Naji Gehchan: Hello, leaders of the world. Welcome to “Spread Love in Organizations”, a podcast for purpose-driven healthcare leaders, striving to make life better around the world by leading their teams with genuine care, servant leadership, and love.
I’m Naji, your host, for this special episode in collaboration with MIT Sloan Healthcare and Bioinnovations Conference, an event that brings the healthcare ecosystem together.
I have the pleasure to be joined by Colin Stultz, who will be Professor in Medical Engineering and Science, Professor of Electrical Engineering and Computer Science, Co Director of the Harvard MIT Program in Health Sciences and Technology, Member of the Research Laboratory of Electronics, and Associate Member of the Computer Science and Artificial Intelligence Laboratory.
Cullen is also a practicing cardiologist at the Massachusetts General Hospital MGH. Cullen’s scientific contributions have spanned multiple fields, including computational chemistry, biophysics and machine learning, for cardiovascular risk stratification. He is a member of the American Society for Biochemistry and Molecular Biology and the Federation of American Societies for Experimental Biology, and he’s a past recipient of the National Science Foundation Career Award and the Burrough Wellcome Fund Career Award in the biomedical sciences.
Currently, research in his group is focused on the development of machine learning tools that can guide clinical decision making. Colin, it’s an honor to have you with me today.
Collin Stultz: Thank you for having me, Naji, and thank you for asking me to participate.
Naji Gehchan: Before we dig in AI as a catalyst for care delivery, which was the topic of your panel at the conference, I’m eager to hear more about your personal story and what’s in between the line of such an inspiring journey.
Collin Stultz: Yeah. Thanks. I wish I had a, um, a simple way to explain getting from getting to where I am today. It was quite a circuitous route. Um, so I’ve been immigrant. I was Born in Jamaica, West Indies. We immigrated to this country when I was four years old. Parents came from poverty and came here to, uh, achieve the American dream.
And one of the things I like to, you know, tell people is, like, my parents, um, religion and education were terribly important to them. So, as a youth, the two things that I remember the most was going to church and studying. I was a big nerd for, for, for most of my life. And it was a good thing because it kept me away from all of the bad things in the neighborhood that I was Surrounded by.
Long story short, came to, um, Harvard, loved mathematics, did, uh, math and philosophy for an undergraduate major. And, uh, my parents wanted me to go to, uh, medical school and I really didn’t, wasn’t keen on it at the time. I wanted to do something in, in pure mathematics, but, uh, pure math really didn’t help people, um, the way that a physician could.
So I went to medical school, wasn’t really fond of that, left for a year, did research, came back, finished up medical school, completed a PhD. And then went on to do an internship and residency and several years ago. Here I am now. So the one thing I can say is what led me to this point is a bunch of different interests throughout throughout time.
And what I have followed more so than anything was, was what I liked. And it primarily dictated what my next, what my next steps were. And the only reason I belabor this point is, uh, because I, I am a little bit of an outlier, given the background, you know, I don’t come from an academic pedigree, parents do not come from money, but, uh, here I am at Harvard and MIT trying to do things I think that can benefit people long after I pass away.
And, um, holding positions that I think my father never would have dreamed. And, um, I think for anybody who’s listening from a similar background, such things are possible, um, with a lot of hard work, some tenacity and a little bit of hubris.
Naji Gehchan: Well, Colin, thank you so much for sharing this. It’s so inspiring.
Is there one thing that you would give your younger self advice as you were going through, uh, through this journey that I’m sure, as you said, it’s built on hard work, belief. Is there one advice that you would give yourself? When you were younger.
Collin Stultz: Yeah, failure makes you better. And, and, and I think that, um, failure is, is very hard to, uh, to, um, understand and process in the moment.
But after the fact, I think, you know, you realize, um, what you could do better when you fail. And that’s really important. So I think we view failure as being a bad thing. It never, it never really is. But it’s hard to get anywhere unless you do fail. What’s important is learning from when you fail. And if I could have Sat myself down and told myself that a long time ago.
I think the road might have been a little bit easier.
Naji Gehchan: So now if we go through, uh, one of the topics you’re fond of and the topic you discussed and led this panel on AI as a catalyst for care delivery. Um, this is obviously AI is a very hot topic. We hear about it a lot these days. Can you share first your high level thoughts?
about artificial intelligence, specifically in healthcare. How do you think about that?
Collin Stultz: Yeah, it’s, it’s, it’s a, it’s a great question. And I think a lot of this is, is conflated with what we see on TV, on science fiction shows, what we read in the, in the newspaper. We think about AI as being a series of, of techniques where you just press a button and then machines and algorithms do things that humans used to do.
Uh, well, that humans can do and I think in the health care for some things, that’s that’s that’s probably okay. But the health care, it’s not, I, I think decisions with regarding, uh, with respect to one’s health are typically done in an informed manner, taking into account the patient’s wishes and their value systems.
Things that are very hard for an algorithm to parse. So the way that I think that AI will, will function in the future in, in healthcare is that it’s going to provide important information that the healthcare provider can factor into their decision making information that leverages lots of data. So.
There’s, if you look at even the average person’s medical record, there’s tons of information there. Tons. Um, and so you can imagine the scale of information for those who have been ill and had many interactions with the health care system. Health care providers don’t look at all of those data because we can’t.
It’s too voluminous. But algorithms can computers not limited by large amounts of data, and it can process every piece of data without ignoring even the most the most minutia or the things like that might seem the most irrelevant and in that way, find new relationships and new new, uh, find out new things about patients that we would never have found out otherwise, because a piece of the data that we don’t look at.
And so I think that’s really the power of being able to uncover. hidden truths about patients, hidden facts about patients that are otherwise hard to decipher, and that the healthcare provider, that really is important for the healthcare provider, um, in, in helping them to discern what’s the right actions to take for a particular patient.
Naji Gehchan: So I think you’re, uh, you alluded to that. And if we want to go into a more use case, uh, specific, how do you see this playing in care delivery? specifically, and are you seeing some applications? I know your research is on this. If you can share some of those.
Collin Stultz: Yeah. Yeah. So I think one of the, so I’m going to focus on, I think I think it’s the most promising for the future.
And, um, it is a lot of care nowadays happens when the, a patient goes in to see their provider in the hospital or God forbid, they get admitted to the hospital with an acute, with an acute illness. And most of the of the therapies that are prescribed are when you’re sitting in front of a health care provider and they’re using the data right in front of them.
How do you look at that point in time? What’s your weight? What’s your blood pressure? What’s your heart rate? And what information they have up until that time in terms of what lab tests have you had with additional studies? But, but. But there’s so much of life that happens outside of one’s precise interaction with the health care system in a hospital setting.
And can you expand the information that the health care provider has to things that are at home? So, Information like, you know, how, how active are you at home? Can you climb up by the stairs or not? And can you record someone’s activity? How many steps they take throughout a day? Are they symptomatic at home?
What’s their heart rate at home? What’s their blood pressure at home? What’s their electrocardiogram? And I think we’ll be able to infer a lot of this information in an unobtrusive way in the future with our smart devices. So, um, case in point, there is a large, uh, Mayo has done a large study. in terms of detecting atrial fibrillation, which is a very common, um, arrhythmia that many patients have.
In particular, it’s very prevalent in patients with, with, uh, with heart failure and, um, from just from an Apple watch. So, you know, several hundred thousand patients have tried to detect when they have this bad arrhythmia from some, some such easily, you know, you don’t have to go into a hospital to get a 12 lead electrocardiogram.
You don’t have to be examined by anybody. It’s just by something rare on your risk. And so can we use similar types of data to infer more things? Can we infer things about the cardiovascular health? What’s the cardiac output? How much blood is your heart pumping per minute? And that’s something that typically you can only get by an invasive procedure by sticking catheters into the heart itself to make these measurements.
What are the pressures inside the heart? And can you use this information to predict are you going to be admitted to the hospital? What your potassium is, what your lab values are. I think we can get the same sort of information that you would get if you were admitted in the hospital and we could come close using simple sensors in the outpatient setting.
And so we like to refer to this as, um, um, like an inpatient information at home or inpatient, uh, uh, uh, monitoring at home. You get the same level of monitoring you would if you were admitted, but it’s in a home environment. And I think that’s a sweet spot for a lot of these things.
Naji Gehchan: This is great. And I saw recently Apple actually did get, uh, through the study, I think some approval, right?
Collin Stultz: Yeah.
Naji Gehchan: Yeah.
Collin Stultz: That’s right. Yeah.
Naji Gehchan: So, so you’re bringing this idea of expanding, I love how you framed it, expanding the information that you can get for in hospital patients outside of home and really leveraging just to have a more holistic view of the patient. When you are treating them and making a decision.
Yeah,
Collin Stultz: that’s right. So you could imagine if like the patient could be admitted to the hospital every day or come into the hospital every day and get information and titrate medications and make decisions based on that all from the in home environment. Because at home, you could get the same information you would as if they were standing in front of you in a hospital or in the clinic.
Naji Gehchan: So as you look into the next decade, right, and more of those connected devices, more data will be analyzed and ultimately having all our data records in one single place that we can probably use. What is the most exciting advancement in medicine that you think we should all be watching for the next 10 years?
Collin Stultz: It’s a great question and I, you know, I am biased. I’m biased. Right. Because I, I am necessarily biased to the things that we’re doing in, in, in my group. That’s the thing that I know most intimate, intimately. There are many things that are important. Uh, and just to be, just to be fair, um, we, uh, in a hospital, you could, you do a cardiac ultrasound and it looks at, it’s a very detailed study, just like an ultrasound looks at a baby moving the mother’s and the mother’s abdomen and pelvis.
You, um, can get an ultrasound to see how the heart is moving. and get a lot of detailed information about the heart function, the valves, and various structures in the heart. And there are data from a number of investigators. To suggest that you can get similar information from the type of, uh, from these, um, ECG monitors.
They are akin to what you get from the Apple Watch. Not precisely the Apple Watch or the Fit Beta, one of these wearable monitor devices. But you can, but from those simple signals, you can get similar types of information that you would get from an ultrasound study. And in our group, what we are working on is sort of the equivalent of digital twins.
For cardiovascular patients. So for a given patient, what can you infer from the electrocardiogram or the simple simple, uh, uh, signals? What are the pressures inside the chest? What’s the cardiac output? All of these things that are normally obtained from an invasive study in a hospital. Can you get them at home with just a patch monitor?
And can you use this to put on, um, to do experiments with patients? So what do I mean by that? So if I have two drugs. that I think, um, would be good for you to take. I could give you one drug, see how you do, take you off the drug, wash out, give you another one, and see how you do. Now that’s clearly not beneficial, right?
Because one of the drugs may have very bad side effects. Maybe toxic, but if I had a digital copy of you in the computer, then I could give that drug to one digital copy, the other drug to another digital copy and follow their trajectories in time to determine who’s going to have an adverse event and who’s not.
And so I can do these, what we call in silico experiments to come up in a reasoned way, the optimal therapy for a patient. And we’ve been putting a lot of effort into that. And I think if for some class of patients. So those certain types of hard failure, I’m very optimistic that we can do quite well in that endeavor.
So that, that’s the thing that I’m the most excited about, um, in the pipeline.
Naji Gehchan: This is phenomenal. Like having our digital twin, as you call them, to run those experiments instead of actually doing it on us. Um, how far are we from doing that? I know some companies, you know, They have research on, um, you have your research on how far away would I be able to test a drug as you said.
Collin Stultz: Yeah, I think, I think a lot of, um, work has been going on to try to make general digital twins. So that’s twins for any patient, irrespective of what their diagnosis is. irrespective of what their medical history is. And that’s a very, very hard problem. Many different diseases are in the textbooks. Um, and and if you’re going to try to do this, generally, you need a tremendous amount of data, a tremendous amount of high quality data to be able to do this well.
And it’s our premise. That if we can narrow this down to specific groups of interest, so we can do this for patients with a particular type of heart failure, can we do this for patients with particular type of diabetes, that we can make, that it lowers the data burden and allows you a better chance of building robust models.
That can predict patient trajectories, how a patient will do in the future over time and response to different perturbations, different medications, different interventions. So, so right. It is true. There’s a lot of work in the space. I think the work that’s going on in space necessarily because how the problems are posed.
are very data hungry, require tremendous amounts of data. Our approach is really a focused one and on specific diseases.
Naji Gehchan: So I love that because it’s a specific use case where you can gather the data you need and deal with it. So I have to follow up questions to this. The first one, um, I was born physician, I’m in drug development.
It is so hard. Up to now to predict once you get a drug, how are you going to react to it? So, um, I would love to get your thoughts about those in silico, the precision about it, how should you think about, and probably even getting it into clinical development, actually, instead of, uh, participation, but even as a predictor of response, as you said, and side effects that you can get, how do you think about, about it again, as science.
several times humbles us once it gets into human.
Collin Stultz: No, it’s good. It’s a good, it’s a good question. And the work is ongoing. So the precise, um, answer is still, is still stay tuned. But I will say that I think that one, all of the information that we have would be there, uh, in a patient’s genetic profile.
Their transcriptomic profile and the various tissues of interest, their medical history, what they eat every day, right, where do they live, their, unfortunately, where their, what their zip code is, all of, all of these things come in and affect clinical outcomes, and much of this information is encoded in, in the medical record, not all of it, but a lot of it, and their proxies for things like the genetic profile.
In the, in the medical records. So my sense is that there’s a, that there’s a lot of this at both the structured and the unstructured data that we’re just not leveraging as much as we can. It’s a hypothesis, admittedly, that all of this information is there. But if you can leverage it, um, leverage it, it is my view that you have a good chance of being able for, for important clinical problems.
You won’t do it for everything. I’ll do it for everything. But, but, uh, to give you a specific example, patients with heart failure, a certain type of heart failure called, uh, heart failure with reduced ejection fraction. There are four different drugs that we like to get all patients on. Not every patient can tolerate all four drugs.
So for the specific question of patients in this class, Can you experiment with these different four drugs and see who will tolerate it and who will not ahead of time? And I think the data are, are, are there and of sufficient quality to answer directed questions like that. So, so I think we can make progress, but like, you know, like most things, it’s for the right question.
And what’s the sweet spot?
Naji Gehchan: I love this. I want to go on the data piece. Obviously, everything is based on data, as you said, and the quality of data that there is in. And I can’t but think about the biases in the data. So I would love to hear your thoughts about this, specifically on the biases, the lack of data we have on certain ethnicities, on women.
You know what I’m talking about. So I’d love to get your thoughts about this.
Collin Stultz: Yeah, life is very unfair. Life is very unfair. And medical care is, unfortunately, suffers from some of the same, uh, unfairness. Um, as I, as I like to say, all individuals are biased. And I’ve said this many times before. Some people choose to recognize that they’re biased and mitigate it.
And some people ignore it. And, and I think, so I think we have to begin from the premise that there is bias in the data. Data are assembled and we decide what to, what to acquire by individuals who are inherently biased. And that’s again, as I’ve said on other occasions, that’s not a nefarious statement. I think it’s just a statement of fact.
So, um, realizing there’s bias, how do you mitigate it? I think the way that people ordinarily think about this is that they identify groups that have not benefited, um, like that others have. So women in cardiovascular care and in many other way, um, areas, people from underserved, uh, backgrounds. And you try to say for these algorithms, how do they perform in these groups that are traditionally underrepresented or where these methods don’t perform well.
And I think that’s a, that’s a necessary thing that we need to do. But the thing that is always in my mind. What groups are we missing? Because if you come to this with a preconceived notion about specific groups, are you missing, um, groups that you would not have, have, have thought of, right? So in the 19, uh, in the early 1900s, 18, I don’t know if anyone would have thought of black women as being a group.
That they should focus on. I don’t, I don’t think that would, it’s an extreme example, right? And I think things about mitigating bias would have excluded that group. And I just wonder, it’s always in the back of my mind about what are we missing that we are not paying attention to because we’re blind to.
So, for me, a part of this is not only coming into With it, with, uh, with some, with some prior notion of what groups are, are, we tend to have, we tend to be biased against, but trying to discover from the data, what groups we are not performing with, what groups, what subgroups are, are we intrinsically biased and learning about our bias from it.
So, I think we can try to mitigate the biases that we know might be there, but also try to discover new biases that originally might, might be blinded to it. And I think both parts of that. It tries to guide our work as well.
Naji Gehchan: This is so powerful. Uh, what, what you said, I always think about what groups are we missing?
This is. This is so powerful and so true. I’m gonna give you now, Colin, a word and I would love your first reaction to it. The first one is leadership.
Collin Stultz: So, my first reaction to a leader, well, I’m gonna, a leader is, the first word that comes to mind is follow. Because I think a leader necessarily is a follower.
Um, um, leadership in part is about listening to others And in a sense, following, or maybe following is the wrong term, but being sensitive and paying attention to the things that are important to them. So I said following, but maybe what I’m trying to convey is, is listening. And I think that’s an important part of leadership.
Um, you know, I don’t want to get on a political rant here, um, because, you know, it’ll get me in trouble. But I will say, in general, politicians, Right. Widely construed. Um, there are leaders, but they’re, you know, but listing is not part of their, of, of their, their portfolio. And I say that broadly and, and, um, for me, there are ports in many of them, many of them, not all, many of them are poor examples of, of leadership.
Naji Gehchan: The second one is health equity.
Collin Stultz: Hmm. So when I, when I hear health equity, the first one that comes to me, it’s a right. Health equity is a right. I believe it is a right of every individual to have good care that is the same as anybody else’s. And that’s, that’s a terse statement that’s categorical and that’s an easy one.
That’s an easy one.
Naji Gehchan: I certainly do believe in that too. The third one is Chad GPT.
Collin Stultz: Oh, let’s see. I don’t know what I want to say. It’s overrated. Overrated. So I, I think, uh, chat GPT for me, the, the really important thing, it’s an example of what some of this technology can do. What can we do with large foundational models?
Uh, you know, large transformer models, the types of things that they can do. Um, It’s, it’s overrated in the sense that we often believe what it says to a fault and in the medical space, there are numerous examples where it is incorrect. Producing, um, data that are, um, incorrect and harmful. So I, and I wish that others would sort of focus this as being an example of the types of things that the technology can do rather than focusing on chat GPT in and of itself.
Naji Gehchan: So before giving you the final word, I see a picture behind you. So I’m going to give you his name. Einstein.
Collin Stultz: Oh
Naji Gehchan: yeah. And I’d love your reaction to, to this.
Collin Stultz: So I love this. I love this quote. Um, I’m going to read it. It says great spirits have always encountered violent opposition from mediocre minds. And I am certainly not in the, in the, in the sphere of, of, of Einstein by any way, shape, or form, but the sentiment is something that I, that I, um, I think is a motivation, as motivates, um, in that, um, and I’ve had, um, many, uh, a number of famous scholars.
Um, tell me about the difficulties that they’ve had in their career when they had a new idea and an inability to get funded and an inability to have people to listen to them. And it took decades later. And they were, you know, my old advisors, the Nobel laureate in chemistry, amongst other things. Other very famous engineers here on the MIT campus.
I think are examples of that. So that for me is a motivational statement from a mind that was much greater than mine.
Naji Gehchan: The last word is spread love and organizations.
Collin Stultz: What was the first one? Organizations and?
Naji Gehchan: Spread love and organizations. Oh,
Collin Stultz: I see. Spread love and organizations. Um, that’s very broad.
That’s very broad. I, I think, um, there are, I think about doing good things in the world. That’s what I think. And, and I think that things that aspire us to learn from one another, communicate, um, and, and, and engage in dialogue, um, about new things. Um, technology and otherwise is, is only a good thing for us overall
Naji Gehchan: engage in dialogue.
So important. Any final word of wisdom, Colin, for healthcare leaders around the world?
Collin Stultz: No, no, I think, you know, I think the healthcare people like, like you and, and, and, and like I, we dedicate our professional lives to the health of, of those who are sicker than us. And, and, um, and I think it’s a, you know, it’s, it’s, It’s a noble profession, and I think as long as we, as healthcare professionals, keep that at the forefront of our mind, we can never go too astray in doing good.
Naji Gehchan: Well, thank you so much, Colin, again, for being with me today and this great chat.
Collin Stultz: Thank you so much. Hopefully I’ll see you soon in person.
Naji Gehchan: Hope to see you soon. Thank you all for listening to Spread Love and Organization’s podcast. More episodes summarizing MIT’s Sloan Healthcare and Bioinnovation Conference are available on spreadloveio.com and wherever you get your podcasts. Follow us to get most recent episodes and spread the word around you to inspire others and amplify this movement our world so desperately needs.
Naji Gehchan: Thank you all for listening to SpreadLove in Organizations podcast. Drop us a review on your preferred podcast platform
Follow us on LinkedIn and connect with us on spreadloveio.com. We’re eager to hear your thoughts and feedback. Most importantly, spread love in your organizations and spread the word around you to inspire others and amplify this movement, our world so desperately needs