“Is omitting race and ethnicity as a predictor in colorectal cancer recurrence risk prediction models associated with racial and ethnic bias?” is the main question Dr. Aasthaa Bansal, an associate professor in the Public Health Sciences Division at Fred Hutch, and her team aimed to address in a recent publication in JAMA Network Open.
“There is ongoing debate around whether race and ethnicity should be included as a risk factor in clinical prediction algorithms” said Sara Khor, a PhD student in Dr. Bansal’s lab and first author of the study. There is currently a lack of consensus on whether and how race and ethnicity should be included in clinical risk prediction models used to guide healthcare decisions, the authors noted. “Many groups have called for the removal of race in clinical algorithms,” Khor continued. This is due to concerns about racial profiling and biased treatment. Yet, it remains unclear whether simply omitting these variables from algorithms will ultimately improve care decisions for patients of minoritized racial and ethnic groups. “Our work contributes to this debate by empirically showing the effect of removing race as a predictor on decision making for minoritized racial groups in the colorectal cancer recurrence (CRC) setting,” Khor added.
To better understand the effects of race and ethnicity parameters on CRC risk prediction algorithm performance, Dr. Bansal’s team conducted a retrospective study on a cohort of 4230 patients who underwent a CRC resection between 2008 and 2013. Patients were binned into the subgroups of Asian/Hawaiian/Pacific Islander, Black/African Americans, Hispanic or non-Hispanic White.
Researchers developed four CRC risk prediction models: 1) the race-neutral model that excluded race and ethnicity as a predictor, 2) the race-sensitive model that included race and ethnicity, 3) a model with two-way interactions between clinical predictors and race and ethnicity and other covariants; and 4) the race-stratified model that separate models by race and ethnicity. To assess the algorithms’ performance, the authors evaluated the models’ calibration (using calibration intercept and slope), discriminative abilities, false positive and false negative rates, and positive predictive values and negative predictive values.
Dr. Bansal’s team found that the race-neutral model had poorer calibration, negative predictive power, and false negative rates among patients from racial and ethnic minority groups compared with non-Hispanic white patients. Conversely, including race and ethnicity as a variable improved model’s accuracy and increased algorithmic performance in calibration, discriminative ability, false negative rates and both positive and negative predictive value. These results suggest that race and ethnicity as a predictor resulted in a more effective CRC recurrence risk algorithm. Omitting race and ethnicity from the clinical algorithm could lead to inappropriate care recommendations for patients who belong to minoritized racial and ethnic groups, ultimately contributing to health disparities.
Going forward, Khor is focusing on understanding “how the inclusion or exclusion of race in risk prediction algorithms impacts long-term health outcomes for minoritized racial groups.” To do so, she will “apply methods that incorporate subgroup-specific benefits, harms, and costs. Some of our group’s ongoing work includes developing complex decision analyses to assess these long-term consequences to using algorithms with and without race,” she concluded.
The spotlighted work was funded by the National Institutes of Health, the Agency for Healthcare Research and Quality, a predoctoral fellowship from the PhRMA Foundation, and the American Foundation for Pharmaceutical Education Predoctoral Fellowship in Health Outcomes Disparities.
Fred Hutch/University of Washington/Seattle Children’s Cancer Consortium members Drs. Veena Shankaran, and Aasthaa Bansal contributed to this study.
Khor S, Haupt EC, Hahn EE, Lyons LJL, Shankaran V, Bansal A. Racial and ethnic bias in risk prediction models for colorectal cancer recurrence when race and ethnicity are omitted as predictors. JAMA Netw Open. 2023 Jun 1;6(6):e2318495.2023.18495. PMID: 37318804; PMCID: PMC10273018.