Sources of Bias in Clinical Artificial Intelligence and Applications in Rheumatology

1. Rheum Dis Clin North Am. 2026 Aug;52(3):545-558. doi: 10.1016/j.rdc.2026.03.005. Epub 2026 May 25. Sources of Bias in Clinical Artificial Intelligence and Applications in Rheumatology. Creasman M(1), Garcia-Agundez A(2), Yazdany J(3), Schmajuk G(4). Author information: (1)Division of Rheumatology, University of California, San Francisco, USA. Electronic address: Megan.creasman@ucsf.edu. (2)Division of Rheumatology, University of California, San Francisco, USA. (3)Division of Rheumatology, Zuckerberg San Francisco General Hospital, University of California, San Francisco, USA. (4)Division of Rheumatology, University of California, San Francisco, USA; San Francisco VA Health Care System. Rheumatology machine-learning models are limited by preexisting, technical, and emergent biases; the interaction of data constraints, design choices, and real-world clinical workflows, rather than from isolated technical errors. Across the model lifecycle, optimization objectives can encode patterns of care, access, and documentation, producing hidden subgroup failures that are obscured by aggregate performance metrics. Given the heterogeneity of rheumatic disease and current disparities in care delivery, addressing bias requires deliberate design choices before, during, and after a model is built, as well as a commitment to transparency, and sustained oversight. Copyright © 2026 Elsevier Inc. All rights reserved. DOI: 10.1016/j.rdc.2026.03.005 PMID: 42409436 [Indexed for MEDLINE] Conflict of interest statement: Disclosures The authors have nothing to disclose.

{
  "language": "lang-en",
  "diagnosis": [
    "ra"
  ],
  "specialty": "rheumatology",
  "study_type": "cohort",
  "evidence_level": "level-3",
  "v6_autopublish": true,
  "clinical_impact": "moderate-impact",
  "practice_recommendation": "monitoring"
}