Introduction
Methods
Results
Discussion
Limitations
Conclusion

INTRODUCTION

Race and ethnicity are social constructs that are associated with meaningful health inequities. To identify and address health disparities, it is necessary to have accurate race and ethnicity data. In 2009, the Institute of Medicine released Race, Ethnicity, and Language Data: Standardization for Healthcare Quality Improvement, with recommendations and best practices for race and ethnicity data collection.¹ However, inaccuracies in race and ethnicity data persist in hospital and administrative databases of adult patients.^2-4

The causes of inaccuracies in race and ethnicity data are multifactorial. Although self-identified race and ethnicity are considered the gold standard by the Institute of Medicine, patients are not always given the ability to provide self-identified demographics.^1,5-7 This may be more likely in the emergency department (ED), where hospital registration staff must find time to collect patient information and sign consents to care without interrupting urgent and fragmented clinical care. Race and ethnicity that are determined based on staff observation may be particularly inaccurate for children, who may not have the same race and/or ethnicity as their caregiver.^8,9

Inaccuracies can be further compounded by limitations in available race and ethnicity categories.^1,9,10 Healthcare systems often restrict race and ethnicity data collection to the minimum standard categories required for federal reporting and rarely give the option to select “other” or to select multiple races or “multiracial.”¹¹ Additionally, although the Institute of Medicine supports the option of presenting race and ethnicity within a single question, healthcare systems often separate these into distinct categories as used in federal reporting.¹ This format can lead to misidentification of Hispanic individuals who do not otherwise identify with the options listed in a distinct race category.^1,12 As part of a quality initiative to improve and standardize accurate demographic data collection, we sought to assess the accuracy of child race and ethnicity data in an academic pediatric emergency department (PED), and to identify risk factors for misidentification.

METHODS

We conducted a cross-sectional analysis comparing child race and ethnicity reported by the parent to that documented in the electronic health record (EHR). This study took place in a single, academic PED with an affiliated onsite children’s hospital. The PED has an annual volume of approximately 26,000 patients. In the EHR (Epic Systems Corporation; Verona, WI), race and ethnicity data are documented separately using the minimum US Office of Management and Budget categories.⁷ Both fields are required. In each field multiple options can be selected, including an option for “other.” If not already documented from a prior visit within the hospital system, PED registration staff obtain patient race and ethnicity from patient or parent self-report or through staff observation.

Between February–May 2021, three trained research assistants (RA) approached a convenience sample of parents or caregivers (“parents”). Parents of critically ill children were excluded. The RAs were in the PED during afternoons and evenings. Sample size was determined by RA availability during the study period. Parents were approached at any time during the PED visit. The RAs explained the purpose of the study and asked parents to complete a brief, tablet-based questionnaire while the RA remained in the room. Parents who used a language other than English completed the questionnaire verbally with video interpretation. The parent was asked to identify the child’s race and ethnicity from a single question (“What is your child’s race and ethnicity? You can choose as many answers as you want to: American Indian/Alaska Native; Asian; Black/African American; Hispanic/Latino; Native Hawaiian/other Pacific Islander; White; other (free-text optional); I don’t want to say; I don’t know”).¹ We chose to use a single question with race and ethnicity presented together, an option suggested by the Institute of Medicine, so that parents were not compelled to make selections within a category with which they or their child did not identify.^1,8 Parents were also asked to identify their own race and ethnicity in a similar, single-question format.¹ Child race and ethnicity were abstracted from the EHR. This study was approved by the University of Florida Quality Improvement Project Registry and determined not to require institutional review board review.

Our primary outcome was EHR misidentification of child race and ethnicity. Child race and ethnicity was considered “misidentified” if the EHR record did not match the parent report. Misidentifications in the EHR included situations in which the parent selected multiple options but not all of those were selected in the EHR, or vice versa. This also included situations in which the parent selected “other” but “other” was not reported in the EHR, or vice versa. Finally, race and ethnicity data are often missing from EHRs, leading to the exclusion of those individuals from equity-focused analyses and research.³ Thus, if EHR race and/or ethnicity was missing but the parent provided a response in the questionnaire, this was considered a misidentification. To assess this approach, we conducted sensitivity analyses in which we excluded those patients with missing race and/or ethnicity. We performed statistical analyses in R (R Core Team, 2021, R Foundation for Statistical Computing, Vienna, Austria). We performed chi-square and bivariate regression analyses to evaluate misidentification by child race and/or ethnicity and age, respectively. We assessed statistical significance at the P = 0.05 level.

RESULTS

Of 219 approached parents, 206 completed a questionnaire (94%). Most parents identified their child as non-Hispanic White (51%) or non-Hispanic Black (26%) (Table 1). Thirty-one parents (15%) identified their child as Hispanic, half of whom did not identify a separate race for their child. Seventeen parents (8.3%) identified their child as multiracial.

Fifty-six children (27%) had misidentified race and/or ethnicity in the EHR. Most misidentifications (89%) were misidentification of race. This includes all 17 multiracial children, most of whom were inaccurately reported as having a single race. Of the 16 Hispanic children who did not have separate race identified by the parent, 70% were misidentified as “other race,” and 30% were misidentified as “White” in the EHR. Six children had misidentified ethnicity, all of whom were identified as Hispanic by the parent and in the EHR as “not Hispanic or Latino.” A full list of misidentified EHR race and ethnicity data is found in Table 2.

Misidentifications were most common among children who were multiracial (100%, 95% confidence interval [CI] 77–100%] vs 15%, 95% CI 10–21%) of children identified as a single race, P < 0.001), children who were Hispanic (87%, 95% CI 69–96% vs 17%, 95% CI 12–23%) of non-Hispanic children, P < 0.001), and children whose race and/or ethnicity differed from that of their parent (79%, 95% CI 60–91% vs 18%, 95% CI 13–25%) of children with the same race and ethnicity as their parent, P < 0.001). There was no association between child age and misidentification (odds ratio 1.04, 95% CI 0.99–1.11). No parents selected “I don’t know” or “I don’t want to say” on the questionnaire, and nine children had missing race and/or ethnicity data in the EHR. The results were not changed when these subjects were excluded in sensitivity analyses.

DISCUSSION

Misidentification of child race and ethnicity was common in our PED, findings that remain similar to pediatric administrative-database analyses from the early 2000s.⁸ Our findings demonstrate a clear need to develop strategies to enhance precise data collection within our EHR and to facilitate self-report of race and ethnicity. Importantly, such efforts to improve precision must be partnered with analyses that consider complex demographic subgroups.⁶

Strategies applied in other healthcare systems have included the following: 1) staff training on self-report of race and ethnicity and education to increase patient awareness; 2) EHR systems that allow the selection of multiple races; 3) EHR alerts when race and/or ethnicity are missing; 4) use of granular race and ethnicity subcategories; and 5) a single-item question for race and ethnicity.^1,4,9,10 Our findings highlight the importance of this multifaceted approach. All multiracial children in our sample were misidentified in the EHR. Our EHR allows for the selection of multiple races, yet in most of these cases multiracial children were misidentified in the EHR as having a single race. Additionally, by intentionally using a single item for race and ethnicity, we found that half of Hispanic parents did not select an additional race option for their child. Most of these children were categorized in the EHR as “other race,” an all-encompassing category that loses precision and is often excluded entirely from data analysis.⁶

We assessed race and ethnicity as a single construct for two reasons. First, as we found in our sample, individuals who identify as Hispanic may not additionally identify with a distinct race category.^12,13 Second, this approach was pragmatic.¹³ Race and ethnicity are often presented as a single construct in health equity research, which requires researchers to collapse the two variable data that are found in administrative and hospital databases.¹⁴ Thus, our approach mirrors the practice of many health equity researchers. By offering choices that reflect the way data will be reported, we allow patients and parents greater self-determination in how precisely they will be identified.

LIMITATIONS

This study is subject to limitations. Questionnaires were completed by parents, which may not reflect child self-identification. Race and ethnicity categories used by the US Census Bureau are themselves limited and do not fully capture individual realities. We were unable to determine how EHR data was collected and could not determine which misidentifications occurred at the level of data entry (ie, if race and ethnicity were determined by staff observation). Demographics may have been collected during prior visits within the hospital system, so our findings are not sufficient to identify misidentifications that are uniquely due to the PED registration process. Finally, we approached a convenience sample, and responses may have been influenced by the timing and methods of questionnaire administration. As part of a quality improvement initiative, our findings are not intended to be generalizable.

CONCLUSION

Despite representing arbitrary social constructs, accurate race and ethnicity data are essential to identifying and addressing health inequities. Although we found that rigidity within race and ethnicity items in the EHR was an important factor in many misidentifications (ie, the requirement for both a race and ethnicity response), we also found that some features of the EHR were not used (ie, the ability to select multiple responses within a category). This work provides the basis for a multifaceted quality improvement effort at our institution. The quality of child race and ethnicity data in the emergency setting warrants further consideration across health equity efforts.

Footnotes

Full text available through open access at http://escholarship.org/uc/uciem_westjem

Address for Correspondence: Colleen Gutman, MD, University of Florida, Department of Emergency Medicine, 1329 SW 16^th St, Gainesville, FL 32610. Email: ckays21@ufl.edu 5 / 2023; 24:547 – 551

Submission history: Revision received October 10, 2022; Submitted January 10, 2023; Accepted February 17, 2023

Conflicts of Interest: By the WestJEM article submission agreement, all authors are required to disclose all affiliations, funding sources and financial or management relationships that could be perceived as potential sources of bias. This study was supported in part by NCATS KL2TR001429 (to CKG). The NIH and University of Florida Clinical and Translational Science Institute had no role in the design and conduct of the study. There are no conflicts of interest to declare.

REFERENCES

1. Race, Ethnicity, and Language Data: Standardization for Health Care Quality Improvement. Available at: http://www.ahrq.gov/research/findings/final-reports/iomracereport/index.html. Accessed July 23, 2020.

2. Lee SJC, Grobe JE, Tiro JA. Assessing race and ethnicity data quality across cancer registries and EMRs in two hospitals. J Am Med Inform Assoc. 2016;23(3):627-34.

3. Lee WC, Veeranki SP, Serag H, et al. Improving the collection of race, ethnicity, and language data to reduce healthcare disparities: a case study from an academic medical center. Perspect Health Inf Manag. 2016;13:1g.

4. Magaña López M, Bevans M, Wehrlen L, et al. Discrepancies in race and ethnicity documentation: a potential barrier in identifying racial and ethnic disparities. J Racial Ethn Health Disparities. 2017;4(5):812-8.

5. Hasnain-Wynia R, Van Dyke K, Youdelman M, et al. Barriers to collecting patient race, ethnicity, and primary language data in physician practices: an exploratory study. J Natl Med Assoc. 2010;102(9):769-75.

6. Jorgensen S, Thorlby R, Weinick RM, et al. Responses of Massachusetts hospitals to a state mandate to collect race, ethnicity and language data from patients: a qualitative study. BMC Health Serv Res. 2010;10:352.

7. Klinger EV, Carlini SV, Gonzalez I, et al. Accuracy of race, ethnicity, and language preference in an electronic health record. J Gen Intern Med. 2015;30(6):719-23.

8. Weikel BW, Klawetter S, Bourque SL, et al. Defining an infant’s race and ethnicity: a systematic review. Pediatrics. 2022:e2022058756.

9. Chaiyachati BH, Peña MM, Montoya-Williams D. The complicated inadequacy of race and ethnicity data. JAMA Pediatr. 2022;176(7):631-2.

10. Pellegrin KL, Miyamura JB, Ma C, et al. Improving accuracy and relevance of race/ethnicity data: results of a statewide collaboration in Hawaii. J Healthc Qual. 2016;38(5):314-21.

11. Cowden JD, Flores G, Chow T, et al. Variability in collection and use of race/ethnicity and language data in 93 pediatric hospitals. J Racial Ethn Health Disparities. 2020;7(5):928-36.

12. Flores G. language barriers and hospitalized children: Are we overlooking the most important risk factor for adverse events?. JAMA Pediatr. 2020;174(12):e203238.

13. Mason LR, Nam Y, Kim Y. Validity of infant race/ethnicity from birth certificates in the context of U.S. demographic change. Health Serv Res. 2014;49(1):249-67.

14. Goyal MK, Johnson TJ, Chamberlain JM, et al. Racial and ethnic differences in antibiotic use for viral illness in emergency departments. Pediatrics. 2017;140(4).