Lucila Ohno-Machado: AI and the Art of Medicine

Harlan Krumholz: Welcome to Health & Veritas. I’m Harlan Krumholz.

Howard Forman: And I’m Howard Forman. We’re physicians and professors at Yale University. We’re trying to get closer to the truth about health and healthcare. We’re excited to welcome Dr. Lucila Ohno-Machado today. But first we always check in on current or hot topics, and I know you and I wanted to talk about someone special to us today, Harlan, so kick it off.

Harlan Krumholz: Yeah. We had a close friend of ours, Irwin Birnbaum, who died recently at the age of 88 after a long and rich life. But I wanted to talk about him a little bit because he exemplified the mentor, and I was thinking about Irwin. I became aware of Irwin, had met Irwin when he became the chief operating officer at Yale when Dean David Kessler had been appointed and he appointed Irwin and Ruth Katz and another friend of ours in leadership positions.

And I didn’t know Irwin so well as a chief operating officer, but when he retired, I invited him to join us at the Robert Wood Johnson Clinical Scholars Program, a program that helped train physicians in areas of statistics and epidemiology, health policy, but also leadership. And Irwin, of course, had had such a wide-ranging career being at Montefiore and then at Yale and actually practiced… a law practice around health policy that he seemed like a natural to bring in and be able to be help and supportive in. And he came in.

I wanted to read the words that Leslie Curry wrote to us in an email. “Irwin was a fiercely optimistic soul, everyone’s cheerleader. Imagine his beautiful fingerprints left on so many.” And I raise it because oftentimes in academia, people are counting the number of papers or grants or how big their groups are, or they’re valuing impact in a lot of different ways. But some of the best ways to make impact and how you’re among the best at this is on person-to-person relationships and the way in which you can influence people’s lives for the better because you actually care deeply about them and are just a mensch, just a good person who seeks to help others through what they’re going through and to make them better. And Irwin was all of that. And I don’t know, I just saw as a moment to pause about people who make impact on others in their lives and what’s a life well lived?

Howard Forman: I really appreciate it because almost the opposite of your experience, Irwin was that person to me. I met him immediately upon arriving at Yale for both good and bad reasons involving my administrative role in radiology. So I got to know him immediately. My funniest anecdote with Irwin, however, is that as soon as I met him, I somehow had to tell him that he graduated with my mother from college. And Irwin was about four years older than I am now, but to my 31-year-old self, Irwin was a much older man who would not be offended by being told that he was my mother’s age and he had never let me live that down every single time we met, right up until if we had a meeting a year ago. He always told people about that story. He always made me smile a little bit about that. He was a kind man. He helped me through some very stressful times early in my career. You did too, Harlan, by the way. Few people kept me at Yale during the ’96 to 2001 period when my job was not fun, and Irwin was one of them and he and I stayed in touch. He attended some of my classes during COVID by Zoom, and I love that man. And everything you said about him is true and more, and I was happy to see how many people have expressed similar opinions on social media over the last few days.

Harlan Krumholz: Yeah, it’s interesting. Someone like that just by his nature could enrich so many other people’s lives. Also in this week, by the way, Howie, there was a perspective piece in the New England Journal of Medicine called “What Is a Mentor?” by Suzanne Koven. And I thought it was interesting. She talks about an experience that she had with going through a period and she sought out a dean of students that she had had, and she just talks about how she wasn’t sure whether that person would still be open to being a mentor to her. And the person said, in her piece, “Nonsense,” that Dean said, “come on over.” And she says, “What he said to me during our brief conversation that day changed my life.” And the power of those moments, just singular moments where you can play such a key and critical role. I just wanted to read at the end of her piece, ’cause I thought it was nice.

She said, “A good mentor makes you feel the way I felt leaving the office of my old dean nearly 40 years ago, crossing the street from medical school back to the hospital, more grounded than before we’d spoken and also lighter than air.” And again, Howie, I know you do this for so many folks, it just is probably nothing better in what opportunities that we have to be able to do that. But I just wanted to remember Irwin today, reflect on that. He did that for so many people, did that for all of us and the importance of mentorship generally. Anyway, just wanted to highlight that.

Howard Forman: I really appreciate it.

Harlan Krumholz: Okay. Hey, let’s get onto our interview.

Howard Forman: Dr. Lucila Ohno-Machado is the current deputy dean for Biomedical Informatics and Data Science at Yale Medicine as well as the Waldemar von Zedtwitz Professor of Medicine and Biomedical Informatics and Data Science, chairing the section of Biomedical Informatics and Data Science at Yale. Dr. Ohno-Machado’s interests lie at the intersection of big data and life sciences, and she works predominantly on modeling and data sharing.

Before coming to Yale, she helped found the UCSD Health Department of Biomedical Informatics and the Halıcıoğlu Data Science Institute at UC San Diego and worked for Harvard and MIT in teaching roles including as Biomedical Informatics distinguished chair at the Brigham and Women’s Hospital in Boston. She’s an elected member of the National Academy of Medicine, among many, many other accolades. She’s one of the most influential biomedical informaticists in the world today. She completed her medical degree at the University of São Paulo in Brazil before going on to obtain her MBA from the São Paulo School of Business Administration.

She later completed her PhD in medical information sciences and computer science at Stanford University. So first I just, on behalf of Harlan and myself, just want to welcome you to the Health & Veritas podcast. And I wanted to start off, I went back and looked at when did you get interested in this, thinking that, like a lot of people, this came to you later in your career somehow, and I discovered that you were on top of this topic from the first paper you wrote when you were basically in college. How did you become so prescient to know that this is going to be this enormous area of scholarship and what drew you to that?

Lucila Ohno-Machado: Yeah, so first thanks for the invite to participate. In our education system in Brazil, you have to select whether you go to medical school or computer science or engineering and so on. And I wanted both. I wanted both computer science and medicine. I decided to do medicine first because I was better prepared for the exams at that time. And once I completed medicine, then I started computer science, and administration was one of the venues to get more exposure to informatics and IT and so on. So I did it in Brazil, but when the opportunities were exhausted there, I sought graduate school overseas, Europe, Asia, and North America, and I was lucky to get into Stanford for the degree.

Howard Forman: Looking through your publications over the last almost 35 years, you were hitting on key topics, 10, sometimes even 20 years before people would talk about them in popular press, natural language processing, artificial intelligence, machine learning. You were working on this really early on, and you were recruited here to help us organize our efforts in this regard in data science and biomedical informatics.

But one of the things that strikes me about Yale is that it is there’s a lot of independence, a lot of investigators working in these areas all over the campus, not even just in the medical school. Harlan and I were talking recently about a center that exists on the main campus that neither he nor I was aware of at that moment. And I’m just wondering, when you landed a place like this, how do you go about accounting and meeting, listening and learning about an institution like this, which is by definition almost different than any other such institution?

Lucila Ohno-Machado: I think I was lucky too, that my coming was very well advertised. So whoever out there knew data science in biomedicine initiative was coming and they were interested. They did come to me rather than having to survey a whole lot. However, I do feel, and I learned actually this February that not everyone came. And when we set up an A.I. symposium for the School of Medicine, we found so many other people doing this that I had not heard from before. We had 600 registered people, we had 41 posters and 16 presentations, ’cause that’s all we could fit.

So even though we did have a lot of people coming to us and getting secondary appointments with us and so on, there was much more out there. And this is just very exciting as well as hard to coordinate. So I see our function mostly of trying to coordinate, matchmake, and create the infrastructure so that all these groups don’t have to build their own everything from scratch and try to use, at least to some extent, an infrastructure that can support initial exploration of models and make the data more available.

Harlan Krumholz: You’re at the nexus of all things exciting that are happening in medicine right now, data science, informatics, the technology adoption, it’s changing so quickly. I think you probably would be in a good position to say, “What do you think it’s going to look like in 10 years?” There’s lots of speculation about change. Change always occurs slower than you might think in medicine. What do you expect to see in a decade?

Lucila Ohno-Machado: I expect more time for humans to be humans and less time on documentation, less time on mechanical things or insurance approvals and other things that are currently taking a lot of time from clinicians and other personnel. So I think if we can automate the more mechanical aspects of things, of documentation and so on, I think we’ll leave more time for actually listening to the patients looking at them.

Harlan Krumholz: Well, just as a follow-up thought about this in the last decade, we’ve had so many advances in technology information. Often when I start my talks I say, “Medicine’s emerging as an information science.” We’ve never had the ability to be able to access information relevant to our patients in such a streamlined fashion to be able to be in control of what is known about a particular patient or condition and be able to get access to that.

I grew up when we were looking at Index Medicus, these large tomes in the library, and you would flip through trying to figure out what had been published, and it was a big deal to go there, lug it around, you’d look into it. Nowadays, anybody can go to a computer. Our patients can go to computers and get access to some information. Juxtapose that availability with our actual outcomes that we’re achieving in healthcare, if anything getting worse, life expectancy is dropping, comorbidity is increasing. On every front we’re losing ground in the United States and actually worldwide. How do you square that? Because it seems like we can do more, but we’re achieving less.

Lucila Ohno-Machado: Mm-hmm. I can tell you it’s not because of technology. In a way I think it could be worse if technology weren’t there, but I do agree that we can do much better, not only in the U.S. but worldwide, because the discoveries, they’re amazing. They have the possibility of extend the life and better quality of life and so on. But somehow the structures that support that delivery of the new discoveries or making it accessible for everyone, the structures are not there. So access to care, access to social services that are needed and so on, that seems to be declining, even recognizing the value of these other disciplines is something we need to work on, I think. So it doesn’t matter. We know how the patterns are and so on. Again, the therapist cannot be delivered.

Howard Forman: One of the things that has changed most in the health system in the last few years has been a greater integration between the medical school and Yale New Haven Health System. And very, very recently, as you know, we’ve named a new president, we’ve named a new chief operating officer. A lot of changes have gone on, including our deputy dean, who we just interviewed on the podcast a few weeks ago, Peg McGovern moving into a physician leadership role for both the medical school and the health system.

And it raises for me the question of, “Where is the governance of clinical informatics going?” and “How much you’re clearly doing amazing work on the research side, on the data side, the access, the ethics?” and all that. “How much are you currently doing and how much will you be doing in helping the health system realize the value of data science, A.I.?” Everything like that?

Lucila Ohno-Machado: Yeah. I’ll start by saying the distinction between clinical informatics and bioinformatics of computational biology is increasingly less, because of, for example, precision medicine, where we want the right therapy delivered to the right patient to personalize treatments and prevention as well. Now, it’s highly interdependent. If we want to be impactful in our region, in our own institution, we need to operate with data from that institution. We have the good fortune to have an electronic health record system that is jointly owned by the university or the medical school and the health system, thus offering the opportunity to develop joint governance in joint responsible data sharing in a way that research can be done in also the products of this research models, A.I. models, and so on, implemented in the health system and measured, because that’s an important aspect of it.

Do we know that these A.I. models are doing what they’re supposed to do? Do we know that they’re doing well for the whole population, or are there underrepresented populations that are actually not being benefited by these models and so on. So those are all important considerations that the marriage of research and the clinical practice becomes so important. So I would say this alignment between the two units is very critical, and it was one of the reasons to come. It’s a very large system with a highly diverse population in which we can discover so many things and apply back. So I would say I see it, the whole informatics data science in medical A.I. to be belonging to both areas, some with more implementation and evaluation of outcomes and the others with the data science and the discoveries that need to be made.

Harlan Krumholz: I know we’re coming to the end, but I have one question around something I’ve been pondering for a while, and I’d be curious what you think as these capabilities increase in information science and the tools to support clinical decision-making grow. What are the consequences for this? For the competencies that we’re seeking in clinicians? I say this as, for example, in my field, cardiology, we’re on the cusp of transforming the competency examinations, what it means to be board-certified, what we expect of people historically, we’ve asked people to sit for all-day tests. We’ve tested their memory and their problem-solving abilities on conventional multiple-choice testing, the pattern recognition.

In many ways, the information science revolution is going to augment the ability of clinicians to do this without the need to having rote memorization and pattern recognition, in fact exceed our abilities at pattern recognition, largely. So do you have thoughts about, as we assess clinicians of the future for their competency for whether or not they’re qualified to be doctors, how is the capability of the information revolution influencing what you’re thinking about, what’s going to be necessary and how should we be assessing performance in that era?

Lucila Ohno-Machado: Oh, I’ll give a personal opinion of that, that as you said, memorization and ability to arrive at the correct diagnosis, given what questions you asked and what the patient tells, will increasingly be automated, no doubt about that. However, the ability to communicate back and to explain to the patient and understand the circumstances of, “Why… even though this is the most efficacious treatment, you’re not in a position to do that because you have this other circumstances.” I think that will all get better. When they say “the art of medicine” and “the empathy” and so on, I would say that might be more valued than the memorization because that will be done by the computer on the side. That’s my prediction.

Harlan Krumholz: You do know that the recent studies have suggested that the artificial intelligence scores higher on empathy and compassion in interactions with patients. So how do you square that with what you’ve just said? Because it may be more important for docs, but I’m having trouble understanding how, when it’s blinded doctors interacting with patients, computers interacting with patients, and now a blinded evaluation of the empathy and responsiveness, the machine sometimes scores higher.

Lucila Ohno-Machado: Well, but remember, this is text.

Harlan Krumholz: Yep.

Lucila Ohno-Machado: It’s not communicating as we are communicating and looking and body language and a whole lot of things. When those comparisons are done in a very controlled manner, I will bet you that the doctors would define, or certain doctors would do—

Howard Forman: That’s right.

Lucila Ohno-Machado: ... better and others—

Howard Forman: That’s, exactly.

Lucila Ohno-Machado: ... not be doing that.

Howard Forman: Yeah. One last question from me. I’m fortunate enough to work a lot with tech transfer at the university to see what we’re doing university-wide in terms of commercialization of new products. And there’s no question that we’re seeing an explosion of innovation in this space. And I’m wondering how do you personally, in your role, foster that innovation so that we both recruit innovators as well as help our current faculty become innovators in this area?

Lucila Ohno-Machado: Mm-hmm. Yeah. Well, I think I’ve seen innovators with all characteristics, but some of them don’t have patience. And those without patience don’t fare as well in academia because we go by grants that take months to be evaluated and so on. And it takes a lot of persistence and so on. Whereas industry can be more agile in many ways.

So I think what we try to do is not prevent people from having ideas and trying to take them out there, but we try to do it in a very systematic way so that when you do have a product, you know it works and you know it’s good enough to be out there. So it is hard because of the funding mechanisms in academia are not as plentiful, I would say, and fast as outside. On the other hand, rest assured that when a product comes out of academia, it probably has been way more vetted and verified than when it doesn’t.

Harlan Krumholz: So I just want to say thank you so much for taking the time with us today. Like I said, we’re so lucky you’ve come. It has been too long before the Yale School of Medicine has invested in a department. You’re building a department, you’ve recruited extraordinary individuals, you’ve built relationships across campus, and already there’s a greater vibrancy to the efforts around informatics and technologic advancements than I’ve seen before. And we look forward to seeing what happens next. So thanks so much.

Howard Forman: Thank you so much.

Lucila Ohno-Machado: Thank you all.

Harlan Krumholz: Well, that was terrific, Howie. It’s nice to hear Lucila. She’s doing so many different things. I can tell you there are challenges here bringing change, but she’s doing a great job. So let’s get to my other favorite part of this podcast, hearing what you’ve got to say this week.

Howard Forman: Yeah. So back in April of last year, episode 78, we talked about vaping and lung injury, if you recall. And e-cigarettes remain controversial due to the harms and particularly the explosion of use among young people, even though it’s down from its peak, as we discussed. But the one-use case that has been promoted as uniformly good as in helping cigarette smokers quit. And there has been some evidence published in this regard, but this week a large, randomized trial has been published that addresses the safety and effectiveness of this approach in comparison with standard of care smoking cessation efforts.

So in this study, a large, randomized control trial, there was greater smoking cessation in the e-cigarette group. In other words, 60% of those given e-cigarettes versus 40% of the control group quit smoking cigarettes, tobacco cigarettes, and this is obviously good, but only 20% of the e-cigarette group quit completely. Whereas 34% of the control group had quit nicotine completely. So in other words, e-cigarettes are good at helping to quit actual smoking… not as good at helping to quit nicotine.

Perhaps one could argue that they enable longer-term use of nicotine in those who might’ve otherwise wanted to just quit smoking. As to safety and health, more individuals in the intervention group, that being the e-cigarette group, reported adverse events at a statistically significant level. But serious adverse events were roughly equal in the two groups. For reasons that I can’t explain, and the authors don’t discuss, COVID occurred more commonly in the e-cigarette group, and the e-cigarette group also reported more episodes of antibiotic use.

There was some slight improvement in respiratory symptoms in the e-cigarette group otherwise. So e-cigarettes in general in most ways are better, better, better. There are a good number of limitations to this study acknowledged by the authors, and this is true of all studies, but it does give us more information of the value of e-cigarettes in a clinical scenario and a clinical setup. And it’s a little mixed. It helps to quit cigarette use, which is great, but might be supporting the ongoing use of nicotine in some.

And while vaping may be seen as just another way to smoke, it does seem, from everything we know, unquestionably a reasonable tool for caregivers to consider when helping their patients quit smoking, and this is not uniformly accepted outside. So one thing I just wanted to raise to you, Harlan, it’s amazing to me how long it’s taken to do this particular study and even others like it. And I wonder, we let so much research get done from an investigator-driven level. Is there a role for more government-instituted research for fostering this type of research as opposed to letting investigators decide to do this research?

Harlan Krumholz: Well, I think there could be a prioritization of research that aligns with needs regarding policy. I’ve felt this for a long time. That can happen because there are organic changes in society that we actually have no idea what impact they’re going to have on health to recommendations that have to be made in the moment when there’s inadequate evidence. But we really need to continue to accumulate evidence in order to help us course correct as necessary. COVID pandemic, I think, indicated that we were doing things, but we needed to be investing in, “Hey, what is the value of this strategy, that strategy?” as opposed to ad hoc bespoke approaches to doing research. There should be ways for us to be able to do that. In industry, of course, with A/B testing, every time they roll out something new, they’ve got something in the background that’s evaluating things and determining what its effectiveness is. I think in the public sphere, we need to be able to do this, but you’re right to point this out, Howie, these are things that we’re exposing people without knowledge of actually what the impact is. And there needs to be much more information about this.

Howard Forman: And this is something that starts off as like a commercial product without too much oversight at all from government or regulatory authorities. And it really needed this type of imprimatur of good peer-reviewed research. And we have it now, but it’s taken a long time, and I think we need to see this incorporated into practice.

Harlan Krumholz: Exactly. Exactly. Yeah. No, I’m glad you brought it up today. You’ve been listening to Health & Veritas with Harlan Krumholz and Howie Forman.

Howard Forman: So how did we do? To give us your feedback to keep the conversation going, you can find me on Threads @T-H-E, number four, M-A-N, that’s @the4man.

Harlan Krumholz: And I promise very soon I’m going to have a social media strategy and I’m going to unveil it on this show.

Howard Forman: That’s right.

Harlan Krumholz: But yeah, right now you can still find me on X @H-M-K-Y-A-L-E. But I got to figure this out.

Howard Forman: And for those of you on LinkedIn, we both are very active onLinkedIn. We read every one of your comments, so let us know. Ask questions, we’ll respond to them here. But you can also email us questions, comments, reviews at health.veritas@yale.edu. Aside from social media and our podcast, I’m fortunate to be the faculty director of the healthcare track and founder of the MBA for Executives program at the Yale School of Management. Feel free to reach out via email for more information on our innovative programs or check out our website at som.yale.edu/emba. If you like the podcast or even if you don’t, please rate and review us on your podcast app, Spotify, Apple. We will always read your reviews, and it really does help other listeners find us.

Harlan Krumholz: Who doesn’t like Howie? Of course everyone’s going to give you good ratings. Health & Veritas is produced to the Yale School of Management, Yale School of Public Health. Thanks to our researchers, Ines Gilles and Sophia Stumpf, and to our producer, Miranda Shafer, extraordinary people to help us every week, they’re amazing. They make this podcast to what it is.

Howard Forman: Very grateful to them.

Harlan Krumholz: Talk to you soon, Howie.

Howard Forman: Thanks very much, Harlan. We’ll talk to you soon.