Comparing EHR-recorded Race/Ethnicity to Self-reported Race/Ethnicity: Insights from the All of Us Research Program

Poster Number: P171
Presentation Time: 05:00 PM - 06:30 PM

Abstract Keywords: Bioinformatics, Data Sharing, Information Visualization
Primary Track: Applications

This study evaluates the accuracy of EHR-recorded race/ethnicity against the gold standard of self-reported data from the All of Us Research Program, aiming to identify factors on accuracy across various groups. Performance metrics such as sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were calculated to evaluate EHR-recorded race/ethnicity data against participant self-reported race/ethnicity separately. The logistic regression was used to examine the association between race/ethnicity agreement (agree/disagree) and EHR-recorded race/ethnicity, also the association between EHR race/ethnicity missing (yes/no) and self-reported race/ethnicity.

Results from 217,546 participants show a 12% discrepancy rate between EHR and self-reported race, with high accuracy for Whites. Non-White race groups were more likely to have race discrepancies and missing EHR race data. Out of 224691participants in the All of Us program with recorded ethnicity data in both EHR and the Basic survey, 4% reported a different ethnicity in the EHR compared to the survey. Hispanic participants were more likely to report ethnicity discrepancies.

Our study reveals high concordance of EHR-recorded race with self-reported race for White participants, followed by Black and Asian individuals. However, significant discrepancies exist for other race groups, including MENA, NHPI, Others, and Mixed. This underscores the necessity for efforts to enhance the accuracy of EHR-recorded race/ethnicity for certain underrepresented groups.

Speaker(s):
Lina Sulieman, PhD
Vanderbilt University Medical Center