A new study published in Science finds that a widely applied medical algorithm used to manage health treatments in large populations suffers from serious racial bias in practice. According to the research, fixing the disparity in this algorithm would “increase the percentage of Black patients receiving additional help from 17.7 to 46.5%.” The algorithm researchers focused on is sold by healthcare company Optum, according to the Washington Post.
The problem seems to be that the algorithm bases its calculations on potential “costs” of patients’ health care and not whether or not people are actually ill. In white populations, those costs are a potentially helpful predictive indicator of who might be in line for “high-risk care management” programs. These are programs that use more robust resources to handle patients with more “complex health needs.” The study used data from an algorithm that is similar to many applied to around 200 million people for this very purpose.
Unfortunately, our entire infrastructure is built on racism and the biases that arise from that poison seed. And as a result, the data being used by the algorithm is affected by that bias. Specifically cost. Black patients make up lesser costs than white patients with similar chronic health issues and this leads the algorithm to target white patients over black patients who exhibit the same medical needs. As the Post points out, the intention of using costs was supposed to make the algorithm race blind in its targeting practices.
Healthcare costs incurred by black patients is lower because there is a history of bias against black patients in general. This includes doctors spending less time with black patients, giving out less pain medications to black patients in pain, underserving black cancer patients, leading to less rigorous treatments … and lower costs.
The researchers hope that their research will give the creators of this algorithm the data they need to revamp a better, less racially biased, version of the tool.