Prevalence of stigmatizing language in outpatient primary care medical records of middle and older aged adults: a cross-sectional study

Abstract Objective To examine prevalence of stigmatizing language (SL) in outpatient primary care clinical notes of middle and older age adults and examine its associated sociodemographic and clinical factors. Materials We conducted a retrospective analysis of free-text notes from electronic health records (EHR) in outpatient primary care visits of adults aged 50–80 years at a large urban health system between January 2018 and December 2023. A natural language processing pipeline incorporating validated lexicons was used to extract SL in clinical notes. Associations between patient characteristics and SL were analyzed using generalized estimating equations. Results A total of 27,652 patients with 45,262 notes were included. SL was documented in 45.3% (20,516/45,262) of notes and appeared in 56.0% (15,497/27,652) of patients. Multivariate analyses revealed significantly higher odds of SL for every 1 year increase in age (OR 1.01; 95% CI 1.01–1.02), male sex (vs. female: OR 1.18; 95% CI 1.13–1.23), Asian race (vs. White: OR 1.10; 95% CI 1.00-1.19, Black race (vs. White: OR 1.29; 95% CI 1.05–1.58), residence in socioeconomically disadvantaged neighborhoods (OR 1.21; 95% CI 1.10–1.33 for highest area deprivation index quartile), and insured by Medicare (OR 1.47; 95% CI,1.35–1.61) or Medicaid (OR 1.18; 95% CI 1.07–1.30). Conclusion SL was present in notes of over half of the patients, with a disproportionate prevalence among historically marginalized groups, suggesting that EHR may reflect or perpetuate broader societal prejudices. There is a need for targeted interventions to mitigate biased language in clinical documentation, as such language may negatively impact healthcare delivery and outcomes.