The growing, widespread use of algorithms to make healthcare decisions for patients could be adding to racial bias against minorities, a new study has found.
Algorithms are the mathematical rules that tell a health care provider’s computer program how to solve problems affecting a patient’s access to medical treatment, quality of care and health outcomes. Doctors increasingly rely on their analysis of a patient’s medical and insurance history to recommend appropriate treatment.
According to a study that seven public health researchers published Friday in JAMA Health Forum, 18 commonly used algorithms flag ethnicity and race in haphazard ways that may reinforce unequal treatment of dark-skinned patients due to a lack of oversight and knowledge of their functions.
The researchers posed 11 questions about the algorithms to the representatives of 42 clinical professional societies, universities, government agencies, health insurance payers and health technology organizations.
“Findings suggest that standardized and rigorous approaches for algorithm development and implementation are needed to mitigate racial and ethnic biases from algorithms and reduce health inequities,” the researchers wrote in the study.
Survey respondents recommended “guidance and standardization from government and others” to purge any bias and prevent the use of race as a “proxy for clinical variables,” the study noted.
“Only 20% of health outcomes are determined by the provision of health care services, and an individual’s ZIP code has more influence on their health than their own genetic code,” an anonymous clinician wrote in the survey, noting that racial data favors patients from better neighborhoods.
Some health care professionals echoed the study’s conclusions, noting that algorithms often pull data from older medical tests that treat skin color as a biological difference.
“Many tests in medicine are based on race, from renal function to lung strength,” said Dr. Panagis Galiatsatos, a faculty health equity leader at the Johns Hopkins School of Medicine. “Right now, we are attempting to change pulse oximeter readings, which are known to cause false reports in dark-skinned individuals, often missing key hypoxemia that would impact medical management.”
Another problem could be the algorithms themselves, said Katy Talento, a former top health adviser at the White House Domestic Policy Council under President Donald Trump.
She now serves as executive director of the Alliance of Health Care Sharing Ministries, a District of Columbia-based association of Christians who work to “rehumanize” medicine by sharing medical costs.
“The study rightly points out that race is a bad proxy for what matters: health history, genetics and social determinants of health such as income,” Ms. Talento said in an email. “Our broken system requires clinicians to use bots, checklists and rapid-fire office visits driven by insurance payment models instead of doctor-patient relationships.”
According to diversity experts, it’s easy to see how mathematical calculations might cause minorities to receive inferior medical treatment.
“Algorithms are scientific tools to analyze data, but the variables are input by humans who of course can have biases toward racial and ethnic groups,” said Tyrone Howard, a Black education professor at UCLA who specializes in racial equity.
Some conservatives cautioned against reading too much into the algorithms. The potential for bias does not prove racist calculations are to blame for unequal health outcomes, they said.
“Certain people are prone to certain diseases based on race, culture, economic status and learned behaviors,” said Gregory Quinlan, a former registered nurse who leads the conservative Center for Garden State Families in New Jersey. “Gay men are at higher risk of HIV-AIDS and monkeypox. That is not bigotry to say that; it’s a statistical, medical fact.”
• Sean Salai can be reached at ssalai@washingtontimes.com.
Please read our comment policy before commenting.