Impact of Preferred Written Language in Patients Discharged With Bronchiolitis From a Children’s Hospital

We conducted a retrospective cross-sectional study of children aged younger than 2 years who were discharged after an inpatient or observation-status hospital admission between May 1, 2021, and April 30, 2022, and had a principal diagnostic code for acute bronchiolitis (International Classification of Diseases, 10th Revision codes (J21.0, J21.1, J21.8, J21.9). Data were obtained from the Pediatric Health Information System (PHIS), an administrative database containing clinical and resource utilization data from >49 children’s hospitals, supplemented by manual chart review. Bronchiolitis was chosen as the subject for this study because it is one of the most common reasons for pediatric hospitalization in the United States¹⁰  and at our institutions, and there are standardized clinical care plans and discharge instructions. We excluded patients with gestational age younger than 35 weeks, complex chronic medical conditions (indicated by database flag for any of the following complex conditions: Premature and Neonatal, Technology Dependent, Respiratory, Complex Chronic Condition; Patient Medical Complexity Algorithm Category of Complex Chronic), or intensive care unit (ICU) stay during the study admission. We opted to exclude patients with ICU admissions because their more severe illness and potential for related complications such as feeding difficulties may have confounded results for length of stay and readmission. The study was performed in a 2-hospital pediatric health system in the southeastern United States consisting of a main freestanding children’s hospital and a separate community children’s hospital in another portion of the state. Both hospitals provide in-person Spanish interpretation, 1 uses 1 Marshallese interpreter, and both hospitals have access to a phone-based language line offering multiple language interpreters.

Data elements extracted from the PHIS database included admission and discharge dates, hospital campus, age, gender, urban status (assigned by PHIS as urban/suburban, not urban, or unknown based on patient’s home ZIP code), race and ethnicity (a combined descriptor assigned by PHIS), payer (private insurance, public insurance, or self-pay; used as a proxy for socioeconomic status),¹¹  and LOS in hours. Additional data were collected by manual chart review in the electronic health record (EHR). As part of the same health care system, both hospital sites use the same EHR, Epic (Epic Systems Corporation). Manually extracted data from the EHR included date of readmission (emergency department visit or hospital admission) if within 7 days of initial discharge date, reason for readmission, preferred written language (obtained at hospital registration), and language in which written discharge papers were prepared. If no preferred language was documented, the patient was categorized in the English group based on the institution’s usual practices. This study included patients from both hospitals; if a patient was transferred from the community to the main hospital, it was counted as 1 visit and discharge instructions from the final discharging hospital were evaluated.

Statistical analysis was performed with GraphPad Prism version 9.5.1 for Windows (San Diego, California, USA). Fisher exact test was used for categorical data and Mann-Whitney test was used for nonparametric continuous data. This study was approved by the University of Arkansas for Medical Sciences Institutional Review Board (IRB #274523).

We identified 501 encounters with a discharge diagnosis of acute bronchiolitis from our hospital system, of which 384 remained after exclusion (Table 1). The study population was compared in terms of preferred written language of English (EP; n = 356; 92.7% of study population) and LOE (n = 28, 7.3%).

The median age of the study population was 7.5 months (interquartile range 3–16 months). Twenty-one patients (5.5%) had readmission visits within 7 days of the initial hospital discharge. The median LOS was 34.3 hours (interquartile range 21.2–49.2). There were no mortalities in the study population.

The majority of patients (71%) had urban/suburban home ZIP codes, were male (58.6%), and identified as non-Hispanic White (57.8%). About two-thirds (67.9%) of patients had public insurance.

There was no significant difference in the median age (P = .93) or gender (P = .33) between the EP and LOE groups. Compared with the EP group, the LOE cohort was significantly more likely to have an urban/suburban home ZIP code (92.9% vs. 67%; P < .05), identify as Hispanic (42.9% vs 6.7%; P < .001), and have public rather than private insurance (96.4% vs 69.4%; P < .01).

The median LOS was longer in the LOE group (37.9 hours vs 34.3 hours; P < .05). Within the LOE group, although power was low, there was no significant difference in the LOS between Spanish-preferring patients and those preferring languages other than Spanish (38.5 vs 37.5 hours; P = .44). There was no statistical difference in readmission between the English-preferring and LOE groups (P = .39) (Table 2).

In the LOE cohort, documented preferred languages were primarily Spanish (13 patients; 46.4%), Marshallese (6; 21.4%), and Chinese (4; 14.3%), followed equally by Bosnian, Burmese, Esperanto, Other, and Tagalog (1; 3.6% each). All patients with English as a preferred written language were given discharge instructions in English, whereas only 46% of patients in the LOE group were given discharge instructions in their preferred written language. Of those, all Spanish-preferring patients received discharge instructions in Spanish but no patients with language preferences other than English and Spanish received instructions in their preferred written language.

This cross-sectional study in a 2-hospital system examined patients with bronchiolitis discharged over a 1-year period to evaluate the potential impact of preferred language on LOS and 7-day readmission. In alignment with existing literature showing worse health outcomes for patients preferring a LOE, our LOE cohort had a significantly longer LOS for acute bronchiolitis than the EP cohort, although there was no statistical difference in 7-day readmission rates. We identified a significant disparity in the use of preferred language in written discharge instructions for patients who preferred a language other than English or Spanish.

Potential reasons for increased LOS in patients preferring LOE may be further classified as (1) communication issues between medical staff and family, (2) difficulty accessing appropriate and timely interpreters and linguistic services, (3) inequity in social needs and access to care, and (4) provider implicit bias or discrimination.^6,12,13  Although our study was not intended to explore what led to increased LOS for our patient population, our results do support that families preferring LOE may experience differential communication and engagement with the medical team compared with EP families.^6,12,14  A multifaceted approach focused on equitable inpatient care for families who prefer LOE should be taken at the system and institutional levels to alleviate these disparities in health outcomes.⁵ 

Our health care system uses in-person Spanish interpreters who are available for written and verbal interpretation of discharge instructions. For languages other than English or Spanish, the provider must use a video interpreting service, which can be hampered by access to equipment, availability of interpreters, and comfort using the service.^15,16  For written translation of languages other than Spanish, there are no formal services available, so providers use free services such as Google Translate or other web-based applications, or only provide instructions in English. Google Translate and other software may not be accurate or consistent for patient instruction and should be proofread by an interpreter proficient in that language.¹⁷  Our EHR does not automatically translate discharge instructions; this is manually performed by our in-house Spanish interpreters.^15,16  Reliance on real-time, individual written translation is time-consuming, costly, and diverts interpreters’ time away from in-person interpretation.¹³  For diagnoses with frequently standard discharge instructions such as bronchiolitis, availability of Spanish and other language versions of instructions in the EHR via dot phrase or other shortcut could decrease burden on interpreters and decrease time to discharge. Research has shown that interpreter services are essential for clear communication between caregivers and their physicians, with in-person services optimal.¹⁸  The role of residents was not evaluated in this study though both their impact and specific training in the use of interpreters can have an effect on equity of communication and care of patients preferring LOE.¹⁹ 

A strength of this study is the use of the manual collection of granular data on language preference and language of written discharge instructions. The study results emphasize that patients who prefer LOE are not a homogenous group and disparities exist among the many subclassifications of languages preferred.

Limitations of the study include the use of a large administrative database for subject inclusion and reliance on reported race, ethnicity, and preferred language. Administrative and hospital discharge data on race and ethnicity has been shown to be prone to inaccuracies, especially in races other than non-Hispanic White and Black.^20,21  The study’s small sample size increases the risk of missing true differences between the 2 groups. Generalizability may be limited given the inclusion of 2 hospitals in a single state; however, these findings add to the existing literature that disparity in health care for patients who prefer LOE is widespread.

We could not assess whether interpretation in a patient’s preferred language was performed verbally at the time of discharge because these data are not captured in the medical record. However, there is literature that parents can have difficulty recalling discharge instructions even in their primary language, and providing discharge instructions that are accessible after discharge may improve recall and understanding of instructions regardless of preferred language. Implementation of methods shown to increase understanding of discharge instructions in LOE patients should be included in improvement work on patient safety.

These data helped identify gaps in equal care in the hospital discharge process for patients who may prefer languages other than English, especially in less common languages where in-person interpretation and written language translation is not readily available. Our results identify opportunities to improve language disparities in discharge instructions by leveraging clinical informatics and the EHR. For example, artificial intelligence is an area of rapidly growing interest and may be applicable to optimizing communication with patients and families, such as automatic translation into the written language indicated in the patient’s preferred language documented in the EHR but need to be tested to ensure accuracy and equity in all languages. There are significant opportunities to decrease disparities related to preferred language in health care settings.