I remember asking in this meeting with all the chairs about, they were talking, they were looking at spreadsheets and they had bad debt. And I'm like, what's bad debt. And that's when, you know, like patients don't pay their bills, but these are things that I potentially should

have known, but I didn't know. And so just feeling like a novice. And it's the same thing coming here. Every time I go into a new role, I feel like I left that comfort and confidence behind, but it opens up new opportunities and I've grown new strengths and been able to do things that I never would have been able to do before. ACM Bycast and AMIA FYI Podcast are available on Apple Podcasts, Google Podcasts, Spotify, Stitcher, and other services. If you're enjoying this episode, please subscribe and leave us a review on your favorite platform. So switching gears just a little bit, health equity is an issue that is in the news and in the world today.

It is a very important issue that we're trying to relate to the world. So what would you say health equity is to you? And then what would be some of the most pressing issues? I think that we get treated differently when we go to the doctor, depending on our sex, our skin color, our financial situation. And if we really had health equity, that wouldn't be the case. Sometimes it's explicit, but most of the time, I think it's this implicit bias that we have. I think that's a huge issue, and it's surfaced now, and I think it's clear to people that it's happening, which is really good because that first step is identifying it. A pressing issue, I think, for us in our field

is that as we build machine learning algorithms from the data that we have, that data was collected in a biased way. And it represents biased workflows and biased processes and biased decisions. And we're using that to train systems to make decisions in the future. But I think, again, that we recognize that's a problem. And so we're addressing it. But we need to be really cognizant not to hard code the biases that exist into our automated systems as well. That's also one of the reasons why we started this series to look into the real-world application between computer science and medicine. We want to see what's the best practice now. And with regards to the health equity issues, did you see any potential problem here that can be

handled by people who are sitting in the middle? Yeah, I think the application of artificial intelligence in healthcare is a perfect place for these two groups to come together. And there's so much work to be done in terms of getting better data, training systems that, like we said, aren't encoding the biases that the system has, but also in creating interpretable machine learning tools so that the developer and the researcher and the clinician can sit together and can look at what's coming out of that machine and talk about it and learn from it and say, oh, I see what's happening. It learned a shortcut here. That's a typical thing. It's trying to see if a patient has COVID on a chest x-ray exam. But what it's really seen is that the font

of the hospital that had more COVID patients is being recognized. And that's what it's predicting. And so I just think we don't know often what our programs are predicting. And having interpretable methods to be able to look at that can help us avoid those types of errors that can be real safety risks. But they can also help us have conversations with the clinicians about what's happening that can give them insight to help make changes into the algorithms. I think Rich Caruana is one of my favorite researchers. He's at Microsoft Research. And he spent two decades working on interpretability. Look him up. He has great stories of building tools that, if he would have really supports people and fits in the workflow.

So I think that's a prime space for our collaboration. Wonderful. As the founder of the Women in AMIA Committee and the Women in AMIA Leadership Program, I'm sure there are many changes that you are proud of. And so what are some of those changes that you have facilitated to make that possible? And are there any new changes you would suggest? What I'm most proud of with women in AMIA is professional organizations tend to be really top-down. In theory, they're member-driven organizations, and that's what AMIA aspires to be. But it's really hard to engage membership. And I think what the Women in AMIA committee did is it created this grassroots movement. And we got a bunch of women together

and started seeing, you know, where are the areas that we can work? And then they got more women to work in those areas. And now, now there's like dozens of women who are working together to do things where they see need. And I just love that bottom up, bringing the things that we care about to the table and then doing them. And our unofficial motto is women coming together and getting shit done. And I love that because that's what I'm most proud of. And I don't have any particular areas or changes that I think I would make just to keep it being this grassroots movement where lots of people are able to be involved. And thank you for doing that. A lot of us certainly benefited from what you have started. Our Leadership Awards, that committee has won.

And I also love the motto, women to getting together to get shit done. This motto has certainly set us up to keep pushing in a grassroots manner. And this podcast series is one of the products out of that as well. And now let's go back to that question, how we can collaborate between ACM and AMIA on health AI. As we know, the maturity of infrastructure has helped us to support large-scale computing now. And the availability of the pre-trend model has helped researchers and commercial organizations to quickly pivot their commercializable systems into something that can potentially be use to improve patient outcome. What's your take on that?

And what did you see as the most important opportunities here? That's a hard one because I think there hasn't been a lot of AI applied yet. But there's two areas where I think there's a lot of promise and can have huge impact. And one, you know, the image recognition to help developing countries. And so being able to have automated systems that can provide feedback about images to clinicians that don't have imaging training or where you don't have a bunch of radiologists to rely on, I think that can have really big impact in the world. And then the second area is identifying patients who have more risk and directing our resources to those patients. And so I just watched a panel yesterday with Claire Sullivan, who's in Queensland here in Australia. and she was talking about, we still do medicine like it's the 1950s, and every person goes into the doctor regularly and sees the doctor one-on-one, and what we really need to be doing,

we don't have enough resources to be doing that model anymore. We need to be identifying the patients who are at risk and putting more resources into them, and then supporting people in their health in a less resource-intensive way. So those are the areas, I think, where there's a lot of potential. And there are applications that are beginning to be put in place. But I would say that a lot of us think that AI is the solution to so many problems. And I think 9 out of 10 times, the problem you need that you're trying to solve doesn't require AI. And we want to go to the fancy thing. And we need to try the simple thing first. That's a very good point. And so

thinking about the past decade and particularly the past few years, it's been transformative for artificial intelligence or AI. And so not so much in terms of what we can do with the technology as far as what we're doing with it. So do you agree with the statement? If so, why do you think the view of society has changed that made AI possible in health enterprise strategy, but not just being seen as IT projects? It seems like in our lives, as we see different technologies in our life outside of healthcare, you know, on our phones and all around us, we see the power of it. And then we think, oh, we should be doing that. You know, why am I not getting that in my healthcare as well? And so I think that's really the power is seeing it around

us. There's so many things we can do. And we on the technical side get really excited about that. But we're doing very little applied artificial intelligence. And an example of that is there is a recent paper in the MIT Technology Review that there were like 400 imaging applications for COVID, like deep learning imaging applications and 250 decision support AI applications, and not a single one of them is being used. And Eric Topol makes that point a lot too. I think that that's the huge gap. Yeah, Dr. Topol talked about that in the AMIA 2021 keynote, and also the Health AI Real World Strategy Panel. He helped chair with us, also with

Dr. Suchu Saria, Dr. Mazia Ghazemi, Dr. Ziad Obermeyer, and Dr. Karan D. Singh. And the consensus now is that evaluation is really key to winning confidence and trust when applying codes to clinics. Is there any particular example that you have seen from the past or in your current role that you feel worth mentioning? Can you exemplify for us what would be the best practice of how people should do things here? Yeah, so this is kind of my passion here at the University of Melbourne. And that's that we build something and we test it out and we make sure it works well. And then we want to put it into and how it's going to fit into the financial system that we're in, who's going to pay for it. And we need to think about those things from the very beginning

so that we design what we're doing to potentially be successful. And so we're developing something called the Digital Health Validatron. And this is a way to take innovators through these steps of thinking through those questions from the very beginning, things like who's going to pay for it in order to go and try to get insurance to reimburse it or the federal government to pay for it, whoever you want to pay for it, you need to demonstrate things to them about the

comparators and the outcomes. But what we do is we build something and then we go say, oh, let's try to get this reimbursed. And here we did this evaluation. Look how well it does. And they say, that's not what I want to know. So we need to have them sitting with us from the very beginning, helping. And then we design what we do differently to make sure that we answer the right questions in the end. And so in the Validatron, we have a simulated digital ecosystem where we have like a simulated version of Epic, a simulated telehealth client, all the tools that you need to accomplish what you're trying to do with this new intervention in a simulated place so that you can start to build you're trying to do with this new intervention in a

simulated place so that you can start to build the connections for the data flow, the fire connections, the interoperability, work on the user interfaces, bring users to the table to test out all the small components and validate them as you go along. And then we have a physical space that's being built that will have a GP office, a hospital room, and a patient living room, and the one-way mirrors so that you can look in and do-designing and validating as you go, you're much more likely to be able to embed this into real practice and have it be adopted. This is a wonderful and very comprehensive framework. Now let's take a moment to think about scalability. Say that we establish the right workflow for one system, how do we then go beyond from one system to multiple systems and eventually across the nation?

How do we reinforce the same set of standards for health AI applications and evaluation? This would be more like a nutrition fact label for the foods perhaps um can you share some insights with us on what you think as a reasonable next step here yeah i think that's always going to be a hard problem no matter what we do because when you are trying to scale something or bring it to a new place it has all kinds of new workflows new cultural issues new data and technology systems. So there's always going to need to be customization. But if we are building on standards, then you have a foundation to be able to just now customize rather than having to start from scratch and do all the work over again. Did you see this more as a government's role to make regulations? Or is it something that might come more like a bottom-up approach that people start choosing their preferred standards and merge at some point?

I think that's a really hard question. But I want to quote Chuck Friedman. So he was at the Office of the National Coordinator. And we had him come give a talk at Utah on learning health systems, of course, and had dinner with him. And he said what his biggest lesson in government is that you can't do things top down, it doesn't work. And I think the same lesson with our women anemia committee, you can't just force people or say, here's what you're going to use. But at the same time, you need to support the grassroots movements to be able to be scalable. And I think that there's a place for legislation. So the 21st Century Cures Act, is that what it's called? And how you can't do

information blocking now, that legislation is going to have huge power to increase the use of standards and then create that connected journey for patients. And so legislation like that, not you have to use X and Y and Z, and here's what X and Y and Z need to look like. And you have to use what we built and only we can build it. Not that, but just saying you can't be blocking information and maybe a few other details and then supporting having government funding to support the bottom up groups to develop out the standards that get, you know, let the market forces choose which ones are working. And so, you know, smart on fire is becoming a really powerful standard. And it's got this kind of bottom up and top down working together. Thank you so much, Wendy. And coming to kind of the end of this particular episode, is there anything that the two strongest professional societies, AMIA and ACM, can partner up to help? Yeah. I always think that having one particular

thing that you work on together is the way to do it because there's just so many areas. But if we were to pick something like interpretability of AI or usability of AI systems in healthcare, some area like that and say, let's have this as a theme for the next five years. And having that focus can open up a lot of opportunities to make progress together that you wouldn't have if you didn't create that focus. So that would be my suggestion. Thank you so much, Wendy. Did you have any parting words for our audience here? I would say that I'm very sensitive when we're

admitting students into programs about their backgrounds. And we can be really quick to judge, oh, they don't have a background in medicine medicine or they don't have the computer science background. And they don't know exactly what they want out of their career. And there are people like me who have no idea what I wanted out of my career, had this very different background and are able to come and make a contribution. And so when I see people with passion that come from other fields like music or philosophy, and if they're smart and they're interesting and they're interested, that we should give them a chance.

And then on the other side of that, that all of us, when we come to these multidisciplinary collaborations, we're going to feel inadequate because we don't know all of the parts that are needed to know to complete the project. And so, but I want everyone to feel like what they bring to the table is needed.

All different types of expertise are needed and to feel confident in being part of the multidisciplinary field. Thank you for listening to today's episode. ACM Bycast is a production of the Association for Computing Machineries Practitioner Board. And AMIA's For Your Informatics is a production of Women in AMIA. To learn more about ACM, visit acm.org. And to learn more about AMIA, visit amia.org. For more information about this and other episodes,

please visit learning.acm.org slash b-y-t-e-c-a-s-t. And for AMIA's For Your Informatics podcast, visit the news tab on amia.org.