Article Text

Download PDFPDF

195 Discordance in race and ethnicity between medical record and self-reported data
  1. Brianna Mills1,
  2. Kelsey Conrick1,
  3. Danae Dotolo2,
  4. Christopher St. Vil3,
  5. Megan Moore2,
  6. Ali Rowhani-Rahbar1
  1. 1University of Washington, Injury Prevention and Research Center
  2. 2University of Washington, School of Social Work
  3. 3SUNY Buffalo, School of Social Work


Statement of Purpose Research into race/ethnicity-based injury disparities requires accurate race/ethnicity data. We compared the race/ethnicity documented in electronic medical records (EMRs) against self-report identity to measure the magnitude of EMR misidentification and potential contributing factors.

Approach Descriptive statistics and misclassification matrices were used to assess sensitivity of EMR to true (self-reported) race/ethnicity within two study cohorts (one including all trauma mechanisms [n=95] and one of firearm assaults [n=217]) at a Level-1 trauma center. Both cohorts oversampled people of color. As processes for data collection vary based on how patients arrive, logistic regression models explored associations between arrival type (transfer, EMS, or walk-in) and accurate EMR identification within a subgroup of individuals reporting a single race/ethnicity.

Results Of 276 individuals who self-reported as a single race that mapped to an existing EMR race, only 65.6% had their race correctly identified in the EMR. More than 90% of Native Hawaiian/Pacific Islanders were misidentified. Patients brought in by EMS and walk-in patients were more likely to be correctly identified than patients transferred from another facility, but this was not statistically significant (OR 1.61, 95%CI 0.59–3.29 and OR 1.39, 95%CI 0.76–3.39, respectively). Although the majority of patients were correctly identified as either non-Hispanic (194/199) or Hispanic or Latino (53/71), the EMR sensitivity for identifying Hispanic/Latino ethnicity was only 74.6%. Multiracial identity was not included in standard EMR racial categories, systematically miscategorizing 18.9% (n=18) of the first and 15.2% (n=33) of the second cohort. More than 30% of multiracial individuals self-reported an American Indian/Alaska Native identity that was not captured by EMR.

Conclusions Large proportions of injured patients are misidentified by race/ethnicity in EMR. Misidentification is most common among indigenous people of color.

Significance Misrepresentation of race/ethnicity in EMR suggests injury disparities among Hispanic and indigenous groups may actually have been be underestimated.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.