Take Racism Out of Medical Algorithms

COVID-19 has wreaked havoc on Black and Indigenous communities and other people of color, and U.S. medical institutions should be doing everything they can to root out and eliminate entrenched racial inequities. Yet many of the screening assessments used in health care are exacerbating racism in medicine, automatically and erroneously changing the scores given to people of color in ways that can deny them needed treatment.

{snip}

A recent paper in the New England Journal of Medicine presented 13 examples of such algorithms that use race as a factor. In every case, the race adjustment results in potential harm to patients who identify as nonwhite, with Black, Latinx, Asian and Native American people affected to various degrees by different calculations. These “corrections” are presumably based on the long-debunked premise that there are innate biological differences among races. This idea persists despite ample evidence that race—a social construct—is not a reliable proxy for genetics: Every racial group contains a lot of diversity in its genes. It is true that some populations are genetically predisposed to certain medical conditions—the BRCA mutations associated with breast cancer, for instance, occur more frequently among people of Ashkenazi Jewish heritage. {snip}

{snip}

A recent study in Science examined an algorithm used throughout the U.S. health system to predict broad-based health risks. The researchers looked at one large hospital that used this algorithm and found that, based on individual medical records, white patients were actually healthier than Black patients with the same risk score. This is because the algorithm used health costs as a proxy for health needs—but systemic racial inequality means that health care expenditures are higher for white people overall, so the needs of Black people were underestimated. In an analysis of these findings, sociologist Ruha Benjamin, who studies race, technology and medicine, observes that “today coded inequity is perpetuated precisely because those who design and adopt such tools are not thinking carefully about systemic racism.”

{snip}