Language Preference and Antibiotic Use Among Hospitalized Older Adults With Dementia

Persons living with dementia are frequently prescribed antibiotics, often for durations that exceed guideline recommendations [1, 2]. Language discordant care is known to contribute to potentially inappropriate interventions for patients with dementia [3] but its impact on antibiotic prescribing is unknown. We examined whether non-English language preference was associated with increased antibiotic use among hospitalized older adults with dementia. We conducted a retrospective cohort study of adults 65 years and older with dementia who were admitted to inpatient internal medicine units at two academic hospitals in Toronto, Canada. A validated abstraction tool was used to identify individuals with dementia among the 1233 screened charts between June 1, 2022, and July 1, 2023 [4]. Demographic information, comorbidities, infection source, antibiotic duration, and 30-day adverse events were collected. The patients' preferred language for healthcare communication is a previously validated variable that was obtained from the electronic medical record [5, 6]. We examined the relationship between non-English language preference and antibiotic use by fitting a negative binomial model to estimate relative risks for antibiotic duration in days adjusted for age, sex, and nursing home residence. A sensitivity analysis was conducted for the binary outcomes of antibiotic use overall, and a subgroup analysis was performed for patients with pneumonia, urinary tract infection (UTI) and dementia severity (categorized by the presence of Behavioral and Psychologic Symptoms of Dementia or BPSD). Analysis was performed using R (version 4.5.1), with statistical significance set at p < 0.05. The University Health Network Research Ethics Board approved the study (#23-5154). Overall, 595 hospitalized older adults with dementia were included with a median age of 86 (interquartile range [IQR], 75–97) and median Charlson comorbidity index (CCI) score of 6 (IQR, 4–8), of which 53.5% were female and 35% were admitted from a nursing home (Table 1). The most common diagnoses were pneumonia (21.7%) and UTI (17.8%). Patients with English and non-English preference had similar proportions of fever (17.3% vs. 21.2%, p = 0.27), hypotension (15.1% vs. 15.9%, p = 0.87), and BPSD (21.2% vs. 26.1%, p = 0.18). Over 60% of our cohort (n = 361) received antibiotics during their hospital stay. The difference in days of antibiotics received was similar by preferred language (3.6 days vs. 3.8 days; incidence rate ratio [IRR] 1.03; 95% CI: 0.88–1.21). The proportion who received an antibiotic was also similar (60.9% vs. 60.4%; adjusted relative risk [aRR] 0.99; 95% CI: 0.88–1.12). Subgroup analysis by BPSD diagnosis showed that patients with non-English language preference had more days on antibiotics compared to English speakers with BPSD (4.2 days vs. 3.3 days, IRR: 1.58; 1.06–2.36), but this was not statistically significant after adjustment (IRR: 1.46; 0.99–2.36), Figure 1. There were no differences within subgroups of UTI or pneumonia or in antibiotic related adverse events (Table S1). Though antibiotics were widely used in our cohort of hospitalized older adults with dementia, there were no significant differences in antibiotic use by language preference. In a subgroup analysis, patients with BPSD who spoke a non-English language may have had more days on antibiotics. A previous study of non-hospitalized adults showed that disparities existed in unnecessary antibiotic use by language, race, and ethnicity [7]. There are several potential reasons why our study in patients with dementia did not demonstrate an association between language discordance and antibiotic use. First, taking an admission history for patients with dementia is often done through a caregiver who may be more likely to speak English. Second, rates of antibiotic use were so high that differences between language groups may have been harder to identify. Third, decisions to prescribe antibiotics in inpatient settings may be more likely driven by investigations, fever, or individual prescribing practices rather than communication with patients [8]. Indeed, a large study from Canadian nursing homes showed that resident communication challenges were not associated with increased antibiotic use [9]. Limitations are the retrospective cohort design, which restricts the ability to establish causality and includes the potential for residual confounding. Variables such as frailty, race, ethnicity, and illness severity were not captured, although fever and hypotension were included and were similar between groups. Furthermore, patients classified as having English as a second language may have had sufficient proficiency to respond to certain clinical questions and patient-level data on interpreter use relied on documentation, which is often incomplete [10]. Patients with dementia who were admitted to hospital and spoke languages other than English were not at greater risk of receiving antibiotics. The general overuse of antibiotics in this population underscores the need for antimicrobial stewardship in hospitalized patients with dementia, especially those with BPSD or advanced illness. All authors listed have contributed significantly to study conception and design and/or data acquisition and/or analysis and interpretation of the data. All authors were involved in the manuscript's preparation, revision, and final approval of version to be published. Study concept and design: Hanbyoul Park, Ahmed Bagit, Shail Rawal, Christopher Kandel, Christina Reppas-Rindlisbacher. Acquisition of subjects and/or data: Hanbyoul Park, Ahmed Bagit, Bernadette McCann. Analysis and interpretation of data: Saeha Shin, Hanbyoul Park, Bernadette McCann. Ahmed Bagit, Shail Rawal, Christopher Kandel, Christina Reppas-Rindlisbacher. Preparation of manuscript: Hanbyoul Park, Saeha Shin, Shail Rawal, Christopher Kandel, Christina Reppas-Rindlisbacher. The authors have nothing to report. The authors declare no conflicts of interest. TABLE S1: Inpatient investigations and 30-day adverse events. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.