In the United States, a major hospital network relies on an algorithm to prioritize kidney transplants. But this race-based tool recently sidelined hundreds of Black patients from the waiting list.
Algorithms in healthcare aim to support clinical decisions and improve outcomes. Yet a study in the Journal of General Internal Medicine (October 15, 2020) reveals pitfalls. Harvard researchers analyzed data from the Mass General Brigham network on 57,000 patients with chronic kidney failure. The algorithm, designed to assess transplant urgency, downplayed risks for 700 Black patients—one-third of Black patients in the dataset. As a result, 64 patients missed priority placement on the emergency transplant list.
This underscores how flawed AI can exacerbate health disparities, as confirmed by rigorous analysis from leading experts.
AI discrimination isn't new. In 2019, another tool—without explicit race inputs—still favored white patients for chronic care access. Here, the algorithm explicitly uses race, amplifying bias.
This case has drawn scrutiny from U.S. federal lawmakers. Rep. Richard Neal (D) called for reevaluating race in algorithms, citing risks of delayed diagnoses and subpar care for Black and Latino patients. He urged industry leaders and the National Institutes of Health (NIH) to probe clinical algorithm impacts. Sen. Elizabeth Warren and colleagues echoed this, pressing the Department of Health for action.
