153 - Demographic Discordance Between Patient Reported and Electronic Medical Records

Poster Number: 153
Publication Number: 153.411

Jasmine Lemmons, Texas Children's Hospital, Houston, TX, United States; Nidhi Singh, Baylor College of Medicine, HOUSTON, TX, United States; Elizabeth A. Camp, Baylor College of Medicine, Houston, TX, United States

Background: Socio-demographics, such as patient race, gender, preferred spoken language, and ethnicity, are commonly used to measure health care disparities, improve quality of care, and evaluate population-specific clinical outcomes. There is conflicting literature on the accuracy of data capture on demographic variables. 

Objective: To assess the consistency of medical staff entry versus patient-reported ethnicity, race, gender, preferred spoken language, and preferred written language within the electronic health record (EHR). We hypothesize that there will not be a discordance between self-reported and staff collected data of patient race, ethnicity, and preferred spoken language.

Design/Methods: This was a cross-sectional study in which demographic data from the EHR and information collected through questionnaire provided by researcher were compared. If there was disagreement between EHR data and the questionnaire, then the patient in-take form (PIF), which is provided by hospital personnel to assist in EHR data entry, was reviewed for comparison. Frequencies and percentages for each category across the three types of records were reported.  To explore test of agreement between patient self-reported race, ethnicity, language, and gender in comparison to what is documented in the EHR and PIF, Krippendorff's alpha along with a bootstrapping procedure was utilized to calculate inter-rater reliability. The results were reported using Krippendorff's alpha and 95% confidence intervals (CI).

Results: There were 204 patients in the study. A majority were Hispanic who preferred to speak and write in English.  There was high agreement for all three types of data collection when documenting ethnicity (α >0.80) and gender (α >0.90).  Race was not concordant and had the lowest Krippendorff’s alpha ranging from 0.04 to 0.49.  Results for reporting preferred written and spoken language were varied with alphas ranging from 0.61 to 0.91. Conclusion(s): Ethnicity and gender are reliably documented between self-report and EHR. Race had the highest discordance for documentation. As race is a common socio-demographic identifier used to determine health care disparities and assess population-specific clinical consequences, careful consideration should be given when using the EHR as the reporting data source.   
CV LemmonsUpdate CV Lemmons PAS.pdf
Table 2.Inter-rater reliability between patient self-reported ethnicity, race, gender, preferred spoken language, and preferred written language and hospital documentation. (Nf204)