LAUSR

Introduction The true extent of racial and ethnic disparities in COVID-19 hospitalizations may be hidden by misclassification of race and ethnicity. This study aimed to quantify this inaccuracy in a hospital’s electronic medical record (EMR) against the gold standard of self-identification and then project data onto state-level COVID-19 hospitalizations by self-identified race and ethnicity. Methods To identify misclassification of race and ethnicity in the EMRs of a hospital in Honolulu, Hawaii, research and quality improvement staff members surveyed all available patients (N = 847) in 5 cohorts in 2007, 2008, 2010, 2013, and 2020 at randomly selected hospital and ambulatory units. The survey asked patients to self-identify up to 12 races and ethnicities. We compared these data with data from EMRs. We then estimated the number of COVID-19 hospitalizations by projecting racial misclassifications onto publicly available data. We determined significant differences via simulation-constructed medians and 95% CIs. Results EMR–based and self-identified race and ethnicity were the same in 86.5% of the sample. Native Hawaiians (79.2%) were significantly less likely than non–Native Hawaiians (89.4%) to be correctly classified on initial analysis; this difference was driven by Native Hawaiians being more likely than non–Native Hawaiians to be multiracial (93.4% vs 30.3%). When restricted to multiracial patients only, we found no significant difference in accuracy (P = .32). The number of COVID-19–related hospitalizations was 8.7% higher among Native Hawaiians and 3.9% higher among Pacific Islanders when we projected self-identified race and ethnicity rather than using EMR data. Conclusion Using self-identified rather than hospital EMR data on race and ethnicity may uncover further disparities in COVID-19 hospitalizations.