To identify patterns of stigmatizing language use, racial bias, and impacted populations in NYC Health + Hospitals to inform mitigation strategies and delivery of equitable care.
Approach:
Lexicon Search Strategy: Implemented a flexible, rapid lexicon search strategy to identify stigmatizing terms, excluding disease-specific terms, and conducted manual chart reviews to refine results.
Data Analysis: Measured prevalence of stigmatizing language in clinical notes, stratified by provider type and care setting, and modeled odds of stigmatizing terms using multivariable logistic GEE.
Key Findings:
Stigmatizing terms were found in 93,107 medical notes (0.6%) representing 26,052 unique patients (3.1%).
Prevalence was highest among notes authored by Social Workers, Counselors, or Psychologists (1.7%) and in inpatient (1.0%) and emergency department (0.9%) notes.
Patients with experience of homelessness (15.9%), SUD/AUD (15.3%), Medicare insurance (7.9%), and ≥ 2 chronic comorbidities (6.3%) had the highest prevalence.
Significant racial variation was observed, with Non-Hispanic White (4.7%) and Black (4.1%) patients having higher prevalence compared to Hispanic/Latinx patients (2.1%).
Interpretation:
Stigmatizing language is concentrated in acute care settings and among patients with behavioral health needs.
Limitations:
The study likely underestimates true prevalence of stigmatizing language due to the use of a rapid lexicon search with uniformly negative terms.
The methodology was less comprehensive than advanced natural language processing or AI techniques.
Conclusion:
Identifying hotspots of stigmatizing language allows for tailored interventions to improve equitable care.
In a target-trial emulation of more than 600,000 veterans, GLP-1 RA initiators saw fewer new substance use disorders—and patients with existing SUDs had fewer overdoses, hospitalizations, and deaths.
