Race Adjustments in Algorithms Boost CRC Risk Prediction

TOPLINE:

Accounting for racial disparities, including in the quality of family history data, enhanced the predictive performance of a colorectal cancer (CRC) risk prediction model.

METHODOLOGY:

• The medical community is reevaluating the use of race adjustments in clinical algorithms due to concerns about the exacerbation of health disparities, especially as reported family history data are known to vary by race.
• To understand how adjusting for race affects the accuracy of CRC prediction algorithms, researchers studied data from community health centers across 12 states as part of the Southern Community Cohort Study.
• Researchers compared two screening algorithms that modelled 10-year CRC risk: A race-blind algorithm and a race-adjusted algorithm that included Black race as a main effect and an interaction with family history.
• The primary outcome was the development of CRC within 10 years of enrollment, assessed using data collected from surveys at enrollment and follow-ups, cancer registry data, and National Death Index reports.
• The researchers compared the algorithms’ predictive performance using such measures as area under the receiver operating characteristic curve (AUC) and calibration and also assessed how adjusting for race changed the proportion of Black participants identified as being at high risk for CRC.

TAKEAWAY:

• The study sample included 77,836 adults aged 40-74 years with no history of CRC at baseline.
• Despite having higher cancer rates, Black participants were more likely to report unknown family history (odds ratio [OR], 1.69; P < .001) and less likely to report known positive family history (OR, 0.68; P < .001) than White participants.
• The interaction term between race and family history was 0.56, indicating that reported family history was less predictive of CRC risk in Black participants than in White participants (P = .010).
• Compared with the race-blinded algorithm, the race-adjusted algorithm increased the fraction of Black participants among the predicted high-risk group (66.1% vs 74.4%; P < .001), potentially enhancing access to screening.
• The race-adjusted algorithm improved the goodness of fit (P < .001) and showed a small improvement in AUC among Black participants (0.611 vs 0.608; P = .006).

IN PRACTICE:

“Our analysis found that removing race from colorectal screening predictors could reduce the number of Black patients recommended for screening, which would work against efforts to reduce disparities in colorectal cancer screening and outcomes,” the authors wrote.

SOURCE:

The study, led by Anna Zink, PhD, the University of Chicago Booth School of Business, Chicago, was published online in PNAS.

LIMITATIONS: 

The study did not report any limitations.

DISCLOSURES: 

The study was supported by the National Cancer Institute of the National Institutes of Health. The authors declared no conflicts of interest.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.