Work

Better Than Nothing: The Local Standard of Care and the Ethics of Clinical AI in Physician-Scarce Health Systems

Year: 2026
Type: preprint
Abstract: : The comparator decides the verdict. Clinical artificial intelligence reaches physician-scarce health systems with evidence that it outperforms local care, which often includes no clinician. The same comparison justified placebo arms in the 1990s zidovudine trials. After that dispute, the World Medical Association reaffirmed in the 2024 Declaration of Helsinki that trials must test new interventions against the best proven care. Clinical AI enters low-income clinics through procurement catalogs and commercial rollouts. Research oversight does not reach those channels. Studies from Nairobi clinics, WHO tuberculosis assessments, Thai retinal screening, a pan-African medical benchmark, and facility surveys test clinical software against real clinic conditions. Trained clinicians and local calibration reduce errors. In Rwanda and across national tuberculosis programs, foreign sponsors still control the choice of baseline and the survival of the service. A *de jure* standard subjects procurement and clinic rollouts to the rules of human research. Keywords: Clinical Artificial Intelligence, Standard of Care, De Jure Standard of Care, De Facto Standard of Care, Declaration of Helsinki, Short-Course Zidovudine Trials, Placebo-Controlled Trials, Research Ethics, Ethics Dumping, Health Workforce Shortage, Computer-Aided Detection, Tuberculosis Screening, Threshold Calibration, Global Drug Facility, Global Fund Procurement, Babyl Rwanda, Horizon1000, Post-Trial Access, Global Health Governance, Low- and Middle-Income Countries
Language: English
Field: Medicine
Open Access status: green
OpenAlex ID: W7226498288
Cites:
Cited by:
Related to: 10