ACM ByteCast - Monica Bertagnolli - Episode 83

Episode Date: March 31, 2026

In this episode, part of a special collaboration between ACM ByteCast and the American Medical Informatics Association (AMIA)’s For Your Informatics podcast, Sabrina Hsueh and Li Zhou host Monica Bertagnolli, a surgical oncologist, physician-scientist, and President Elect of the National Academy of Medicine—the first woman to hold that position in NAM’s history. She previously served as the 17th Director of the National Institutes of Health and the 16th Director National Cancer Institute (NCI), as well as President of the American Society of Clinical Oncology. In the past, she was the Richard E. Wilson Professor of Surgery in surgical oncology at Harvard Medical School, a surgeon at Brigham and Women’s Hospital, and a member of the Gastrointestinal Cancer Treatment and Sarcoma Centers at Dana-Farber Cancer Institute. In the interview, Dr. Bertagnolli shares her unique journey from Princeton engineering to cancer surgery and national leadership. She emphasizes collaboration, system thinking, and bringing an engineering mindset of “pilot, test, scale, and continuously improve” to AI in healthcare. She highlights her role in founding mCODE, an initiative to improve patient care through oncological data interoperability, and how NAM's six core commitments and ten guiding principles for responsible AI address issues of bias and equity. Dr. Bertagnolli also offers insights on the growing erosion of trust in science and medicine—and how to restore it.

This episode is part of the special collaboration between the ACM Bycast and the AMIA for Your Informatics podcast series. ACM, the Association for Computing Machinery, is the world's largest educational and scientific computing society. Amia, the American Medical Informatics Association, is the world's largest medical informatics community. In this series, we feature leaders in AI in medicine. including researchers, practitioners, and innovators who are shaping the future of healthcare at the intersection of computing, artificial intelligence, informatics, and life sciences.

Our guests share their experiences in their interdisciplinary career paths, the lessons learned on AI in medicine, and their own visions for the future of computing. Hello, and welcome to the ACN-AIMA, and AMA joint podcast series. This joint podcast series aims to explore the intersection of AI and medicine with both the practitioners of AI and wealth builders and stakeholders in the healthcare systems.

I'm Dr. Sabrina Shea, AI Enablement and External Innovation Lead Advisor, representing the Association for Computing Machinery by Cass Series here. Co-hosting with me today is Dr. Lee Joe, Professor of Harvard Medical School, representing for your Informatics podcast series with the American Medical Informatics Association. We also represent the Women in AIMA community and the AIMA AI Discussion Forum, previously AI Evaluation Showcase and will be turned into an AI work group later.

This joint podcast series highlights women's leadership and career paths, pathways cross interdisciplinary fields and create space for collaboration. We are exploring how organizations like the National Academic Medicine, the Association for Computing Machinery, and the American Medical Informatics Association can work together to bring clinicians and technologists closer at the intersection of medicine and AI advancing responsible innovation that improves patient outcomes, while safeguarding safety and trust. So today, we are very honored to welcome Dr. Monica Padaganli.

She is a surgical oncologist, physician scientists, and also a visionary leader in biomedical research. Dr. Pardagin Lundi is the newly elected president of the National Academy of Medicine. She's also the first woman to hold this prestigious pose in its 54 years of history. Dr. Monica particularly, is also the first woman to be elected previously as the director of NCI and also the second woman to serve as NIH director. Her career is a testament to the power of bridging disciplines. But what makes her the perfect guess for our audience here today is her origin story. Dr. Pertaganoly studied her academia journey, not in a biology lab, but with an engineering degree from Princeton. We will hear more from her about her journey, how she served as the founding chair

of N-Code, and how she make those a center for cancer data interruptor, but he works. And now, Dr. Padaugoneli, welcome to the show. Oh, thank you so much for inviting me. I'm just thrilled to be here. Your career seems to be uniquely spent across engineering, surgical oncology, and national health leadership. And now you are still one of the most influential institution in medicine, NAM. So our audience here, including clinician, informatician, computer scientists alike, would like to learn more about your perspective in this promoter moments. So while you are stepping in this role, can you tell us a little bit about what has any

that drives you the most. Wow. So my approach, when I look back, has always been, try to tackle big questions. And in tackling big questions, one person alone can do very little. But there's so much talent in the world. There's so much talent in biomedicine. And I've always been so fortunate to be able to reach out and gather colleagues who have joined in trying to tackle these really big questions. And so that's another reason why I'm thrilled to be among your community here today, because, I mean, you said it perfectly when you said that what your groups are trying to do is to create space for collaboration. So if I think about what anything I've been able to accomplish in my career, it's always been about having that space for collaboration and doing

whatever I can to bring that kind of environment about. Yeah, and your career certainly highlight that intersection, how collaboration can work. But candidly, there must be something that kept you up at night that worries you, right? Even we have this good collaboration across our communities. Is there something that really worries you that even we collectively work as a community we won't be able to solve. Well, it's a terrible worry for everyone right now, and I think that is just loss of trust. The big loss of trust we are experiencing,

maybe it's not so new, maybe it's even been around for a long time, and we just didn't recognize it as much as we had, but it feels like it's much more severe and amplified than it has ever been before. because without trust, we can't accomplish anything in medicine. It's impossible. And that's what I think about at night. I think about how do we earn trust? You do not just, our trust is not bestowed upon you. It has to be earned constantly. How do we earn trust? And then that's not going to be enough. How do we also find ways to counteract what I think are some pretty toxic force? in our world right now that are fostering mistrust, that really aren't helping anyone. So two things. How do we earn our trust, but also how do we counteract these new forces

that are eroding trust in what science and biomedicine and research have to offer? Thank you, Dr. Bertagher. I love your story that you started with engineering and Christian and not exactly start with medicine, and when you look at your career from surgical ontology to leading the NIH to fund the M-code and now step in as president of the National Academy of Medicine, there's this clear thread of system thinking running through it. So for our listeners, including clinicians,

information, computer science, how has this engineering medicine shaped the way your approach, big, messy healthcare problems. I think you just mentioned entrust with the new forces from different angles. And how you take on this new rule, how is it influenced the way you think about where medicine and the National Academy of Medicine need to go next? Yeah, so big, messy, challenging problems. Engineering is a discipline designed to do exactly that.

And I guess, let me, a strategy that I think about, it's from engineering principles. Pilot and then scale. Pilot, test, and scale. You know, you start with tackling a problem. You have to build something. You can't just sit there and think about it. You do your best, you bring the best minds, you build something, and then you test it in small environments, and then you expand those environments. And the other huge principle that, is imbued in engineering is that nothing is ever perfect. You continue to improve and modify and make better for as long as you can until someday you then hit another inflection point that changes everything. And then you have to start all over. Think about the problem again from fundamental

principles and begin the process of pilot scale improve. And guess what? That's where we are with the application of AI in health and medicine. This is a big inflection point where that is asking all of us to think completely differently about how we approach our problems in health and medicine. So it's unsettling, but it is incredibly exciting. Yeah, indeed. We are so lucky to be here at this moment, Bill, especially those working in this intersection. We see those needle being moved. But in the meantime, we know that they are also biased that exists behind the thing of those AI. And you're having a passionate advocate in the past for bringing research and high-quality care to rural and underserved communities. So are you worried, about those bias that will be further introduced? And did you see National Academy, Mathematousine here to play the role in the future to tackle this persistent health inequities that might be further introduced by emerging technologies? I spent a lot of my career being a clinical

trialist and doing prospective randomized clinical trials, you know, which are supposed to eliminate bias to the fullest extent possible. But also being someone who realizes that that approach is very limited, and we have to adopt other ways. We have to bring data in from observational studies and case studies, and these, our community has always thought about the biases that those kinds of studies bring in. So bias has been, it's been a critical problem. The clinical research community has been wrestling with forever. Now, when you add these powerful tools from machine learning, We again have to think about bias much more deeply, but it's the same old issue. Do you have a very representative sample that will allow you to understand all of the conditions that are at play?

So the sample itself needs to have all the necessary conditions. And then the second part is you have to ask the right question. And we are learning this. We've known this forever from the world of trials. But if you don't ask the right question, you're going to get an answer that can be very biased. But now it's even more obvious than usual with these powerful new techniques that machine learning are bringing. So you mentioned there's general liability of AI and also how to make it fair and transparent. And you also mentioned that like,

there's no perfect, right? Solution in the beginning, you need a pilot test, employment, and refine it, right? So recently the National Academic Medicine, healthcare AI conduct, lay out six call commitment, and 10 guiding principles for responsible AI in healthcare and biomedical science.

If you had a name two or three commitments that are truly non-negotiable at this moment, which would they be? And where do you see the biggest tension right now between speed and safety, innovation and equity, transparency and competitive pressure, how should healthcare leaders navigate those treatoffs responsibly? You're only giving me two or three, huh? Okay.

Okay. I'll give you, okay, I'll pick my top three then, if you're forcing me into that little model. Number one, responsible. And let me just start. All three of the things I picked. Tell us why we can trust it. It's all about trust.

Responsible, accountable, and consulted. So we cannot do this alone. We cannot do this without transparency and accountability. And we cannot do it without a very, very clear-eyed view of the responsibility. ability we have to use it where health is in the equation to achieve the very best health for all people. So there's my three. Wonderful. And it's also a wrong thing that trust issues we talked about earlier.

Right. And that also bring us to our next topic. I don't know if you have read the recent MIT study that shows the AI productivity paradox. So what has gone into POC that is so successful with the gray model performance doesn't seem to be so useful in real-world scenarios. And this is not just for healthcare, but overall, people only see 5% of the even successful POC actually have yielded any positive outcomes in real-world scenarios. But let's just have that in mind when we talk about AI productivity paradise.

And now coming into the healthcare space, we have heard so much about the potential of AI can reduce cost and improve efficiency, for example, for the revenue lifecycle management. But there is certain that it is fear, the trust issue you talk about, that has hinder its usefulness, adoption, and also people are afraid it will increase the clinician runouts. So from your perspective, how did you see this AI productivity paradise evolving, in particular in AI for medicine, in the healthcare industry? We're at early days here, right? And so there's lots of confusion. What you've presented here is a mix of so many different things. And so as I'm listening to you, I'm thinking about, okay, what questions are we asking?

Remember, questions and data. Do we have the right questions? And can our data answer those questions? And so just what you've presented is a whole bunch of different questions. It's how do we make our clinicians' life easier and let them not get burnout? How do we have our clinicians trust in the information that algorithms are giving them? How do we think about the health care? I think you mentioned cost and effectiveness.

How do we figure out how to do the very, very best care? But do it in a very cost-efficient way because we don't have infinite dollars to spend for people's care. And then finally, the fundamental one that we have to start with at the very beginning. And all doctors will buy it, all clinicians. I have to remember. There's so many people who are critical to the clinical community who don't have an MD after their name, but have so much other tremendous expertise. all clinicians, their goal is to take the very best care of their patients.

And some of the burnout, frankly, is that that is their ultimate goal. And all the clinicians I know will just drive themselves into the ground, just making sure they're taking such good care of their patients. That is the incentive. You've got a whole bunch of questions there. And I think right now, and then we have to figure out how to, gather the data to answer those questions. I have only one solution for all of this. We really need to take not only an individual question level, but a complicated, complex,

system level to this problem. This is one of the reasons why I'm so excited about where we are today, because for 20, 26 years, 25 years, the National Academy's, have been talking about building a learning health system and how what a learning health system could do. And my definition of a learning health system is, and some of these words are really important, an integrated health system, integrated into all it needs to be, which harnesses the power of data and analytics to learn from every patient. and every clinician's practice, and then feed that knowledge of what works back to clinicians, health professionals, patients, and everybody else so that we make cycles of continuous improvement.

Now, if you think about having a system that focuses on that, it's about learning all the time. The goal is knowledge that works best for the clinicians that are just. delivering this care and health systems that are delivering this care. That to me is how we begin to then ask the right questions, gather the right data, and do this continual testing and improvement that we need to make progress for everyone. Yeah, wonderful. In Amia, we actually have a learning health system working group as well, that our AI working group when established will work very closely. This kind of learning health system is continuous learning.

But without being able to measure those human costs, we still might not be able to build those integrated system effectively and efficiently. As you mentioned earlier, we are looking at so many different things that AI can do, including reducing or losing clinician burnout. But here, how are we going to start introducing with these measures, we can start really accounting for these very complex landscape

what I'm looking at and bring them all together in a system method way to measure, to start with, and then to manage them. Yeah, I think use case by use case, as we build anything in a complex environment. So I'll give you one example that is pretty exciting right now. There is a group of clinicians right now that I actually have the privilege of working with They're at the University of Nebraska. They have, University of Nebraska and Nebraska medicine covers a huge part of the United States that has all these rural communities. And they're serving patients in rural communities.

And they are working on a project that is trying to make sure that patients with cancer get guidelines care. We know that guidelines care, not everybody gets guidelines care. and it's a complicated issue. The other thing we know is that the evidence that has been used to develop those guidelines, I've participated in developing those guidelines and getting that evidence, is not as good as we need it to be. Nobody is saying the guidelines are perfect. It's just this is the best we can do with the knowledge we have.

So this team is tackling both sides of that problem. They're tackling the, here's the guidelines. Some people don't get the guidelines for what. whatever reason, they don't have insurance or they choose to do something different, whatever the reason. And then the other side of the problem is, well, maybe the guidelines aren't good for everybody. It is often that a doctor or a nurse practitioner or somebody will be sitting in clinic and they'll say, well, this is what the guidelines say, but it doesn't work for you because you have this, that, and the other condition that won't make it possible for you to

follow what is a recommended guidelines. So what are they doing? They're gathering the data. And they're doing two things. They are building, working with a team to build an AI tool that will look at all the data on a patient and say, okay, here's what we know about your, here's what the guideline says. We should give to you. And then they're giving that information to the clinician and the patient together to discuss. So they're doing all that data gathering for the, for the clinician and the patient, synthesizing it, coming up with the guideline. But then this clinical team is very smart. that clinician and patients sit together and they look at the data again and then they come up with a decision and it might be in accordance with the guideline because the guidelines do make a lot of sense

and it might be something different and then that data will be captured and they'll also measure did this help did this help make the patient feel more more informed the clinician feel more informed. And then last of all, they're going to be doing some looking at how did this really change, going back and seeing what's happened. How did this really change what we see in how clinicians and patients have been making decision together? And then last of all, I take a deep breath here because it is the most critical. The team is going to follow what happened to those patients. How did they do? Did they survive their cancer better?

Did they have better quality of life? All of these kinds of issues. Now, remember I said, everything we need to do in health is pilot and scale. Will a team just, this team is over Nebraska Health, University of Nebraska. Can a team just there develop the data we need for all of us to really answer this question? We don't know. We think maybe not. But we think they will develop a model of working together that makes the lives better for the patients, the clinicians, and gets us better tracking results.

And then we could scale that everywhere. And then finally, the output of this in, I don't know, three years, 10 years, however long we make our learning cycle, goes back to the people who are making the guidelines. And the guidelines get better and better and more specific to patients. So this is just a little example of an effort that's going on right now that I think is going to do a lot of what you're talking about. It's not just about making the clinicians' lives better, meaning, oh, I don't have to do so much administrative work today. It's about making the clinician's life better because not only are they getting better, is it making it easier for them to look at data for patients and to follow the data for patients, it's also making it easier for them to take better care of their patients. and we need to make it part of this learning health system. I keep telling this group, I'm so thrilled to be working with them

because what they are doing is a learning health system method and approach. And that's the way they're thinking about it. So I think this is very exciting. And I know there are others. There are others out there that are doing similar things. And that is very exciting. What's my job at the National Academy of Medicine right now is to make sure that all out there who are doing these really,

exciting things can bring their knowledge to what we're doing because building a learning health system and using these exciting new tools is a grassroots effort. If we're going to make it work, it's got to go to the bottom line. And the bottom line that are fundamental, you know, you talk about in engineering about the fundamental units. In my view, the fundamental unit is the clinician and the patient in clinic together. And so anything we do to bring about a learning health system and use AI tools, the thing it has to serve primarily is that fundamental unit, in my opinion.

This is an exciting example. My next question about how we leverage AI, like AI-driven clinical trial matching and create a better real-world evidence to support, right to build a guideline to support those clinical trials. So from your perspective, how do we ensure that AI and rework data we collected actually reduce disparity in here rather than amplify them? What's your suggestions for the informatics and computer science community?

How we build a better, powerful, Goddrell, initiatives, and data status that are most critical. So if we want to modernize the hereselaneous systems or clinical trial ecosystems, that is not only efficient, but fair, transparent, representative of the population it means to serve. So, yeah, this is a critical question, and it's one where the informatics community has a huge role. Now you're getting to the data. What are the data? Everybody knows the classic horror stories about how if you gather data for one specific use, you try to repurpose it into something, you're going to get a bad, bad, and perhaps even very harmful increasing disparities answer by repurposing data.

And that's number one. Number two about the data is the data might be fine when you first got it trained something, But then things change. And now you're doing a different type of care based on, and then care changes. And so you can actually, things can actually happen and evolve over time. My point is there needs to be a continual challenging the data to do checks that allow you to identify when the data is not delivering what you need. And there are many system checks when we're at data scale. These are very hard to, you know, the clinicians can't do this.

There are system checks that you can do for representativeness, for missingness. You challenge the system. We've got a tool now that helps us understand what are the optimal treatments for, I'm just going to pick something, for low back pain in a patient. And we've trained it on sets that really do emphasize the capabilities, the patients in the set that we are applying it to, but, you know, things might change over time. And then you challenge that data set. Maybe someone, maybe with people who have low back pain, but they were found to have it was caused by a tumor. And that low back pain algorithm took them,

you know, took them to a place that didn't let their tumor get diagnosed very quickly. And what I'm doing here is giving an extreme example or a rare example, but it's those rare examples that we always have to be aware of. So we challenge the system with rare examples. We see what happens. Then last of all, how do we do this? These tools need to be developed with the maximum intellectual input from the broadest possible group of people.

So take this low back pain. If we only gave our low back pain algorithm to orthopedists who are treating back pain with surgery or to physical therapy programs or to, you're going to get one set of answers. But we may create algorithms that don't allow us to do what we would need to do for these exceptional cases. So we have to make sure we have cancer experts on the team to say, what about my patient? And bringing that in at the design phase of these kinds of, especially when it's decision support, is really critical. Yeah, this is definitely a model disciplinary.

Get people together and looking at the data is their data shift and also computer scientists. I see there's an args degradation from one data site to another to one scenario to another. So definitely, I totally agree with you. Your advice continues monitoring the AI solutions in healthcare. ACM Bycast and Amy for your informatics are available on Apple Podcasts, Google Podcasts, Spotify, Stitcher, and other major podcast platforms. If you are enjoying this episode, please subscribe and leave us a review on your favorite platform.

We also want to ask questions for our ASEAN audience who might just come into this feel new, right? They are not with health care background, therefore challenging the data with sociality, with expertise, might not be exactly their strong suits. But in essence, they still want to help, right? So this is the area collaboration work the best. How would you advise to further how with responsibly and safely integrating AI into healthcare? Something working with the clinicians is a starting point. Or are there other things you think, like as a professional society, of this very strong community that we can do together?

Yeah, it needs to be a very collaborative activity. I have seen experts in machine learning and data science come, into medicine completely from outside the field and seen them succeed tremendously, but here's how they did it. They said, okay, I'm just picking a random disease. Let's pick the disease diabetes and say, we're working on an application for some aspect of either diagnosing or taking care of people with diabetes or understand. And what the new informatics and data science team did is They didn't say, okay, I'll go get a big data set of all the patients with diabetes and start answering questions with it. Instead, they sat down with the clinicians in clinic, the nurses, the doctors, literally everybody.

The nurses, the doctors, the administrators who were making phone calls to find people. They sat down with the whole team and said, okay, how do you diagnose diabetes? What are the essential data elements that are required for diabetes? What kind of tests do you do? And they just kept asking questions that got them to the data that they would need to interrogate from fundamental principles. And what happened with that, it was wonderful when I've seen this happen because someone coming in with a brand new view was they were also uncovering a lot of ways that are thinking. as clinicians is already skewed and forced just by the way we've been learned how to take care of patients. Do you see what I mean? So there is a tremendous value to having people who understand

data and data science coming in completely new and taking a fresh look. And the only thing I'll say, the thing that makes it work is just they ask a tremendous number of questions. And it's been fun to see, you know, in these rooms and this is happening when they'll ask a question and the clinician will turn to them with this puzzled expression on their face, like, what do you mean? And then it'll dawn on them. Oh, my goodness, what a great question. So these are the kind of things that I think are exciting and fun, and I hope that your community understands. That's what they need to do. Big questioners, because that's where you're going to teach us what we need to learn to do better. I can totally vouch for this as a computer scientist to start with. That's why I do.

Being humble and being able to sit down with everyone to ask questions is definitely the most important step. Yeah. Yes, that's from the computer science, information, what they need to do. And then from the point of view of clinicians, right? Nowadays, where you're seeing, like, Ambien documentation tools, drafting clinical nodes during, provide a patient encounters in real time, and also there's AI systems like supporting clinical decision making. So there's also early forms of agentic AI that carry out multiple-step tasks for our clinicians. So they are coming and they are entering clinical practice now. We talk about building those AI health care folks workforce and also the next generation.

So what does that mean in this reality? How do we train our clinicians and next generation clinicians to work thoughtfully and safely alongside those AI technologies? What are your thoughts on? Every clinician today needs to understand that these tools are powerful and can be so incredibly beneficial to our patients and to society overall. So that's number one. Number two, they won't get there unless the clinicians are participating and actively engaged

in developing as well as using these new tools. Also critical. I mean, this has been true of all technology, but even more so now today. When you get a busy, busy clinician who has, you know, 30 patients to see in an afternoon and they don't want to have to do. And anything that slows them down, anything that makes their day harder is not something they're going to be interested in. So the development of this has to be placed within a framework that allows them the space

to be able to innovate and contribute. We've got to take that into consideration. Ultimate design of anything also has to make their life better in the clinic. And there are two things that will motivate clinicians, doing better for their patients and being able to deliver much more effective and efficient and cost-effective care. So anything that can be done that lets the clinician do what they do best, what they have been trained to do, is something that clinicians are going to want to embrace and collaborate with. So what I'm saying here is that training, even our training of the next generation,

needs to make sure that every new clinician knows that being part of this process is what we do. It's what every single clinician needs to do throughout their entire. career. Last I'll say, as doctors were used to knowing that, you know, I've got to update my knowledge constantly. I've got to update my knowledge. Well, part of updating knowledge and part of making medicine is also being a part of being in a learning health system and using these new tools in the way that their contributions are essential. Yeah, participate in college, test out and hopefully the AI too will improve their workflow and make our patient safer. Absolutely. And you know, I'm a surgical oncologist. I did not invent a CT scan or an MRI. I did not

read them. Expert radiologists read those for me. Now maybe expert AI tools are going to, will read them for me. I had to understand how to use those tools to take care of my patient. It's a different tool. What's new now is that it's so immediate and so all-encompassing of everything that we do that our clinical feedback needs to constantly inform these tools. Okay, so you're asking me to go one other place that I want to go with your team here, with your listeners here. We get an MRI scan. Okay, now the MRI scan's working, and we kind of, we know what the MRI scan does and we use it reliably and we know what its problems are. and we just go right ahead. When we start doing these methods, they constantly change and evolve as we go, right? I don't use a CT scanner that changes every, we might get an upgraded new fancier tool every 10 years or something. You know what I mean? But it's not something that literally the learning cycle is as incredibly rapid as I believe we will start to see with,

these new approaches. So that's yet another thing that this feedback loop that we're talking about in the learning health system, that feedback loop is something that has to be done very carefully with the clinicians very much a part of it. And so that's another thing that wrapping our heads around and training the next generation to think about is how that feedback loop is safely and responsibly implemented because it's faster than ever before. And no surprise, anybody who's practicing medicine knows right now that just keeping up with the rapid pace of volume is something that very soon we're going to need new tools to help us do. We can't just read every journal article right now and be absolutely sure we're taking the best care of patients. So we need learning

health systems powered by AI. It's very powerful. And I think it resonated a lot with our own discussion. Last year in November, we have this annual symposium, and we hosted this Detroit workforce panel with people in the field doing AI in medicine. And seeing one thing that clearly comes out on that is that AI is a powerful assistance. It's not replacement. You will change our job discretion.

You will not replace our job. But we all need to learn how to better use it. And also the system designer need to make sure this continuous learning feedback loop is in place. So we can continue using those measures we put in the workflow to help us to get better. I couldn't agree more. And I will say this. It is impossible to me to think that we would have, let's say, a decision-sum. support tool that would say, just like I describe about the guidelines with a patient, because there's this

other part here, and that is the patient and a patient sitting down with a clinician, that human contact, to discuss what are the needs and desires of the patient always have to be part of this equation, and we know this as clinicians, it doesn't matter what I think is the very, very, very best science in the world. What matters is me sitting down with the patient explaining that, and then the patient deciding what they want for them. A.I will aid that because it will help, I think it will help us all understand maybe what the best options are, but it is never going to replace that human interaction because these are life and death decisions. I couldn't agree more. And then finally, we're all patients and I can't imagine. And we all know that the last thing we want

to do is do our own medical care, even though as smart as we think we are. So we all need to have those interactions to know that we are getting what we need as human beings. But one thing that also comes out on the discussion, if I may ask, is that people are quite afraid of de-scaling. Because AI as getting introduced into a workflow, people are afraid they may start forgetting how to do their original job, right? Certainly not a part about interacting with patients, but if you realize on technology, sometimes too much you do forget how to drive, how to sing, I guess. That's the kind of general fear. What was going to say about that?

Did you think it would change our way to train our own clinicians in the future? This fear has been around forever. Okay, I remember when I remember when I was in training, it was critically important. Believe it or not, because I've been around for a long time. It was absolutely important that I, if I wanted to diagnose a pneumonia, I got a specimen from the patient and I did my own gram stain to see if it was a gram positive or a gram negative organism and whether it was a bacillus or a coxite. Do I remember how to do that?

No. I mean, is that a skill that our clinicians, it's kind of nice to know what happens about it, but is that a skill that I would trust myself to do any more? No. I mean, it's just an example that illustrates that this is the evolution of technology and of anything. And no, I don't know how to do a gram stain anymore, but I hope I would know that how I need the information that it can give me to be able to guide care. So we're going to see the same thing with what we find with these powerful new tools. We're at a time, though, where if we don't build them together, they're not going to do.

what we need them to do. Thank you. Thank you. Now, that's a switch gear a little bit to talk about our women leadership experience and perspectives. As you know, that in AIMIA, in ACN, we constantly survey our own members to see how the women leadership comes about. In AIMIA, we have this woman in AIMIA community that surveyed the leadership position. Right now, we are seeing 28% of they make and 70% of a clinical leaders are women, right? And in the NIH, you lead research project grantees. It's now currently, I think, approximately 40% female to 60% male.

So what did you think that's happening here, considering that more than 50% of the workforce in healthcare, a woman? What happened here doesn't seem to be proportional to the potential. What do you think we can do here? And is there any fundamental issue here we can hold together? So this is a very, very important issue. And it's, and I'll broaden it. I'll broaden it beyond women. We need the best minds. We need the very best minds. But medicine is not, the other thing about it is medicine is not just about your intellectual

capabilities. It's also about who you are as a human being, what your perspectives are, how you think about society, how you think about people. And so, you know, it's that more human aspect of medicine and it's also key. So bringing understanding and including in our work, both of those key values is incredibly important. So that, you know, there's always an issue of any time you're trying to do anything important, looking under the lamp post for your keys, you know, who's ever, is ever around and they are the ones you're going to focus on. And it's harder to do a big search for who has, who can bring these special capabilities to what you're trying to do. but we have to do that. We have to continually ask, who's the best minds? The best minds and the best

hearts. How do we get them here? That's number one. Number two, I also think especially about medicine is if we're going to succeed, it cannot be medicine on the two coasts of the United States. It cannot be medicine at the big academic medical centers. It has to be about little primary care communities or public health nurses that are in the schools. And it's got to be literally everywhere. And the kind of brilliance and contribution it takes to be a public health nurse. My mom was a public health nurse. The type of contribution it takes to be a public health nurse and the knowledge you need and the skills you need are very different and critical, that also need to be included to what we're doing.

So as you can see, I take a very, very, very broad view of what it takes to deliver health to people. I think that also helps think about where the best minds are, who can do the best caring. So, yes, I'd like to see more women in all of these roles, and we want them there because they're bringing what we need. Yeah, we want to empower the best mind in the right place. Yeah.

And then I love people who are agitators who stand up and say, hey, wait a minute. I have something to offer. Don't sit quietly and say, well, nobody lets me in. Well, make a big noise. Yep. Because sometimes that's what's needed to pay attention to people who really deserve to be there. Yeah. And I'm sorry.

In health, everybody does. deserves to be there. That's how broad health needs to be. Exactly. Yeah, we have a lot of transportation and nurse informatics in the audience here. They will resonate a lot with what you share as well. Thank you. Totally great. Should be very inclusive and also collaborative. And talk about collaboration. I want to hear your vision and thoughts about how different organizations like Amya,

National DEM and the ACM can work together the emerging AI tools and now has been employment in their practice, clinical practice, and the research areas. How we work together to make it more cost-effectively, how efficiently are more safer, transparent, and all those different aspects. How we work together. I'm going to go back to the learning health system approach because now you're back to my engineering routes. Pilot scale, pilot scale. We want to see the work that we are doing, which your teams have the firepower we need to make this all work. We need to have you integrated

into an overall system that allows us to get the ultimate benefit, which is the very, very best health care for everyone with all the constraints that we have in our world that need to be optimized for where the person lives, what their overall health state is, what kind of resources can be delivered because resources are not infinite, and the incredible diversity we have. So how do we integrate your community into that learning system that can address these issues and then provide the effective feedback loops that allow the learning to continue. Let me say, it's very sad to me

when I see wonderful ideas, tools created, and then it just sits there. It doesn't happen. It doesn't go anywhere. Maybe it just didn't meet its initial design criteria. And why did it just stop dead? Because there wasn't that feedback loop that would allow it then to be improved

and integrated and move on. So I'd like, how do we get rid of it? I'd like to see many, many more people working in partnerships and initiatives that have this continual learning, long-term view. That's something, and again, that is something that I think the National Academy of Medicine is working really hard to bring about and to bring that about, have that be something that we can make a commitment to as a society, that this is what we're going to need to improve care. And there is no bigger time than now because now we have the kind of capabilities that your community can bring to us, which is incredibly special. Exactly.

Yeah. And that's why we earlier have this AI showcase, as we might have introduced you earlier. So we curated all this example to see how people continue making this happen, starting from systematic evaluation, and then going into the workflow to measure things that do not usually can measure on the human side, and then being able to link it to the longer-term outcome and see how this can become learning loop by cell later with the processing measure to give early indicators. so how AI will work or will not work down the road. So, Marika, so thank you so much for the great insight today. We in the audience have a lot of younger generation, early stage scientists and practitioners. So did you have any parting words for them when they are navigating this new area of AI in medicine? Any advice you will give to those youngsters?

Oh, yes. So your contributions are so critically important. And you are at an incredibly exciting time because the world is changing and it's changing more rapidly than it ever has before. I have two pieces of advice for you. First of all, I hope that you will follow your passion for doing this kind of work. everyone always talks about that advice is something really essential to being a fulfilled human being. But the other is don't do it alone. Look for others. Reach out as much as you possibly can to interact with everyone who can help you because none of us ever, ever accomplish anything alone.

And the ones that really go far and really succeed are the ones who are not afraid to ask questions, not afraid to put their ideas forward, and definitely have the humility to not think, the humility and the courage to not do it alone. It takes both. You have to be brave sometimes to say, hey, I need your help. And I say that at least 20 times a day. So that's the way you will also be able to have a great career and make an incredible contributions. Thanks you for this great suggestions and advice for the next generation.

We are so honored to have you. Yeah. Well, I'll just say again, I'm very excited and incredibly privileged to now be moving on in July to take over. as president of the National Academy of Medicine, I will say that what we've been talking about here today is what that organization was created for 60 years ago. And it is to bring together evidence, the evidence that allows us to build a world where everyone has better health and long and healthy lives. It is also committed to science and research that are going to tell us how to be better. Science and evidence are the North Star of that organization. So everything we've been

talking about today are what NAM aims to promote. And then finally, it's an organization that's made up of the world's most prominent leaders in science, medicine, and very important health policy. How do we also try to promote needed actions? So I encourage everyone to look at the activities of the National Academy to get involved with them. Our members are very actively involved. They make a commitment to public service by getting involved in the work of the academies. And we very much hope that your community will be prominently represented and very active contributor. Yeah, we're very much looking forward to the future collaboration opportunities.

Please do let us know our audience is eagerly awaiting those opportunities. And we will put the National Academy Medicine website of that in our share resources. Thank you. Thank you very much. There's two documents there on artificial intelligence and health and medicine that I hope can be of assistance to you. One of them is on just the Code of Conduct that the National Academy has put on. and the other one specifically addresses generative AI in health and medicine. And if you have the reference pointer for the University of Nebraska, we would love to

get that. Well, they're just in design phase right now. They're just getting started. We'll stay tuned as we see them. But it is really just lovely to see this group coming together. And I know they're going to have an amazing impact. Absolutely. Thank you so much.

ACM Bycast is a production of ACM's practitioners' 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 and visit the community of women in AMIA.