Lack of racial diversity in rheumatology image banks perpetuates physician blind spot
This past spring, two studies shed light on how some of the largest clinician organizations and publishers in rheumatology, including the American College of Rheumatology and EULAR, have failed to provide adequate training materials or clear guidelines when it comes to caring for patients of color.
However, as they say, knowing is just half the battle. Now, these and other organizations are determining just what to do about it.
The first study, published in March in the ACR’s own Arthritis Care & Research, found that darker skin tones are “significantly underrepresented” in rheumatology clinical image banks, including the ACR Image Library, UpToDate, The New England Journal of Medicine Images in Clinical Medicine and Clinical Cases, and the ninth edition of Kelley’s Textbook of Rheumatology.
Two months later, another study, also published in Arthritis Care & Research, examined 65 ACR and EULAR practice guidelines and recommendations, and found a worrying lack of clarity regarding race information, an “urgent need” for standardized race reporting, and, in at least one case, research with insufficient racial diversity that was used to make race-based recommendations for Black patients.
For its part, the ACR, having published both studies in one of its official publications, stated it is committed to addressing all the above inadequacies, with steps already in place aimed at improving its clinical image library, according to officials.
“The American College of Rheumatology has been making active steps to rectify this, and the image bank is the tip of the iceberg,” Noelle A. Rolle, MBBS, rheumatologist at Piedmont Columbus Regional in Columbus, Georgia, and chair of the ACR’s Image Bank Diversity Task Force, told Healio Rheumatology. “While to some extent ACR’s diversity, equity and inclusion efforts are in their infancy, ACR’s broader efforts around equity are not.”
‘Chicken or the Egg’
According to the authors of the March study, led by Adrienne Strait, BA, of the University of California, San Francisco, School of Medicine, 13.4% of the 1,043 included clinical images portrayed dark skin, while 84% demonstrated light skin and 2.6% were indeterminate. This, despite U.S. census figures showing Asian, Native American and Black individuals comprising 20.6% of the population.
Dark skin was “underrepresented considerably more” in images of patients with systemic lupus erythematosus. Asian, Native American and Black individuals represent approximately 44.2% of all prevalent SLE cases in the United States. Yet, overall representation of dark skin in SLE images was only 22.6%, according to the researchers.
Among the included image banks, UpToDate had the largest proportion of images of dark skin — at just 17.4%. Kelley’s Textbook contained the smallest proportion, with 7.1%, followed by the NEJM Clinical Cases, at 8%. Meanwhile, 13.3% of images in the NEJM Images in Clinical Medicine bank portrayed darker skin tones.
Among the 538 images from the ACR included in the study, 13.6% represented darker skin.
“Underrepresentation of dark skin tones among educational materials may perpetuate health disparities by increasing the risk for misdiagnosis or diagnostic delay when diseases present in people with skin of color,” Strait told Healio Rheumatology in April following the study’s publication. “In addition, it may contribute to implicit bias and structural racism present in medical education by promoting white-only models of disease. We should work together to develop more equitable educational tools.”
According to Rolle, who is also a member of the ACR Collaborative Initiatives Special Committee, the ACR image bank was initially established “out of an awareness that a lot of our diseases are relatively rare,” with health care providers potentially unable to easily identify images.
In the past, the ACR has invited providers to submit pictures of interesting rheumatological diseases to its image bank and, once deemed appropriate, they are added to the library. This system has ultimately led to a distinct underrepresentation of images portraying people of color, however, stemming from — according to Rolle — what could be a variety of reasons. This could include providers being unable to identify these cases due to lack of training with images showing people of color — representing a “chicken or the egg paradox” that continually perpetuates the same problem over time, Rolle said.
“When we analyze why there is an underrepresentation of images of people of color, it’s multifactorial and can include, but is not limited to, patients not having access to rheumatological care to first identify these cases; an inability to identify the disease in people of color, because it wasn’t taught or seen in the first place; a lack of submitted images in people of color; and a lack of diversity and representation where critical decisions are being made,” she said.
“There is a general default to making one demographic the prototype for ‘normal’ and all else is considered ‘other’,” Rolle added. “Unfortunately, this extends beyond rheumatology and the image bank into other areas within and outside of medicine, and the ACR is making a concerted effort to be the change we need to see. This precedent was set decades ago, and it’s not going to be an easy overnight fix, but we are committed to doing the work it requires.”
Training the Trainers
Publishers of the other image banks included in the study have also issued statements in response. Asked for comment, Jennifer Zeis, director of communications and media relations for the NEJM Group, said, “We acknowledge that considering images with varying patient skin tones wasn’t historically a priority for us, but it is now. This week, for example, we published a report on delayed large local reactions to mRNA COVID-19 vaccines in recipients who are Black, Indigenous or people of color.”
Zeis added that the NEJM Group has also partnered with the Skin of Color Society and VisualDx to cosponsor a series of virtual panel discussions on health disparities and structural racism in medicine.
“The four-part webinar series, ‘The Impact of Skin Color and Ethnicity on Clinical Diagnosis and Research,’ delved into longstanding problems in education, research and patient care, and introduced ways to improve outcomes in marginalized patients,” Zeis said. “This webinar series began in October 2020 and ended in December.”
Ted W. Post, MD, editor-in-chief for UpToDate, which is published by Wolters Kluwer, told Healio Rheumatology: “We are heartened to see UpToDate had some of the highest rates of variety in skin tones — phototypes — in the study but there is more work to be done. We have recognized this challenge ourselves and are already exploring different sources of photographs of physical findings across a range of specialties, including rheumatology, depicted on a range of skin phototypes.”
“We also plan to increase our visual representation of different ages and body types,” Post added. “In addition to graphics, we are modifying our language to be more inclusive and taking care to re-examine discussions of risk factors and other issues that often reflect systemic racism. These efforts are part of a broader commitment to health equity across Wolters Kluwer. We’re committed to transforming our content to ensure patients can receive the best care everywhere.”
Lastly, a spokesperson for Elsevier, which publishes Kelley’s Textbook of Rheumatology, said: “Elsevier recognizes that inequities in academic research are manifested in many different ways, and are committed to making a positive impact to enhance diversity and inclusion in research across gender, race and ethnicity. I wanted to let you know that we have shared the concerns and we are investigating.”
According to Rolle, the ACR’s own efforts to improve health equity include the creation of its Collaborative Initiatives Department in 2014. The member-led working group that supported this department, tasked with “advancing health equity, eliminating health disparities and improving the quality of patient lives,” would later be established as a special committee in 2018.
“I recall being a fellow at our 2018 annual meeting and being excited about listening to all the presentations for research that was being done regarding increasing diversity and eliminating health care disparities,” Rolle said. “I had no clue people were doing studies like this. Now 3 years later, I’m a member of this committee where we collaborate on advocacy, training and educational initiatives to improve in this area.”
It was in this committee where members last year began seeking ways to increase representation in the ACR’s image bank, Rolle said. Their answer: Asking providers directly.
In previous years when the ACR has called for images, the announcement has come with a theme, typically an area where officials believe images are lacking. Previous years’ themes have included sarcoidosis, vasculitis and adverse effects of rheumatic medications. This year’s theme will be “Rheumatic diseases in people of color.” Details about how to enter images are available at the ACR’s website.
The ACR is also applying for a grant that will support efforts to collect a more representative album of training images, according to Rolle.
“As a task force, we are working on other ways to increase these images in our library, being careful to be sensitive to our patient population in our approach,” Rolle said. “This includes putting in an application for grant money to support collecting diverse images. We feel along with getting more images, it will shine a light on the needs, importance, and impact of this issue.”
In the second study, researchers Rose McKeon Olson, MD, and Candace H. Feldman, MD, ScD, both of Brigham and Women’s Hospital, in Boston, analyzed 42 clinical practice guidelines from EULAR, and 23 from the ACR, all published between 2010 and 2020, searching for race and ethnicity terms, how they were used and what data they included to support their recommendations.
According to that study, 16 of the 65 included guidelines and recommendations used race terms in their text. However, none clearly defined race, and race was often used interchangeably with ethnicity or genetic ancestry. In addition, racial categories varied considerably by guideline, and often used classifications that oversimplified and excluded non-white races.
In one instance, a study cited by EULAR in its lupus nephritis guidelines concluded that mycophenolate mofetil (MMF) may be more efficacious than IV cyclophosphamide in Black patients. This, despite sample sizes of just 26 for MMF, and 20 for IV cyclophosphamide, among Black patients.
Furthermore, efficacy results for Black participants alone were not statistically different. However, when researchers combined data from Black patients with a racial group defined as “other,” there was statistical difference. According to Olsen and Feldman, this led the study authors to conclude that MMF may be more efficacious than IV cyclophosphamide in Black patients.
“These data were used to support claims found in several rheumatology guidelines that there are, ‘possible ethnic/racial differences,’ suggesting that MMF may be more efficacious in African–Americans,” Olsen and Feldman wrote. “As a result, this is broadly accepted in clinical practice and part of the teaching provided to rheumatology trainees.”
This use of race — a social construct that changes based on time and place, with no basis in biology — in clinical decision-making can lead to racially biased care, they added.
Robert B.M. Landewé, MD, PhD, of the Amsterdam Rheumatology Center, and of the University of Amsterdam, in the Netherlands, and EULAR’s chair of Quality of Care, responded to a request for comment stating that EULAR does not currently have specific rules mandating how race should be handled in studies. However, he added that EULAR “welcome the results of the study and will be considering its findings carefully as we go forward in forming our procedures and task force activities.”
“We at EULAR we have been informed about this remarkable study that has systematically investigated how race-information is reported in the literature, especially in guidance documents,” Landewé told Healio Rheumatology. “We agree with the authors that it is relevant for many diseases to take race-information seriously, since disease presentations and outcomes may differ. It is crucial that reporting race-information is done in a sensitive, sensible and non-offending manner. “
Susan Goodman, MD, of Weill Cornell Medicine and the Hospital for Special Surgery, in New York, and chair of ACR’s Practice Guideline Subcommittee, noted that while “this is clearly an area of importance,” the ACR is limited by the available information when it reviews literature for guidelines and criteria.
She added that the ACR plans to discuss how to improve its handling of race in clinical practice guidelines at its next meeting.
“There are many areas that were not apparent until viewed through the lens of George Floyd and the pandemic,” Goodman said. “We should aspire to more specificity in our descriptions of race and can describe limitations in the reviewed literature regarding race and ethnicity.”
In addition, Goodman characterized many of the recommendations included in the study — such as using accurate definitions of “race,” “ethnicity” and “genetic ancestry,” as well as adding country-of-origin information to demographic data and increasing racial diversity in research — as “achievable.”
“Using accurate definitions of race, and standardized categories of race and ethnicity, is an achievable goal,” Goodman said. “In addition, noting limitations in the available literature regarding race and ethnicity is important and may be useful to encourage guideline authors to be more precise. Incorporating country of origin can also be useful. We agree that rigor should be applied to any race-based recommendations — this is frankly better medicine. Our goals as guideline authors should include precision regarding race and ethnicity and attention to evidence of structural racism when present.”
Addressing Systemic Racism
For the last recommendation in their study, Olson and Feldman called on researchers to collect and report data on exposure to structural racism, including residential segregation and access to stable housing, transportation, health care, healthy food and voting sites. The goal, they said, is to identify the actual reasons for reported racial disparities in health, rather than simply hypothesizing about the issue in the discussion section.
“Once differences between racial groups were identified, guidelines largely stopped at reporting the racial difference without further analysis of what caused the disparities seen,” they wrote. “Guidelines often placed commentary about racial differences in the discussion section where a variety of reasons for differences were postulated, such as yet-to-be-discovered biological differences or hypotheses about varied social exposures.”
“This common academic practice attempts to absolve researchers, guideline authors, and clinicians alike of further investigation that may reveal the root cause of the inequity,” they added. “Once the structural inequities are identified, such as lack of access to appropriate housing, transportation, healthcare, healthy food and voting centers, these areas can be targeted to relieve racial health disparities and improve quality of care for patients of color.”
According to Goodman, recent studies have begun to examine these factors, spurred in part by disparities that have been heightened during the COVID-19 pandemic.
“More recent studies have been collecting and analyzing more complex data regarding social determinants of health,” she said. “Increased awareness of health care disparities was heightened during the pandemic and may serve as a stimulus to understand the impact of social determinants of health including race and ethnicity on outcomes. The recognition of the impact of social determinants of health on outcomes has increased, and in fact we have learned that traditional health care may contribute as little as 20% to health outcomes, with a significant proportion of the variance in outcomes attributed to social determinants of health.”
Rolle, too, drew a direct connection between the ACR’s historic dearth of people of color portrayed in its clinical image bank and the perpetuation of systemic racism. This lack of adequate training specifically related to patients of color can and has led to poorer treatment for these patients, she said.
However, Rolle cautioned that she doesn’t believe that the ACR directly “contributed” to systemic racism, stating that such language suggests deliberate intent.
“I do think a lack of images of people of color in general contributes to systemic racism as a whole,” Rolle said. “Not being able to efficiently identify a clinical manifestation in a person of color because we aren’t taught it does indeed contribute to missed and delayed diagnoses and treatment. However, I don’t think ‘contribute’ is the fitting term to use as I feel it is multifactorial and more complex than that.”
“I wouldn’t go so far to say that the ACR is contributing to systemic racism as there’s a connotation that comes with that, which suggests this lack of images in people of color was deliberate or an intentional act of omission,” she added. “It’s more complex than that, and there are a lot of variables that brought us to this point. Oftentimes, we all can be accused of some level of ignorance when it comes to correcting issues that perpetuate a type of injustice. We all need to unlearn some things, including myself as a Black woman, and none of us knows what we don’t know.”
According to Rolle, the important thing is, “once you know better, you must do better.”
“To do that, we must push pride aside, identify and face the issue, be open-minded, ask questions, actively listen, reevaluate what we thought we knew with a level of humility, take action where necessary once we become aware and not be afraid to change,” she said. “The ACR has done that, is doing that and will continue to do that. The work we are doing with our image library is one very important example of this with a lot more to come.”
For more information:
Noelle A. Rolle, MBBS, can be reached at 2000 10th Ave Suite 225, Columbus, GA 31901; email: email@example.com.
Susan Goodman, MD, can be reached at 535 East 70th Street, New York, NY 10021, email: firstname.lastname@example.org.
Robert B.M. Landewé, MD, PhD, can be reached at Spui 21, 1012 WX, Amsterdam, The Netherlands, email: email@example.com.