Regulation
Will AI transparency make a difference?
One of the most significant regulations on health artificial intelligence just went into effect.
As I report in a new story, as of January 1st, electronic health record vendors like Epic, Athenahealth, Meditech, and Oracle must disclose to end users, like hospitals, a broad swath of information about predictive models and algorithms in their software. This includes: what variables go into a prediction, whether a tool was tested in the real world, what the tool's developers did to address potential bias, warnings about improper use, and more. Epic told me they had to detail over 100 tools, like their algorithm for predicting sepsis.
But while experts told me the new transparency efforts are an important step forward, they do not ensure the quality of products — regulators are hoping that, for now, industry can figure that part out. (Some of the products may be subject to FDA review as well.) Moreover, there are concerns about the scope of the regulations. Some argue that it should be limited to only the riskiest algorithms used in care. Others say it should be broadened to include algorithms developed independently by third parties. Still others say the information should be disclosed to the public.
Do you have opinions? Let me know.
Read my whole story here
policy
FDA drops guidances on AI, pulse oximeters
Seems like regulators at the Food and Drug Administration didn't take any time off over the holidays, as they dumped a series of anticipated guidance documents this week.
- The agency released draft guidance on improving the accuracy of pulse oximeters for people of all skin tones. In part, FDA wants way more data, for example, by asking companies to increase the size of pulse oximeter studies from 10 people to 150 or more. STAT's Usha Lee McFarling has been following this issue for years; read her story with Anil Oza on the new guidance here.
- Another draft guidance offers recommendations on AI in drug development focused on risks that might be introduced by the technology. It does not cover issues around AI used in drug discovery. STAT's Katie Palmer explains here.
- And FDA offered draft guidance on AI medical devices. The agency recently crossed 1,000 cleared AI devices and the guidance focuses on what developers ought to include in marketing submissions to FDA. The agency also offers recommendations for how manufacturers can mitigate risks from AI devices throughout their lifecycle, for example, by testing them in all relevant demographic groups and following up on their performance in the real world.
clinical trials Actigraph acquires Biofourmis to expand clinical trials offerings
With more drug companies using devices and sensors to study the impact of their treatments, Actigraph is acquiring the life sciences business of Biofourmis to beef up its offerings. Actigraph, which makes wearables and software for trials, is known for its ability to measure physical activity in ways that meets the rigorous demands of some of the largest drug makers. With Biofourmis it adds the ability to help developers collect data from other devices, as well as a suite of software and algorithms.
Biofourmis previously raised hundreds of millions of dollars to develop not only a clinical trials business but also offerings around delivering care at home. The company ran into issues a few years ago, and last year, it merged with chronic disease management company CopilotIQ. Actigraph is acquiring the assets from the new combined entity. I'm told the cash deal was "eight figures," meaning over $10 million.
Read my whole story here
No comments