Accuracy and Adequacy of Preferred Language Data in a Pediatric Electronic Health Record

Effective pediatric care depends on understanding and supporting the language needs of patients and their caregivers.1–3 Challenges in capturing the full range of family language preferences in the pediatric electronic health record (EHR) can lead to data inaccuracy and incompleteness, hindering organizations’ ability to provide adequate language services and undermining confidence in pediatric research using administrative language data. Although language data accuracy has been studied in adult settings,4–6 pediatrics presents unique challenges because of family communication dynamics. We sought to describe the accuracy of language preference data in our organization’s EHR and the adequacy of the pediatric EHR to reflect families’ communication needs.Caregivers of patients aged 0 to 17 years and patients aged at least 12 years reported spoken and written language preferences and self-reported English proficiency via an in-person survey from November 2021 to February 2022 in 4 pediatric settings (ie, inpatient, emergency department, urgent care, primary care). We compared language preference data from the EHR with those from our survey (prioritizing the preferred spoken language of the primary caregiver bringing child for care) to calculate concordance (number correctly assigned/total number), inter-measure reliability (Cohen’s κ statistic), and sensitivity and specificity for the EHR detecting caregiver non-English preference. We calculated teen-EHR and teen-caregiver spoken language concordance (teens being aged 12–17 years), as well as rate of English proficiency among caregivers preferring a language other than English. Language needs reflected in the survey were compared with data collection options in the EHR to determine EHR adequacy to fully reflect families’ needs. See Appendices A and B for methodology and survey details. This study was approved by our organization’s institutional review board.Caregivers for 561 out of 601 patients (93%) approached for the study completed the survey; 116 patients aged at least 12 years completed the patient version. Survey locations and participant demographics appear in Table 1.Caregiver and patient language preferences from the survey and EHR appear in Table 2. Language preference concordance was 93.9% when prioritizing the surveyed spoken language preference of the caregiver most likely to bring the child for care. Inter-measure reliability yielded a κ statistic of 0.84 (95% CI: 0.79–0.89; SE 0.03), characterized as “very good” agreement per Altman.7 The EHR had a sensitivity of 87.8% and specificity of 96.0% for detecting non-English spoken language preference. Teen-EHR and teen-caregiver concordance were 75.9% and 73.8%, respectively, with English-speaking teen/Spanish EHR (24 out of 25, 96%) and English-speaking teen/Spanish-speaking caregiver (24 out of 28, 86%) being the most common mismatches. About one-third of caregivers preferring a language other than English in the health care settings reported speaking/understanding (85 out of 271, 32%) or writing/reading (88 out of 275, 32%) English well or very well.Although our pediatric EHR showed good accuracy for one caregiver’s spoken language preference, it failed to fully reflect family language needs because of missing key data fields (ie, patient spoken language preference, patient and caregiver written language preferences, English proficiency). An oversimplified, incomplete language preference picture in the EHR can lead to confusion regarding language support needs in the pediatric setting. For example, when 1 parent is bilingual but the other is not, or when a teen prefers 1 language and the caregiver another, the full range of family needs does not appear in EHRs limited to 1 language field. One-third of caregivers in our study preferred a non-English language but reported speaking or reading English well or very well. Not recording English proficiency in the EHR can lead to interpreter services being mistakenly ordered for bilingual families who do not want them, resulting in frustration for family members, inefficiency for interpreters, and an erosion of trust in the EHR data.Inaccurate/incomplete language preference data also compromise research and quality improvement efforts focused on language-related health equity. This problem occurs at the local hospital level and with networks of hospitals contributing data to shared data sets.8 For such data sets to be trustworthy, organizations contributing language-related data must specify whose language preferences (ie, patient, caregiver 1, caregiver 2) and what type of preferences (ie, spoken, written, English proficiency) are reflected.2,9Progress toward language equity in pediatrics depends on improving the ways that we collect, store, and use language preference data. As EHR companies evolve to provide better support for pediatric organizations, understanding the current problems with accuracy and completeness of family language demographics can inform our approach to EHR, research, and quality improvement design.Thank you to the Office of Equity and Diversity and the Children’s Mercy Research Institute (especially Vince Staggs, statistician). We appreciate the graciousness of the clinical staff who partnered with us for data collection, including the Patient Access department. The support of Luis Sainz, bilingual surveyor and research assistant, and the many interpreters who partnered with us was essential for clear communication with the patients and families who spoke languages other than English. Above all, we want to thank the hundreds of patients and families who allowed us into their rooms and lives, providing us their personal stories and invaluable information.