Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 47 of the 49 references it cites were found in a registry or an index; 13 of the 20 sources its claims cite were checked for support.

Evaluating and addressing demographic disparities in medical large language models: a systematic review

Open access

Published version of a preprint

Abstract

Abstract Background Large language models are increasingly evaluated for use in healthcare. However, concerns about their impact on disparities persist. This study reviews current research on demographic biases in large language models to identify prevalent bias types, assess measurement methods, and evaluate mitigation strategies. Methods We conducted a systematic review, searching publications from January 2018 to July 2024 across five databases. We included peer-reviewed studies evaluating demographic biases in large language models, focusing on gender, race, ethnicity, age, and other factors. Study quality was assessed using the Joanna Briggs Institute Critical Appraisal Tools. Results Our review included 24 studies. Of these, 22 (91.7%) identified biases. Gender bias was the most prevalent, reported in 15 of 16 studies (93.7%). Racial or ethnic biases were observed in 10 of 11 studies (90.9%). Only two studies found minimal or no bias in certain contexts. Mitigation strategies mainly included prompt engineering, with varying effectiveness. However, these findings are tempered by a potential publication bias, as studies with negative results are less frequently published. Conclusion Biases are observed in large language models across various medical domains. While bias detection is improving, effective mitigation strategies are still developing. As LLMs increasingly influence critical decisions, addressing these biases and their resultant disparities is essential for ensuring fair artificial intelligence systems. Future research should focus on a wider range of demographic factors, intersectional analyses, and non-Western cultural contexts. Graphic Abstract

Overview

CoChat Summary

TL;DR

This systematic review of 24 studies found that demographic biases are pervasive in medical large language models, with 22 studies (91.7%) identifying bias. Gender and racial/ethnic biases were most common, while mitigation strategies such as prompt engineering and debiasing algorithms showed promise but remain inconsistently effective.

Key findings

  • Among 24 included studies, 22 (91.7%) identified biases. Gender bias was reported in 15 of 16 studies (93.7%), and racial or ethnic bias in 10 of 11 studies (90.9%). Only two studies found minimal or no bias in certain contexts.
  • Other biases included age, disability, socioeconomic status, sexual orientation, language, and cultural factors. Cultural, age, and intersectional disparities were apparent in all evaluated studies (100% of 3, 2, and 3 studies, respectively); socioeconomic and language biases were noted in 50% of 2 studies each.
  • Mitigation strategies were implemented in 7 studies (29%): 4 used prompt engineering and 3 used debiasing algorithms. Six of the seven reported reduced disparities; Bhardwaj et al. reported a 63.9% reduction in gender bias metrics for BERT models.
  • Examples included GPT-3.5-turbo predicting lower death rates for White patients (56.54%) compared with up to 62.25% for Black patients, and GPT-4 showing lower rates of advanced imaging recommendations for patients from underrepresented racial groups.

Methods

  • Systematic review following PRISMA guidelines, registered with PROSPERO (CRD42024578467). Searched PubMed, Embase, Web of Science, APA PsycInfo, and Scopus for publications from January 1, 2018, to July 31, 2024.
  • Included peer-reviewed studies evaluating demographic biases in LLMs applied to medical or healthcare tasks, focusing on gender, race, ethnicity, age, and other factors. Excluded non-LLM models, performance-only studies, and non-peer-reviewed materials. Two reviewers independently screened and assessed studies.
  • From 863 identified articles, 257 duplicates were removed, 539 excluded through title/abstract screening, 67 underwent full-text review, and 24 met inclusion criteria. Quality was assessed using JBI Critical Appraisal Checklists (Diagnostic Test Accuracy for 3 studies; Analytical Cross-Sectional for 21 studies). Due to heterogeneity, a narrative synthesis was conducted.

Limitations

  • The paper states potential publication bias, as studies with negative results are less frequently published. It also notes limited quantitative data on mitigation effectiveness, lack of standardized metrics for measuring bias reduction, and that most studies focused on cisgender men and women, with almost no work on non-binary gender identities.
  • The paper states that few studies address sexual orientation, non-binary gender identities, and intersectional identities; geographical concentration in Western countries limits understanding of biases in diverse cultural contexts; and current literature mainly focuses on GPT models.
  • Quality assessment found that 8 of 24 studies (33.3%) met all applicable criteria. Weaknesses included identification and handling of confounding factors (7 studies, 29.2%, failing or unclear) and appropriateness of statistical analysis (3 studies, 12.5%, not meeting). The text also makes evident that heterogeneity prevented meta-analysis, and it remains unclear whether legitimate sex-based variations were captured.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1186/s12939-025-02419-0

Cites

47 of 49 references verified2 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 PRISMA flowchart
Fig. 1 PRISMA flowchartCC BY · PubMed Central
Fig. 2 A numeric overall analysis of the detected bias and mitigation strategies. Gender bias manifestations in LLMs
Fig. 2 A numeric overall analysis of the detected bias and mitigation strategies. Gender bias manifestations in LLMsCC BY · PubMed Central
Fig. 3 Racial and ethnic biases manifestations in LLMs
Fig. 3 Racial and ethnic biases manifestations in LLMsCC BY · PubMed Central

Verification

Claims the paper makes

11 Valid · 0 Not valid · 4 Not checked · 37 Own results and interpretations

18 more claims

Create a free account to see all 52 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 19
  • It may state two separate claims as one. 13

Create a free account to see these claims.

Citations

Cites

Show all 42 references

Research works printed but not linked: 2

  1. Lefebvre C. Chapter 4: Searching for and selecting studies [Internet]. [cited 2024 Aug 10]. Available from: https://training.cochrane.org/handbook/current/chapter-04
  2. Yeh KC, Chi JA, Lian DC, Hsieh SK, Evaluating Interfaced LLM, Bias. In 2023 [cited 2024 Jun 20]. Available from: https://www.semanticscholar.org/paper/Evaluating-Interfaced-LLM-Bias-Yeh-Chi/be4adc35746c179eb4e660894f7af0ced88b6bdb

Cited by

See all 110

Metadata

Authors
Omar, M., Sorin, V., Agbareia, R., Apakama, D. U., Soroush, A., Sakhuja, A., Freeman, R., Horowitz, C. R., Richardson, L. D., Nadkarni, G. N., & Klang, E.
Affiliations
  • Icahn School of Medicine at Mount Sinai (US)
  • Mayo Clinic (US)
  • Tel Aviv Sourasky Medical Center (IL)
Corresponding author
Omar, Mahmud
Venue
International Journal for Equity in Health
Published
2025-02-26
Volume
24
Issue
1
Pages
57-57
ISSN
1475-9276 (ISSN-L)
OpenAlex
W4407965332
Cited by
110
References
42
Citations by year
2026: 72 · 2025: 34
FWCI
16.40
Type
journal article
Language
en
Open access
Licence
Locations
  • International Journal for Equity in Health (publishedVersion, cc-by)
  • medRxiv (acceptedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (acceptedVersion)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Social Sciences › Gender Studies › Diversity and Career in Medicine
Keywords
large language models, demographic bias, health disparities, gender bias, racial and ethnic bias, bias mitigation, prompt engineering, algorithmic fairness, artificial intelligence in healthcare, bias detection
MeSH terms
Large Language Models, Demography, Ethnicity, Female, Humans, Healthcare Disparities, Sexism

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Evaluating and addressing demographic disparities in medical large language models: a systematic review. https://app.cochat.ai/papers/10.1186/s12939-025-02419-0
The paper
Omar, M., Sorin, V., Agbareia, R., Apakama, D. U., Soroush, A., Sakhuja, A., Freeman, R., Horowitz, C. R., Richardson, L. D., Nadkarni, G. N., & Klang, E. (2025). Evaluating and addressing demographic disparities in medical large language models: a systematic review. International Journal for Equity in Health. https://doi.org/10.1186/s12939-025-02419-0

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.