Penn researchers find evidence of racial bias in models used to allocate limited resources

A study by Penn researchers found evidence of racial bias in models used to allocate limited health care resources during the COVID-19 pandemic.

Health care practitioners follow crisis standards of care during health crises that are intended to save the most lives and do so fairly, which suggests prioritizing patients with the best chances of survival. The most common mortality prediction model, however, overestimates mortality among Black patients and underestimates mortality among white patients, putting Black patients behind white patients in the line to receive scarce resources, Penn Leonard Davis Institute of Health reported.

The researchers found that the Sequential Organ Failure Assessment score, which is used in many hospitals to predict how likely a patient is to survive their hospital stay, is racially biased. According to the study, 81% of Black patients from lower priority CSC groups and 9% of all Black patients would be erroneously left out of the highest priority CSC group.

"Both scores were miscalibrated within racial subgroups in ways that could systematically divert critical care resources away from Black patients,” the authors wrote in the study.   

The study was conducted by Perelman School of Medicine instructor George Anesi, Perelman School of Medicine second-year fellow Christopher Chesley, Perelman School of Medicine assistant professor Nwamaka Eneanya, Perelman School of Medicine assistant professor Gary Weissman, Perelman School of Medicine assistant professor Michael Harhay, Perelman School of Medicine professor Scott Halpern, and researchers from Duke University, the University of Chicago, and Kaiser Permanente. 

"We must develop equitable models that accurately predict mortality among all patients lest we exacerbate the unacceptable racial disparities that have become so evident during the COVID-19 pandemic,” study author Deepshikha Charan Ashana wrote in Penn LDI.