Algorithms creating discrimination in the US healthcare system
Researchers looked in nearly 50,000 records of patients.
Algorithm effectively low-balled the health needs of the hospital’s black patients.
Software was designed to predict patients’ future health costs, as a proxy for their health needs.
Algorithm’s bias reduced the proportion of black patients receiving extra help by more than half.
Discriminatory algorithm is used in the care of 70 million patients and developed by an insurance company.
Software was selecting patients likely to cost more in the future—not based on their actual health.
Challenge is to create software which will identify patients with complex care needs.