The article titled “A scoping review and evidence gap analysis of clinical AI fairness” by Liu et al. (2025) offers a comprehensive examination of how fairness in artificial intelligence (AI) is approached within the clinical domain. The authors emphasize that although AI has transformative potential in healthcare, it risks exacerbating health disparities if fairness is not systematically addressed. AI fairness refers to both mitigating bias in AI systems and promoting equity in healthcare outcomes. The study identifies several disconnects between the technical advancements in fairness-aware AI algorithms and their practical clinical implementation.
The authors conducted a scoping review of 467 studies from five databases (MEDLINE, Web of Science, Embase, IEEE Xplore, ACM Library) and performed a quantitative evidence gap analysis. Their findings underscore several critical gaps. First, AI fairness research is unequally distributed across medical domains—fields like anesthesiology, oral health, and family medicine are notably underrepresented. Second, the focus has primarily been on a limited set of bias-relevant attributes (ethnicity/race, gender/sex, and age), while other important attributes (e.g., skin tone, institutional affiliations, and clinician expertise) are often neglected. Third, fairness methods are predominantly group-based (parity and performance metrics), with insufficient attention to individual or distributional fairness, which are crucial for clinical ethics and resource allocation.
Furthermore, the review critiques the limited involvement of clinicians in the development and validation of AI fairness models, warning that such exclusion may undermine contextual applicability. The study also highlights a lack of fairness benchmarks across public datasets and the need for better standardization. Emerging approaches like explainable AI and federated learning are promising but remain underutilized.
To address these challenges, the authors propose actionable strategies: encouraging cross-disciplinary collaboration, improving dataset diversity and transparency, integrating individual fairness metrics, and developing decision-aware fairness frameworks. These steps are essential to ensure that AI in healthcare does not only enhance efficiency but also promotes equity and ethical integrity.
Reference:
Liu, M., Ning, Y., Teixayavong, S., Liu, X., Mertens, M., Shang, Y., … & Liu, N. (2025). A scoping review and evidence gap analysis of clinical AI fairness. npj Digital Medicine, 8(360). https://doi.org/10.1038/s41746-025-01667-2
Podcast Link: https://notebooklm.google.com/notebook/0f9e2078-5fad-49a7-ac45-27ecc9236a86/audio