Enhanced Identification of Hispanic Ethnicity Using Clinical Data

Abstract

Background: Collection of accurate Hispanic ethnicity data is critical to evaluate disparities in health and health care. However, this information is often inconsistently recorded in electronic health record (EHR) data. Objective: To enhance capture of Hispanic ethnicity in the Veterans Affairs EHR and compare relative disparities in health and health care. Methods: We first developed an algorithm based on surname and country of birth. We then determined sensitivity and specificity using self-reported ethnicity from the 2012 Veterans Aging Cohort Study survey as the reference standard and compared this to the research triangle institute race variable from the Medicare administrative data. Finally, we compared demographic characteristics and age-adjusted and sex-adjusted prevalence of conditions in Hispanic patients among different identification methods in the Veterans Affairs EHR 2018-2019. Results: Our algorithm yielded higher sensitivity than either EHR-recorded ethnicity or the research triangle institute race variable. In 2018-2019, Hispanic patients identified by the algorithm were more likely to be older, had a race other than White, and foreign born. The prevalence of conditions was similar between EHR and algorithm ethnicity. Hispanic patients had higher prevalence of diabetes, gastric cancer, chronic liver disease, hepatocellular carcinoma, and human immunodeficiency virus than non-Hispanic White patients. Our approach evidenced significant differences in burden of disease among Hispanic subgroups by nativity status and country of birth. Conclusions: We developed and validated an algorithm to supplement Hispanic ethnicity information using clinical data in the largest integrated US health care system. Our approach enabled clearer understanding of demographic characteristics and burden of disease in the Hispanic Veteran population.