 Good afternoon, and thank you for joining us for our day two keynote with Dr. Kadejah Ferryman, Race Matters and Health Data. Kadejah Ferryman is a cultural anthropologist who studies the ethical, social, and policy dimensions of digital health. Specifically, her research examines how technologies, including genomics, digital medical records, and artificial intelligence impact health injustices, such as racial health disparities. Dr. Ferryman is an industry assistant professor at NYU's Ten in School of Engineering, where she has redesigned the ethics and technology core course. Before her training as an anthropologist, Ferryman was a policy researcher at the Urban Institute in Washington, DC. She's an affiliate at the Data and Society Research Institute and at the Center for Critical Race and Digital Studies. She also serves as the Institutional Review Board for the National Institutes of Health, all of us research program. Dr. Ferryman received her BA in anthropology from Yale University and her PhD in anthropology from the New School for Social Research. She has published research in journals such as Journal of the American Informatics Association, Healthcare for the Poor and Underserved, and Genetics in Medicine. Please join me in welcoming Dr. Ferryman. Awesome. Thank you so much for that lovely invitation and hi to everyone out there virtually. Okay, I am going to share my screen with you all. Okay, so hopefully everyone can see the screen and I'll just get started. Okay, so thank you so much to the organizers for inviting me to give this talk. Even though we're not all together in person, I really am honored to be in your presence today. And so even though we are virtual, I would like to begin my talk with a land acknowledgement. So I'd like to ask you to join me in acknowledging the Lenape community and the exclusions and erasures of many Indigenous peoples, not just the Lenape community on whose land I am located and what we now refer to as New York City. This acknowledgement demonstrates a commitment to beginning a process of working to dismantle the ongoing legacies of settler colonialism. Acknowledgements such as these invite us to ask and think about what does it mean to live in a post and neocolonial world? What did it take for us to get here and how can we be accountable to our part in history? So I'd like to start my talk today with the title Race Matters in Health Data. One meaning of the title Race Matters in Health Data is that race is important in health data. But I want to parse out another meaning of race matters that brings us closer actually to the word matter as in substance or material. So when we do this, we can think about the title as referring to how race becomes matter and how processes of racialization become material in health data. It is this interpretation that I will draw on strongly for this talk. I also want us to note that the two meanings of the word matter as in substance and the other meaning as in importance are linked together. And that's because matter is substance and substance must be contended with. So when we bring these two meanings together, matter as substance and matter as being important, we can think of this title Race Matters in Health Data is also saying that race, because it is made material in health data, must be contended with and demands attention. I'd like you to keep these two threads in mind as I go through the talk today. So I want to just share with you an outline of the talk today. I'll begin with the way that race is often raised as being important or mattering in health data. And that is through the absence, missingness or non-representativeness of health data sets. I will then take us to a different view of race matters in health data, not as a lack of representative data or of missing data, but how and where we can see processes of racialization in health data. And by racialization, I mean the processes by which racial hierarchies are formed, circulated and reinscribed in society. I will then conclude with some thoughts and provocations on how we can contend with these race matters in health data. So not too long ago, I was at home in front of my laptop watching a conference online as many of us are doing these days and as you are doing right now. The topic was artificial intelligence and health, and the speaker was discussing the development of a machine learning tool to detect prostate cancer. During the question and answer portion of the talk, I asked how health disparities in prostate cancer had been considered during the development of this tool, since it hadn't been mentioned in the talk. I knew that race mattered in prostate cancer, since I myself have a number of relatives who are dealing with this disease, and I know that racial disparities exist in this area. For example, black men have higher incidence and mortality rates for prostate cancer. And research has also shown that black men are less likely to receive treatment that follows prostate cancer care guidelines. They are also less likely to receive diagnostic MRIs, and as this article shows, also experience a longer time between diagnosis and the start of treatment than white men. And although there is some research on biological contributions to racial differences in prostate cancer, there's also evidence showing that black and white men who receive similar care have similar outcomes. So for example, there was a study showing that black and white men who receive care through the Veterans Administration actually, because they have equal access, have very similar outcomes. So when I asked this question about health disparities in prostate cancer and how that was considered in the development of this AI tool, the response that I received was that the team had used a representative data set, and that was how they had attended to issues of racial disparities, health disparities in developing the tool to make sure that their data set that they used to train the tool had been representative. This response was satisfactory, but I also found myself unsettled by this answer. Now, don't get me wrong. I know that representative data sets matter, especially for developing AI tools for medicine. Clinical trials are notorious for being not representative, and the emerging field of genomic medicine includes a number of examples of how a lack of representativeness in data led to harm for racial and ethnic minority groups. So for example, these are just headlines about two cases of genomic medicine. One where due to a lack of representation in clinical trials data that black individuals were told that they had a pathogenic genetic variant, when in fact, if the data had been more representative, it would have come out earlier that that variant is actually not pathogenic and actually found quite often in people of African ancestry. There was another case of plavix, which is a blood thinner, and again, due to a lack of representativeness in the study data that it showed that there were different outcomes for plavix for Asian and Pacific Islander individuals. So it's clear that having representative data, having representative health data matters. It's clear also that having, that a lack of representation matters for other kinds of health data too, not just genomics, such as images used, images of skin used in dermatology and oncology. There is evidence that a lack of diversity of images in skin is leading to bias in AI tools that detect skin cancer. The lack of diverse and representative images of skin conditions, and especially the lack of images of skin conditions on dark skin, has prompted the grassroots effort called Brown Matters to make these images more available and to make image data sets more representative and therefore useful for more people. Also yes, data representation matters in health, but is that enough? My answer is no. And that is because it is not enough if racial and ethnic groups are present in health data, but it matters how they are represented in that data. It also matters how this data is understood and interpreted. As I mentioned before in the previous example of prostate cancer, black men are less likely to receive care that adheres to the official clinical guidelines. So this example shows us that race matters when it comes to the kind of care one receives. It's also important to note that clinical guidelines and the care that is given, that that is then represented in health data. Thus, another way to think about it is that guidelines and care shape how people are actually represented in clinical data, not just whether they are represented proportionally or not. This became clear for me a few years ago when I was conducting research for the fairness and precision medicine study along with my colleague, Michaela Pitkin. Our research focused on understanding how this emerging field of data-centric medicine, precision medicine, might lead to disadvantage and or might follow along existing pathways of marginalization and discrimination in healthcare. During this study, one of our interviewees explained to us how race matters for clinical guidelines and how that in turn shapes clinical data. And I'm going to share an excerpt from our interview with him with you. So he says, here are the guidelines for lung cancer screening. You must be 55 to 80 and you must have 30 pack years of smoking. This all came from a study that was done previously of 53,000 people. Of those 53,000 people, only 4% were African-American. So we now have lung cancer screening guidelines based on that 4%. And then he went on to say, you don't have to smoke a pack a day if you smoke menthols. So now when I do lung cancer screening in the community, most of the community members that we screen don't qualify because they don't have the smoking history, meaning that they don't have the 30 pack years of smoking. So we can use this excerpt to tease out why representative data sets might not be enough. His argument here is that because of the way guidelines were developed, some black people were not screened. The guidelines did not affect whether or not people got cancer, but instead affected how far their cancer progressed before detection or diagnosis. Thus the clinical data reflects this lack of early screening, since those who are not able to be screened early would present with cancer at more advanced stages. And this would of course be reported in the clinical data. So here we see it's not that black individuals are not in the data or missing from the data. It's a matter of how the clinical data itself represents patterns of racial discrimination. A data set that is representative by population groups is still stamped with this history of racial discrimination. And without knowing why the data looks the way it does, a number of seemingly reasonable patterns could be identified and conclusions could be drawn. For example, it might be reasonable to conclude once looking at this data that has been shaped by these lung cancer screening guidelines, it might be reasonable to conclude that there is some biological or genetic difference that causes black people to present clinically with lung cancer at more advanced stages with although they have fewer pathways of smoking. But knowledge of the data's history and context would not lead you to those conclusions but would lead you down a different path. We were told this history of the screening guidelines and how they disadvantage black patients almost three years ago. Our interviewee had known of his history for much longer and this was because he was on the ground doing research in black communities. But it wasn't until and we conducted that research in 2017-2018. But it wasn't until about two months ago that this issue was addressed and lung cancer screening guidelines were changed. The US Preventative Services Task Force recently lowered the number of eligible smoking years from 30 to 20 and the age for screening from 55 to 50. And this was based on additional research that was done. And what did this research show? It showed that there were racial disparities in lung cancer screening. This was actually a study that was done by Vanderbilt University. And this research done, this is the study, the evaluation of the screening guidelines, showed a number of differences. But one, according to this research, of those diagnosed with lung cancer, 32% of African Americans had been eligible for screening versus 56% of whites. This screening example shows us that race matters in health data, again, matters in terms of the substance that you can see racialization happening and made real in the data. And that part and that patterns of marginalization literally, quite literally, materialize in health data. It took years for this racial gap in guidelines to be addressed. And we can only think about how many people were harmed due to not being able to be screened and also what kinds of conclusions, again, were drawn from this kind of data without a knowledge of the history. And this, I argue, is not something that a representative dataset can fix. So there are other examples of how clinical data is racialized. Historian Lundy Braun's book, Breathing Race into the Machine, The Surprising Career of the Spirometer from Plantation to Genetics, explores how and why this tool, the spirometer that measures lung capacity, includes a racial correction for Black people and Asian people. That is, when someone breathes into the spirometer, their measurement is changed depending on whether the patient is identified as Black or Asian. This correction to spirometer data often happens invisibly and unbeknownst to the patient and often unbeknownst to the clinicians as well, that this racial correction is happening. Lundy Braun, this book came out, I believe, in 2008. And in the book, she traces how beliefs about the supposed smaller lungs of Black people help justify plantation labor as suitable, even helpful for the supposed weak lungs of Black people. So this book raised the question of why this kind of racial correction was still being used. And then a decade ago, in 2011, clinicians also began raising questions about other racial corrections of clinical data, specifically the EGFR that is a calculator of kidney function. This calculator uses a race multiplier to adjust for the, again, supposed higher muscle mass of African Americans. This article here published in 2011 raised questions about the clinical data that informs this multiplier and whether this clinical data is scientifically valid. This came out in 2011. I did my dissertation research a few years later, which was an ethnographic study in an academic medical center. And during that time, one of the trainee physicians, a Latina woman, shared with me her perplexities and exasperation with the EGFR. She told me that she asked her attending physician how to use EGFR for Latinos when she saw there were these different factors for White individuals and Black individuals. He said that what he does is look at how dark the patient is to determine how much African ancestry they have, and then he would base his use of the race multiplier on that determination. She shared with me then her shock and disappointment at how this calculator was being used in what she called was an imprecise and unjust manner. So almost 10 years later, Vias and colleagues published the first comprehensive examination of racial corrections used in medicine. This article is called Hidden in Plain Sight, Reconsidering the Use of Race Correction in Clinical Algorithms. And I'll show you here, this is an excerpt of a longer table that they include in this article of the multiple racial corrections that are used in multiple specialties in medicine from oncology to cardiology. So in this article, the authors question the scientific evidence undergirding these corrections. And here's an excerpt from the article where they discuss a common risk score that's adjusted by race. The American Heart Association Heart Failure Risk Score predicts the death, risk of death in patients admitted to the hospital. It assigns three additional points to any patient identified as non-black, therefore categorizing all black patients as being at a lower risk. The AHA American Heart Association does not provide a rationale for this adjustment. Clinicians are advised to use this risk score to guide decisions about referral to cardiology and allocation of health care resources. Since black is equated with lower risk, following the guidelines could direct care away from black patients. So the authors are arguing that these corrections shape what could be seen as objective clinical data. And it's important, one, because as they mentioned, there's often not a rationale or the scientific evidence undergirding these corrections is outdated or questionable. That's one problem. But they also highlight that these racial corrections are a matter of justice because these corrections can further health disparities and health inequities. And it wasn't until this past June that major hospitals in Massachusetts decided to stop using the EGFR race multiplier. And here's a tweet from med student Lash Nolan announcing the change. And she notes here that this is a major win. And it's the result of years of hard work, advocacy, and research done by black students and scholars. So again, the article, the color of kidneys that I showed earlier was from 2000, I believe from 2011. And here we have the change happening and only happening in some medical centers in 2020. So this is an important step to have this major institution drop the use of the EGFR multiplier. And you can see some of the reasons given here in given by the institution as to why they dropped this race multiplier. First, that race is a social, not a biological construct. And that research studies have not provided an acceptable scientific rationale for making these decisions based on the social construct of race. But although this academic medical center has taken this step, many others still use not only the EGFR correction, but some of the many other racial corrections that were identified in the bias and colleagues article. And in addition, there was another commenter on this thread of this tweet that noted that not only are these corrections still used by many, many hospitals, but they are also used by apps that that use these risk calculators. So for example, there are apps such as these that use some of these risk correction risk calculators to provide people estimates of their risk. And as is kind of mentioned in this tweet, there's a couple of institutions that are called out. Apps like MD Calc and other apps to, you know, say, you know, when are these other apps going to remove race based weights from their platforms? And it's unclear if these apps and apps such as these and many others will continue to use these racial correction racial corrections. And if there is really a way to enforce them or make them kind of not use these kind of calculators. So I want to share another example of how race is made material and clinical data. And this time in clinicians notes and electronic medical records. So in this example, Chen and colleagues analyzed clinicians notes and found differences in how mental health was documented for patients of different races. And again, here's an excerpt from this article. And one of the things that the authors note, and they were looking at clinicians notes, right, where clinicians write in about patients in the electronic medical record, and they found that white patients had higher topic enrichment values for anxiety and chronic pain top topics, while black Hispanic and Asian patients had higher topic enrichment values for the psychosis topic. And they also find some other differences by gender and also by insurance type. And so this is an example, this example shows that essentially doctors talk about, in this case, mental health of their patients in different ways, depending on their race. And in another more recent example, researchers from Johns Hopkins University analyzed clinicians notes and found that black patients were, again, discussed differently than others. So as this headline shows, an analysis of the medical records, specifically the clinicians notes, showed that physicians are more likely to doubt black patients than white patients. And what does that mean, doubt patients? And this comes from the study that was done, was called testimonial injustice linguistic bias in the medical records of black patients and women. And what this article shows is that doctors were more likely to use doubting language such as claims as in a person claims to have pain or a person claims to have soreness, when referring to black patients reports of their health. So hopefully it's not hard to see racial hierarchies at work here in these examples. And we can see that clinical data mirrors and reinforces age old and harmful stereotypes about black people being more unstable and less trustworthy than others. And again, this is not a problem that representative data sets can fix. And I want to share another example of race matters and health data. This time when race is not even present as a variable. So in the previous examples, these were examples of how black people were talked about in this case, talked about in medical records how their clinical measurements were changed because of explicit racial corrections. But we can see this even when race is not used specifically. So we know, hopefully, hopefully by now we all know that the COVID-19 pandemic has not affected, has not affected everyone equally, and has mirrored the previous patterns of health disparities in the US. In York where I am, the hardest hit areas during the peak pandemic times here, overlapped almost exactly with areas that had the lowest life expectancy pre-COVID. So we know that race matters for COVID-19. We know that it is racism, not race, that is a risk factor for dying from COVID-19. But recently, Schmidt and colleagues explored the ways that racialization affects clinical care, even when race is left out of the data. Their examination focuses on clinical risk calculators that are used to determine the allocation of ventilators, a critical issue during the COVID-19 pandemic. These calculators use creatinine, which is an indicator of kidney health, to compute scores that indicate the patient's chances of survival in the near term, as well as their overall life expectancy. The authors argue that the use of creatinine in these calculators disproportionately and unfairly puts Black people at a lower priority for receiving ventilators. This is because, as they argue, that creatinine is not just a physiological measure, but that it also, quote, measures social disadvantages that may cause higher creatinine. So higher levels of creatinine result from chronic conditions such as kidney disease and high blood pressure. And the prevalence of these conditions in racial and ethnic minority groups can be attributed in large part to social determinants of health. And can be, as they argue as well, quote, best understood as the consequences of health inequities and structural racism. So again, their argument is that even when race is not present, when a measure of creatinine is used, that that is still bringing in with it histories of social determinants of health, histories of health inequities. So again, because this is an example where race matters in health data, even when race is absent, a more representative dataset, a more representative dataset by race and ethnicity would not do much to address this issue of racialization in this data. So now that I've taken you through these examples of racialization in health data, from screening guidelines to clinical notes to risk calculators, the question might be, well, if data representation is not enough, what else should we be doing? And this is where we get to the contending with race matters part of the talk. So I do have a few suggestions in this regard. The first is to shift our language, to shift our language around data problems, because how we talk about problems, shapes and structures are universe of solutions. As my former colleague at the Data and Society Research Institute, Kenjil Davi wrote, when we stop overusing the word bias, we can begin to use the language that has been designed to theorize at the level of structural oppression. By using the language of bias, we may end up overly focusing on the individual intents of technologists involved, and even of technologies involved, I'd say, rather than the structural power of the institutions they belong to. We need to move away from a language of bias, she argues, since it implies local technical fixes for data, when instead we should be looking at how institutional and systemic factors shape our data. Similarly, Stevens and Keyes argue that we should not just focus on, quote, biased data sets, because the data sets themselves are not so much biased as they are reflective of their sites of use. And we must instead focus on the logics and systems of inequality that lead to the data sets themselves and lead to the data sets looking the way they do. And one common refrain that's often mentioned in the kind of growing attention to bias, and again, just like with representative data sets, I'm not saying that they're bad, I'm saying they are not enough. And in the conversations around bias, there have been some positive things to come out of the conversations around data bias, what the impacts of data bias are. But they do leave out this issue of structural power of systemic factors that are shaping the data the way they do. And one common refrain is often, well, garbage in, garbage out. And the idea is that if the data are bad or garbage, trash, then the result, when those data are used, how they're used is going to lead to a kind of garbage or trash result. But I would actually argue that that garbage that's going in is not actually garbage, right? That garbage data actually is very informative of these social forces, of these patterns of marginalization, right? Of these histories that if we did more investigating, did more digging of the data, that we could actually learn a lot from. So those data aren't garbage. They actually are really important political and social artifacts that we should be attending to and understanding what they mean. So I suggest then that we, when we shift our language from bias, it does get us to this focus on institutional and systemic forces. So moving away from bias data to thinking and talking about our data as racialized data. And so it's my argument that when we approach and kind of think about our data as racialized data, then we can begin to ask different questions about the data. Then we can begin to see and understand and interpret the data in new ways. And so some questions that could be asked about the data include, what histories does this data represent? As well as what current processes of marginalization does this data represent? Not necessarily that it's not representative or it's bias, but what is it actually showing to us? And I believe that we can actually learn a lot from asking these kinds of questions. And so these are questions that individual data analysts can ask themselves and can discuss with their teams. It may also prompt critical reflections on how some of these same histories and current discriminatory and exclusionary practices factor into who is in the room of data analysts and who is not there and whose values, worldviews and experiences, not just technical know how are shaping the analysis, the analyses that are being done. So for example, going back to the quote from Dr. Karim Watson, he is a scientific researcher and he knew many, many years ago because of his experience, because of the kind of work that he did, that lung cancer screening guidelines were disadvantageing the community. But it took many years and again, formal research studies to begin that process of changing the guidelines. And so what happens when we bring in people who have different experiences and different interpretations, different views of the data and what histories they actually represent. So this brings me to this suggestion, which follows from the reframing of data from bias or not representative to racialized data. That is that technical data analysts can ask themselves these questions, but they may also need to expand their network of expertise to include anthropologists, sociologists, historians, and others who can provide essential information about the data histories, contexts, and social processes that animate the data. For those data analysts who are already doing this work, bringing in collaborators from other disciplines, I would encourage you to share your stories of how seeing race mattering outside of data represent, representativeness and data bias has improved your work. So what are race matters in health data and why does race matter in health data? I tried to show that health data are political and social artifacts, meaning that they are marked by reflect and shape systems of power in society. When we shift our language and our frames, we open up new possibilities for understanding and for action. We open up new paths for the data work that is so needed and so important for it to matter for everyone. Thank you. And here are my references and email and Twitter to reach out. Thank you so much for this great presentation and it's really evoked a lot of conversation in the chat and there's a number of questions. Great. Should I keep my sharing screen up or stop sharing? Whatever you prefer to do. If you want to keep it up or if you want to take it down, that's fine as well. Okay, let me take it down for now. I might put it back up just to put my email address and Twitter handle back up there, but I'll stop sharing for now. So we have some questions from our audience and the first question is from Nirwan Chatterty. What if we want race-based corrections? I am South Asian and new research shows that my community's heart, community's heart health risks are different from those of white patients. Most doctors don't follow their research and treat using white norms, which puts families like mine in danger. Why shouldn't we be demanding that our doctors consider research-based racial corrections? That's a great question and thank you for that. I think the goal of my discussion of race corrections is kind of sharing the story about hearing the comments from the fellow about the EGFR and showing that table of some of the racial corrections and the quote from the American Heart Association is to begin to question the racial corrections that are being used. And so it's not to say that we should not have racial corrections or we should not have adjustments, risk adjustments that are more tailored that are more precise. This is actually part of the goal of precision medicine to gather more data about risks for smaller and smaller sub-populations of people to really understand health risks because up until this point they have been based around the white male as the norm. But with the corrections that I showed, for example, the EGFR, the corrections to the spirometer, those are based on what is really being understood as outdated, incorrect clinical research and corrections that are really drawing upon their kind of age-old racial stereotypes. And even in the case of the spirometer and the racial correction there, correcting and that correction is used for Black and people who are Black and Asian, that correction again was sort of drawing on racial stereotypes, but it was also quite conveniently as Lundy Braun describes used as a way to justify existing racial hierarchies. So literally these measurements were used as a way to justify why Black people were better suited to plantation labor, especially and even as factory labor, there was more factory jobs available. It was again used to say that no Black people, they can't kind of handle factory labor because they don't have their lung capacity, can't handle it. So it was a way to exclude Black people from factory jobs that were offering higher wages. So I think that is the critical question that we have to ask about when we're thinking about different populations and their risks and when to include that is what kind of racial hierarchies are they drawing on? Do they fall in line with maintaining existing power hierarchies or not? The other interesting thing about sort of thinking about South Asians and having different risks for heart disease, let's say, again even that category because there has been research showing even looking at breast cancer in Asian women, the risk for breast cancer in Asian women is very different for Asian women in the United States than it is for Asian women living in Asia. So even when we say something like there's research showing that South Asian people have different risks for heart disease, which South Asian people are we talking about? Are we talking about South Asian people in the United States? Are we talking about which countries? So there has to be and these are these kinds of generalizations that are used that can then again question what the evidence there is and sort of think about if these definitions are valid or not and are they being used to kind of improve the health and lives of people with these calculators. Thank you. A second question from Alan asks, what are the mechanisms within institutions like AHA, med schools and others to revise race-based corrections? Do they have teams or committees to hear and respond to this research? I guess I'm wondering if there are even enough structural or institutional starting points for advocacy. Yeah, that's a great question and it's an empirical question I think and it I think the evidence showing that this is all very recent, so that article identifying that there are all these race-based calculators just came out not only a year ago and as I mentioned, there are many clinicians who didn't know that these racial calculations were as widespread as they are. If you're working in one specialty, you might not know that there's a racial correction used in other other specialties. So the first step I think was this kind of awareness raising, but as Lash Nolan discussed in her tweet, people have been raising this issue and talking about it in different specialties in different contexts for years and it really did take advocacy from medical students, physicians, at institutions to say let's really think about how we're using these calculators and if we should be using these calculators. And if you can imagine something like a calculator that has a correction that's literally built into the health IT infrastructure, it can be very costly and take quite a while to turn back to undo that kind of decision. So it's not surprising that a very well-funded hospital like I believe was Brigham Women's was able to be one of the first to step out and say that they were able to do this because of the sort of time and funding and everything that's involved with changing that kind of correction. So I think it depends on the politics of particular institutions. I learned from doing my dissertation research that healthcare medical centers and hospitals have their own politics there as well. So it will sort of depend institution by institution, I think in some cases, how these kinds of changes can be made. But I think taking, having a very well-known, well-regarded hospital take the public step to say that we have found that this is not useful, that it's not scientifically valid is really an important one. And I think there are these conversations happening at more and more institutions about do we continue to use these corrections and how do we begin to turn back and kind of take these out of our systems because it literally, they literally quite literally become systemic, right? These are things that are programmed in and it can be quite difficult to remove them from systems. So I think we have a sort of related question or at least one that in my head is related. So Candace Ubrea was asking states, I'm wondering if there's, if social determinants of health are habitually considered by a doctor during a typical doctor's visit? Yes, oh my goodness, who asked that question? Candace Ubrea. Candace Ubrea, thank you so much for asking that question. That is a fantastic question. And again, this is a sort of emerging area of clinical medicine and there are a number of healthcare institutions that again after years of advocacy and research showing that social determinants of health are real, that they really do translate into different health outcomes for people that one zip code is more determinative of one's health than one's genetic code, for example. So it has become accepted that social determinants of health are real and they should be contended with. And so there is actually a movement afoot to a number of hospitals around the country and my knowledge is kind of limited to the US are trying out essentially kind of social risk screeners as part of routine clinical visits. So how do we, how do clinicians build in as part of, you know, the usual sort of questions that you might be asked as part of your routine visit, you know, your weight, this whatever, to ask about things like housing instability, to ask about things like food insecurity, because again, of the wealth of data showing that social determinants of health really do kind of determine, really have a great influence on people's on people's health outcomes. So there's there's been a lot of movement around that they're also in terms of data and sort of data infrastructures and architectures, which I am really interested in as these political artifacts. There's also some work happening again at a number of institutions around the country, sort of thinking about how to link clinical health data infrastructures with social and kind of community health data infrastructures, right. So we have hospitals on the one hand who are providing health services that again may not have a full picture and or an idea of what what else is happening in that person's life, but then there are actually a network of community institutions or other kinds of institutions that do have information on this individual and some of these social social determinants or social information. So how do we begin to literally kind of stitch these data architectures together so that in addition to maybe asking some questions that that data can kind of be at the fingertips of the clinicians on one hand and of you know on the on the side of the the community organizations as well, of course, being privacy protecting and things like that. One thing I will say that although it is encouraging to see these efforts to create and I believe they're called community care networks that's the linking of these disparate forms of data and things like training clinicians to ask about social issues that might impact people's health outcomes, although I do think those are important, I think we still have to be critical as we're sort of moving in this direction and ask critical questions about that because research has shown that other kinds of screening tools and screening questions are again can be used in ways that mirror existing patterns of racial discrimination. So for example, there are screening questions or ways that physicians can screen for if a mother is addicted to drugs and research has shown that black mothers are asked those questions more than white mothers are. So if you know questions and incorporating social determinants of health are going to be done, they should be sort of applied equally and they shouldn't become yet another tool to marginalize, discriminate, oppress people who are already facing social and health inequities. So that would be my only kind of caveat and note that we should still be sort of critical as we're seeing movements on these two fronts to incorporate social determinants and social issues more into the clinical space. So with your last example, sort of links to another question from the audience from William Lager who writes, I'm very interested to know of similar assumptions about data have been shown to be a cause and other health outcomes such as maternal health. Yeah, so there is a crisis of black women's maternal health in this country, maternal and infant health in this country. So black women have I think at least three times higher incidences of maternal complications and infant mortality than white women do. And this is even after adjusting for socio-demographic information like income, like education, etc. So this is one of those cases where it's not poverty, it's not access even to healthcare, that when all those things remain equal, there's still a big disparity. And so there has been, there's a lot of research sort of looking into how racialization, how processes of racialization sort of actually happen in during reproductive care for black women and during prenatal care for black women. And there's a great ethnography by a cultural anthropologist called reproducing race pregnancy as a site of racialization by care of bridges. And she literally sort of talks about how racial hierarchies happen sort of black women get racialized as soon as they step into the door to receive prenatal care. So she has some really compelling interviews with clinicians who again sort of think they in this her ethnography was done actually in a community hospital. So these are clinicians who are helping some of the underserved, right, have experience with treating a diverse array of patients, but hold these paradoxical views. So for example, one of the kind of paradoxes that she highlighted from her, from through her ethnography was that black women were treated on the one hand as wily patients who were trying to gain kind of gain the healthcare system and get access to benefits that they weren't necessarily entitled to, but also at the same time uneducated. And so, you know, not able to understand health information the same way that other patients do. And so it was the sort of paradoxical thing, like, you know, that they're, you're saying that they can kind of use their smarts in some ways, but are actually not smart in others. So it's these kind of processes of racialization that I think we really have to attend to when thinking about how we see again holding other, you know, other variables, you know, the same that we see these, these different outcomes. There's also some great work being done sort of looking at, and I'll put a quick, quick plug in here for a documentary called Listen to Me that's being shot and directed and produced by black women. This sort of comes out from the Serena Williams story of not being listened to when she was receiving care, when she was having her child that she was experiencing, you know, a blood clot. And even though she was, again, wealthy, it wasn't an issue of not having access. She was there in the best hospital that she could have been, but she literally wasn't listened to while she was there. And so that documentary sort of looks at and explores these instances of black women not being listened to about their care. And there is also research, again, showing, like I showed in the talk, where physicians, you know, don't trust what, you know, are more or less likely to trust what black people are telling them. So I think these are some of the things that we can look to and conduct more research on quite frankly, to begin to identify some of these processes of racialization that happen in the clinical space. So I think we have time for a couple more questions. And Danielle asks, do you have any insight into how people interested in the Quantified Self movement, as was talked about yesterday by some speakers yesterday, can interface with precision medicine if they choose to share their personal health data openly? Yes. And Quantified Self is a really interesting, I'm sorry I didn't get to see some of those presentations. Quantified Self is a really interesting kind of social and cultural movement that a number of anthropologists and science and technology scholars have done some really interesting work analyzing. But I think what's interesting about some of that work is again, looking at thinking about where race, racialization, racial hierarchies come to play, right? So as I understand it, some of the people who are sort of would call themselves part of the Quantified Self community, you know, are able to gather data, troves of data about themselves and sort of use it in share the data with their clinicians or also kind of use their data to come up with their own kind of hacks or practices to alter their health. And so again, sort of thinking about in that Quantified Self space, how being a, and I don't know if I'm using the right terminology, but a Quantified Self or looks different and is reacted differently to if one is let's say black, right? And there's another, for example, there's a, there's a great book by Eric Topol who's done some really great work talking about the future, I think it's called The Doctor Will See You Now. And it's a great book because it sort of highlights, you know, how in the future, when we have all this data and we have AI, that people, you know, will be able to kind of be managers of their health and doctors will actually take this consultant role. But that book doesn't mention race at all. And it doesn't mention how that opportunity to sit in and tell your doctor, here, this is all the data I have, is not likely, again, based on real evidence that we have, that, you know, scholars have collected is not going to be treated the same way if you are black than if you are white, right? So it's, I think it's really important for people involved in the Quantified Self movement to bring in, to sort of bring the subject of racialization into some of their discussion, some of their plans for how to bring in data, how to share data to think, okay, let's not just think about this in terms of the default of a white male who is going to be respected, listened to with their data, and, you know, be able to assume these different kinds of social positions, change, you know, shifts in relationships between clinicians and patients, right? That's not going to work the same for everyone. So I think that would be my kind of first suggestion to people interested in those sort of questions around the Quantified Self movement and what people will do with, with their data and this kind of really, I think, important thinking about the future and sort of future making. And, but imagining that that future may look different depending on what you look like, and how we can actually change that so that the future can be, that we can't imagine a future where race won't play a role, but I think we have to sort of think about that at the outset and we can't just think about, you know, sharing data or, you know, different ways to make medicine more precise without bringing in these issues of racialization. Thank you so much. I think that is all the time we have for questions.