Minimizing the Impact of Bias in Healthcare Algorithms

Last week, researchers published a special communications article in JAMA Network Open outlining principles to address and minimize the impacts of racial and ethnic bias in healthcare algorithms. The article emphasized the critical importance of minimizing bias to achieve health equity and ensure marginalized populations are getting quality care.

As the article mentions, healthcare algorithms are used in nearly every component of patient care, including diagnosis, treatment, prognosis predictions, risk stratification, and research allocations. However, many of these algorithms favor White patients, creating an inequity for the already marginalized racial and ethnic minority groups.

The report outlines several cases in which healthcare algorithms used to manage patient care differentiated between racial and ethnic groups, disadvantaging minority communities.

For example, in chronic kidney disease (CKD), the estimated glomerular filtration rate (eGFR) is used to assess the severity of kidney disease by evaluating kidney function. Lower eGFRs usually indicate more severe disease, signaling providers to evaluate patients for a kidney transplant. However, some facilities use a biased algorithm that yields higher eGFRs for Black patients, delaying the cascade of events required for a transplant.

This is just one example of biased algorithms in healthcare. The researchers postulate that racially and ethnically biased algorithms have been applied in heart failure, cardiac surgery, rectal cancer, breast cancer, and more.

To combat these biased algorithms and mitigate the resulting health inequities, the Agency for Healthcare Research and Quality (AHRQ), the National Institute on Minority Health and Health Disparities (NIMHD), a subset of the National Institutes of Health (NIH), the United States Department of Health and Human Services (HHS) Office of Minority Health, and Office of the National Coordinator for Health Information Technology (ONC) convened a panel of nine diverse stakeholders to establish principles for elimination bias in healthcare algorithms.

The panel outlined five guiding principles for using algorithms in healthcare. The first principle emphasizes the importance of promoting health equity across all phases of the algorithm lifecycle.

Additionally, the second principle prioritizes using transparent and explainable algorithms. The other principles include engaging patients and communities, identifying fairness issues and trade-offs, and establishing accountability for the outcomes perpetuated by algorithm inequity.

“Multiple stakeholders must partner to create systems, processes, regulations, incentives, standards, and policies to mitigate and prevent algorithmic bias,” concluded researchers in the publication.