Disability and algorithmic fairness in healthcare: a narrative review

Background and Objective: Algorithmic fairness is a growing field with promising improvements to healthcare system, equity and accessibility, yet there is limited literature on how AI-driven healthcare applications affect people with disabilities (PWD). This narrative review synthesizes current literature regarding the problems and potential solutions of algorithmic (un)fairness in healthcare with a focus on PWD. Methods: PubMed Central, Springer Link, Google Scholar, and Science Direct were queried using key terms related to AI, fairness, healthcare and disability; relevant English-language papers from January 2019 to December 2024 were identified, and issues as well as improvements in algorithmic fairness for PWD were collected. Key Content and Findings: Out of 66 papers, 26% discussed disability to a greater extent, whereas 74% examined broader issues of fairness in AI and healthcare. Five key areas of concern with possible solutions were identified: socio-ethical factors—(I) fairness definitions and measures, (II) guidelines, law, and rights; and data-driven algorithmic factors—(III) design and use, (IV) training data, (V) algorithm structure. Conclusions: PWD are excluded throughout the design and implementation of medical algorithms and require better representation in training data, while also facing wider societal and legal inequalities, which calls for action.

Yvonne Vogt