Philadelphia hospitals drop race-based algorithms
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Natalie Peeples/Axios
A group of 13 Philadelphia-area health systems is ending the inclusion of race in screening algorithms that help make clinical decisions for lung, kidney and OB-GYN care, saying they incorporate outdated assumptions about biological differences between races.
Why it matters: It's part of an intensifying push to root out systemic biases in clinical tools as AI models take on a bigger role in diagnosis and treatment.
Driving the news: The Regional Coalition to Eliminate Race-Based Medicine is removing race adjustments from four commonly used tools that gauge lung function, anemia during pregnancy, the quality of donor kidneys and the opportunity to have a vaginal birth after a previous C-section.
- Officials said removing the race adjustments will speed up diagnoses for Black patients with chronic kidney disease, Black and Asian patients with lung disease and pregnant Black people, among other subgroups.
- Multiracial people and those who identify outside of a race category that was previously included in one of the tools will no longer be excluded or receive unclear guidance, they said.
- Changing the kidney function test calculation already helped more than 720 patients move up on kidney transplant lists, with 63 receiving a kidney transplant this year.
What they're saying: "By eliminating race adjustments in clinical decision support tools, we can help to dismantle harmful stereotypes and ensure that everyone receives the care they deserve, regardless of their race or ethnicity," said Rosangely Cruz-Rojas, vice president and chief diversity & equity officer for Main Line Health, one of the participating health systems.
Catch up quick: AI algorithms can sort through massive amounts of data quickly and identify patterns that might otherwise go unnoticed.
- But medical research has historically focused on men, and publicly available data is likelier to come from people of European descent. That can allow biases to creep into health care AI and result in misdiagnoses based on patients' race, ethnicity or gender.
- The calculator for assessing vaginal birth after cesarean delivery has led Black patients to get more C-sections than were necessary, by erroneously predicting that minorities were less likely to have success with a vaginal birth after a cesarean delivery, per Information Week.
What's next: The Philadelphia health systems are advocating to remove race from the Kidney Donor Risk Index, a formula used by the Organ Procurement and Transplantation Network to assess donor kidney quality.
