NYU Langone Tackles Implicit Bias in Clinical Algorithms for Health Equity

NYU Langone is taking the next step towards realizing health equity by refining clinical algorithms to remove race-based adjustments and focusing on social determinants of health instead.

This comes as the healthcare industry grapples with an implicit bias problem in its algorithms.

Most medical specialties utilize clinical decision tools, with over 90 percent of hospitals incorporating medical algorithms into EHRs to enhance patient outcomes. However, when these tools create biased assumptions based on flawed data, they can lead to poor health outcomes, especially for minority patients.

“Clinical calculators are used as a proxy for the gold-standard treatment of patients,” said Ogedegbe, Adolph, and Margaret Berger, Medicine and Population Health professor. “But when the calculations adjust outcomes based on race, the formulas can result in different treatments and procedures for Black patients than White patients, often resulting in worse outcomes.”

A minimum of 15 clinical algorithms in use by hospitals and supported by many top medical associations have race-based adjustments, noted Olugbenga G. Ogedegbe, MD, founder of NYU Langone’s Institute for Excellence in Health Equity.

“Race is a social construct, and yet it has been used as a substitute for genetic and biological factors for decades,” said Kathie-Ann Joseph, MD, MPH, professor of surgery and population health and vice chair for diversity and health equity in surgery and at the NYU Langone Transplant Institute. “These calculations assume inherent differences instead of digging deeper into the social determinants that explain why Black people have worse outcomes.”

NYU Langone said it has only recently begun prioritizing efforts to address racial bias in clinical algorithms.

A recently corrected calculator once estimated that Black and Hispanic women were 20 percent less likely to have a successful vaginal birth following cesarean delivery than White women. In the past, the biased correction would have impacted birth plan choices for individuals in these groups leading to harmful health outcomes since cesarean deliveries carry a higher risk of complications.

“The problem was with the assumption that the data reflected inherent or biological differences,” said Dana R. Gossett, MD, professor at the Stanley H. Kaplan and chair of Obstetrics and Gynecology. “The truth is that where a patient gets prenatal care, who delivers the baby, and how much counseling and support they receive are far more pertinent than the color of their skin.”

NYU Langone previously utilized a kidney function assessment formula that assigned Black patients a higher filtration-rate score, indicating the kidney’s capacity to eliminate creatinine, a waste product. The historical GFR score graded Black patients higher due to an incorrect assumption about increased muscle mass and creatinine production.

“The bad assumption baked into this tool is that Black patients have more muscle mass and therefore higher levels of creatinine,” said Dr. Ogedegbe. “This is just false.” For Black patients with early-stage kidney disease, the adjustment could mean a missed opportunity to receive specialty care, and for those with end-stage kidney failure, the score correction could render them ineligible for the kidney transplant waitlist.”

Ilseung Cho, MD, NYU Langone’s chief quality officer, emphasized that doctors have an ethical and moral responsibility to address bias to provide equitable care.

“As doctors, we should be asking why race is a part of these guidelines or of our care,” said Ilseung Cho, MD, NYU Langone’s chief quality officer, who is leading the effort to review clinical calculators for implicit bias and remove or correct them. “We have an ethical and moral responsibility to close any gaps.” He notes that simply pointing out race-based corrections to doctors is often enough to move the needle. “Everybody here aspires to provide equitable care,” Cho said.