Letter to The Hon. Gopal Khanna, M.B.A. Director of Agency for Healthcare Research and Quality - Warren, Wyden, Booker, and Lee Question the Use of Race-Based Algorithms in Standard Medical Practice

Letter

Dear Mr. Khanna:
We write to request that the Agency for Healthcare Research and Quality (AHRQ) conduct a
review of the use of race-based clinical algorithms in standard medical practice. The ongoing
coronavirus disease 2019 (COVID-19) pandemic and its disproportionate consequences for
communities of color have starkly revealed that racism itself is a public health crisis.1
In order to
reduce health disparities among communities of color, we must ensure that medicine and public
health organizations take a staunchly anti-racist approach to medical care and reevaluate the
ways in which current practices, including the use of race-based algorithms, could be worsening
outcomes for people of color.
Race-based clinical algorithms, which include a patient's race among their inputs, inform how
clinicians adjust medical test results based on their patient's race. These results then inform
treatment regimens as well as overall assessments of health. These race-based algorithms risk
embedding racism into medical practice. One such algorithm has recently come under fire after it
was reportedly used by the National Football League (NFL) in a way that cut benefits for Black
players. The Wall Street Journal reported that, because the NFL was using a clinical algorithm
that assumed Black players had lower cognitive functioning compared to white players, Black
players were entitled to lower settlements for concussion-related injuries.
2 But this overtly racist
assumption and subsequent adjustment to players' test results is just one high-profile example of
race-based clinical algorithms. There are many others.
Another example in which race-based assumptions are embedded in medicine is in the estimated
glomerular filtration rate (eGFR), an algorithm that allows clinicians to indirectly measure a
patient's kidney function.3 The eGFR adjusts test results for all patients identified as Black.4 This
race-based adjustment to eGFR originated from a 1999 study which found differences in a
measure of kidney health among Black study participants compared to white participants.
5

1 Bloomberg, "Dozens of City Governments Declare Racism a Public Health Crisis," Brentin Mock, July 13, 2020,
https://www.bloomberg.com/news/articles/2020-07-13/dozens-of-cities-dub-racism-a-public-health-crisis.
2 Wall Street Journal, "Lawsuit Alleges NFL's Concussion Settlement Discriminates Against Black Players," Louise
Radnofsky, and Andrew Beaton, August 25, 2020, https://www.wsj.com/articles/lawsuit-alleges-nfls-concussionsettlement-discriminates-against-black-players-11598371843.
3 The Journal of the American Medical Association, "Reconsidering the consequences of using race to estimate
kidney function," Nwamaka Denise Eneanya, Wei Yang, and Peter Philip Reese, June 6, 2019,
https://doi.org/10.1001/jama.2019.5774.
4
Id.
5 Annals of Internal Medicine, "A more accurate method to estimate glomerular filtration rate from serum creatinine:
a new prediction equation." Levey, Andrew S., et al., March 16, 1999, https://doi.org/10.7326/0003-4819-130-
6-199903160-00002.
2
The 1999 study that forms the basis of the eGFR's race-based adjustment is based on the notion
that differences in measurements of kidney health may be attributable to high muscle mass.
However, rather than include a correction for muscle mass in the eGFR, a broad, race-based
adjustment was included. In a larger follow-up study in 2009, a new measure was developed that
had a lesser adjustment by race; despite this new measure being recommended for use, many
hospitals in the U.S. still use the older algorithm.
6 Dr. Mallika L. Mendu, a nephrologist at
Brigham and Women's Hospital, recently found that close to a third of Black patients would have
been reclassified to reflect a more severe form of kidney disease if their eGFR had not been
increased because of the original algorithm.7 This standard practice in medical care thus runs the
risk of worsening existing health disparities.
Across the U.S., some practitioners are challenging the use of race in this algorithm. Beth Israel
Deaconess Medical Center in Boston, for example, moved away from race-based adjustments to
eGFR, beginning in March 2017.7
In 2019, Zuckerberg San Francisco General Hospital began
replacing race with a measure of muscle mass in the algorithm.8 As of this summer, University of
Washington Medicine,9
the Vanderbilt University Medical Center,10 and the Massachusetts
General and Brigham health systems7
have asked their staff to no longer include race in their
calculations of eGFR.
Several other instances of race correction have also come under scrutiny.11 The American Heart
Association's (AHA) "Get with the Guidelines--Heart Failure Risk Score," is another example of
a measure used in medical care today that treats race as a proxy for biological and genetic
differences. This algorithm categorizes all Black patients as being at a lower risk of in-hospital
death for heart failure.
12 However, it is well-documented that Black and Latinx patients
presenting with heart failure are actually more likely to have negative health outcomes, possibly
attributed to lack of access to care.
13 The STONE score--which predicts the likelihood of kidney
stones in patients experiencing extreme pain--uses an algorithm that assigns a higher risk score

6 Advances in Chronic Kidney Disease, "Estimated Glomerular Filtration Rate; Laboratory Implementation and
Current Global Status," W. Greg Miller, and Graham R. D. Jones, January 01, 2018,
https://doi.org/10.1053/j.ackd.2017.09.013.
7 Medscape, "Dropping Race-Based eGFR Adjustment Gains Traction in US," Mitchel Zoler, July 06, 2020,
https://www.medscape.com/viewarticle/933418.
8 STAT, "A yearslong push to remove racist bias from kidney testing gains new ground," Theresa Gaffney July 17,
2020, https://www.statnews.com/2020/07/17/egfr-race-kidney-test/.
9 University of Washington Department of Medicine, "UW Medicine to exclude race from calculation of eGFR
(measure of kidney function)," May 29, 2020, https://medicine.uw.edu/news/uw-medicine-exclude-race-calculationegfr-measure-kidney-function.
10 Vanderbilt University Meidcal Center, "Trainees lead an effort to end race-based adjustment to eGFR," July 29,
2020, https://discover.vumc.org/2020/07/eliminating-race-as-a-variable-in-estimating-kidney-function/.
11 The New England Journal of Medicine, "Hidden in Plain Sight -- Reconsidering the Use of Race Correction in
Clinical Algorithms," Darshali Vyas, Leo Eisenstein, and David Jones, June 17, 2020,
https://doi.org/10.1056/NEJMms2004740.
12 Id.
13 Circulation: Heart Failure, "Identification of racial inequities in access to specialized inpatient heart failure care at
an academic medical center," October, 29 2019, Lauren A Eberly, et al.,
https://doi.org/10.1161/circheartfailure.119.006214.
3
to any patient identified as "nonblack."14 The STONE algorithm, by assigning a lower risk score
to Black patients, may dissuade clinicians from evaluating these patients for kidney stones,
despite the fact that the developers of the algorithm did not provide an explanation for why Black
patients would be less likely to have kidney stones.
11
The use of spirometers, devices that measure lung function, presents yet another example of
race-based medicine--this time, with clear roots to slavery and the eugenics movement.
Spirometers apply a "racial correction" for patients identified as nonwhite; because this
"correction" is programmed into the spirometer by the manufacturer of the device, it can be
difficult to disable.15 This racial adjustment to the measured lung function can be traced back to
slavery, when enslavers asserted, with no basis, that Black people had a "smaller lung capacity"
than white people, and to eugenics-era medical texts that reported differences between white,
Black, Chinese, and Filipino populations' lung capacity.
16 Despite its history, and a lack of a
global consensus or standard, the race-based adjustment of lung function measures are still
used--with the power to impact the lives of Black patients beyond the doctor's office. Black
Americans, compared to their white coworkers, had a more difficult time proving the existence
of negative lung health effects resulting from dangerous working conditions--in part, because of
these algorithms.
17
Other race-based clinical algorithms may be exacerbating health disparities by assuming that
patients of color will experience worse health outcomes than their white counterparts, and then
pushing those patients towards higher-risk procedures. For example, the Vaginal Birth after
Cesarean (VBAC) algorithm, which predicts of labor for people who have previously undergone
a cesarean section (C-section), automatically calculates a lower likelihood of successful labor for
patients identified as Black or Latinx.
18 The study used to create the algorithm also found VBAC
varied by marital status and insurance type, but opted not to include these in the algorithm.19
Women of color already have higher C-section birth rates compared to white women--a factor
contributing to the maternal mortality crisis. The VBAC may thus be exacerbating this
disparity.20

14 British Medical Journal, "Derivation and validation of a clinical prediction rule for uncomplicated ureteral stone
-- the STONE score: retrospective and prospective observational cohort studies," Christopher Moore, et al., March
26, 2014, https://doi.org/10.1136/bmj.g2191.
15 Canadian Journal of Respiratory Therapy, "Race, ethnicity and lung function: A brief history," Lundy Braun,
2015, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4631137.
16 For example, "Vital capacity of the lungs: A handbook for clinicians and others interested in the examination of
the heart and lungs both in health and disease." J.A. Myers, 1925.
17 The Baltimore Sun, "Racial basis for asbestos lawsuits?; Owens Corning seeks more stringent standards for
blacks," Erin Texeira, March 25, 1999, https://www.baltimoresun.com/news/bs-xpm-1999-03-25-9903250041-
story.html.
18 Obstetrics and Gynecology, "Development of a nomogram for prediction of vaginal birth after cesarean delivery,"
William A. Grobman, et al., April 2007, https://doi.org/10.1097/01.AOG.0000259312.36053.02.
19 American Journal of Obstetrics and Gynecology, "The MFMU Cesarean Registry: factors affecting the success of
trial of labor after previous cesarean delivery," Mark B. Landon, et al., September 2005,
https://doi.org/10.1016/j.ajog.2005.05.066.
20 Women's Health Issues, "Challenging the Use of Race in the Vaginal Birth After Cesarean Calculator," Vyas,
Darshali A.,et al., May 06, 2019, DOI: https://doi.org/10.1016/j.whi.2019.04.007.
4
Race-based medical care and algorithms must be reassessed. These algorithms may be overly
simplistic in that they apply population standards to individuals and treat heterogeneous groups
as biologically homogenous.
21 It is important to understand not just that there is bias in these
algorithms, but also to better understand what historical factors led to the creation of these
biases, whether they are based on unscientific, racist assumptions, or if they merely reflect the
effects racism has already had on the health of people of color.
AHRQ's mission is to "enhance the quality, appropriateness, and effectiveness of health care
services, and access to such services, through broad-based scientific research and initiatives,"
with the goal of transforming research into practice.22 In order to better understand disparities in
health, Congress passed the Healthcare Research and Quality Act of 1999, which mandated that
AHRQ produce annual reports on national healthcare quality and on healthcare disparities in the
U.S.23 In its annual National Healthcare Disparities Report, the agency details the disparities in
care experienced by different racial, ethnic, and socioeconomic groups and ongoing and recently
completed AHRQ activities that address the health of these under-represented populations.24
These racial disparities in healthcare outcomes stem from a wide variety of factors. As part of
AHRQ's work to address any disparities stemming from race-based assumptions embedded into
medical care, we ask that you conduct a review of the use of race-based clinical algorithms that
answers the following questions:
1) To what extent are race-based clinical algorithms used in medical practice, recommended
as medical standards, and taught in medical curricula?
2) Are the race-based algorithms used in medical practice based on scientifically sound
studies that include a large enough sample of participants? Have the studies been
replicated? If not, what is the basis of these algorithms? How often are these algorithms
updated?
3) Are the current race categorizations most commonly used (Black vs. non-Black, or White
vs. African American vs. Hispanic) an appropriate means of distinguishing human
populations? Are these categorizations internationally recognized in medical care? Do
these algorithms advise how multiracial individuals should be classified? Are patients
made aware of how these algorithms are being used to influence their care and, if so, can
they opt out of racial classifications?

21 WBUR News, "Dorothy Roberts: What's Race Got to Do with Medicine?" February 10, 2017,
https://www.wbur.org/npr/514150399/what-s-race-got-to-do-with-medicine.
22 International Journal for Quality in Health Care, "The US Agency for Healthcare Research and Quality's activities
in patient safety research," Gregg S Meyer, James Battles, James C. Hart, and Ning Tang, December 2003,
https://doi.org/10.1093/intqhc/mzg068.
23 National Healthcare Quality & Disparities Report Chartbooks. Agency for Healthcare Research and Quality,
Rockville, MD. Content last reviewed May 2020.
https://www.ahrq.gov/research/findings/nhqrdr/chartbooks/index.html.
24 "2018 National Healthcare Quality and Disparities Report. Agency for Healthcare Research and Quality,
Rockville, MD. Content last reviewed April 2020.
https://www.ahrq.gov/research/findings/nhqrdr/nhqdr18/index.html.
5
4) Among algorithms where factors besides race were predictive of the outcome (e.g.
socioeconomic status, muscle mass, type of insurance, height and weight), what were
these factors? What was the justification for not including them in the algorithm?
5) How do these algorithms currently affect patients? Is there evidence that they worsen
access to care, quality of care, or outcomes for patients of color? Is there evidence of
provider bias? If so, can this harm be quantified?
6) Different hospital and medical systems have taken different approaches in their shortterm solutions to removing race from these algorithms. What approaches have different
medical systems taken, when, and to what degree? What has been the effect on patient
outcomes among those hospitals and medical systems no longer using race in these
clinical algorithms compared to those still using the measure?
7) How should medical standards and curricula be updated to move away from instances in
which the inclusion of race in these algorithms is harmful to patients?
Thank you for your prompt attention to this matter.
Sincerely,

Source