AI Diagnostic Tool Outperforms Clinicians

Sequential reasoning with language models could improve diagnostic accuracy while reducing unnecessary testing, according to new research

1

Diagnostic Orchestrator (DxO) uses large language models to ask questions, request tests, and refine diagnoses as evidence emerges.
2

In 304 clinicopathological cases, DxO achieved 83% diagnostic accuracy, compared with 36% among physicians with reference materials.
3

In 325 emergency department encounters, DxO achieved 80% accuracy, compared with 47% among physicians.
4

For clinicopathological cases, DxO achieved 81% accuracy at an estimated testing cost of $2,241, versus $4,725 for physicians achieving 36% accuracy.
5

The evaluation used controlled, text-based cases and estimated testing costs, excluding factors such as imaging, pathology slides, incomplete records, and multidisciplinary decisions.

1