LLMs Reproduce Sociodemographic Bias in Nursing Care
Researchers from Wuhan University and collaborators (published in J Med Internet Res, 2025) used GPT-4 to generate 9,600 nursing care plans for 96 simulated patient profiles varying by sex, age, income, education, and residence. Multivariable analyses and expert review of 500 plans found systematic sociodemographic disparities—e.g., reduced Environmental Adjustment for low-income profiles and differing completeness and safety by residence—raising equity and deployment concerns.
Key Points
- 1Generated 9,600 GPT-4 nursing care plans across 96 sociodemographic patient profiles, revealing eight thematic categories
- 2Found systematic disparities: low-income profiles less Environmental Adjustment; older profiles more Pain Management and Family Support
- 3Requires human oversight because expert review showed variable completeness, applicability, and safety across sociodemographic groups
Scoring Rationale
Strong empirical evidence and peer-reviewed publication; limited to nursing context and single LLM (GPT-4) evaluation.
Sources
Public references used for this report.
Practice with real FinTech & Trading data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all FinTech & Trading problems
