A new study challenges the growing trend of removing race from medical algorithms, revealing that race-adjusted tools for predicting colorectal cancer risk may actually help reduce health disparities. Researchers from Cornell University, the University of Chicago, and the University of California, Berkeley, found that including race in these predictive models leads to more accurate risk assessments for Black patients.
The study, published in the Proceedings of the National Academy of Sciences, analyzed data from 77,836 diverse participants in the Southern Community Cohort. By developing two algorithms – one race-blind and one race-adjusted – the team compared their effectiveness in predicting colorectal cancer risk for Black and white participants.
Results showed that race-blind algorithms tended to underpredict cancer risk for Black participants while overpredicting risk for white participants. In contrast, the race-adjusted algorithm more accurately predicted risk for both groups. Importantly, when using the race-adjusted model, 74.4% of participants ranked in the highest risk category were Black, compared to only 66.1% with the race-blind approach.
The disparity stems from less reliable family history data for Black patients, a crucial factor in cancer risk assessment. By accounting for race, the algorithm compensates for this data quality issue, potentially identifying more Black patients who would benefit from cancer screening.
Emma Pierson, senior author of the study, emphasizes the importance of carefully evaluating the impact of removing race from medical algorithms. While acknowledging the need to improve data quality in the long term, Pierson argues that race-adjusted algorithms can serve as a crucial stopgap measure to ensure accurate predictions and appropriate care for patients today.
“Considering race in colon cancer prediction algorithms reduces disparities, researchers find” (September 19, 2024)
