The World as Our Laboratory: Natural Experiments and the Science of Causality with Dr. Claire Boone

Dr. McGavock: [00:00:00] Welcome back to the Emerging Scholar Podcast. I'm your host, Dr. Jon McGavock from the University of Manitoba, and I'm thrilled to have you back for season two. Last season, we took you on a journey through your first year of research training, covering everything from crafting great research questions to understanding the foundations of both quantitative and qualitative methods.

Dr. McGavock: We met incredible scholars, explored fundamental concepts, and built a roadmap to help you navigate your early research career. This season, we're going deeper. If season one was about laying the foundation, season two is about mastering the craft. We're taking those methods we introduced and really unpacking them, diving into the nuances, the challenges, and exciting new possibilities that each method has to offer.

Dr. McGavock: Throughout this season, you'll meet a remarkable team of both established and emerging scholars with expertise across a diverse array of approaches, from natural experiments that reveal causal insights into real-world [00:01:00] settings, to qualitative inquiry grounded in social justice, the complexities of implementation science, and yes, my personal favorite, causal inference.

Dr. McGavock: Each episode will introduce you to researchers who are pushing the boundaries, asking bold questions, and making novel and important scientific discoveries. First, we're celebrating Canadian research excellence, showcasing incredible work happening right here at home. Second, we're emphasizing the critical importance of patient engagement in shaping research that truly matters across a range of scientific methods.

Dr. McGavock: And finally, we're championing curiosity, the essential spark that drives all great science and keeps us asking what if and why. Whether you're designing your first research study, wrestling with methodological challenges, or simply passionate about rigorous and meaningful research, this season is for you.

Dr. McGavock: So let's dive in, stay curious, and continue building the skills that will shape your own scientific journey. What if the most powerful experiments in science aren't designed in the lab, but by life itself? From a wartime sugar shortage in Britain to a [00:02:00] bureaucratic glitch in a Chilean health system, the world is constantly running experiments on us.

Dr. McGavock: The question is, are we paying attention? Today on the Emerging Scholar, we sit down with someone who is, Dr. Claire Boone, an assistant professor in economics and public health at McGill University, to explore the fascinating world of natural experiments. We touched on this in season one with Dr. Bapu Jena, and today we do a deeper dive on what natural experiments can teach us about health, health behavior, and how to design research that changes lives.

Dr. McGavock: I'm your host, Dr. Jon McGavock, and this is the Emerging Scholar Podcast

Dr. McGavock: Welcome back to the Emerging Scholar Podcast. I'm really fortunate today to have Dr. Claire Boone from McGill University with us to talk about what I would call natural experiments. Claire, can you introduce to us who you are and what your areas of expertise are?

Dr. Boone: Sure. Thank you so much for having me. I am a new assistant [00:03:00] professor at McGill.

Dr. Boone: I'm in both the Department of Economics and the Department of Equity, Ethics and Policy, which is in School of Population and Global Health. Super long title. I am in those two departments 'cause I'm a health economist, and in general in my work I use natural experiments to study decisions and decision-making in health and healthcare.

Dr. McGavock: Awesome. And that's why you're here, because I wanted to do one episode or two on what I call natural experiments, but in the economic world I guess you also have different terminology where you might call them quasi-experiments. Just for terminology sake, what do you like to use and why?

Dr. Boone: I like to use natural experiments when I'm talking to, like, my family or friends when I'm trying to describe what I do.

Dr. Boone: I think within the academic world, in econ we normally say quasi-experimental to differentiate it from experimental research where there's, like, an actual randomized trial going [00:04:00] on.

Dr. McGavock: In the epi world, a quasi-experiment would also be anything without a randomized control trial. In your mind, though, what makes something a natural experiment?

Dr. Boone: I think of it as there's some... Okay, here's another economics term which is exogenous versus endogenous.

Dr. McGavock: Okay.

Dr. Boone: Something that is exogenous is almost like a bolt of lightning. Like, it was unpredictable. Out of nowhere this thing just happened, and it affected, like, some people and not other people, or it affected everyone starting at a certain time.

Dr. Boone: And so as economists, I think we're always trying to look around for these bolts of lightning that have happened in the world and, like, leverage that exogenous variation to do a study that we think is getting at, like, a causal... being able to answer a question causally.

Dr. McGavock: Yeah, a quasi-experiment. And, and for me, the bolt of lightning is also that we had no [00:05:00] control over it.

Dr. McGavock: What I wanna talk to you a little bit about is, like, in the ec- economics world it feels like it's the standard approach. Mm-hmm.

Dr. Boone: It's

Dr. McGavock: like economists when they're trying to answer a question, they're trying to think causally about it, and they're like, "Okay, let's go find some bolts of lightning." Sometimes a natural...

Dr. McGavock: A quasi-experiment could be the investigator had control- Yeah ... and they just did a pre-post evaluation of something, right? Which I think you and I both agree that's not really a quasi-experiment, but it definitely does not fall in the natural experiment world.

Dr. Boone: Yeah, so we We're always looking for these bolts of lightning so that we can do a study as...

Dr. Boone: that answers a causal question. I think this is the difference between health economics research and, like, non-econ sort of health services research or public health research. Health economics is, like, really focused on answering causal questions. You know, what is the effect of X on Y? We're not, as health economists, usually [00:06:00] trying to write papers that are like, "What is the association between X and Y?"

Dr. Boone: Mm-hmm. And I think lots of your econ training is, like, spending a lot of time thinking about these quasi-experimental and experimental methods so you can make sure you're answering causal questions and not just doing an, an associations paper.

Dr. McGavock: It's like the whole field has committed themselves to causal questions.

Dr. Boone: Mm-hmm. And I don't know that it's- Good. I think you can learn a lot from like very good descriptive papers and descriptive evidence. I think what we're trying to avoid is wanting to answer a question that is causal in nature, but falling short of that and doing an association-based paper because you know the variation wasn't there.

Dr. McGavock: Right.

Dr. Boone: Yeah.

Dr. McGavock: And then in the epi world though, it's different. And so you've straddled both worlds. You started a little bit in epi, and then you went to economics. What are the differences, and [00:07:00] what is the commitment to causality or causal inference?

Dr. Boone: I have no idea where it it comes from, but for me it was a factor.

Dr. Boone: So I started by doing a master's in epidemiology and biostatistics, and I was really excited about learning to code, like using big data sets to answer research questions. Very excited about lots of different topics in health Had also majored in biology, economics, and international development in undergrad, so I was already extremely multidisciplinary.

Dr. Boone: There's a program at McGill called Arts and Science where you just kind of do everything if you're excited about a lot of different things and you can't pick. Try '

Dr. McGavock: em all.

Dr. Boone: I was very excited about data analysis and this, like, health-based research, and I loved a lot of the classes [00:08:00] in my epi biostat masters, but I had this background in economics, and so I had this interest in behavioral economics specifically.

Dr. Boone: I realized during my master's that I wanted to be able to think not just about counting diseases and thinking about exposures and outcomes, but I really wanted to think about human behavior. And to study human behavior really carefully, like, one way that you can do it is do a bunch of classes in economics.

Dr. Boone: So my primary motivating factor for switching over was that, and I kind of learned later that it came with this difference between, you know, hearing about causality. And moving into econ, I realized, like, okay, guess things are done differently over here. We're very focused on quasi-experimental and experimental, and I really also ended up liking that.

Dr. McGavock: Yep. So yeah, person, place, and time is epi, and counts are epi. But then moving into econ, which, [00:09:00] like, for me, as someone who thinks about economics in a different way, I think more on a business side of things. But when you think about behavioral economics, how did you make that jump? How did you even know that there was this field of behavioral economics, and if I move into that, I can merge, I can combine my love of biology, my love of behaviors or psychology, and actual international development equity, if you will?

Dr. Boone: I took behavioral economics class in undergrad.

Dr. McGavock: Oh.

Dr. Boone: And I remember it so clearly. It ended up being my favorite class that I ever took, and it just made so much sense to me in, like, how I saw myself behaving and my friends behaving, especially, you know, as 18-year-olds, 19-year-olds. Yeah. I have a very clear memory of this moment where I learned about present bias and the fact that in behavioral economics, one of the theories is that we basically, like, w-weight the present higher than we weight the future, and that can lead to things like procrastination because, uh, you know, [00:10:00] right now I kind of wanna keep scrolling on my phone.

Dr. Boone: I know I should study, but I don't want to because it's painful to study.

Dr. McGavock: Yeah.

Dr. Boone: And so there were a few instances where it just felt like that theory really clicked for me, and then a few years later started being interested in health research. I saw this link, which I then realized many other people had already seen as well.

Dr. Boone: I saw this link between, like, lots of behavioral econ theories and health behaviors. So, like, going back to the present bias example, with diet and exercise, it can be kind of painful to exercise and eat healthy when your body is kind of craving sugar and sitting on the couch, and all these things that feel good in the moment but are bad for you long term.

Dr. Boone: I started to think, okay, there's probably a cool way to combine this behavioral econ interest with this data health analysis interest that I have, which ultimately led me to work on this [00:11:00] paper that I know you wanna talk about.

Dr. McGavock: Oh, that's awesome.

Dr. Boone: Yeah. And

Dr. McGavock: I, I find that super interesting that the origins of behavioral economics as the Freakonomics, uh, podcast would be the hidden side of everything.

Dr. Boone: Mm-hmm.

Dr. McGavock: Right? So in one generation, the guys who are bringing this in were like, "We're gonna show you that people do not act in logical ways." Mm-hmm. "And that their behaviors are not predictable by standard economic theory."

Dr. Boone: Right.

Dr. McGavock: But then the next generation comes in and you're explaining behavioral economics, and we're just like, "Oh my God.

Dr. McGavock: That explains everything." Right. "This totally makes sense." Yeah. So you had this feeling that you were maybe the first in class to tie behavioral economics to health outcomes? No,

Dr. Boone: no, no. Okay. I said I'm definitely not the first. I think it was, you know, one of those moments when you're early in your career and you think you have a genius idea, and you realize someone has figured that out, like, 15 years ago.

Dr. Boone: But for me, you know, it was a good moment 'cause it helped me kind of, like, realign what I wanted to do with, with my career.

Dr. McGavock: And so for trainees listening, doing the pivot worked out for you.

Dr. Boone: Mm-hmm. [00:12:00]

Dr. McGavock: And you're loving it?

Dr. Boone: Yeah, I think I come very full circle. I did my undergrad at McGill in this weird three-discipline topic, and now I'm back at McGill, where they value, like, multidisciplinary research, and I'm in two very different departments.

Dr. Boone: And it... That allows me to keep doing this econ plus health kind of research.

Dr. McGavock: Awesome. And for listeners who are following, in season one we interviewed Bapu Jena. He wrote a book about natural experiments and how to capitalize on them, and I think he's done some very cool behavioral economic health research using these natural experiments.

Dr. McGavock: What I liked about that interview with him is how he came up with the different questions. So I'm gonna start with this one. You had this question around nudging patients- Mm-hmm ... living with chronic disease, so type 2 diabetes or hypertension, nudging them to do the kind of things that we know they should be doing in terms of their preventive care, right?

Dr. McGavock: You had a question, and what [00:13:00] was your question? And then you ended up with this paper here that we'll put in the chat, but it's called Encouraging Preventive Care to Manage Chronic Diseases at Scale. So you wanna tell us a little bit about that question first, and then where you found the data to answer the question?

Dr. Boone: Yeah. So this was maybe a little bit backwards compared to a lot of how research starts. The program was already existing. Okay. So a Chilean co-author, Pablo, had heard a whisper that there was a staggered rollout of an appointment reminder nudging program. This is one of these, like, lightning bolts that economists are always trying to find.

Dr. Boone: As soon as you hear staggered rollout, like, we get very excited. Yeah. From a

Dr. McGavock: methodological standpoint, why is this staggering important? And then what was the program?

Dr. Boone: Yeah. So if something is staggered, like in this case, the program was implemented at primary care clinics across the country at a bunch of different times.[00:14:00]

Dr. Boone: If you can show that the reason it was taken up at different times across different clinics has, like, nothing to do with the outcomes you're trying to study, then you can move into a research design called difference-in-differences analysis or also adding event study analysis into that. And it's one of these quasi-experimental methods that can help you get at a causal research question.

Dr. Boone: And so we heard that in the past, the Chilean Ministry of Health had offered their hundreds of primary care clinics that exist across the country the option to start doing this nudging program, where anyone at their clinic who had been already diagnosed with diabetes or hypertension would start to get a text message reminder about their upcoming visit.

Dr. Boone: So very, very simple. You know, every time you go to the doctor, you're already getting one of these reminders. [00:15:00]

Dr. McGavock: Mm-hmm.

Dr. Boone: The reason that we got super excited about... There were two reasons we got really excited about this one. One is that it was on a very, very large scale. So they're different to Canada. They have pu- public primary care system, but it's federally administered, so we were able to go in and collect data on the whole country of patients.

Dr. Boone: The other reason we were really excited about this is we had this idea that- If the reminder induces people to use primary and preventative care more, we might be able to use the nudge as an instrumental variable to look at the effect of seeking additional preventative care on future health and behaviors of the patients.

Dr. McGavock: Amazing. You're answering two questions.

Dr. Boone: Yep. The- First question is, what is the impact of this nudging program on patient behavior? And the second question [00:16:00] is, what is the impact of seeking additional preventative care on patient behavior, healthcare received, and patient health?

Dr. McGavock: On patient health outcomes, right?

Dr. McGavock: 'Cause they're getting measurements for their blood pressure, right? Or they were getting measurements for their weight. So it's kind of health-seeking, but then it's also, like, following up- Mm-hmm ... and getting the, getting the value. Mm-hmm. So, um, that's awesome. So yeah, it's a double whammy. Beautiful. So to help trainees, a difference-in-difference you would define as what?

Dr. McGavock: Difference-in-difference analysis. And then the second question would be an instrumental variable analysis that's allowing you to answer the second question. Mm-hmm. What's an instrumental v- variable analysis?

Dr. Boone: Yeah. So difference-in-differences is you have two groups. So in our case, there are some clinics that started running this program, and there's other clinics that didn't.

Dr. Boone: And you compare the trajectory of outcomes of those two groups over [00:17:00] time. So, you know, it's possible that- The clinics that took up the program are in, like, wealthier places with healthier people, and then there's less healthy people in the other clinics. That's fine. We just wanna make sure that the difference in trends is constant over time, and the second difference is the treated versus control group.

Dr. McGavock: So just like in a randomized control trial, the difference is actually the intervention. Your baseline differences- Mm-hmm ... are gonna be there, but the only thing that changed, it's not like because of the nudging program, patients moved into that area to get the care. Right. Everything stayed the same beforehand.

Dr. McGavock: Then the implementation happened, and now you're measuring the change- Mm ... between the groups or the difference in the differences. It's as close as a randomized control trial as you're gonna get. And the staggering ensures that it's relatively random in terms of how it was implemented, so [00:18:00] the comparison of the two groups is as close to individual randomization as you're gonna get.

Dr. McGavock: Because the concept in the natural experiment is that individuals are not getting randomized to that lightning bolt. The lightning bolt- Right ... is happening. It happened to you, and it didn't happen to your neighbor. We wanna see- Mm-hmm ... is there a difference?

Dr. Boone: Mm-hmm. And there's an extra layer in this particular study, which is people who use the public healthcare system in Chile get assigned a clinic to go to based on where they live.

Dr. Boone: And so we can be quite confident that if you just happen to live near this clinic that implemented the program, you're gonna be exposed, and it's unlikely that you're gonna go and move across town to go to that other clinic that didn't implement the program or something like that.

Dr. McGavock: Right. So one of the assumptions for causal inference if we're using epi language would be, like, consistency.

Dr. Boone: Mm-hmm.

Dr. McGavock: So that there isn't... There's only [00:19:00] one intervention that happened, and if the intervention in one area led to people changing their behavior in the control area, then you have multiple things happening.

Dr. Boone: Yeah. I

Dr. McGavock: think that's how I would translate it into an epi world. You want that intervention to just also only affect the group that you're saying is in the intervention group.

Dr. McGavock: Then so an instrumental variable analysis, so in this case you're also using the nudge itself as an instrument for something.

Dr. Boone: Yeah. Yeah, so here we are using The nudge as an instrumental variable to study the impact of preventative care on a bunch of other things. So there's kind of like two stages in the regression.

Dr. Boone: The first is we need the nudge to induce people to go to preventative care more, and then if they do, we can study the impact of preventive care on all these other downstream outcomes.

Dr. McGavock: Well, and some of the assumptions would be that, like, the IV has to [00:20:00] be really correlated to the exposure that you're interested in.

Dr. McGavock: The nudge should have had an effect, right? Now, in terms of you're using the nudge, you're using the text message as an instrument for seeking preventive care.

Dr. Boone: Right.

Dr. McGavock: There's gonna be variation in that. Not everyone who gets the nudge- Mm-hmm ... is gonna grab it. And then you say that the more you engage in preventive care, the better your health outcomes are gonna be.

Dr. McGavock: Mm-hmm. The better adherence you're gonna have to the strategies that your physician is asking you to engage in. Right. So the instrument can't be related to the outcome directly. Right. It can only be related to the outcome through the exposure variable that you're interested in. So in this case, the text message going to the patients was what?

Dr. Boone: It was sent one or two days before their scheduled preventative care visit, and it just said, like, "Hi. A reminder you have a visit tomorrow. It's at this time with this doctor in this location."

Dr. McGavock: So the instrument that [00:21:00] we're all familiar with, I think, or could be familiar with in the epi world, is the randomization.

Dr. McGavock: Mm-hmm. Right? Like- Yeah ... randomization's the best instrument. When you get randomized to a group, you're not always gonna adhere to it, right? So the control group isn't gonna adhere to everything in the control group. The intervention arm isn't gonna adhere to everything in the intervention arm. So there's gonna be variation in there.

Dr. McGavock: And so the randomization to the exercise isn't gonna affect your blood pressure at the end of the intervention, but it just serves as an instrumental variable saying that the people who we randomized to the exercise intervention arm, they should look just like the people in the control arm because of the randomization.

Dr. McGavock: We're not gonna analyze the data based on who's exercising. We're just gonna analyze the data based on the outcome. So in your case here, you're saying the text message that you got to go to your appointment is gonna be the randomization scheme or the instrument, and if it has an effect on you ful- filling your blood pressure medications or measuring your blood sugars [00:22:00] or whatever the outcome's gonna be, it happened through that primary care appointment.

Dr. McGavock: So what did you find?

Dr. Boone: So we find that the text message induces about like a 10 or 12% increase in preventative care use. So it's not that big. It's the magnitude you would expect from this, like, very light touch intervention, which is sending this text message that some people might ignore. One thing I liked is we look at heterogeneity in that analysis.

Dr. Boone: And as you would expect, like the very old people, you know, there's no effect of the nudging intervention for them. They may not even have a phone potentially. There's also very little effect for younger people. I think if you're in your 20s and you've been diagnosed with hypertension or type two diabetes, you might just be the type of patient who needs a lot more help than simply this little nudge.

Dr. McGavock: Mm-hmm.

Dr. Boone: Um- Huge

Dr. McGavock: socioeconomic [00:23:00] differences.

Dr. Boone: Yeah. Younger and older. We don't have real- really good SES measures in this paper, unfortunately, but I'm sure that they're there. And another cool one I thought was we also looked at heterogeneity in what was your blood pressure or hemoglobin A1C at the time of diagnosis.

Dr. Boone: So... And we find, as you would expect, if you have very high levels of either of those things when you get diagnosed, the text messages don't help you go to preventative care more. Again, you might be this kind of person who needs more help than this little light touch nudge if you've allowed your blood pressure to become like 180 over something and you're just getting diagnosed for the first time.

Dr. Boone: So if you're like middle-aged and have medium high biomarkers and you get diagnosed, it worked quite well at about like 10 or 12%. So this is actually the variable we use as the instrument is clinics [00:24:00] compliance with the nudging intervention.

Dr. McGavock: Hmm.

Dr. Boone: So to give some background, like the first version of this paper, we found very, very tiny effects on everything.

Dr. McGavock: Mm-hmm.

Dr. Boone: And we presented it to the Ministry of Health. We were discussing why do we think this might not be working that well. And someone there, I spent a lot of time living in Santiago doing this work. Someone there said, "Oh, you know, we actually have phone records from the software that sends out the text message, and you could maybe look at those and see if there's something interesting."

Dr. Boone: And I was able to-

Dr. McGavock: Awesome. Like, did your eyes just light up like, "Oh my

Dr. Boone: God"? Yes. So it was- Oh my God. ... such a incredible overall like incredible research partner, very willing to provide data. Like- So cool ... such a good experience.

Dr. McGavock: Yeah.

Dr. Boone: Um, which is why I have four papers now with Joey and Duga. Yeah. So I looked at the phone records and lo and behold, average compliance was [00:25:00] 50%.

Dr. Boone: So about half of people who are supposed to be getting text message reminders and in our sort of intent to treat style approach- Yeah ... where you assume someone actually got the intervention if they were assigned to the intervention arm. We were like throwing out half of the real variation that was happening because in reality, only half of people who came into the clinics actually got a reminder.

Dr. McGavock: So that's like you're randomized to an exercise intervention arm and then your exercise arm is you're supposed to go to the Y to do- Mm-hmm ... your exercise training and someone tells you not to go or someone forgets to tell you to go to the Y.

Dr. Boone: Right.

Dr. McGavock: So like half of your intervention arm doesn't even get the intervention.

Dr. McGavock: So then phone records allow you to dial in who actually got the

Dr. Boone: Yeah, it's not- The

Dr. McGavock: intervention ...

Dr. Boone: we weren't allowed to link the phone records to electronic health record data, which is fine. We ended up making a variable to capture compliance [00:26:00] at the clinic level and track how it changed over time. So in most clinics, it improved during the study period, which was, I think, five years.

Dr. Boone: And you could imagine, like, all of a sudden your clinic has this nudging program. You know, maybe you didn't have my phone number on record, but when I come in in six months, you start to write my phone number down.

Dr. McGavock: Mm-hmm.

Dr. Boone: And so once we added that into our analysis, essentially the effects doubled in size because we're dividing by 0.5.

Dr. Boone: Okay. And I do think in itself that's a pretty interesting finding. Like, it's extremely challenging to scale programs up.

Dr. McGavock: Mm-hmm.

Dr. Boone: And text messaging is probably, like, the most scalable possible program. And if you read SMS reminder papers, like, they all say at the end, "This is a great intervention 'cause it's extremely scalable," and then we can show that even, even though it's so easy to send a text message reminder, when you [00:27:00] implement things at scale, like, it's just very hard.

Dr. Boone: Yeah. And we shouldn't expect the effects to be as large as if a small scale, like, efficacy style study was run.

Dr. McGavock: Right. Like a smaller scale, super well-funded RCT that only has 5,000 people in it- Mm-hmm ... instead of 300,000 people in it. Right. Where you can monitor everything that they get, you know? Which is probably not translatable.

Dr. McGavock: You're gonna have a hard time transferring whatever you get in that randomized trial to the population. Mm-hmm. One of the advantages of this difference-in-difference model or natural experiment model is that it usually happens at a population level, so you can say, "Yeah, like, this is the real world. We did this in the real world.

Dr. McGavock: This is what we saw." And when you do stuff in the real world, it's not gonna change the whole population by 50%, right? It's a nudge. Yeah. You're only gonna get a certain effect. But you got, like, 50% is still a big effect at the population level, where the ministry must have been like, "Yeah, this is great.

Dr. McGavock: It's fairly easy, scalable intervention, [00:28:00] and we're seeing that more people are going to their appointments," right? And then the instrumental variable says, "Yeah, and if they go to their appointments- What happens? And so the answer to the second question was?

Dr. Boone: If they go to their appointments, they are way more likely to get blood pressure measured, blood glucose measured, and weight measured.

Dr. Boone: That's what we would hope would happen if you're someone with one of these two conditions. Yeah. Because we're u- in the instrumental variable analysis, we're measuring the effect of going to primary care on these outcomes. Our effect sizes are massive. They're like

Dr. McGavock: Yeah.

Dr. Boone: 90 percentage points, and that's reflecting the fact that, you know, these are pretty high-quality clinics, so if you show up with hypertension, like, they will almost certainly measure your blood pressure.

Dr. Boone: So those were, I think, more than anything, just like reassuring results. One that's a little more interesting maybe is [00:29:00] we were able to get medication prescriptions and fills data-

Dr. McGavock: Mm-hmm ...

Dr. Boone: from the national pharmacy database in Chile. And we were able to measure like actual medication adherence at the individual level.

Dr. Boone: And we find, I think it's about a 20 percentage point increase- Mm-hmm ... in medication adherence, which is very big. Like, there are lots of people who spend a lot of time trying to do medication adherence interventions, and I think one of the best interventions we could do is get people to go to primary care.

Dr. Boone: 'Cause if you don't go to primary care, you can't even have a prescription. Right. And if you don't have a prescription, you definitely can't adhere.

Dr. McGavock: And if you don't adhere- Mm-hmm ... you're not gonna get the health benefit. So huge- Yeah ... effect, very relevant. Answered two questions in one beautiful natural experiment using two causal approaches.

Dr. McGavock: Mm-hmm. Which is why I like this. That's why I was hoping we could talk about this. [00:30:00] It was like a double bang for your buck. Very long. Uh, what was the uptake from the health authority? Was the government excited? 'Cause now they have a causal ques- they have a causal answer. Like, one- Right ... text messages work.

Dr. McGavock: Two, if they go to care, they adhere to what we're asking them to do, we should fund this, you know? Or we should build on this.

Dr. Boone: Yeah, so this is quite a disappointing story. Oh. They canceled this program.

Dr. McGavock: No.

Dr. Boone: Yeah. And I think... I don't know. So we were working with the Ministry of Health, which includes a lot of political appointees.

Dr. Boone: Mm-hmm. And because, you know, one thing about economists is our papers are super long, and they take a very, very long time. And there were two elections during the time that we were working on this paper, and a lot of turnover in the people that we were working with. And last I heard, they had just scrapped this program.

Dr. McGavock: However- Yeah ... there will [00:31:00] be a new election coming up at some point. Mm-hmm. And hopefully, and more importantly, you did publish this paper, and it got out there, and it's open access, so any government out there looking for any nudge unit in any kind of government can point- Mm-hmm ... to this and say, "Hey, we've got great experimental data."

Dr. Boone: Yeah.

Dr. McGavock: And a plug out for a future episode, we are gonna talk about implementation science. So you talked about two really important things there that affect the implementation of the interventions, which is, like the primary care provider has to adopt it, right? Like, if they're not sending out the SMS to people, or if people don't have phones, or the political climate or like the external forces that could potentially affect the intervention are gonna potentially affect it.

Dr. McGavock: So on a cool side, because you're doing it at, at scale in the real-world setting, you could also study some of those factors, and then the information goes back to the health authority, like, "In this population with this health outcome at baseline in these conditions, this is gonna work super well," right? I [00:32:00] think that's awesome.

Dr. McGavock: Like, that's great information, and it's causal. It's not associative. So- Mm-hmm ... they can really lean into it. So we got a difference-in-differences as an example of a causal question. We've got an instrumental variables as a causal question. Mm-hmm. Hopefully, we'll do an episode on like instrumental variable analysis, I think, even though that goes into some deep assumptions, and it's challenging to find a really good instrument, so we're gonna find some good ones.

Dr. McGavock: The more recent paper that you published in Science looked at exposure to sugar rationing in the first thousand days of life. I would also call this a natural experiment. Mm-hmm. But in contrast to the one that you studied in the paper we just talked about, that was prospective, right? You got a hint that this was gonna happen, so you guys went down there, tried to capture it, set it up in Chile?

Dr. Boone: No, it was totally retrospective. We heard... It was still going on while we started collecting the data, but-

Dr. McGavock: Okay ...

Dr. Boone: we started studying it, I think, in 2018, and it had begun in 2015.

Dr. McGavock: So it was in the middle of the experiment. You were trying to capture data- Mm-hmm ... work with the government. [00:33:00] Thankfully, they had a nudge unit that was really open-minded about data access- Mm-hmm

Dr. McGavock: which is really cool. Now, this one, the experiment happened like in the '50s. So let's talk about this. We'll put a link to this paper in the show notes, but the title is Exposure to Sugar Rationing in the First Thousand Days of Life. I am very interested in some developmental origins questions, so this totally piqued my interest, and it went against everything that I thought.

Dr. McGavock: So that's another reason why I was interested in this. I was like, "I don't believe in this," and then someone sent me this paper, and they're like, "Aha, maybe you should," 'cause it protects against chronic disease. So you were looking at sugar rationing early in life, like- Mm-hmm ... while the fetus is in the womb, and then you looked at disease later on.

Dr. McGavock: So walk us through that design and that experiment and the data that you used to try to tackle that, and maybe just talk to about your colleagues too that were involved in this paper.

Dr. Boone: Sure. So this is another case of three economists looking for a lightning bolt, [00:34:00] and my colleague Tadeja found... She does a lot of work on sugar generally, sugar taxes, things like that.

Dr. Boone: And one day, as she tells the story, she was pregnant and lying in bed.

Dr. McGavock: Where good

Dr. Boone: ideas

Dr. McGavock: come from.

Dr. Boone: Yes. She started to think about how there was food rationing in the UK, and there actually was sugar rationing in Canada as well.

Dr. McGavock: Mm-hmm.

Dr. Boone: If someone is hoping for a new research topic. Yeah.

Dr. McGavock: Boom, replication study.

Dr. McGavock: Let's go, Manitoba.

Dr. Boone: Yeah, I would love to see that. And she realized that there was sugar rationing in the UK that lasted well beyond the end of World War II. So that's important because we wouldn't wanna be doing a study where we're just, like, picking up the effects of the war on health.

Dr. McGavock: Mm-hmm.

Dr. Boone: We want to know, you know, absent the war-

Dr. McGavock: Mm-hmm

Dr. Boone: what is the effect of sugar on health. And so she realized that sugar rationing ended in 1953. And if you look that up, [00:35:00] there are all these news articles about how people lined up around the block to get their hands on sugar as soon as rationing was lifted. So during rationing, they had rationing books and little coupons that allowed them to buy a certain amount of a bunch of different foods, one of them being sugar, every week.

Dr. Boone: And so basically, people couldn't eat as much sugar as they wanted. They were stuck at this, like, healthy, low level of sugar consumption. And the news articles sort of cued her in to think, you know, maybe consumption really changed as soon as rationing was lifted, right? Mm-hmm.

Dr. McGavock: People love sugar.

Dr. Boone: People love sugar.

Dr. Boone: So the first thing we did was to look for food surveys and food sales data from, like, the 1950s to understand if consumption did change a lot right after rationing ended. Turns out it did. [00:36:00] So we can look at meat and fish and vegetables and all these other things. The line is very flat. Mm-hmm. But in September 1953, as soon as sh- sugar rationing was lifted, consumption doubled right away.

Dr. McGavock: It's an insane graph.

Dr. Boone: Yeah. Yeah. It's like

Dr. McGavock: people were desperate for sugar.

Dr. Boone: Right.

Dr. McGavock: You don't even need to do stats on that graph. No. Like, when you look at figure 1A-

Dr. Boone: Mm-hmm ...

Dr. McGavock: it just skyrockets by 60% within the first quarter of the year. So you got

Dr. Boone: proof- So the- ...

Dr. McGavock: that your experiment elicited a behavior change.

Dr. Boone: So, like the effect of the nudge in the previous paper, we think of this as the first stage. You know, to see anything down the line, you need to have the fact that rationing was binding. Like, it was keeping people at levels they didn't want to be at. And so right after it was lifted, they changed their behavior to start eating a lot more sugar.

Dr. McGavock: Where'd you get the data on the survey?

Dr. Boone: There were dietary and food surveys that had been done [00:37:00] in the 1950s that we were able to access. Um-

Dr. McGavock: Which is, this is the best part of these natural experiments, right? It's like the Sherlock Holmes-ing of- Yes ... finding the data. Like, it's so exciting. Totally. It's so fun.

Dr. Boone: Yeah. We had to digitize a bunch of food survey and sales- So ... records from the past in order to create data sets to make that figure that you're referring to. So fun. It, it's a lot of work.

Dr. McGavock: So fun. A lot of work, yeah. But it- Did you have to do it or was it, like, an undergrad who was doing it?

Dr. Boone: Yeah, I didn't do it myself.

Dr. Boone: But speaking of not having to do statistics, I was doing the data analysis on this project initially, and Just comparing the mean age of onset of diabetes and hypertension between people exposed and not exposed to rationing.

Dr. McGavock: Yeah.

Dr. Boone: You know, it's not something we would publish, but it was such a big difference that we just thought, "Okay, there has to be something here."

Dr. McGavock: Mm-hmm.

Dr. Boone: It must be that the sugar rationing had an effect on long-term health. [00:38:00] And so that was our que- our cue to go in and start to do this, like, more formal analysis.

Dr. McGavock: That... So season one, Bapu Jena talks about that, I think, with the marathon study and time to hospital or adverse outcomes after an MI. That he said, "We just looked at the baseline characteristics," and w- it was so huge.

Dr. McGavock: He's like, "We got something here." Yeah. Um, so that must've been very exciting when you did that.

Dr. Boone: Yeah. I think all of my decent papers like kind of start with that. Like, e- there's something very undeniable from the get-go in the data.

Dr. McGavock: Yeah.

Dr. Boone: You avoid being worried, too worried about your P values or all these different things.

Dr. Boone: Um-

Dr. McGavock: So yeah, when you can eyeball it, you know it's gonna be good. And then so you don't have a difference in differences like we discussed in the other one, right? So you do need... Like, everyone was exposed in all of England, right? So what's your comparison group? Like, how'd you compare the rationing to non-rationing in the same group?

Dr. McGavock: And then where did you find the data to [00:39:00] get diabetes and hypertension outcomes, which you just talked about?

Dr. Boone: So here when you have the lightning bolt, it's something that happened very quickly in a point of time. The, we don't do a regression discontinuity design in this paper, but the intuition is the same.

Dr. Boone: And in regression discontinuity design, the idea is that there's a value that completely determines whether you're exposed to something or not exposed. And so here the value is date of birth of people in the 1950s. And what we ultimately do is compare people conceived- Yeah ... a couple of months before rationing ended versus a couple of months after.

Dr. Boone: And what you wanna do is kind of zoom in as close as you can to that threshold so that hopefully everything else about their experience is the same. If you were born a week ago, you were conceived during [00:40:00] rationing, and if you were born one week from now, you're conceived after rationing ended in this more sugar-rich environment.

Dr. McGavock: Through those nine months, your mom was waiting in line for sugar.

Dr. Boone: Mm-hmm.

Dr. McGavock: And then there's a cohort of people that could be living on the same block as you. They were conceived a year later, and they were exposed to as much sugar as they wanted, or m- mom having expos- as much sugar as she wanted, right? And then you're gonna compare those two trajectories.

Dr. Boone: Exactly.

Dr. McGavock: So you had to do in the '50s because in the '40s you didn't have consistency assumption. The consistency assumption is it has to be one intervention, right? So you had to be-- make sure that it was just sugar rationing, and then when sugar rationing went up, boom, that was the effect. If you went beyond the '50s, there's a bunch of other stuff happening.

Dr. McGavock: But if you keep it- Mm-hmm ... really tight around that policy change, that lightning bolt, and in this case, the lightning bolt is actually a removal of something. It's like you lifted something, which is also kinda cool. And then [00:41:00] one of the things that you might be worried about is that are there other factors that would make these two groups different?

Dr. McGavock: So you had data from people who are, are now in their 50s, and they're being followed up over time, 50 years after being conceived. So were they balanced? Like in a randomized control trial, the randomization balances those two. Right. So in this case, it's like an exchangeability assumption. So what did they look like in terms of their biological and socioeconomic makeup 50 years later?

Dr. Boone: Yeah. So we went out and got data from the UK Biobank study. What we needed was lots of people born in the 1950s who were surveyed much later in life, and we ideally wanted to have a bunch of objective health measures like blood pressure, blood glucose, et cetera, and then also their medical records. So in UK Biobank, they survey people to ask, "Have you ever been [00:42:00] diagnosed with X?"

Dr. Boone: But they also link the NHS medical record then to get the first date of diagnosis with something.

Dr. McGavock: Mm-hmm.

Dr. Boone: So we were looking for people in the 50s with these outcomes and, like, ideally lots of pe- as many people as possible so that we could zoom in as close to that threshold as we could. And so once we collected that data, we compared people conceived on either side, and they're just identical in every way except for their date of birth, which determines whether they were exposed to rationing.

Dr. McGavock: Which is amazing. And how excited were, was Tadeja or yourself when you realized UK Biobank data was there? Which for those of you listening, it's a publicly available data set on hundreds- Mm-hmm ... of thousands of people between the ages of, uh, 40 and

Dr. Boone: 70. Who are living in the UK, which is also-

Dr. McGavock: Who are living in the UK where this intervention happened.

Dr. McGavock: It's like, oh my God.

Dr. Boone: Yeah.

Dr. McGavock: That must have been so exciting to see, that we can test- Right ... this hypothesis using data [00:43:00] that's already there. Like- Mm-hmm ... this lightning bolt is a studiable experiment. Definitely. Yeah. Amazing.

Dr. Boone: Yeah. So this isn't in this paper, but one of my favorite things about this idea of using sugar rationing as a natural experiment is there's actually lots of data sets that fit the criteria I just mentioned.

Dr. Boone: So they have people born in the '50s, and they measure different outcomes in the present day. We've been able to show, I think we have four, maybe five different data sets where we can actually replicate this exact same finding, and it just makes me feel very confident that it is a real effect that we're seeing.

Dr. Boone: Yeah.

Dr. McGavock: It's not... The UK Biobank is a dif- it's not the whole population, right? So

Dr. Boone: if you're- Yeah. It's kind of selected on income, and people are a little healthier. They're whiter.

Dr. McGavock: Yeah.

Dr. Boone: So we've tried it on, there's like this data set called ELSA, Longitudinal Study of Aging. Yeah. There's just like, there are lots of different data sets that are measuring older people who presently live in the [00:44:00] UK, and we can show that it works every time.

Dr. McGavock: The effect sizes are gigantic. Like, the age at which the hypertension and type 2 diabetes start to emerge, and then once you hit a certain age, like if you're enrolled in ALSPAC at 64, you're four times more likely to have hypertension and four times likelier more to have type 2 diabetes. If you were exposed during the window where sugar rationing was lifted, then- Right

Dr. McGavock: your colleagues, your neighbors that were conceived during the sugar rationing.

Dr. Boone: So it both reduces the likelihood. So being conceived and then growing up, we look a bit from sort of the point of conception to age two. Living your early life in a low sugar environment can both reduce the probability that you get diagnosed with one of these chronic conditions, and if you do get it, you get it later in life.

Dr. McGavock: Right.

Dr. Boone: Of like more years of being healthy before the diagnosis. [00:45:00]

Dr. McGavock: Totally goes against what I would've said.

Dr. Boone: You thought it would've been the opposite?

Dr. McGavock: Yeah, which that is for another episode. But because of the data that informed a lot of this DOHaD work was ecological, like it was poorly done. We should do this where we compare the Barker hypothesis data to your data.

Dr. McGavock: Yours line up and, you know, Barker will be very happy that it lines up with the hypothesis. I would've said because of the low quality of the original data and then the subsequent inability to replicate that, we wouldn't necessarily have seen it. Mm-hmm. But this is way more robust. And the kicker for me here and the message to trainees, it's because you approached it using a different design.

Dr. McGavock: Yeah. You just thought about analyzing it or tackling it from a causal a- approach.

Dr. Boone: Mm-hmm.

Dr. McGavock: Using more rigorous, and I even think more exciting and like what I love about this work that the economists are doing, that you're doing, it's like you g- you gotta think through it. Is there a better way? And then when you find the better way, the answer is so much more [00:46:00] satisfying to me.

Dr. McGavock: Mm-hmm. Positive or negative, you've just done a better scientific experiment.

Dr. Boone: Yeah. I will also say it takes a very long time. Yeah. So the means that had a big difference that I was describing before, that occurred in 2020. Yeah. It's hard to do these like extremely rigorous natural experiment style papers and do them well, but I do think it is worth it.

Dr. McGavock: So what's your next lightning bolt?

Dr. Boone: I mean, I'm generally interested in anything related to chronic or cardiovascular disease. The link between these two papers is the outcomes which are... Or the conditions people have. For me, that ties back to behavioral economics a little bit. Some things I'm working on now are related to hypertension and blood pressure.

Dr. Boone: It's very, very hard to measure blood pressure, which means there's a little bit of discretion in terms of whether a [00:47:00] physician believes that a positive test for hypertension is really positive or maybe you just happened to test over 140 today and you won't tomorrow.

Dr. McGavock: Mm-hmm.

Dr. Boone: So I've been doing some work that uses a different and less well-known quasi-experimental method called bunching.

Dr. McGavock: Mm-hmm. I read about that on your website.

Dr. Boone: Which is, yeah, so-

Dr. McGavock: I like that ...

Dr. Boone: blood pressure, for example, it's like this biological parameter and we know what it should look like on a population scale, which is a nice normal distribution. What you actually see in a few different clinical data sets is there are some abnormalities around the diagnostic threshold that suggest that some patients who test just over 140 are being reassigned from positive to negative by their healthcare provider.

Dr. McGavock: Hmm.

Dr. Boone: So I've been doing some work to try to understand if this is a good thing. Like maybe the providers are correct and we're avoiding [00:48:00] diagnoses that we didn't need to do, or is it bad and we're now delaying diagnosis in some people who should be on medication, which could be bad for their health.

Dr. McGavock: Mm-hmm. Sign of an awesome question, doesn't matter what the answer is. It's clinically relevant. Yeah. Very cool. Mm-hmm. Very cool. Yeah. Our lab's doing work on the natural experiments of new cycling infrastructure in different cities. So- Oh, wow ... the lightning bolt of like a new five-kilometer trail that's allows you to get to work and back.

Dr. McGavock: Um- Very cool ... a lot of them have counter data or if you're into Strava, we can download Strava data. Yeah. So I can, we can see who's using what. So some trails are different than others. There's heterogeneity there. It's a natural experiment. It's just dropped in a neighborhood- Mm-hmm ... at random. We didn't drop it because there's high rates of hypertension there.

Dr. McGavock: It just was there. So then we're working with five different cities to look at hundreds of different trail natural experiments and then we can- Wow ... use health data on either side of the trail to [00:49:00] see what happened to Are

Dr. Boone: leaving Montreal. We have so many bike lanes that have opened recently

Dr. McGavock: Not yet, no.

Dr. McGavock: We've got to find the data. So right now our colleagues are in Sherbrooke and Quebec, but yes, it would be amazing if we could get health data for Montreal too

Dr. Boone: And one other idea, I don't know if you heard this, but Toronto is- Yeah ... removing bike lanes.

Dr. McGavock: Yeah, yeah. So

Dr. Boone: you've heard. It can be nice sometimes to have un-adopters in your data.

Dr. Boone: It can help with power and stuff. You could add that.

Dr. McGavock: That's sad. We're all very pro cycling, so it's like-

Dr. Boone: Yeah ...

Dr. McGavock: when we hear that, it's like, oh yeah, that's a really good counter to our, you know, we could triangulate our data. Mm-hmm. Um, so last question for you is you probably get asked to review papers all the time.

Dr. McGavock: Like how do you determine what is a good looking natural experiment versus one that maybe needs a little bit of work? What are some of the criteria for you that constitute very good natural experiments?

Dr. Boone: Yeah. So each of the different methods, like diff in diff, [00:50:00] IV, et cetera, they have essentially checklists that go with them that can help determine whether it's a valid experiment.

Dr. Boone: So we've mentioned the one in our sugar paper. We were hoping that the people were comparable on either side of the threshold. There's three or four of those that relate to each of the methods that I go in, like right away before looking at the results. I would check to see if they passed each of those tests, and if they don't, you can ask for them.

Dr. Boone: It's possible they just didn't put it in the paper.

Dr. McGavock: Like you could do the analysis on any data set, right?

Dr. Boone: Mm-hmm.

Dr. McGavock: It's only valid if they meet the assumptions.

Dr. Boone: Exactly.

Dr. McGavock: So are you going through that first to say, like check one, check two- Yeah ... check three? Okay. Is this a list that's published- Um ... that I could go find or

Dr. Boone: I find myself going back to like my PhD course.

Dr. McGavock: To

Dr. Boone: check books and stuff, but I think it, it should-- If it doesn't exist, it should exist. As

Dr. McGavock: a- Let's

Dr. Boone: make

Dr. McGavock: it exist. Let's put it together. Yeah. [00:51:00]

Dr. Boone: I'm interested. Yeah, me too. The Columbia School of Public Health has some pretty good websites about this that every time I click them, my Google says, "You visit this page frequently."

Dr. McGavock: Me too. It's at the top, and it says, "You've been here like 50 times in the last two weeks," because it lists very clearly the assumptions on difference in difference analysis. That's a good paper. We should do analyzing and have the different models, difference in difference- Mm-hmm ... instrumental variable, segmented regression.

Dr. McGavock: I don't know anything about bunching, but if bunching falls into a causal approach, then that's really cool. Okay, so you line those up, make sure that they meet those other criteria?

Dr. Boone: I think, yeah, if they, if they don't meet those, it's not worth spending a lot of time thinking about the results because I don't know if I believe the results if those aren't met.

Dr. Boone: But if they do, then you have a great natural experiment, good to go, and then you can review the paper as you would any other paper. Like, is it clear? Did their literature review good? Are they interpreting well but not [00:52:00] overstating the results? Things like that.

Dr. McGavock: Right. And thinking about it from the beginning, talking to Bapu Jena about this too, they c- sort of had that checklist, right?

Dr. McGavock: Like this is a good idea. Can we meet these assumptions? We can't meet them. It's not even worth it. If I can't do it well, I'm not gonna do it. So if you can't find the data, it's not worth it. If two of the assumptions aren't gonna be met- Find a different question, find a different lightning bolt, 'cause you're not gonna answer it right now.

Dr. Boone: Yeah. I mean, we've talked about two papers here, but I have, like, the iceberg of ones that you don't see where I thought it was a cool natural experiment and the assumptions weren't met, so I didn't pursue it further. So I think that's how economists approach research a lot. Like, there's a very high quality standard for meeting these different assumptions.

Dr. Boone: And so I think it's really good to try to fail fast by just, like, very quickly checking if those work out, and then only if they do [00:53:00] should you go forward and spend four or five years on this question.

Dr. McGavock: Totally. Absolutely. Because it's not worth d- going that far on something that from the beginning you knew right away you weren't gonna be able to answer the question.

Dr. Boone: Mm-hmm.

Dr. McGavock: So having a high causal standard I think is really, really key. I agree with that. Right on. Anything else we didn't talk about that we should think about f- in terms of, like, natural experiments or the different approaches to looking at lightning bolts?

Dr. Boone: I think there should be more people doing this kind of research, like quasi experimental research in health and especially healthcare.

Dr. McGavock: Mm-hmm.

Dr. Boone: There's so much interesting variation in healthcare that we could be using to do this kind of work. I don't see most of it 'cause I'm not a, a clinician. Mm. But I think we need more Bapus in the world, people who are physicians and have this kind of training, or people who are physicians who partner with economists like me.

Dr. Boone: Yes. [00:54:00] I would just-

Dr. McGavock: Yes.

Dr. Boone: Yeah. I encourage anyone who's interested to try to get training in these methods, and not just take one class, but take many. It's hard to grasp on your first go around.

Dr. McGavock: Go find the Dr. Boones- Okay ... in your environment and be like, "I got a question," and then work through. Like spend some good time like going back and forth on things you've observed in the healthcare world, right?

Dr. McGavock: And then- Yeah ... for Claire it'll, you know, it'll be like, "Oh man, this is a great question," and then it's Sherlock Holmes-ing finding the data that can help you answer this in the right way. Honestly, there's this disconnect right now in health epi and health economics-

Dr. Boone: Yeah ...

Dr. McGavock: where people are in separate camps, but I think the more that they can work together, the better- Yeah

Dr. McGavock: these projects are gonna be. So yeah, if you're in Montreal or in Quebec, come find Dr. Boone's lab. Brand new lab. Thank

Dr. Boone: you. Yep.

Dr. McGavock: Tackle some cool questions. Any favorite paper or favorite thing, like something that essential reading?

Dr. Boone: Honestly, all of Bapu's papers And are what I think of as [00:55:00] like the most exciting difference in different papers in medicine specifically.

Dr. McGavock: Yeah.

Dr. Boone: Uh, yeah, I would start there.

Dr. McGavock: Agreed.

Dr. Boone: And if you're interested in behavioral economics in medicine, I would check out Kevin Volpp, who is at Penn, and he has, I think, pioneered a lot of the using behavioral econ to do cool interventions in healthcare-

Dr. McGavock: Oh, cool ...

Dr. Boone: field.

Dr. McGavock: Cool.

Dr. Boone: Yeah.

Dr. McGavock: Yeah. We'll link to that Kevin Volpp's lab, and we'll take a look at what they're doing there.

Dr. McGavock: Awesome. Thank you very much, Claire. Really appreciate your time. We'll put a link to all your work here and then some of that other work there, and hopefully some residents are gonna come your way, and trainees are gonna come your way to do this kind of work.

Dr. Boone: Yeah, I would love that.

Dr. McGavock: All the best in your work, and we look forward to crossing paths one day.

Dr. McGavock: A huge thank you to Dr. Claire Boone for joining us today and for making science of natural experiments feel so alive and so relevant. Whether it was the sugar rationing study that [00:56:00] Dr. Boone published in the journal Science, the work on preventative care, from behavioral nudges in Chile to lifelong echoes of early childhood nutrition, I hope today's conversation opened your eyes to just how much we can learn when we treat the world as a laboratory.

Dr. McGavock: If there's a takeaway for this episode from me or for you as an emerging scholar, it's this: the next great study might already be happening in your hospital, in your city, in your country. You just have to know where to find it, how to look for it, and how to find data to study it. Links to all the papers that we discussed, all the resources we mention, are in the show notes today, including Dr.

Dr. McGavock: Boone's own work, Dr. Bapu Jena's book, Random Acts of Medicine, and some classic texts on behavioral economics. If you enjoyed today's episode, please subscribe, leave us a review, and stay tuned for next week, where we interview Dr. Deborah Lawlor about triangulation and do a deeper dive into other methods that could inform causal inference.

Dr. McGavock: We'll see you next week The Emerging [00:57:00] Scholar podcast was recorded in the UMFM studios located at the University of Manitoba in Winnipeg, Manitoba, Canada. This is Treaty 1 territory, the original homeland of the Anishinaabe, the Ininew, and on the national homeland of the Red River Métis. Manitoba is also homeland to the Anishininew, the Dakota Oyate, and the Dene peoples.

Dr. McGavock: We acknowledge that our water is sourced from Shoal Lake 40 First Nation. As a settler scientist working in health research, I acknowledge TRC Call to Action number 18, that health inequities between Indigenous and non-Indigenous people are the direct result of colonization. Winnipeg is the center for reconciliation in Canada, and where we are speaking from today is at the crossroads of Canada's history and Canada's future.

Dr. McGavock: We are three kilometers away from one of the most sacred meeting grounds for Indigenous people on the plains and within Turtle Island. We are also several kilometers away from one of the last residential schools in the province. We are working collectively on [00:58:00] reconcili-action with youth, elders, and community leaders to heal from the trauma of colonization, past and present, honor the original treaties, and create a Canada that addresses and eliminates the inequities and injustices of the past and present to create a future for Indigenous children to live and thrive.

Dr. McGavock: On behalf of the team at the Emerging Scholar podcast, we want to thank Osani Balkaran for the great production for our opening and closing themes. We also want to thank UMFM 101.5, especially our executive producer, Jared McKetiak, for technical assistance and production on this project. We also want to thank all of our guests for the discussion and the insight that they've provided.

Dr. McGavock: Once again, thank you for listening. Be sure to subscribe to the Emerging Scholar podcast on Apple Podcasts, Spotify, and umfm.com.