Equal Risk

This led to tables of race-specific normal lung functions from the 1850s until January 2023, when the American Thoracic Society finally retired them. Similarly, the American Heart Association removed race from its risk predictions last November, and the National Kidney Foundation removed race from its kidney function estimator.

When you include race-specific norms or “adjust” for race, you systematically bias the detection of disease. Hypertension is one of the sharpest examples. For decades, we’ve been looking for the “Black gene” for hypertension, and it just doesn’t exist. You see huge disparities, but they are socially, not biologically, determined. The reality is being a Black person in our society contributes to much-earlier onset, worse hypertension and all the downstream consequences.

The way we provide health care contributes to these problems. Medicaid, which is used in greater numbers by Black and brown people, is another example of structural racism leading to poor health outcomes for people of color. The amount of money that Medicaid reimburses both physicians and hospitals is much less than the actual cost of providing care and services. We’ve disincentivized the average physician to accept patients from historically underserved communities.

There’s a teachable moment right now and recognition that we have to do better. While ancestry is a real issue — there are heritable disorders like Tay-Sachs disease in Ashkenazi Jews — having dark skin does not put you at risk of anything except social aggressions. Many medical societies are concluding that we should not have race in risk-prediction equations, but it doesn’t happen overnight. After a 2017 study of racial bias in pain assessment and treatment recommendations, the Association of American Medical Colleges added teaching standards that address racial disparities in treatment to medical school licensing criteria. More work to root out additional, deeply seated racist practices in health care is long overdue.

It can be very difficult to find a data set that includes all the necessary social, environmental and genetic factors that you would want to tease apart and accurately predict disease risk. As a cancer epidemiologist interested in health disparities, I’m particularly focused on prostate cancer, where we see very large disparities by race and ethnicity. Incidence is much higher among Black men, and much lower among Asian men, with white and Hispanic men in between. Studies have shown that disparities in prostate cancer survival are entirely preventable. Black men and white men who present with the same type of tumor at the same age, with all else being equal, should have the same outcomes. But due to U.S. health-care system inequities arising from structural racism, Black men do worse.

We’re an increasingly multiracial society. People don’t fit into neat boxes, and these risk calculators require neat boxes. It’s a much more complicated web using genetics, social, environmental and lifestyle factors to try to understand which of these factors is leading to the observed differences that we see by race and ethnicity.

This is a solvable problem with, yes, enormous challenges in pulling together different data types, but it’s getting easier. There have been some instances where taking race and ethnicity out of a model does not diminish the model’s performance, while in other instances, it has. In those latter circumstances, we really need to prioritize research to figure out what underlying causal effect is being captured by that variable within the model, since it is almost certainly not an effect of race. Policymakers and those developing clinical guidelines also need to think about equity considerations in how these models are used.

Not recognizing variations across diverse patient populations has led to gaps in medical education and research. When treating a dermatologic condition [that] disproportionately affects patients with skin of color, research on causes and therapeutic approaches is far more limited. Clinical textbook examples of common skin disorders are mostly individuals of European ancestry with lighter skin complexions. We need more funding for studies that specifically address dermatologic conditions in specific populations to ensure research is representative of our U.S. population. There is a low percentage of dermatologists who self-identify as an underrepresented group. Clinician diversity benefits cross-cultural awareness and provides patients with options of clinicians who have some insights into their cultural background.

The Skin of Color Society is providing guidance to researchers and clinicians on appropriate population descriptors used in research and clinical contexts and is actively expanding education on dermatologic disorders and populations with skin of color. The American Academy of Dermatology also is expanding education and diversity within dermatology. At Weill Cornell Medicine, I work closely with Dr. Eva Kerby and the Center for Diverse Skin Complexions on patient care, clinical research and education related to diverse skin types. Our residents have a robust set of didactic and practical educational modules on best practices for treating these various conditions across diverse patient populations.