Doctors are using AI for more than diagnosis, study finds

A new study published in the Journal of the American Medical Association (JAMA) revealed something surprising about how doctors are using AI. They’re asking it more than twice as many questions about how to treat patients as about how to diagnose them.

Why it matters: For years, much of the excitement around the use of AI in medicine has focused on its potential to diagnose and spot clues physicians might miss. But the study suggests that AI is playing a different role in clinical practice: it’s helping physicians navigate treatment decisions, including questions outside of their own specialties.

Driving the news: Research led by Harvard Medical School and Brigham and Women’s Hospital analyzed nearly 39 million queries submitted by more than 900,000 US clinicians to OpenEvidence, a medical AI tool, over two and a half years. OpenEvidence provided the data, drawn from more than 270 million queries, and three of its employees co-authored the study.

• Treatment and management accounted for 33.1% of physician questions, compared with 13.9% for diagnosis.
• Doctors frequently asked questions about areas where they don’t have specific expertise. Emergency department physicians had the highest share at 88%, compared with 35% among oncologists and hematologists, the lowest share.

The bigger picture: Access to specialized information isn’t the same as having the expertise to apply it. And that’s a distinction with implications far beyond healthcare. This week on Mixture of Experts, panelists discussed how that distinction is playing out across professions.

• Aili McConnon, a Tech News Writer at IBM, highlighted the difference between diagnosis and treatment. “It’s less about ‘What does this patient have?’ and [more about] ‘What should I, as the doctor, do about it?’”

Aili McConnon, a Tech News Writer at IBM, highlighted the difference between diagnosis and treatment. “It’s less about ‘What does this patient have?’ and [more about] ‘What should I, as the doctor, do about it?’”

• Sandhya Iyer, Associate Partner for AI and Analytics at IBM, pointed out that emergency department physicians may not always have time to consult specialists as new patients arrive. AI could help them make decisions faster, provided the information is relevant enough to guide their judgment, she said.

Sandhya Iyer, Associate Partner for AI and Analytics at IBM, pointed out that emergency department physicians may not always have time to consult specialists as new patients arrive. AI could help them make decisions faster, provided the information is relevant enough to guide their judgment, she said.

• Principal Engineer Jacob Repp shared an example from his own team at IBM. One designer has been using AI to access highly technical information about IBM Z mainframes to help build demos and modify user experiences. But specialists still have a role in reviewing the work, he said.

Principal Engineer Jacob Repp shared an example from his own team at IBM. One designer has been using AI to access highly technical information about IBM Z mainframes to help build demos and modify user experiences. But specialists still have a role in reviewing the work, he said.

Between the lines: The study showed what doctors ask AI, but not what they do with the answers. Researchers couldn’t determine whether doctors use AI in place of a specialist or as a first step before consulting one. Vishal Patel, the study’s lead author, said more research is needed on the tools’ accuracy and effects on patient care.

The broader implication: As AI puts more specialized information at people’s fingertips, how will they know when to act on it and when to turn to an expert?