What Medicine Gets Wrong About Race

ENEANYA: The reason I became kind of a central figure in this controversy is not because I’m an equation person. I have family members that were affected by the use of this.

That’s Dr. Amaka Eneanya. She’s a nephrologist, or kidney doctor. And the controversy she’s referring to has to do with an equation that doctors have used for decades to estimate kidney function. That number determines crucial things about a patient’s care — including whether they’re eligible for a spot on the waitlist for a kidney transplant. But the problem is: there are actually two equations — one for Black patients and one for everyone else.

If you took two patients with identical health profiles, one Black and one white, the Black patient’s estimated kidney function would look better than the white patient’s, even if it wasn’t. This can have all sorts of implications but perhaps most critically, the Black patient might have to wait longer for a spot on the transplant wait list.

ENEANYA: I have kidney disease that runs in my family. And I saw explicitly how that number affected an individual’s care in my family. And so I just really could not walk away from this controversy. It’s deeply personal and I think it’s deeply wrong.

There are huge disparities in health care and health outcomes across racial groups in the United States. Some of those disparities are explained by socioeconomic factors, access to healthcare, and treatment from individual doctors with conscious or unconscious biases. But that’s not the whole story. Because in medicine, racial bias has been built into some of the tools and measurements that should be the most objective.

From the Freakonomics Radio Network, this is Freakonomics, M.D. I’m Bapu Jena. Today on the show: we’re going to talk about a medical device that’s critical to patient care, but doesn’t seem to work the same on everybody.

SJODING: If all I had was the pulse oximeter, I probably would not have recognized that those treatments were necessary.

But first, I’ll dive deeper into the kidney risk calculator controversy with Dr. Amaka Eneanya to understand how race became a factor in the equation at all — and the work she’s doing to take it out.

ENEANYA: We’re even widening those disparities if you’re making Black patients wait a longer period of time for them to get “sick enough”.

 *      *      *

A family history of kidney disease isn’t the only reason Dr. Amaka Eneanya wanted to specialize in nephrology.

ENEANYA: My father is a nephrologist.

BAPU: There, there you go. So if I were to paraphrase why you went into nephrology, it’s because your dad threatened to disown you?

ENEANYA: Yeah, I’m Nigerian American. And it’s like you have three options: Physician, engineer, or lawyer.

BAPU: Among Indians it’s electrophysiology, interventional cardiology, or general cardiology. You can pick one of those three.

Amaka is currently the head of strategy and operations for the global medical company Fresenius, which provides services to patients with kidney disease. Before that, she was a researcher at the University of Pennsylvania, where she studied how race is used to diagnose and manage kidney disease.

ENEANYA: Let’s kind of just rewind and think about what the kidneys are. So their primary function is to rid the body of waste and excessive fluid — your urine. So 24/7, your kidneys are working hard to make sure that they’re filtering the blood in your body of extra waste and fluid.

BAPU: What causes things to go wrong?

ENEANYA: The number one cause of chronic kidney disease in the United States is diabetes, followed very closely by high blood pressure. Chronic kidney disease in the mild stages is pretty, silent. You don’t really have a lot of symptoms from that disease. It’s when you get to the very advanced stages when you’re approaching complete kidney failure, where there’s a need for dialysis or kidney transplant, do you really start to feel symptoms. All of that fluid that your kidney works to get rid of starts to accumulate in your body. You can get very puffy. It’s very uncomfortable. And the toxins that are normally cleared also build up in your body so you can feel nauseated, really out of it.

BAPU: Who gets kidney disease? Are there particular groups that are at higher risk?

ENEANYA: Patients that have a high proportion of the comorbidities that I’d listed, such as diabetes and high blood pressure are particularly at high risk of chronic kidney disease. And in the United States, we know that those populations skew heavily towards racial and ethnic minorities. So Black patients, Hispanic or Latinx patients, Native American patients all have high risk of developing kidney disease because they also have a large burden of diabetes and high blood pressure in those populations.

BAPU: How do you diagnose kidney disease and then monitor the function of the kidneys?

ENEANYA: The gold standard way is actually quite intense, and we don’t do that in clinical practice. You’d have to collect a series of urine and blood measurements over time. Sometimes that requires like a hospitalization. So an estimate of doing that is to use what we call estimated glomerular filtration rate, eGFR. And that number, the way that we describe it to patients is the percentage of how well your kidneys are working. So if your eGFR is 40, we say your kidneys are working about 40 percent.

Remember the number I talked about at the beginning of the episode that measures kidney function? And that’s calculated differently in Black people? That’s eGFR.

ENEANYA: The first eGFR equation in the United States was developed in 1999. They had thousands of patients where they did collect blood and urine over time to measure directly how their kidneys were functioning. And then they did some statistical modeling and said, “What kind of variables can we put in this equation that will best predict somebody’s kidney function?” They looked at age, they looked at a number of clinical tests, and they also looked at race. And they said, “Oh my goodness. The Black population, specifically, when you use their race in this equation, it makes the equation more accurate. So therefore we’re going to have two equations, one for Black individuals, and then one for people who are not Black.” And that is how race was introduced into eGFR equations.

The rationale, at the time, had to do with creatinine, which is the main lab test used in the eGFR equation to make sure that somebody’s kidneys are filtering waste appropriately. Generally speaking, the higher the creatinine, the worse the kidney is functioning. But creatinine also depends on muscle mass: the more muscular an individual is, the higher their baseline creatinine level will be when their kidneys are functioning normally.

ENEANYA: And so we were taught in medical school like, “Oh, by the way, Black people are more muscular than every other race group, and therefore we need to kind of adjust for that using this different eGFR equation.” And many people didn’t feel like that made sense. I was one of those people. But medical students, they really are what drove the big change for eGFR equations. They were constantly pressing their professors in medical school, like, “There’s anthropology studies that show that biologically we are no different between racial groups. So why are you saying that Black people, have more muscle mass just inherently?” And professors couldn’t answer that question.

This is a crucial point: race isn’t a biological category — it’s a social and cultural category. It can be tied to certain biological traits, like skin pigmentation, but it’s not that clear cut. For example, there are lots of African Americans who have much lighter skin than many South Indians, but South Indians may not identify as Black. There’s no medical test that can identify a person’s race. And the Human Genome Project found that race has no genetic basis. In fact, the amount of genetic diversity that we observe within races is actually larger than what we observe across races. So what does this mean for how we measure kidney function?

ENEANYA: There was a group of medical students in 2017 at Harvard Medical School that really challenged the faculty and the faculty said, “You know what? You’re right. You’ve produced all this evidence. You’ve gone back to anthropological studies, sociological studies. We’re not going to use the Black eGFR equations anymore. We’re going to use one equation for everybody.” And so after they did that, my colleagues and I at the University of Pennsylvania decided to write this article in JAMA that was really challenging it. Like, why are we using this? This doesn’t make sense. And we should get rid of it. That was published in 2019. And, you know, I gave some talks here and there and we were trying to make changes at Penn, and then the pandemic hit and George Floyd was murdered.

And that was really another racial reckoning in this country of how we’re treating racial and ethnic minorities differently. The same group that pushed out the eGFR equation at Harvard Medical School and Beth Israel Deaconess wrote a New England Journal perspective where they detailed all of these racialized algorithms, not just in nephrology and kidney care, but in cardiology and pulmonology. And that is where we start to see some movement. Institutions across the country just started to stop using this race-based eGFR equation because it’s wrong.

BAPU: When you say it’s wrong, what do you mean by it being wrong?

ENEANYA: We have international guidelines. And for you to be actually put on a transplant wait list, your eGFR should be equal to or less than 20. So again, you have Black individuals that have to wait longer than everybody else for their eGFR to fall to these cut points. And we already know there’s significant and longstanding disparities in terms of Black patients receiving kidney transplantations compared to other racial groups. So we’re even widening those disparities if you’re making Black patients wait a longer period of time for them to quote unquote get “sick enough”.

BAPU: So you wrote this article in 2019. What’s happened in the last few years as it relates to this problem?

ENEANYA: I’m proud to say that I served on this task force that was convened by our two largest nephrology associations, the American Society of Nephrology and the National Kidney Foundation. And that process took over a year, but what came out of those deliberations was the development of a new equation that was published in the New England Journal. I was grateful to be an author on that paper where we have an equation where race is not one of the inputs, and it’s still clinically acceptable and very accurate to use. And so the recommendations that came from this task force was that all institutions, academic institutions, laboratories should switch their eGFR equations in their systems and use this new — what we call the 2021 C.K.D. Epi- equation.

BAPU: How widespread has the implementation been?

ENEANYA: Many major academic institutions in fact had moved forward with using a race-free equation even prior to the recommendations. Major medical institutions have made this change — the Veterans Affairs system has made it, as well as a number of commercial laboratories. The reason why it’s not cut and dry to many of the naysayers is that There was a difference that was seen in the 1999 study, it was seen in the 2009 study. So I think a lot of the discomfort comes from like, we don’t know exactly why those differences were seen. But what research has shown that goes back generations is that it’s really social inequities that cause differences in health outcomes — not that individuals are born a certain way, but over time oppression, lack of access to healthcare, poor control of certain illnesses, that is what’s causing the change in kidney function. And you can’t generalize that to everybody in one race.

BAPU: Just purely from a prediction perspective, does the science suggest that including race is more accurate than not including race and including other things that we currently do model?

ENEANYA: The same eGFR equations that were developed in the United States, that statistical accuracy does not ring true when research studies outside of the United States examine Black populations. So there has been a study of populations in West Africa, Black Europeans for instance, and that modifier does not make the equation more statistically accurate in those individuals. So that just tells you that it’s capturing something specific to the United States. I think we have some work to do to see what race is a proxy for.

Coming up, we look at the flip side — a case where there are inequities in care because a medical technology didn’t take differences in skin tone into account.

SJODING: The pulse oximeter wouldn’t work as well in those patients, simply because it wasn’t designed as well to work with those patients.

I’m Bapu Jena, and this is Freakonomics, MD.

*      *      *

SJODING: My name is Michael Sjoding. I’m an associate professor of pulmonary and critical care medicine at the University of Michigan, I care for patients with lung disease and patients who are critically ill in the ICU.

Michael is also a researcher, with a focus on what happens in the intensive care unit. You’ll find lots of monitoring devices in the I.C.U. that run nonstop to make sure patients are stable. One of the most fundamental measurements is a patient’s oxygen levels because, obviously, humans need oxygen to survive. Our organs can’t function without it.

SJODING: For a normal person, oxygen saturation of 97 to 99 is quite normal. Most people live above 95. Once we’re down to 92, that could be a sign that there’s something going wrong. And then anyone who’s in the 80s has a serious problem that immediately needs medical attention. Eventually if your oxygen gets too low, your heart will get injured and will not have enough oxygen to work, and you will have a cardiac arrest.

To directly measure the oxygen saturation in a patient’s blood requires an invasive and time-consuming test called an arterial blood gas. But because we need to take these measurements constantly, most of the time doctors estimate blood oxygen levels using a device called a pulse oximeter — like how the eGFR equation allows nephrologists to estimate kidney function without performing the invasive, gold-standard test. Instead of sticking a needle in your artery to draw blood, a pulse oximeter simply clips on your finger.

SJODING: What it’s doing is it’s shining light at two wavelengths through the finger, and when you have a lot of oxygen in your blood, one of the wavelengths absorbs more, and when you have a lower level of oxygen in the blood, the other wavelength gets absorbed more. And so basically the pulse oximeter is measuring how much of that light is shining through the finger versus how much is getting absorbed. And from there, it can do an estimate of how much oxygen is in the blood.

But there are things that can affect how much light gets through your finger — like the color of your skin.

SJODING: The pulse oximeter tends to overestimate the true oxygen level in your blood in patients who have darkly pigmented skin.

In 2020, Michael and some colleagues published a study in the New England Journal of Medicine that looked at just how often pulse oximeters gave inaccurate oxygen levels in Black patients, and how these inaccuracies could impact care.

SJODING: We looked at patients who were hospitalized at the University of Michigan who self-reported as either being a white race or Black race. And we took opportunities where it just happened that after a pulse oximeter reading that follow-up arterial blood gas reading was performed to compare what the reading on the pulse oximeter was showing versus the true gold standard reading on the arterial blood gas. And we found that in Black patients, the pulse oximeter tended to overestimate the oxygen saturation compared to white patients. And we ask this really important and relevant clinical question, which I cared a lot about as a physician: When the pulse oximeter was telling me the oxygen level on the patient was normal, how often was it getting it wrong? How often was that follow up true arterial oxygen saturation low, dangerously low? And what we found was in white patients, when the pulse oximeter was reading in a relatively normal range, 92 to 96 percent, the true oxygen level was low really rarely, 3 percent of the time. But in Black patients, that same thing was happening 11 percent of the time.

JENA: How does that manifest in terms of the differences in treatment decisions?

SJODING: If I were to have known that a patient that I’m caring for who is Black had a low oxygen, I would’ve changed how I was caring for that patient. For example, if that patient had COVID-19, I might have added additional treatments to help treat their COVID-19. And if I had no idea that the oxygen level was low because all I had was the pulse oximeter, I probably would not have recognized that those treatments were necessary.

JENA: So is this an issue about race or is this an issue about people with darker versus lighter skin tones?

SJODING: Other work in the past has taken people into a laboratory setting and tried to quantify their skin tone level and found that with a darker skin tone level, the pulse oximeter is less accurate. We didn’t have the luxury, in some of our work, to be able to say with precision what someone’s skin pigment level was. So we relied basically on a surrogate, which was their self-reported race. And that was a criticism of our study, which is within a group, people will have variation in their skin pigment.

JENA: Is the problem that this technology was developed primarily in white patients, or how did this come about in the first place?

SJODING: That is one potential reason why the problem exists. When you’re developing a device like the pulse oximeter, all that light absorption has to be calibrated and it has to be calibrated on real patients. If there wasn’t a lot of patients with darkly pigmented skin when these calibrations were developed, it is likely the case that the pulse oximeter wouldn’t work as well in those patients, simply because it wasn’t designed as well to work with those patients.

JENA: So what’s happened as a result of the research that you and others have done about this issue?

SJODING: The Food and Drug Administration, the F.D.A., has really gotten serious about the problem. They just conveyed a panel to discuss the issue of bias in pulse oximeters. And are likely going to change the way that pulse oximeter testing occurs. Right now when a device is tested, it needs to be shown to be effective on 10 patients and two of those patients should be darkly pigmented. but there isn’t a lot of specifics about testing the device in both lightly-pigmented patients and darkly-pigmented patients separately, and demonstrating that the device is working the same in both groups. This is called disaggregating the data. You know, instead of showing that overall performance, it’s important that the performance is demonstrated in each group individually.

Michael’s study wasn’t the first to find that pulse oximeters are slightly less accurate in Black patients than white patients. What was novel about his research, though, was how it showed that even small differences in oxygen saturation readings can have a big impact on patient care when they occur in this critical range between normal and low. There’s another reason Michael’s pulse oximeter study, which came out in 2020, made a bigger splash than its predecessors: timing.

SJODING: Everyone was recognizing how important the pulse oximeter was in caring for patients with COVID. At the same time, the summer before our study, the Black Lives Matter movement, people were sort of primed in a way that hadn’t been before about potential inequalities that Black people in America experience. I do wonder if in a different era this study would’ve gotten such attention.

JENA: Were you working on this before the pandemic started? Or did it happen during the pandemic?

SJODING: It’s kind of a crazy story. I had never thought about doing a research study on the accuracy of the pulse oximeter before the pandemic. But, I work at the University of Michigan, which is in Ann Arbor and Detroit is about 45 minutes to an hour away. And at the beginning of the pandemic, the Detroit hospitals were just completely overwhelmed. And so we were getting lots and lots of patients who were Black with COVID — more than I would say is typical at Michigan. And on a couple of occasions, the pulse oximeter was reading normally, and the arterial oxygen level was low when we were caring for these patients, and we were really mystified by it. And then once we discovered some of this prior work on the subject, it was sort of like an aha moment that like, “Oh my gosh, this is still happening and we’re seeing it in front of us caring for patients with COVID.” And so that’s really the motivation for the study, how it all started.

JENA: Are there other examples of devices that are perhaps titrated to lighter versus darker skin tones and which may not give accurate information as a result?

SJODING: There was a real recent study that came out in JAMA this year describing the use of the forehead temperature sensors, so basically a thermometer shooting out infrared light. And there does seem to be this suggestion that that type of a device could be impacted by skin pigment level. Identifying someone who has a fever in the hospital is critically important and you could be missing fevers in darker pigmented patients.

SJODING: One of the lessons from this pulse oximeter story is that when a new technology, a diagnostic technology, is developed, it has to be developed in a diverse population. If from the outset these pulse oximeters were truly developed in patients with a broad range of characteristics, I think maybe this problem would’ve been recognized and addressed sooner.

JENA: When you’re in the I.C.U. now, or when you’re in your clinic, how has the research that we’ve been talking about changed the way that you approach clinical care?

SJODING: You know, I think the pulse oximeter is still important and provides us a lot of information. And when the pulse oximeter is reading low, that’s a clear sign that there’s a problem and you need medical attention. The problem is when it’s reading normal and you’re feeling poorly, I won’t necessarily be reassured by a pulse oximeter that’s reading at 91 or 92 percent. And in fact, if I’m caring for someone with darkly-pigmented skin, I’ll be particularly more concerned that maybe this value could be slightly off. And, if it was slightly off, would I do anything differently. I have to have that mental exercise now when I’m caring for these patients. And I am performing more arterial blood gasses for better or for worse.

So, what does the future hold for pulse oximeters?

SJODING: I’m actually pretty optimistic after the F.D.A. meeting. And I even heard at this panel, some of the manufacturers coming out and saying, “We’re committed to ensuring our devices work the same in everyone.”

Amaka Eneanya is similarly optimistic when it comes to risk calculators and kidney disease.

ENEANYA: People are starting to think about What are the other questions they can ascertain in the clinical encounters that can really individualize somebody’s clinical pathway. I think what’s really promising is the younger generation of scientists and clinicians. They are all over this. And I feel inspired that that will change over time.

It’s already changing, because researchers like Amaka Eneanya and Michael Sjoding are focusing on these issues, sounding alarm bells — and being heard.

It may seem on the surface like their findings are at odds; Amaka is working to remove race from the eGFR equation while Michael says that ignoring differences in skin color is what causes pulse oximeters to be inaccurate. But I actually think they go hand in hand; medicine in the U.S. has a long history of both making unscientific assumptions about Black people and ignoring Black people altogether. So it seems that the path towards treating people equally is not for doctors to be color-blind per se, but rather to separate the measurable, objective, biological differences between patients — like skin pigmentation — from socially and culturally determined differences — like race.

That’s it for today’s show. I’d like to thank my guests, Amaka Eneanya and Michael Sjoding. And here’s an idea for you based on my conversations with them. After Michael’s study on pulse oximeters was published, a different study found that Black patients were receiving less supplemental oxygen than white patients in the hospital, largely because the Black patients’ pulse oximeter readings were overestimating the level of oxygen in their blood. It’d be interesting to see whether the levels of supplemental oxygen given to Black patients is increasing since the publication of this study — especially after it got so much press. With electronic health record data, it wouldn’t be too hard to track whether or not that’s happening.

So, what ideas do you have? Did you hear about either of these studies in the news? Did it change the way you thought about risk calculators or medical devices? Email me your thoughts, because I’d love to hear them. I’m at bapu@freakonomics.com. That’s B-A-P-U at freakonomics.com.

Coming up next week:

VASAN: Mark Zuckerberg was my college classmate, and so I think I’m the 1100th person on Facebook.

Ever since Facebook was created, people have speculated it could impact mental health.

MAKARIN: There are a lot of papers that document, significant association between mental health problems and social media use^.

But is Facebook causing these problems? And if social media is making mental health worse — can we do anything about it?

VASAN: In medicinewe take this oath to do no harm. And that was really the idea that we had around how do we change the user experience

That’s all coming up next week on Freakonomics, M.D. Thanks again for listening.

*      *      *

Freakonomics, M.D. is part of the Freakonomics Radio Network, which also includes Freakonomics Radio, No Stupid Questions, and People I (Mostly) Admire. All our shows are produced by Stitcher and Renbud Radio. You can find us on Twitter at @drbapupod. This episode was produced by Julie Kanfer and Lyric Bowditch. It was mixed by Eleanor Osborne, with help from Jasmin Klinger. Our staff also includes Neal Carruth, Gabriel Roth, Greg Rippin, Rebecca Lee Douglas, Morgan Levey, Zack Lapinski, Ryan Kelley, Katherine Moncure, Jeremy Johnston, Daria Klenert, Emma Tyrrell, Alina Kulman, Elsa Hernandez, and Stephen Dubner. Original music composed by Luis Guerra. If you like this show, or any other show in the Freakonomics Radio Network, please recommend it to your family and friends. That’s the best way to support the podcasts you love. As always, thanks for listening.

ENEANYA: Yes. That was a fantastic summary. I give you an A plus for today.

BAPU: You know what? I didn’t get A pluses at the University of Chicago, so I’ll take them now.