Healthcare providers and researchers have combed through clinical guidelines and decision-support tools to uncover debunked racial biases that contribute to inequity and disparities in care. Expunging those problematic elements from electronic health records systems, however, has proven to be vexing.
The task is technically simple. Health systems such as Chicago-based CommonSpirit Health and Philadelphia-based Penn Medicine are slowly working with EHR vendors, medical societies and other healthcare organizations to identify and fix the problem..
Related: Racially biased medical device results raise patient safety concerns
However, their initial findings reveal a ubiquitous dilemma: Racism is baked into the central systems they use to track and treat patients. And it's been there for a long time.
"The surprise to me has been the degree that it's insinuated in the electronic health records," said Dr. Joseph Wright, chief health equity officer for the American Academy of Pediatrics. "That's been really eye-opening at the health system level."
For example, calculators for diagnosing kidney disease and anemia and algorithms for organ transplants and cesarean sections recommendations have racial adjustments that result in worse care for Black patients.
Physicians are inclined to rely on these systems because they are integrated into the workflow, said Dr. Nav Persaud, co-chair of the American Heart Association's advisory panel on bias in clinical algorithms. "If you ask clinicians why they use race-based algorithms, they say, 'That's what's recommended. That's what I see my colleagues use. That's what's built into the electronic health record.' It's not that they went out and chose it themselves," he said.
Absent directives from regulators, unwinding how race is embedded in these tools is a long game. Electronic health records need to be thoroughly reviewed for explicit racial biases and replaced with more accurate measures of biology, said Dr. Jaya Aysola, executive director of the Penn Medicine Center for Health Equity Advancement.
Providers must understand the biases in these tools and make necessary updates, Persaud said.
Health systems may face operational challenges aligning stakeholders to make system-level changes and educating clinicians, said Dr. Ankita Sagar, system vice president for clinical standards and variation reduction at CommonSpirit Health. Including clinical teams in the process is critical to avoiding resistance from potentially dubious professionals who are accustomed to doing things a certain way.
"Any change that's going to have to happen around a certain guideline involving spirometry, for example, has to happen with and by the pulmonologist," Aysola said.
Identifying racial bias
Detecting racial bias in tools is straightforward, Aysola said. Many have explicit race-based metrics that are identifiable and can be easily removed or replaced.
The healthcare system needs to take time to review them. Researchers, medical societies and health systems have led the effort, in partnership with insurers, EHR vendors and clinicians. Medical societies are revising calculators and adopting new clinical guidelines to influence broader change.
For instance, the Doris Duke Foundation recently distributed $10 million to the American Academy of Pediatrics, the American Heart Association, the American Society of Hematology, the National Academy of Sciences and the Coalition to End Racism in Clinical Algorithms (which represents New York City-based health systems) to build infrastructure to review clinical algorithms and guidelines for racial bias and generate alternatives that improve health outcomes.
The American Heart Association is reassessing equations and clinical calculators that incorporate race to predict the likelihood of a patient having a heart attack or a stroke. The patient advocacy organization convened 16 internal councils that cover areas such as cardiovascular disease and stroke to prioritize what algorithms are most commonly used and should be reviewed.
Penn Medicine began researching and eliminating tools with racial bias after receiving a grant from the Agency for Healthcare Research and Quality in 2021 to support an internal review team that evaluates race-based clinical tools.
This month, Penn Medicine also joined a coalition of Philadelphia hospitals, in partnership with Independence Blue Cross, to investigate 15 commonly used clinical support tools with explicit race-based corrections. Together, they are exploring tools used for cardiology, oncology, nephrology, obstetrics and gynecology, urology, and orthopedics.
Providers should focus their research efforts on areas with the most significant disparities in care, Sagar said. CommonSpirit Health recently phased out a race-based correction in the eGFR calculator, which is used to measure kidney function and leads to delayed interventions for Black patients. The health system prioritized updating this tool because of the massive disparities in health outcomes related to kidney failure, she said.
"We needed to change the way we're diagnosing kidney disease because we are capturing people too late. We need to capture them earlier, specifically African Americans," Sagar said.
Instituting change
Some systems are easier to change than others. Clinical tools such as the eGFR calculator are used across various disciplines and require system-level solutions.
CommonSpirit phased out the race-based correction in the eGFR calculator at its 140 hospitals in two years. The health system began implementing the changes after the National Kidney Foundation revised its medical guidelines in 2021.
The nonprofit company coordinated with its laboratories, informatics team, clinicians and health equity leaders, Sagar said. CommonSpirit started at its largest hospitals, then moved on to the rest. "It's a matter of due diligence, details and ensuring that there was no missed opportunity or miscommunication between clinical teams and other teams," she said.
Including clinicians in the review process and articulating the broader role of race and ethnicity in medicine is key to promoting buy-in, Sagar said. Medical schools need to update curricula to ensure race-based corrections aren't still being taught, Aysola said. Health systems need to re-educate those already in the workforce, including an older generation of practitioners who have incorrectly used race as a proxy of biology for their entire careers.
"There's going to be clinicians that are not going to understand. They're genuinely not going to believe that race is not some indicator of biological differences that needs to be corrected for. That's a huge problem," Aysola said.
Medical societies hope to leverage their relationships with clinicians to provide education and guidance, Persaud said.
"I know this is this could be viewed as a controversial thing to do, but we just have to confront the fact that we live in a society where racism is a problem," Persaud said. "It was a problem when these clinical tools were being developed. The developers of these tools believed that it was appropriate to include race and they were wrong, and we were all wrong to have continued to use them."
Additionally, health systems advancing efforts to eliminate race-based medicine are closely watching the consequences for quality and safety. There needs to be more data on how specific algorithms affect clinical outcomes, but Aysola believes removing antiquated science will naturally improve care.
"When you equalize whether patients are being referred to transplant earlier or referred to nephrology earlier or just getting the exact same algorithm as a white girl for pediatric UTIs, you've eliminated disparities," Aysola said.
