4

Efforts to Promote Race-Conscious Medicine in Health Organizations and Systems

Reginald Tucker-Seeley, vice president for health equity at ZERO Prostate Cancer, moderated a discussion among Michelle Morse, chief medical officer for the New York City Department of Health and Mental Hygiene (NYC DOHMH); Rohan Khazanchi, resident physician in the Harvard Internal Medicine and Pediatrics combined residency program at Brigham and Women’s Hospital; and Carmel Shachar, faculty director of the Health Law and Policy Clinic at the Center for Health Law and Policy Innovation at Harvard Law School. Tucker-Seeley asked Morse to discuss how her department prioritizes race-conscious solutions to address health equity and who her department brings to the table, both externally and internally. Morse replied the first step has been to define race consciousness in a public health framework through workshops and discussions with the department’s evaluation, research, and program leaders. She noted opinions vary widely about whether race consciousness is good or not, and the recent Supreme Court decision on affirmative action created confusion about what race-conscious decisions are allowed or disallowed. Morse added, “However, as Justice Sotomayor said, racial equity is impossible with race blindness. There is no way to get racial equity without race consciousness, and that has to be clear for everyone.”

One action NYC and New York State took when Paxlovid became available to treat COVID-19 was stating that non-White and Hispanic ancestry could serve as social risk factors in deciding who would get Paxlovid when the drug was in short supply. Morse said that decision unsettled many people, even those who support health equity. It was

hard for some clinicians to bring race consciousness into the exam room. Then, in autumn 2021, her department launched a coalition to end racism in clinical algorithms. It comprised 11 health systems, including NYC’s six largest health systems. One coalition charge is to end race adjustment in clinical algorithms for estimated glomerular filtration rate (eGFR), pulmonary function tests, and vaginal birth after a prior cesarean section (VBAC). A second task is to measure how racial inequities change when eliminating race adjustment in those three algorithms. A third task is to develop a patient engagement plan that addresses discussing these changes with patients and ensures patients whose care was delayed or harmed by race adjustment receive an expedited form of the care they always deserved.

Tucker-Seeley then asked Khazanchi about the role policy plays in influencing the medical community’s adoption of new practices and facilitating a paradigm shift to race-conscious medicine. Khazanchi said he sees policy playing a key role in addressing health inequities through promoting the design and implementation of antiracist systems and structures that clearly reflect the need for racially unbiased care and clinical algorithms. “We know from decades of research on provider behavior, implementation science, and quality improvement just how hard it is to combat heuristics, biases, and societal narratives that are ingrained in medical practice,” said Khazanchi. Therefore, interventions cannot solely rely on individual practitioner actions; they must structurally influence the actions of health and social systems that serve these communities, he concluded.

In 2020, Khazanchi and a nationwide coalition of trainees pushed the AMA House of Delegates to pass landmark resolutions that advocate for addressing racism as a public health crisis, reconsidering the essentialization of race as biology in medical education and practice, and abolishing the use of race as a proxy for biology, genetics, or ancestry (Khazanchi et al., 2022). When Khazanchi and his team began drafting these resolutions in 2018, the AMA House of Delegates had only two adopted policies that even mentioned the word “racism”: one addressed racism in music, and the other supported the American Public Health Association’s Campaign Against Racism. AMA passed the new resolutions in the aftermath of George Floyd’s murder. For Khazanchi, this emphasizes the inextricability of social movements and movements for racial justice in medicine.

Simultaneously, Khazanchi added, without narrative change, data and policies are often left as standalone declarations that lack downstream influences on systemic reform (Khazanchi et al., 2021). He highlighted AMA’s subsequent actions have helped shift these narratives; for example, congressional and regulatory testimony and letter-writing continue to influence hundreds of proposed regulatory rules and pieces of legislation

by explicitly framing racism as a public health crisis. Even congressional hearings have begun explicitly emphasizing the role of structural racism in shaping health and social inequity. “Policies change narratives, and narratives get people to act, so as we look to expedite the implementation of race-conscious solutions for clinical algorithms that have done harm, we need to think about what policy levers we have at our disposal,” said Khazanchi.

Shachar said regulatory agencies can have a huge impact by setting parameters for developing and using race-conscious algorithms. For example, the Office of the National Coordinator for Health Information Technology (ONC) recently issued a call for developers of artificial intelligence (AI) applications to release what Shachar characterized as “nutrition labels” for their products. Such labels would detail the population developers used to test their application, enabling physicians and patients to assess whether the product represents their patient panel. The Food and Drug Administration (FDA) has powerful leverage because it regulates software as a medical device; it can demand that products be unbiased and not perpetuate health inequities, Shachar said.

Tucker-Seeley asked Morse to explain how the NYC DOHMH is implementing race-conscious medicine. Morse replied that leadership support for this work is critical and that trainee advocacy has been key to facilitating needed changes (Heffron et al., 2022), additionally, having health equity and racial justice already in an institution’s strategic plan was crucial. She noted that it has been more difficult to remove race from these algorithms than it historically was to insert race into them. Removing race, she said, requires changing the EHR, clinical practice, and medical education.

Morse said the 11 coalition members are at different stages of implementation. Some have completed the transition, are now measuring changes and outcomes, and are even seeing improved rates of kidney transplantation for Black patients. Currently, the coalition is targeting safety-net hospitals since they often see more patients of color and have fewer operational resources to make the systems changes required to implement the new algorithms. In recent months, this work received a boost from the American Thoracic Society, which provided new guidance on how to cease using race and ethnicity in pulmonary function tests and spirometry calculations.

Khazanchi emphasized the need for clinicians and health systems to create mechanisms for accountability to their patients when considering opportunities for de-implementing race in clinical algorithms. Concurrently, he said, regulatory interventions may be required if the impetus sparked by narratives, patient advocacy, and evidence fails to create meaningful change. Another opportunity is to link de-implementing race-based

algorithms to mitigating health liabilities. He noted that some patients who recognized the way race-based medicine adversely affected their care have filed lawsuits to draw attention to this issue. “This concern about liability may help press organizations, medical societies, and hospitals to move forward,” said Khazanchi. He also highlighted the importance of involving patients in designing, implementing, and evaluating work to introduce race-conscious medicine and phase out race-based algorithms.

Tucker-Seeley asked Shachar how the Office of Civil Rights (OCR) can expand the scope of antidiscrimination law to health care to further protect patients from harms described by the workshop’s speakers. Shachar said Section 1557 of the Affordable Care Act, which the OCR enforces, states that health systems must work to end discrimination in their operations or lose federal funding. Until recently, gender and sexual orientation received the most focus. Now, however, the OCR has issued proposed rules to hold health systems accountable for using race-based algorithms (Shachar and Gerke, 2023). Shachar noted that dealing with simple race-based algorithms such as the VBAC one will be straightforward, but dealing with complex AI-powered algorithms that lack transparency about how they account for race will be more difficult.

Shachar said ONC’s move to demand greater transparency will be useful for AI algorithms. Greater transparency can provide the tools physicians and hospital organizations need to understand what they get when they use a particular algorithm and whether an algorithm is suitable for their patients. At a time when clinicians are overburdened, AI-backed algorithms could augment medical provider efforts. “We do not want to shut down the use of this technology completely,” said Shachar, “but it would not be a good thing if we put all the responsibility on providers to become computer scientists so they can understand the products they will use to provide medical care.” While FDA has engaged this issue, it has done so in the context of whether a product is safe and effective and not in the context of how health care systems implement a new algorithm. What is needed, she said, is greater coordination among regulatory agencies to determine what incentives are available for developers to avoid releasing race-based products and for users to flag problematic algorithms.

Tucker-Seeley noted that acting on inequity is a political process and asked Morse what her experience with that process has been as she advances efforts focused on race-conscious medicine. Morse acknowledged challenges but said George Floyd’s murder and the COVID-19 pandemic created conditions for the conversation to happen in more places. However, she believes conversations about race consciousness in medical practice will continue to be challenging. “On the one hand, we all agree that equality cannot be achieved with race blindness, but what does

that mean in terms of my behavior as a clinician?” asked Morse. “What does that mean in terms of my conversations in the way I approach my patients, and what does it mean to the local health department’s policies and guidance about who gets what medications and who should be prioritized for what treatment?” In NYC, a political opportunity arose to institutionalize some of these conversations and resources for work within the Board of Health’s resolution declaring racism a public health crisis. Morse added that events in 2020 created a space where city voters passed several ballot initiatives, including one that solidified and institutionalized a racial justice commission. The questions now are whether these efforts will continue to be supported and if the commitment to racial justice will remain as strong as it is today.

Tucker-Seeley said de-implementing race in these processes requires a multilevel approach involving providers, organizations and systems, and communities. He asked Khazanchi for his thoughts on how providers and health care systems can act within that context to implement race-conscious approaches and to dismantle race-based approaches. First, Khazanchi replied, it is important to acknowledge that no two algorithms are identical, and, as such, no two algorithms will necessarily be amended in the same way. He then offered a possible solution: ask whether an algorithm exacerbates or reduces health inequities and then weigh quantifiable harms and benefits alongside the values of impacted patients. From Khazanchi’s perspective, the role of convening bodies and regulators is to address algorithms systematically; many medical specialty societies have adopted declarations that racism is a public health threat, and some have begun issuing race-conscious rather than race-based guidelines.

Addressing next steps, Shachar said medicine is at an inflection point regarding medical practice and algorithms. The vast majority of practicing physicians predate the era in which machine-learning, decision-making software is the norm, and they believe algorithms are suggestions, not dictates. The new generation of physicians, however, will have to work hard to unpack why particular algorithms use racial assumptions that lead to negative outcomes for certain patients.

Morse noted the importance of remembering that not everyone has deep experiences in health equity and therefore, they may struggle to speak directly with researchers about considering health equity in their work and with professional societies about prioritizing health equity in their planning and decision making. She highlighted a major implementation gap related to how Black, Latina, Latino, and Asian patients benefit from the necessary policy changes that need to happen to ensure they receive high-quality care. Morse also called for further interventions, actions, and implementation research that describes what happens when algorithms are changed.

Khazanchi said medical curricula should not only reflect the future reality of medical education and practice but also uplift and center the voices of experts who have long thought about these issues. For him, addressing race-based clinical algorithms fundamentally returns to focusing on who benefits and who is harmed. It also involves learning from clinicians who use these algorithms and seeking their insights on associated pros, cons, and ethical conflicts. Finally, it entails including patient and community voices in the process during every step of the journey.