As large language models (LLMs) enter clinical decision support, concerns persist about sociodemographic bias. We assessed whether LLM recommendations for dizziness vary by patient descriptors and clinical detail.MethodsWe conducted a cross-randomized in-silico vignette study....
✦ The floor
Discussion
Signed responses from readers of the wire.
Audiologists and clinicians using AI decision-support tools for dizziness evaluation should be aware that LLM recommendations may vary by patient sociodemographic characteristics — independent validation of AI outputs is prudent before clinical reliance.
If AI clinical decision-support tools encode sociodemographic bias, their adoption in audiology and vestibular medicine could systematically worsen health disparities for already underserved patient groups.
- 01Case-control study assessed sociodemographic bias in LLM (AI chatbot) clinical decisions about dizziness.
- 02LLMs produced differing recommendations based on patient age, sex, race, or socioeconomic status.
- 03Bias in AI tools could widen existing healthcare disparities in vestibular care.
- 04Published in Journal of Vestibular Research, 2026.
- 05Findings argue for bias auditing before deploying LLMs in clinical settings.
Large language models used in dizziness clinical decision-making exhibit sociodemographic bias in their recommendations.
studypartially supportedAI-generated clinical guidance for dizziness may differ systematically by patient race, sex, age, or socioeconomic status.
studypartially supported- PMID
- 42569795
- DOI
- 10.1177/09574271261474218.
- Journal
- Journal of Vestibular Research
- Publication type
- research_article
- Evidence level
- 3
- Population
- Simulated clinical dizziness cases with varied sociodemographic characteristics input into LLMs
- Intervention
- Large language model (LLM) clinical decision support for dizziness
- Comparator
- Sociodemographically varied case presentations (e.g., differing age, sex, race, income)
Primary outcomes
Differences in LLM clinical recommendations by sociodemographic group; Detection of bias patterns in AI-generated dizziness management guidance