1.5. Putting the Social Back in Social Epidemiology with Dr. Whitney Robinson

I think you have just changed the way that I'm going to review applications and things. You're welcome future students in that box. Welcome back to Sirius Epidemiology, a new podcast from the Society for Epidemiology Research. I am Haley Bannock from the University at Buffalo and I am pleased to be co-hosting this podcast with Dr. Matt Fox from Boston University. Today we are delighted to welcome to Sirius Epi, Dr. Whitney Robinson, who is an associate professor at the University of North Carolina, Chapel Hill, Gilling School of Public Health. Whitney's research is focused on why black women in the US experience worse health and health care than other Americans, in particular when it comes to gynecologic issues, cancer and obesity. Dr. Robinson focuses on Epidemiologic methods for health disparities research. She focuses on how and why the processes of obesity and cancer development differ by sex, race and ethnicity. Recently I know she's been doing some very interesting work investigating how racial and ethnic well-associated economic inequalities in gynecologic surgery among US women affect disparities in fertility and health disparities overall. Whitney also co-hosts the podcast "Acadames" and if you have not listened to it, I very highly recommend it. Welcome Whitney and thank you for joining us today. Thanks. It's great to be here Haley. We're very excited to have you here to talk with us about social epidemiology. But before we get into the hard stuff, we always like to begin with some lighter questions so some guests can get to know you better. So first, can you tell us what your favorite movie is or a movie that you could re-watch over and over again? So I'm actually not a movie person, I'm more of a TV person. And I recently watched "Never Have I Ever" on Netflix. Oh yes. It's like a high school show. It's so good. And I just finished it and I can't wait to re-watch it. I see. Yeah. I think that I should be watching with my kids or I should be watching myself. You can watch it with your kids. You have good conversations. I think you're like so. Okay. This might be awkward, but it's good. Okay. All right. Matt has teenager. So this is important stuff for him to know. It's a great show. Also, Haley recommend. All right. I'll look into that one. Yeah. I have a second one that is a specific episode of a TV show. There's a Netflix show called "Master of None" and there's a Thanksgiving episode. That's amazing. It's just the most beautiful piece of television I've ever seen and I have yet to re-watch it, but I think I might put that on as a re-watch soon. Good to know. All right. I have not seen that. So now I'm going to add it to my list. Tell us, what is your guilty pleasure? My guilty pleasure is podcasts. I listen to podcasts. Incessantly. I listen to podcasts while I'm out walking my children and I'm ignoring them and they're talking to me and I'm like, mm-hmm, mm-hmm, and I'm just like listening to my podcasts. So it's great. I've perfected this one-year-bud in, one-year-bud listening to the children and hear emergencies but also learn something at the same time. When I need to concentrate, I says, "Mom, mom, I need some quiet time." Which does have some quiet time. What are some of your favorite podcasts? I listen to so many podcasts. I was just listening to one that is personal finance for PhDs. I also love personal finance and I moderate a personal finance Facebook group. And you talk about it a fair bit on your podcast, which is fantastic, I'd say. So I just listened to one about somebody who is an academic. He moved from his PhD to his first job in Savannah and bought a house and it ended up being a money pit. And to the whole episode is I'm talking about how his first house was a money pit. And I love it. That's interesting because I know nothing about finance and I feel like I should know more than I do. So I need to really, maybe I'll join your Facebook group, it's really good. And the third question before we get into the real stuff is, tell us, what do you think is the coolest bias out there? So I personally think the coolest bias is selection bias as a find, as like structurally defined by Miguel Hernán and his work, which also leads into one of my favorite papers. I just think it's fascinating. I always have to sit and think about it and think about the direction of the bias and it just pops up everywhere if you're looking for it. It is my favorite bias. I'm with you. Haley is all about that bias. I love it too, but not as much as you and Haley do apparently. And that paper, which I've referenced on previous episodes about a paper that everyone and I read all the time. Okay, great. Well, I'm glad we got to know a little bit about you and that we share some TV preferences. We like to start off a bit general with our conversation and work our way to the more specific. This podcast is intended for trainees or practicing epidemiologists who want to learn more about a topic. And certainly I would say I'm not an expert in social epidemiology. I would say probably Matt would say the same about himself. And so we're here to learn some more about social epidemiology and I think most people have a general idea of what is social epidemiology, but maybe I wish they would tell me. So I guess this actually might be a tough question, but but the question I was going to start out with is how do you define social epidemiology? So there are many definitions of social epidemiology and it's really contested and then kind of a whole meta issue is whether there should be any social epidemiology because people like Steve Wing, one of my colleagues who's deceased now, always said that all Epi should be social Epi. And so he was a social epidemiologist who refused to be called a social epidemiologist so there's that whole thing. A definition that I actually like is that it's basically all the Epi that separates epidemiology from just being veterinary science because we're all animals, right? And you know, we could just think about epidemiology as just veterinary science of humans, but it's more complicated than that because humans are more complicated than that. And so for years, I've been using this in class and saying that I got it from the Coffman and Oaks methods in social epidemiology first edition, but I never actually pulled the quote. So when I was preparing for this, I'm like, I should look to make sure it's actually, I don't like look through the book. And I think I found the part that inspires that definition and it's not exactly what I thought, but could I read a paragraph? Yes, please. Okay, so this is from Methods in Social Epi Demiology by Jay Coffman and Michael Oaks. Social epidemiology is different from the bulk of traditional epidemiologic practice, which tends to operate with a model based on the fictitious Robinson Crusoe. Recall that this character is someone in an environment devoid of social context, whose health depends only on biological relationships and the vicissitudes of island weather. Social interaction and thus political and economic power play no role in Robinson's health. Although the same is perhaps not so true for his quote unquote "friend" Friday. Large interactions are central to social epidemiology, however. Without any attention to social arrangements and institutions, epidemiologic research on humans is almost indistinguishable from an application to say livestock. It is the incorporation of propulsive human interaction and agency that is social coordination and conflict that links social epidemiology to the social sciences and raises enormous methodologic obstacles to inference. So that's basically it. That's how I think about it. That's a great definition and I really like that it rolls into the definition all of these different components that you really need to consider because we're not just interested in studying very specific risk factor X and its relationship to why we need to consider these broader social forces. So I think that's a great definition to consider. So I'm curious. SER brings together epidemiologists from all kinds of different disciplines and yet many of us so I do HIV research and we have our own HIV meetings that are separate from the SER meetings. Does social epi have a home outside of SER? We all think we're like special snowflakes. I don't think we could all agree on one meeting. I was in the Robert Whit Johnson Health and Society Scholars program which has a lot of people who would kind of loosely affiliate with social epidemiology and when they had their annual meeting that would be the place I would go to get my like dose of social epi. Now another meeting is IAPHS, Association of Population Health Sciences. I always forget what it stands for and it's a newer meeting that's kind of in that spirit kind of demography, social science, social epidemiology and health. So for me, IAPHS feels some of that but people who have different niches in social epi probably have different meetings they go to. It's a very complicated field because it overlays so many other topics. I mean for Matt's HIV researcher, my obesity research there needs to be social epidemiology in each of those different domains which makes it a very complicated problem. So that I guess brings up our next question which is why do you think it's so important to study social epidemiology? I think part of that came from the definition that you mentioned but I'd love to hear a bit more about your thoughts on that. I just think it's involved with everything. I think all epi should be social epi but for better for worse the tradition especially in the US hasn't necessarily been that way. And so I think to have a subfield that really focuses on things like inequality, racism, social support, social relationships, built environment is really important. Under failing and links, fundamental causes theory of socioeconomic status and mortality they have this argument that disparities by socioeconomic status are always going to keep replicating themselves. So let's say the main disparity in a culture is about dysentery. And then you find some magic bullet that gets rid of dysentery. So you think okay all our inequalities are going to go away but it doesn't happen. Whatever is the next cause of death is going to find an inequality as well as there gets to be increased knowledge or resources to avoid mortality. Those are going to be distributed unequally in society and you're just going to have new disparities replicate themselves over and over again. So I think kind of a whack-a-mole focus on okay I'm just going to focus on this disease and once I figure out this disease everything's going to be great misses the boat. Without an attention to the larger structures that give rise to inequalities they're just going to keep replicating themselves and you're just going to keep chasing after inequalities and it'll health over and over again without an attention to the fundamental causes that kind of a portion risk and mortality in the first place. Wow. That's a lot to think about. I mean it's hard for me to think about that and I very much agree with what you're saying but also reconcile the traditional epidemiology, the traditional methods that many of us use and do in our research and how is it possible that I don't include these concepts in my research because you're completely right on exactly what you're saying yet it's not something that I incorporate on a regular basis and so I'm left feeling like my research isn't complete or I need to consider more than I am considering because it's not the risk factor in the disease necessarily that we're interested in with the dysentery example. And some people would say well even if I'm just focusing on my one disease and I manage to drive down the overall rates in a population even though the inequalities will still be there and maybe pop up in other things at least I have diminished the suffering from that disease and there's something to that but I just think the absolute burden of health can't be adequately addressed without thinking about social forces. It does feel to me like we have made a big mistake as you said in the beginning of not taking this view that all epi is social epi because it does feel to me like we teach you know an infectious disease epi course there is social epi undergirding all of the problems that we're trying to deal with and yet we don't talk about it that way. We have methods classes where we teach students different approaches but we're not teaching them how to think in a social epi context and I think we end up making a lot of really fundamental mistakes because we don't do that and I will be the first to say I'm quite guilty of this and I'm only starting to realize a harm that is done by not viewing everything to the social epi lens. I mean I think it also gives social epidemiologist kind of a superpower because we think about the big picture patterning of disease by sex, by race, by SES, by geography and so often I can take a quick look at a table one or read a paper and I can know a lot you know I don't do a lot of my own coding anymore I rely on students and programmers and you know ideally we'd have like double and triple coding and stuff and so I'm often relying on my own intuition and analysis to identify if something's right or wrong with data and having a good sense of kind of what you'd expect to see in the distribution and patterns helps a lot. Like I was just reading about one of the recently retracted COVID papers in the Lancet that had gone through review and then people are reading it and they're like this doesn't make any sense. So one of the things was that like it's this multinational 90,000 person study and like race has no missingness and across like all these different countries and I'm like first of all do they even collect race in the same way like they don't and like what's going on and they're just all these things that basic demographics they were talking about like the smoking patterns didn't vary like they should you know if you know something about smoking by a country's level of development I think when you really have a good understanding of kind of social patterning of a lot of behaviors and health outcomes and things you also can just kind of quickly pinpoint issues and anticipate issues. Since you brought it up I just have to say I'm so glad I was never asked to be a reviewer of one of those papers not that I ever would have been because I would have missed all of those things. I never would have picked up despite the fact that there were some very obvious problems. I would have been one of those reviewers said yeah it makes sense I'm so glad I wasn't. Well there's still anonymous right so we'll never know. At the moment. Yeah. So I don't know I think it's valuable. For the field I think having more social happy perspective would be valuable and also as an individual I think it can kind of give me an edge in anticipating issues with analyses that are already done or in thinking ahead to issues that might come up and in analysis. Like I can think about dissertation committees I've sat on where I'm like you know well what will you do if the results come out this way knowing because of the evidence and the strong social patterning of his outcomes that it might go a certain way and then if the dissertation defense he's like well what Whitney said would happen is exactly what happened you know. So yeah. So this idea that all epidemiology, social epidemiology I've never heard that quote before but it's really got me thinking and do you think that there's sort of a two-step process that many people could or should think through which is first we need to know if your exposure is linked to your outcome associated causely and then start exploring what I would maybe call effect modification by some of these or interaction by some of these social factors or can you do it all at once. I'm just trying to think through how people might start to think about how to incorporate these types of variables into analyses that they're doing because I don't really know how to do that. I mean I know some of the methods and whatnot but the broader considerations. I think some of it's iterative. I think COVID is a good example. So here we have a novel disease. And so I think the first thing we do is like well who was it affecting? I think one of the first things I noticed about COVID was the age distribution that was different than the age distribution for mortality among like flu or something. And so I think I kind of start with looking at the big picture basic epipatterns. Who's affected? Thinking about age. Thinking about sex. Thinking about SES. Thinking about geography. So that's how I approach it. And then from there I can use that to hypothesize about some causal factors and also to kind of vet hypotheses that come about. So when people are thinking about a transmission pattern and whether we should open schools and like well, kind of these age distribution is maybe giving us some clues about who transmits more or less. So I think for me is that first want to know the epidemographics by any social strata that tend to be correlated with health for even other diseases that helps generate hypotheses about risk factors use that first base of information to kind of screen the plausibility of different risk factors to know what you should prioritize as far as kinds of investigations. Then as you get to know more about causal mechanisms and risk factors then circle back and say well how are these distributed in populations to be able to anticipate what kind of inequities are going to arise. So early on people were like a COVID's equal opportunity but I work loosely with a group of black epidemiologists and we're like nope, it's going to start this way because it's being introduced by pretty wealthy international travelers who are seeding it in different places like that's how it got into South Africa, that's how it got into the U.S. but ultimately from what I know about fundamental causes and what I know about how this is spread, you can just see how the inequities are going to arise. It's a fascinating example of public health and epidemiology playing out in real time you know with this idea that it's I forget you just use the quote you know it's equal opportunity or whatever that quote was and no obviously not that is not going to be the case even if the very very first bit of evidence might suggest that you're right the social epidemiology lends the you know investigative superpower that you possess you know very clearly gives you an indication that that is not how it's going to play out for better for worse for worse mostly. Well not your superpower I know the worst for population health. Then we also like have tools so the theory also says like fundamental causes the flip side of the theories that there are some kinds of interventions that are going to be more powerful for reducing inequities than others so it's not all bad news but there's sometimes to be political will to do those kind of things. And so what would be some examples of those? So people talk about behavioral interventions or expensive interventions that rely on individuals to have to use resources to buy health are going to be ineffective for reducing inequities. So if it's that you know you have to pay for some expensive treatment like that's not going to do something that is going to reduce inequities like that's kind of the extreme often kind of policy solutions that affect the whole population like seat belt laws go a lot towards decreasing those inequities or vaccination campaigns. And so I think a lot of social epidemiologists are interested in policy because when you have population wide policies those often reduce inequities versus having to rely on people to use their own time and capital and privilege to be able to access limited resources. So the policy things some targeting can be effective and trying to also tailor interventions that are culturally relevant like often a first wave of treatments and interventions will be those that are most accessible to people who have the most privilege. You know like even something like refrigeration of a medication you know if you're trying to go to a lower middle income country where everybody doesn't have refrigeration that's not going to be the best thing but I think people who might come from a rich wealthy country think refrigeration it's fine and you know like even just the process of formulating what interventions are feasible and effective to deliver if you start from your own context you might miss doing something that is actually feasible in a different context. This conversation has gone towards this COVID topic and you know related to your interests in personal finance I've been struck throughout this process about how inextricably linked public health and epidemiology are with economics. In particular in this time where unemployment is really skyrocketing schools are closed so how do parents go to work these are all issues that strongly affect the health of the population and I don't know how we're going to get out of it and this is not a question I assume you have an answer too I wish you did that would be fabulous but I'd love to hear some of your thoughts about you know the current situation we're in and how the social factors like economics affect that. Yeah and I mean I think this also gives us a different lens it's interesting because when this started among my social epidemiology friends you know we thought there's going to be big inequities that's going to be a problem with the current course of how we're handling this in the US but also there's going to be the spillover effects and we were immediately thinking about the long-term effects of economic contraction and especially about kids like from the beginning like we can't keep kids out of school for two years or I mean who knows maybe we can we live through disasters but like hopefully not because I think we think broadly so we don't just think about via immediate disease like we also think about socio-emotional development and how profound that is and so that's one reason why that was immediately on my radar when I didn't hear other people talking about that because I'm like this is something that's not going to look like a crisis but is a crisis is going to be brewing under the surface. So life course is also part of social epidemiology that the things that happen when you're young will manifest later in poor health and we have these critical periods. So I think for us because we think about populations broadly and not just like one disease but kind of the range of health and the whole population I think we do tend to think in a more nuanced way about risks benefits but also we're depressed because we see all the problems coming down the pike that will be with us for decades literally like under a best case scenario there's probably going to be lasting trauma on people that experience this during certain critical periods and we will maybe literally see in 20, 30, 40, 50 years. We are largely guided by this medical model and oftentimes as we've talked about we don't fully consider the social factors that influence our health and I just wondered if you have thoughts on maybe more of a historical perspective or how we ended up at this point where we really focus on one more than the other replace I suppose more importance on one compared to the other in our research. There are people who are really good like the history of FB I mean some of it's just funding in the US a lot of our funding comes from National Institutes of Health which is just a biomedical focus I'm not sure how that translates for our international listeners. I'm also really influenced by Shonda Ford's Public Health Critical Race Praxis which is like a multi-pronged theory but one of it is about kind of interrogating who we are as researchers and what we're drawn to and if you have a research community that tends to be high SES and pretty advantage they're not thinking a lot about the social forces that shape their health at advantage they are thinking about well the things that tend to affect me or maybe cancers that kind of spring out from family history or you know a genetic commutation and so I think sometimes we don't think about the ways in which our own experiences shape what we think is important and this is one of the reasons I think it's very important to diversify the body of researchers who do work because it actually changes the questions that are asked and valued. I think that it's also just a larger issue with public health people take for granted deaths that are prevented I mean this is just the thing with vaccines and the things with all public health and our aim is to do interventions at a population level that mean that people don't get sick in the first place which is powerful and it's the reason I'm drawn to this field and I love this field but it's not sexy people literally forget about it. Like I think the alleviation of poverty and the elderly through Medicare and Social Security and Medicaid is like one of the great triumphs of the US and people just take it for granted they're like all this is just the way it is but if it was removed tomorrow I mean the amount of suffering and the elderly that we would see would be just tremendous and so I think one people just take our successes for granted versus a medical innovation you know you have cancer you go in and they treat you and you live and that just like is powerful and it captures your imagination and it makes you want to like donate all your rich person money to the canvas center having to have a named after you. Yes or the congressperson who like comes up with like this disease is something we need to put money into so some of it's just who's in charge of deciding what's important or not and I think some of it's a normal human tendency to value the narrowly averted crisis versus the crisis that was totally prevented. So one of the things that has really struck me over the past few years as people have been talking about this on Twitter and various places is the point that you just raised a minute ago which is that we tend to think of these decisions that we make around what to study as being value free. You know we're studying cancer because cancer is bad and you know I don't have to make a decision to study cancer over studying something else because everybody agrees that cancer is something that we all want to eliminate but you're making a value statement in the decision to study whatever it is you study and making the decision not to study these larger social forces is doing the same it seems to me I'm curious your take on that. I mean I see it with my students I see the students who are most passionate about thinking about the intersection of race and poverty on cancer like anything has social determinants like they're a lot of their passion for asking these questions is because they have seen it play out in their lives and I think sometimes when I'm reviewing applications if somebody's like I want to study this because somebody in my family had this I can't kind of be skeptical because I think that won't get you all the way but it can inspire you to have insights about a process that other people simply don't have like that experience of saying I went through this process with my grandmother my mother and I saw how she was treated in the hospital and it was bad and where I saw this or I saw this gap in the insurance like there are people who literally will not know that's happening because they haven't experienced it and nobody in their family has experienced it but it is happening to a mass of other people. I think you have just changed the way that I'm going to review applications and think about you know those comments because I've always found those motivations to be really interesting but I never really thought about it in a way you just described it which is that it provides an insight that I think I would be lacking because I hadn't gone through that. Nobody cares about my perspective on on incoming students at this point. However what you said previously really resonated with me in terms of diversifying our perspectives and why that is so important the things that I think about might be different than the things that someone else thinks about and that is why we need a diversity of backgrounds a diversity of perspectives in our field not just to play lip service to the fact that let's be more diverse and let's be more inclusive but it will fundamentally change the research that we all do you know as an SER we could see differences in the future in terms of research priorities by changing the makeup of the people that are part of the society so I think there's a really a very real world implication to what you are what you're talking about. I mean one of the things that I've been convicted and is my huge gap about native health and indigenous health and working with students who are indigenous and who want to do this kind of work has opened my eyes to a part of US history that I did not know enough about even though I think of myself as being well read and this like you know woke social epidemiologist like I have my own gaps and I just really appreciate being educated. That's great. I'm going to shift gears a little bit to talk more on some methodsy topics that I know the three of us really enjoy talking about I want to make sure to leave space and time for these topics so one of the issues that I often think about when I think about social epi is measurement and measurement concerns and how difficult it is to accurately and reliably measure the constructs that we're talking about you mentioned earlier are they even asking the same questions across the different countries and probably not so there's so many different measures of what SES is or what you know neighborhood SES is or you know are using education or income to represent one of these two concepts. How do you think that these measurement concerns have impacted the field of social epi broadly? I speak a lot to the US because I think this will vary. I think one of the reasons we're lazy with our SES measures is we can soak up so much variation with race in US because we're such a racially stratified country so I think that's in the background I think other countries do do better with SES although they have gaps about their own kinds of stratification I think it's a country we have myths about class and economic mobility that actually make us not want to invest a lot of public money and really measuring these markers of wealth and class because you know I get a lot of my data from the US government and the US federal categories for race and guide a lot of things so I think a lot of the tools that we have emanate kind of down from federal government priorities also I think our federal government downplays divides by class and has not made it a priority to have better measurement and so then a lot of other studies don't and I think another thing is that people don't like to talk about it so I always tell people don't bother trying to ask people about income 20% of people are not going to tell you you know like just a straightforward income question people just will refuse to answer and I think some of it is just about like a US mentality you know I love talking to some of my Chinese colleagues really just like so blunt about money and I love personal finance I love talking about money so I appreciate it so much but I think there's a cultural thing like if it's rude to talk about money and I think that plays out even when you're serving American population so I was a piece of volunteer in Turkmenistan in the 90s yeah you'd meet people and they would say what do you do for a living and then how much money do you make love it and like it fell so rude to me but it's just a normal question to them it's not rude at all yeah but I think the fact that it was so rude to you I think that's deeply embedded in the US so it's actually hard to do this work because literally people don't want to tell you right and increasingly that's an issue with race so more and more there's white people who don't want to do race and so in the past race was this like really easy variable everybody would report it and it like soaked up a lot of variation so we kind of got lazy and so I think topics about yeah there's a whole thing so I think there's even going to be more missingness with race so you know some of the most profound data I've seen is not directly at B but like economists who are using federal tax data to really think about economic mobility and where people live like Raj Chetty is one of the authors on this work and it's been talked about in the New York Times so I think increasingly we're gonna have to get creative about the way we get data because I think just asking people isn't working that well and maybe there's a way around it I think sometimes people will use um like social status letters or they will ask about assets but I think you do have to be creative and it's hard because you have just the conceptual thing about what's the construct I want to measure because income is different than wealth it's different than education it's different than assets but then also the social context about people not wanting to answer do you think that differs from lots of other self-reported variables that we use is reporting your weight any different is reporting your smoking status any different you know there's lots of variables that one might say there's some stigma attached yeah you know with reporting those variables yeah weights a good example and I don't like using self-reported weight because it also has all these patterning of self-report by gender and races so I'm like measure them but because you can't measure people you know and I think in a lot of studies I don't know if people refuse to be measured that could be a thing but I think most people will just let you measure them if you don't make them look mighty right but in many cases there's no analogous objective measure to the social unless you go to the tax data or something and people cheat their taxes but you know it's if you do you have trade-offs right exactly another question related to methods is about you know when we read papers that talk about including race either as some kind of binary variable or some kind of categorical variable in a regression model to talk about quote controlling for the effects of race what does that even mean I struggle with understanding what that means and it's in so many papers you know it's just kind of a sentence that's often a clue we control for age and race can you help us understand a little bit to you what it means to control for the effective race hey you asked a good question you're like oh just explain this um it's complicated okay so like everything it's complicated just explain the history of the universe just just give us that yes we have 10 minutes so go ahead what does it mean so I mean you could think about first principles like in a lot of those intro bio stats classes they start teaching you about a nova as the basis of linear regression and I know a lot of that like flew over my head because it's all decontextualized when you learn it but a lot of these models you're trying to soak up variation and for better or worse so many things track with race that it ends up soaking up a lot of variation so just in a purely statistical what is it doing in my regression model it's just soaking up a lot of variation but it means so many different things as we've talked about I think that people do it a lot as it does soak up a lot of variation because it's going to be a significant variable because of fundamental causes and the social racial patterning of life in the US so you know it feels good when you see that it's statistically significant it's making your estimates move and you're like oh I'm doing something you know like that feels nice and it's just kind of easy I think often the assumptions about the mutual independence of the race quote unquote affect from other things like I think those assumptions are violated but you know their models converge and it works so I'm interested in this topic and you know I've done some work I did a paper with Tyler van der Will and done some other work thinking about this I am a person who's not really into the language of race affects so I talked about that in some of my work so in the absence of thinking about it as an effect if it's just a covariate it's just soaking up variation if it's your main independent variable then I think about it as a marker of disparity in the population just what's the gap in the outcome given different adjustments which is its own thing I don't think that answers your question exactly no it does because you know like you said when you include race it often changes your estimates right and so you get this good feeling like oh I satisfied that 10% criteria that everyone has to have for a covariant to be included in a model but nobody really knows actually what you are controlling for well maybe some people do I don't really know what you are doing most people who do it don't or they have an idea that's wrong which could be another podcast yes exactly but it's sort of a composite variable you may call it race in your data set but it's really a composite variable that's sort of melding in a whole bunch of different things okay I think the last question we'll talk about is about causal inference in the context of social epidemiology and how you think about evaluating cause and effect in the context of social epidemiology because it's so hard to think in terms of counterfactuals and variables that you can actually manipulate in terms of exposures in many times in social epidemiology yeah I think there are some social epidemiologists who really push back against the growth and influence of this potential outcomes framework the idea of everything needed to be a well-defined intervention but I actually don't I really like these frameworks I think that it has clarified some important things for me like the idea that I don't think of race and sex as like effects comes directly from this idea of potential outcomes and thinking like what will be the well-defined intervention is there a well-defined intervention and so a lot of my work has been influenced by this perspective and clarified by it and kind of pushing to think about okay what would be the intervention but I think one thing that is a misconception about these frameworks and the people who are really into them would agree with this is that people think oh it has to be something that I could really imagine doing in real life and that that's not what it's about it's not just what you could plausibly imagine doing as a randomized critical trial at a hospital when I think about counterfactuals I think about well what if the civil rights law of 1964 had never been passed or what if reconstruction had actually been implemented to its full force what if there were a national policy of some kind that changed resource allocation you know like there's so many what ifs where you can maybe imagine the world as a different place you are not going to be able to estimate that using a regression model in a cohort of people who currently exist so I think there's a distinction between imagining the counterfactual you want and what could actually do and what you can actually get out of a regression model with a traditional like epi study design but I think the practice of imagining the counterfactual is very powerful because it helps you clarify what actually is my research question that thing with putting race in the model and being like we adjust it for it where people actually have no idea what they're doing like I think pushing the idea of potential outcomes and well-defined interventions pushes us as researchers to ask better questions or realize when we're asking bad questions and I also think that it helps you say well for the question I really want to ask I'm not going to be able to get it out of multivariable adjusted regression in this cohort of people so I'm going to need to use different study designs I love beautiful study designs I think about migrant studies like some of our earliest work and understanding that cancer has environmental underpinnings was you know these studies where you took Japanese Americans with low levels of breast cancer and people thought oh it's just the genetics their Japanese they don't get breast cancer but then you look over three generations of migration they've been living in California for three generations and oh their breast cancer rates are like the rates of white women in California who are in their same SES you know like something like that is not the kind of epi studies we're doing now but it's really powerful for imagining a counterfactual of genetically similar people under very different environmental conditions there've been similar studies about race and birth outcomes where people compare black Americans whose ancestors were enslaved in this country to African immigrants and Caribbean immigrants and then look over generations and you see the pattern changing and it's saying it's not just the genes like there's other stuff going on and so that's not giving you like one point estimate causal estimate but it is giving you causal inference that okay like it's not just this thing and you know there's studies about the gender gap and math scores from epidemiology and how it varies over time and space so I think that sometimes asking a bigger question might feel sad because you realize okay I'm not going to be able to do this one study that's going to pop the estimate out of SES I'm going to have to depend on more novel study designs and triangulating across studies but sometimes that's what you need to answer the actual important question that is of interest that that's fantastic we're putting it in I chair your feelings about the counterfactual model being a really nice way to be able to concretize your question and it sheds so much light although I do think it has limitations and I'm curious your thoughts on whether you think that the counterfactual model has in some ways limited us in that I do think most people do think of it as in order to use this model you have to have something that you can actually do or is it a consequence of the model that we have this model that tells us something and therefore we think small because we think that that's the only way to answer a question I think it's a few things I think that sometimes people confuse the concept with the analytic methods so a lot of people are like causal inference is doing marginal structural models and I'm like no so that's just a conflation and I think people are just attracted to fancy analytics and that's another issue one of the arguments that I made in a response to an HA commentary to Miguel Hernán and Sandra Balea was that I think a lot of the innovation in causal inference has come from people who were doing biomedical stuff like infarmical epi or kind of HIV treatment I just think that's where a lot of the development was so I'm actually hopeful I think if there was an investment in thinking about social epidemiology and social forces like I think it's possible I think some of it is just like these methods have developed in settings where kind of an individual level biomedical focus was the main focus and they've been like super powerful but they've developed with an error or scope that's my hypothesis is just based on intuition and nothing else that's fabulous thank you so much Whitney for joining us today I learned a huge amount can I just say though I think you have to promise us that you're going to come back yes I was just going to say something like that this is about a quarter of the conversation that I would love to continue to have so promise us that you'll come back I'll come back okay good so let's call this Whitney Robinson Part 1 and please join us at some point in the future for Whitney Robinson Part 2 and yes it was a pleasure and thank you very much for joining us it was great thank you just one last thing for those of you who are not members of the Society of Epidemiologic Research I strongly recommend you consider becoming a member membership gets you discounted fee for the annual meeting which is coming up in December this year in Boston it also gets you access to the SCR library which gives you access to some great learning materials seminars and activities find out more at Epidemi Research Hub you really appreciate listening and look out for our next episode coming up next month thank you