Breaking News

How much time do doctors spend correcting AI?

February 4, 2025
Health Tech Correspondent

Good morning health tech readers!

Lots of action, including a scoop about digital health unicorn Sword Health's expansion plans. Let's get to it.

Reach me: mario.aguilar@statnews.com

Artificial intelligence

DeepSeek and Chinese biotech, OpenAI and deep research

A little-known Chinese company called DeepSeek caused a stir in technology circles by releasing artificial intelligence models that rival those of U.S.-based competitors and were developed at a fraction of their cost. The move highlights China's interest in the cutting edge tech, including around drug development. In a new story, Brian Yang speaks with executives at XTalPi and other companies about the competition overseas. 

"There are more Chinese scientists in AI now than non-Chinese scientists." said Alex Zhavoronkov, the CEO of Insilico, which has operations in the U.S., China, and Hong Kong.

Not to be outdone, OpenAI over the weekend announced a new product, called deep research, that crawls the web and produces sourced research documents on-demand. The company writes that deep research is "built for people who do intensive knowledge work in areas like finance, science, policy, and engineering and need thorough, precise, and reliable research." The product is currently only available to people with a "pro" ChatGPT account. I haven't shelled out $200 to try it out, but I might! What science should I have it research to test its potential?


policy

FDA director overseeing digital health and AI departs

Troy Tazbaz announced on Friday that he's leaving his post as the director of the Food and Drug Administration's Digital Health Center of Excellence. During his two-year tenure, he helped reorient the agency's devices center's artificial intelligence efforts to include generative AI. Important milestones relevant to digital health include the FDA's recently finalized pre-determined change plan guidance that provides a way for AI medical device developers to have their software updates approved by the agency in advance. The agency also recently published draft guidance to AI developers in which it described what information it would like to see in marketing submissions.

Along with former ONC-chief Micky Tripathi, Tazbaz was a key government ally to the Coalition for Health AI, an advocacy organization representing Microsoft, Amazon, CVS, Mayo Clinic, and many others. Both regulators  backed away from CHAI after backlash that accused them of being too cozy with industry.

No word yet on where Tazbaz is headed, but he's a former Oracle executive, so it's worth wondering if he's going back to the mothership to work on the company's EHR business — or if maybe Larry Ellison beckoned him for Stargate-related plans. If you know, let me know.


business

Scoop: Sword teases AI mental health expansion

Sword Health is planning to expand to more kinds of care, beginning with a mental health offering that could be formally announced as early as this month, CEO Virgílio Bento tells me. Sword is known as one of the leading providers of digital musculoskeletal care — mostly physical therapy — to employer health plans. As part of the new expansion, the company hopes to ditch the "MSK" association and become known as an "AI care company that is going to reinvent all care delivery models that are 100% labor intensive, with our AI-first model," said Bento. Beyond talk therapy, Bento also mentioned speech care, GI care, and cardiac care as ripe for reinvention with AI.

Bento presented some of this behind closed doors last month at the J.P. Morgan Healthcare conference, where he told attendees that health care costs have gotten dramatically higher despite the adoption of new technology because innovation has only created marginal improvements to the human clinician's efficiency. To really improve the system, Bento believes, health care needs to move away from the "human first" model of care. This is Bento's first interview with a reporter on the broader vision for Sword.

The company's primary offering is digital physical therapy delivered through a proprietary tablet. People complete sessions in their homes, at their convenience, instead of going to a physical therapy clinic. A soothing voice guides users through a program giving feedback and pointers along the way. The program has been approved by a clinician but it is supervised by technology that's using cameras in the tablet to analyze movements. After a session, the company's AI sends a report back to the clinician and may also propose changes to someone's plan.

Using this model as a template, Bento believes Sword can address all kinds of needs. Looking at talk therapy for low severity conditions like anxiety, Bento said that, "for me, it's not an efficient way to address mental health issues." He didn't elaborate on the plans except to say that it's "very disruptive." From his comments I can infer that it's higher-frequency than traditional care and, of course, will use automation to reduce the reliance on human labor. We've seen a lot of automated agents delivering CBT or similar in the past, so I am ready to be surprised by anything truly new.

Asked whether the company had the clinical expertise necessary to stand up whole new verticals of care, Bento pointed out that the company successfully added an offering for pelvic health, which is technically distinct from its core physical therapy offering. 

"That's in our DNA. Look, I'm an engineer, right?" he said. "We start by surrounding ourselves with the top experts in the field."

Sword Health last year raised $130 million at a reported $3 billion valuation. 



cybersecurity

FDA warns about 'serious' security failings in patient monitors

Katie Palmer writes: The FDA put out a safety communication last Thursday warning of serious cybersecurity vulnerabilities in two patient monitors that track vital signs, from Contec and Epsimed. The monitors' software includes a hidden backdoor that could compromise the device or the network that it's connected to. And if the device is connected to the internet for remote monitoring — a mode it wasn't cleared for by the FDA — it can expose patients' protected health information. 

"What's most striking is the severity of two violations: the use of uncleared wireless communication and a hidden backdoor," said Kevin Fu, health care cybersecurity expert at Northeastern University. "Frankly, I'd expect to see tactics like a backdoor from a nation-state threat actor, not a medical device manufacturer. Someone either made a deliberate design choice, or they failed to check that a backdoor was embedded in the device." Adding wireless communication, he added, can put patients at risk by increasing the ways their information can be attacked. 

"In my opinion, this case feels like the cyber equivalent of picking your nose before examining a patient," said Fu. "These aren't just regulatory oversights; they're serious patient safety issues."


doctors

How much time do doctors spend checking on AI?

g9l6e-clinicians-using-ai-say-they-have-to-review-or-correct-irrelevant-information-generated-by-ai-tools

Brittany Trang writes: In a recent survey from health care AI company Corti and market research company YouGov, 30% of clinicians using AI in their day-to-day work said that they spend between 1 to 3 hours "reviewing or correcting irrelevant information generated by AI tools." STAT asked Corti for more details, and you can see the full breakdown above.

The survey, conducted in November 2024, questioned 500 U.S. doctors, physicians or nurses, 188 of which indicated that they used AI regularly. How are you using AI tools in your work? Do you find that you're having to spend significant amounts of time correcting their output? Let us know.


More around STAT
Check out more exclusive coverage with a STAT+ subscription
Read premium in-depth biotech, pharma, policy, and life science coverage and analysis with all of our STAT+ articles.

What we're reading

    • The young, inexperienced engineers aiding Elon Musk's government takeover, Wired
    • How Trump's tariffs may raise health care costs, from medical devices to prescription drugs, STAT
    • Noom cuts coaches and customer service staff as fewer sign up, Endpoints
    • He ruled the operating room and then daytime TV. Now, Dr. Oz is set to take over a $1.5 trillion health agency, STAT
    • Chief VA watchdog who helped expose flawed computer system in Spokane speaks out after being fired by Trump, The Spokesman-Review

Thanks for reading! More on Tuesday - Mario

Mario Aguilar covers how technology is transforming health care. He is based in New York.


Enjoying STAT Health Tech? Tell us about your experience
Continue reading the latest health & science news with the STAT app
Download on the App Store or get it on Google Play
STAT
STAT, 1 Exchange Place, Boston, MA
©2025, All Rights Reserved.

No comments