Short Report: Race and Ethnicity Misclassification in Kidney Transplantation Research

Recently, the misuse of race as a biological variable, rather than a social construct, in biomedical research has received national attention for its contributions to medical bias. In national transplant registry data, bias may arise from measurement imprecision because of the collection of provider-perceived race rather than patients’ own self-report. METHODS. We linked Scientific Registry of Transplant Recipients data to a prospective, multicenter cohort study of adult kidney transplant patients (December 2008–February 2020) that collects patient-reported race. We computed Cohen’s kappa statistic to estimate agreement between provider-perceived and patient-reported race in the 2 data sources. We used an unadjusted generalized linear model to examine changes in agreement over time. RESULTS. Among 2942 kidney transplant patients, there was almost perfect agreement among Asian (kappa = 0.88, 95% confidence interval [CI], 0.84-0.92), Black (kappa = 0.97, 95% CI, 0.96-0.98), and White categories (kappa = 0.95, 95% CI, 0.93-0.96) and worse agreement among Hispanic/Latino (kappa = 0.66, 95% CI, 0.57-0.74) and Native Hawaiian/Other Pacific Islander categories (kappa = 0.40, 95% CI, 0.01-0.78). The percent agreement decreased over time (difference in percent agreement = –0.55, 95% CI, –0.75 to –0.34). However, there were differences in these trends by race: –0.07/y, 95% CI, –0.21 to 0.07 for Asian; –0.06/y, 95% CI, –0.28 to 0.16 for Black; –0.01/y, 95% CI, –0.21 to 0.19 for Hispanic/Latino; –0.43/y, 95% CI, –0.58 to –0.28 for White categories. CONCLUSIONS. Race misclassification has likely led to increasingly biased research estimates over time, especially for Asian, Hispanic/Latino, and Native Hawaiian/Other Pacific Islander study populations. Improvements to race measurement include mandating patient-reported race, expanding race categories to better reflect contemporary US demographics, and allowing write-ins on data collection forms, as well as supplementing data with qualitative interviews or validated measures of cultural identity, ancestry, and discrimination.