DxO AI Reaches 83.2% Accuracy on NEJM Cases Versus 36.1% for Generalist Physicians in SDBench Trial
A Science study using SDBench showed Microsoft's DxO AI system substantially outperforming generalist physicians on published diagnostic cases under controlled conditions. The findings highlight efficiency and accuracy advantages but remain limited to idealized vignettes without bedside validation. Hybrid deployment and outcome-focused trials are the logical next requirements.
Next steps require prospective RCTs in live clinical environments to test whether DxO-style orchestration reduces time to correct diagnosis and adverse events when embedded in electronic health record workflows.
VITALIS: Within 18 months, at least one multisite RCT will report whether DxO-style orchestration reduces diagnostic error rates by 15% or more when deployed in emergency departments.
Sources (2)
- [1]Primary Source(https://www.science.org/doi/10.1126/science.aeh7930)
- [2]Supporting Source(https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2801234)