Implicit bias in the patient descriptor "homeless" and its association with emergency department opioid administration and disposition

Background: Biased language in provider documentation of marginalized patient populations has been shown to negatively influence patient management. There has been debate over the use of "homeless" as a descriptor of people experiencing homelessness (PEH), as it is a potentially biased term with negative connotations. This study explores the relationship between the use of the word "homeless" in Emergency Department (ED) provider documentation and admission rates, as well as intravenous (IV) vs. oral (PO) opioid administration rates.

Methods: This single-center retrospective study analyzed electronic medical record (EMR) data from 2 academic EDs in Southern California. ED physician encounter notes from the calendar year 2021 were included if the patient had a history of unstable housing documented elsewhere in the medical record in the past year. Patients were categorized as HIN+ if the term "homeless" was featured in the note. Multivariate logistic regression analysis was performed to determine if HIN+ was associated with admission vs discharge for 7 common ED diagnoses, as well as for the administration of IV vs PO opioids.

Results: The cohort consisted of 2751 PEH with 5049 unique ED encounter notes. Of these, 920 patients with 1333 unique notes were HIN+. Regardless of diagnosis, charting "homeless" in the note (HIN+) was a strong predictor of discharge rather than admission (OR: 0.70, CI: 0.60-0.82). This was especially true for those with a diagnosis of heart failure exacerbation (OR: 0.45, CI: 0.23-0.87) and alcohol withdrawal (OR: 0.47, CI: 0.24-0.90). HIN+ patients were also less likely to receive IV opioids (OR: 0.41, CI: 0.24-0.69).

Conclusions: ED physician use of the term "homeless" in medical documentation was associated with a lower likelihood of being admitted to the hospital for certain conditions and more likely to receive PO rather than IV opioids compared with those patients not labeled as "homeless". This study reinforces prior studies in how biased language may alter medical decision making.