Artificial intelligence can help clinicians navigate medical information faster, but its growing role in healthcare also raises a basic question: Can doctors see the evidence behind what the machine tells them?
Wolters Kluwer Health says they should.
The company is emphasizing evidence, transparency, and accountability in clinical AI as healthcare providers increasingly explore the technology for decision support. The approach is central to its UpToDate Expert AI, which is designed to let clinicians inspect the basis for AI-generated responses rather than simply accept them at face value.
“The important principle is the transparency and the accountability. So if I did want to inspect, I could,” Ihtimam Hossain, RPh, MBA, MPH, head of Clinical Advisory International at Wolters Kluwer Health, said during the “Wolters Kluwer Health: Leading Healthcare in the AI Era” forum at the Philippine General Hospital.
Hossain, who presented a study with Sheita Bond, MD, head of Clinical Artificial Intelligence at Wolters Kluwer Health, said the system allows users to compare AI-generated responses with human-written content and examine the studies supporting clinical claims.
The company evaluated the system using more than 3,000 clinical queries covering over 20,000 criteria. Its latest unpublished evaluation found that essential criteria were met about 15 percentage points more frequently than with general-purpose AI models, while omissions were about 15 percentage points lower.
Another evaluation involving nearly 2,000 queries found that almost 98 percent of claims generated by UpToDate Expert AI could be mapped to UpToDate sources, with about 2 percent coming from the model’s broader knowledge.
Wolters Kluwer also conducted more than 250 hours of adversarial testing to probe the system for weaknesses.
“This technology is very, very powerful. It would be unwise to ignore it,” Hossain said.
But power, he added, needs guardrails. Clinical AI should ultimately “preserve the discipline of evidence-based medicine,” keeping human judgment and verifiable medical evidence at the center of care.
