Background: Use of stigmatizing language in clinical note documentation is a recognized problem, but such notes may be viewed disparately and irregularly. Whether such language is also used in highly visible behavioral alerts (BAs), which are presented forcefully through push notifications and viewed repeatedly whenever a chart is opened, is unknown. Further, whether BAs with stigmatizing language are used disproportionately in already vulnerable populations is unclear. We studied the patterns of use of BAs with stigmatizing language.

Methods: UCSF is a 785-bed hospital with approximately 33,000 inpatient admissions each year. We focused our study on patients admitted to or discharged from the UCSF Division of Hospital Medicine (Internal Medicine) service between January 2016 and June 2021. We assessed each chart for BAs in which stigmatizing language might have been used. We defined potentially stigmatizing language as the use of specific words or word fragments, including “abusi”, “abuse”, “junki”, “aggress”, “pain-seeker”, “user”, “frequent flyer”, or “narcotic”. We assessed whether BAs with stigmatizing language were used with varying frequency in patients based on age, gender, race/ethnicity, marital status, primary diagnosis of psychiatric illness, or type of insurance.

Results: Of the 25,675 total patients with 44,243 patient encounters during this time period, 261 patients had BAs added with potentially stigmatizing language. Patients with any psychiatric diagnosis were more likely to have BAs with stigmatizing language (odds ratio [OR] 6.55; 95% CI 4.26-10.06). Patients with Medicare (OR 2.55; 95% CI 1.52-4.26) or Medi-Cal insurance (OR 3.85; 95% CI 2.42-6.13) were more likely to have BAs with stigmatizing language compared to patients with private insurance. There was no statistically significant relationship between self-identified race and stigmatizing language in BAs; self-identified Hispanic ethnicity was protective for stigmatizing language (OR 0.45; 95% CI 0.27-0.73), as was female gender (OR 0.48; 95% CI 0.37-0.63), being married (OR 0.40; 95% CI 0.27-0.57), and being older (OR 0.89 per ten years; 95% CI 0.82-0.97).

Conclusions: We found that BAs with stigmatizing language were more likely to be used in non-Hispanic, unmarried, younger patients on public insurance; moreover, patients with a psychiatric diagnosis were far more likely to have BAs with stigmatizing language. Our findings are consistent with previous literature on the stigmatizing effect and biased application of BAs; given their prominence and semi-permanence in the patient chart, our research suggests that BAs may further discriminate against already vulnerable groups. BAs should be written with these effects in mind, and their authors prompted to evaluate and supported to remove any stigmatizing language.