Race/Ethnicity Accuracy in Electronic Medical Records at Pediatric Hospitals

Abstract

Introduction

Numerous studies have examined health inequities. Inherent to the validity of these studies is accuracy of racial and/or ethnic designations. Without knowledge of the degree of misattribution, studies risk missing disparities that exist and finding some that do not. This study examined the accuracy of electronic medical record (EHR) racial/ethnic attribution across the three largest pediatric hospitals in Michigan.

Methods

For each hospital the specific options for the classification of race/ethnicity available in their EHR were identified (ranging from 6 to 47 for race and 2 to 9 for ethnicity). Parents were approached in outpatient clinics and asked to select from a list of the options at their hospital, the race and ethnicity of their child. This was used as the gold standard for comparison with the information in the EHR.

Analysis for matching occurred in three stages. 1) Exact match; 2) “Prioritizing” of non-white component of combinations (e.g., any combination with Black designated as Black); 3) “Prioritizing” plus consolidation (e.g., putting all Asian groups together).

Results

Approximately 1500 parents participated from each hospital. Across the 3 hospitals, exact matching for race ranged from approximately 50% to 78%. Lower matching occurred for hospitals using more categories. After consolidation and grouping of categories, match rates improved markedly.

Conclusions

Accuracy in race and ethnicity data is essential for valid assessment of inequities. Hospitals with expanded categories may need to consolidate/group race/ethnicity data for accurate analyses. Efforts to improve EHR accuracy and to statistically account for current error rates are urgently needed.