JMIR Publications

ABSTRACT

Introduction: Large Language Models (LLMs) are increasingly applied in medical contexts, offering benefits for clinical decision-making, education, and patient communication. However, bias in LLM outputs may exacerbate healthcare disparities and compromise trust. This systematic review will examine how bias is identified, measured, and mitigated in healthcare use cases of medical LLMs. Methods and Analysis: A systematic search will be conducted in EMBASE, MEDLINE, PsycINFO, PubMed, ACL Anthology, ACM Digital Library, ArXiv, MedRxiv, and BioRxiv. Studies will be included if they investigate bias in LLM applications within healthcare, report experimental findings, and are published in English from 2017 onwards. Grey literature with adequate methodological detail will also be considered. Findings will be synthesised using a narrative approach due to anticipated methodological heterogeneity. Ethics and Dissemination: As a secondary analysis of published literature, ethical approval is not required. Results will be disseminated through peer-reviewed publications, academic conferences, and open-access repositories to inform responsible LLM deployment in healthcare. Registration Details: This protocol has been registered in PROSPERO (ID: 638943) https://www.crd.york.ac.uk/PROSPERO/view/CRD420250638943 and OSF.