News Science & Society Bias in a common health care algorithm disproportionately hurts black patients Simple tweaks to the machine-learning program could eliminate the disparity, researchers say One widely used machine-learning program that helps hospitals identify patients who need extra health services refers fewer black patients than white patients, a study finds. aldomurillo/E+/Getty Images Share this:EmailFacebookTwitterPinterestPocketRedditPrint By Sujata Gupta October 24, 2019 at 2:00 pm A widely used algorithm that helps hospitals identify high-risk patients who could benefit most from access to special health care programs is racially biased, a study finds. Eliminating racial bias in that algorithm could more than double the percentage of black patients automatically eligible for specialized programs aimed at reducing complications from chronic health problems, such as diabetes, anemia and high blood pressure, researchers report in the Oct.