P2.171: Pediatric race, ethnicity, and gender identity data: accuracy of the electronic health record and agreement between patient and caregiver

Background: The accuracy of pediatric patient race, ethnicity and gender identity (GI) data in the electronic health record (EHR) has not been well characterized. Agreement on these demographic characteristics between pediatric patients and their parents/guardians/caregivers (hereafter ‘caregivers’) is also not well understood. The objectives of this study were to: 1) measure the accuracy of patient race and ethnicity data in a single hospital’s EHR; and 2) determine level of agreement of patient demographic data between pediatric patients and their caregivers.

Methods: In this prospective, observational study, we surveyed adolescents 12-17 years old and their caregivers presenting to an urban pediatric emergency department from May-December 2024. Our primary outcome was concordance of caregiver-reported race and ethnicity data for the patient with the EHR. Our secondary outcomes were concordance of patient demographic data between: 1) patient report and EHR, and 2) patient and caregiver report. We calculated descriptive frequencies for demographic characteristics as reported by patients and caregivers, and as documented in the EHR. For concordance, we calculated correlation coefficients, reporting kappa statistics with 95% confidence intervals (CI).

Results: There were 295 patient-caregiver dyads enrolled in the study (Table 1). For the outcome of concordance between caregiver report of patient race and the EHR there was substantial agreement for the most commonly-reported races: Asian (κ = 0.65, 95%CI 0.44, 0.86), Black/African American (κ = 0.76, 95%CI 0.66, 0.86), and White (κ = 0.66, 95%CI 0.58, 0.75). There was low-fair agreement for “more than one race” (κ = 0.27, 95%CI 0.03,0.52) and moderate agreement for Hispanic/Latino ethnicity (κ = 0.53, 95%CI 0.44, 0.61) (Table 2). There were similar results for patient report and the EHR: Asian (κ = 0.56, 95%CI 0.35, 0.77), Black/African American (κ = 0.73, 95%CI 0.63, 0.84), White (κ = 0.67, 95%CI 0.58, 0.75) race, and “more than one race” (κ = 0.21, 95%CI 0.09, 0.41). There was poor agreement for Hispanic/Latino ethnicity (κ = -0.17, 95%CI -0.22, -12) (Table 2). For the concordance of patient demographics between patient and caregiver, there was strong correlation for Asian (κ = 0.89, 95%CI 0.78, 1.00), Black/African American (κ = 0.90, 95%CI 0.83, 0.96), and White (κ = 0.84, 95%CI 0.77, 0.91) race and for Hispanic/Latino ethnicity (κ = 0.86, 95%CI 0.79, 0.92). There was also strong concordance for patient gender identity between patient and caregiver report (κ=0.93, 95% CI 0.90, 0.97).

Conclusion: Accuracy of patient race and ethnicity data in the EHR compared to caregiver and patient reported data was variable. There was a higher concordance between caregivers and patients. Accuracy of EHR race and ethnicity data should be interpreted cautiously and best practices should be implemented to improve data quality.