Fairness in multimodal machine learning applications in clinical decision support: a systematic review
Only one in ten clinical AI studies measure fairness across demographic groups, highlighting a critical gap in ensuring these tools work equally well for all patients.
This PROSPERO-registered systematic review (3059 search results, 160 articles with fairness evaluations, 29 fairness measurement techniques) reveals that multimodal fairness in clinical AI is measured in only 11% of relevant studies, and in high-stakes chest X-ray/sepsis applications it drops to 8% despite the majority of studies possessing the demographic data needed for fairness analysis. The findings highlight a systematic gap in AI health equity assessment with direct implications for regulatory and clinical deployment frameworks.
What the study was
- Study design
- systematic_review
- Category
- ai_ml_clinical_diagnostics
- Maturity
- Validated
- Journal
- npj Digital Medicine
Why it surfaced
NPJ Digital Medicine systematic review exposing the scope of unmeasured algorithmic bias in multimodal clinical AI; signals an emerging regulatory and ethical risk relevant to AI diagnostics pipeline. Score 7 (N=2, D=2, P=1, E=2) for systematic review with actionable equity and policy findings.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.