Effect of race and ethnicity on advanced breast cancer risk prediction model performance

Race-naive prediction modeling is proposed as a method to address algorithmic bias. We evaluated model performance by removing race and ethnicity from the Breast Cancer Surveillance Consortium 6-year cumulative advanced breast cancer risk model used for decision-making on screening frequency and supplemental imaging. Excluding race and ethnicity from the model resulted in overestimation of risk in Asian women compared to the original model [expected/observed = 1.28 (95%CI = 1.05-1.66) vs. 1.00 (95%CI = 0.82-1.29)] and underestimation of risk in Black women [expected/observed = 0.61 (95%CI = 0.53-0.70) vs. 1.00 (95%CI = 0.88-1.15)]. Advanced breast cancer among Asian women classified as intermediate/high risk increased from 6.1% to 16.7% while among Black women decreased from 75.3% to 47.5%. Fewer Black women with advanced breast cancer were identified as intermediate/high advanced breast cancer risk, with the model excluding race and ethnicity due to worse calibration. This could result in suboptimal implementation of tailored screening strategies to reduce advanced breast cancer diagnoses and breast cancer mortality.