Artificial intelligence
A closer look at FDA's new AI guidance
In new draft guidance earlier this week, the Food and Drug Administration advised developers of medical devices that use artificial intelligence that it would like to see more information about sources and demographics of data used to train and validate their products, and disclose blindspots and potential biases that might impair performance. Specifically, FDA asked for this information to be included in marketing submissions. The new guidance comes shortly after FDA crossed 1,000 AI devices authorized.
As STAT's Casey Ross notes, the guidance does not specifically call for more testing in live clinical settings as a condition for getting approvals. But it does recommend that device makers submit a plan for monitoring the performance of their devices in clinical settings after they are cleared for commercial use.
Read Casey's full analysis here
business
Transcarent spends $621 million acquiring 14 million members
Greg Allen/Invision/AP
Yesterday, I wrote about Transcarent's huge acquisition of Accolade. The deal makes sense on many levels. Both companies have a similar mission of helping employers deliver better health care at lower cost for their workers. Transcarent CEO Glen Tullman, through his own funds and close ties to venture firm General Catalyst, has access to lots of capital. And Accolade, which went public during the pandemic digital health frenzy, has struggled to turn a profit, and like other companies in the sector, its stock has suffered.
In an interview, Tullman pointed to things Accolade offers and that Transcarent could use, like advocacy services and virtual primary care. But my take, which is loose and not in my story, is that Transcarent took a half-billion dollar shortcut to jump from 4.5 million in the U.S. to 18 million members. While we don't know how much Transcarent has spent so far, we do know the company has raised over $400 million to get where it was. It's now quadrupling its member base and adding more services for $621 million more.
Will it work? Will Tullman reinvent the health care for workers as he has promised? That's the hard part.
Read my story and interview with Tullman here
research New study sheds light on hospital algorithm use
A new study provides an interesting window into how hospitals are using clinical algorithms — and what steps they are taking to ensure the technologies are accurate and unbiased. To paraphrase Health Affairs:
- 65% of U.S. hospitals used predictive models.
- 79% of those used models from their electronic health record developer.
- 61% of hospitals that used models evaluated them for accuracy using data from their health system.
- 44% percent of hospitals reported local evaluation for bias.
The data is based on 2,425 hospitals that answered questions about AI in the 2023 American Hospital Association annual survey IT supplement.
What's your take? Email: mario.aguilar@statnews.com. If I get interesting responses, I'll publish them next week.
No comments