You’re studying the online version of STAT Well being Tech, our information to how tech is remodeling the life sciences. Signal as much as get this text delivered in your inbox each Tuesday and Thursday.
Business fumes as FDA will increase AI instrument regulation
Well being tech corporations are in a tizzy. After years of letting corporations roll out software program instruments to information affected person care with little oversight, the Meals and Drug Administration is taking a harder stance.
In September, the FDA announced its intentions to regulate many of those AI-powered scientific determination help instruments as gadgets — which regulators say has at all times been their plan and inside their purview. However the trade says it was blindsided by the transfer. Earlier this month, the Scientific Determination Assist Coalition filed a petition asking the FDA to withdraw its last CDS steerage, arguing regulators are overstepping their bounds. “Honestly, the result’s that extra merchandise are regulated,” stated Bethany Hills, a companion at DLA Piper who advises FDA-regulated machine corporations. “That’s onerous for people to digest.”
I’ve bought the complete particulars on the battle here.
Stroke danger instruments might not work properly for Black sufferers
Stroke danger prediction instruments don’t work as properly for Black sufferers, in accordance with a brand new research led by researchers on the Duke College Faculty of Drugs. Docs use these algorithms to suss out whether or not sufferers are at risk of affected by strokes, with newer fashions counting on machine studying. The analysis crew discovered these newer algorithms weren’t a lot better at predicting stroke danger than less complicated algorithms — and had been worse at detecting danger for Black sufferers in comparison with white sufferers.
“We bought the surprising outcome — to me, surprising — that the measure of discrimination, the flexibility to rank them, was a lot better… for white contributors than Black contributors,” Michael Pencina, director of Duke AI Well being and a lead researcher on the research, informed Ambar Castillo. Read more.
Amazon closes $3.9B buyout of One Medical
Amazon has formally closed its $3.9 billion deal to amass main care firm One Medical, which it first announced in July. New U.S. prospects will be capable of purchase an annual membership on the discounted worth of $144 (the service is usually $199 a yr).
The deal has bolstered Amazon’s rising well being care portfolio, which embrace its Amazon Pharmacy prescription business and the telehealth platform Amazon Clinic. The One Medical deal rounds out Amazon’s ambitions to streamline primary care and well being care supply. The acquisition rang alarm bells for antimonopoly teams and the Federal Commerce Fee, which started a evaluation of the deal in September however hasn’t blocked it. “The FTC’s investigation of Amazon’s acquisition of One Medical continues,” FTC spokesperson Peter Kaplan stated in an announcement. Read more here.
De-identifying wearable machine information isn’t sufficient
Wearable gadgets usually de-identify consumer information earlier than sharing it with third events, like researchers, to guard privateness. However is that this follow efficient, or is wearable machine information so in depth and personalised that it’s potential to re-identify people? In a paper published last week by The Lancet, researchers from Duke College analyzed 72 research that examined whether or not people had been in a position to be reidentified primarily based on information from wearable gadgets. Most research used electrical coronary heart alerts, electrical mind exercise, and physique motion to determine contributors.
Eight research had been in a position to determine people primarily based on coronary heart alerts, 17 research primarily based on mind exercise, and 13 primarily based on a person’s strolling gait. The researchers concluded that sharing deidentified information might not be ample to guard affected person privateness. “These findings name for concerted efforts to rethink strategies for information sharing to advertise advances in analysis innovation whereas stopping the lack of particular person privateness,” they wrote.
- On Thursday, Atmosphere Healthcare introduced the launch of its AutoScribe know-how, which summarizes patient-doctor visits by means of synthetic intelligence. The startup has raised $30 million in funding from traders equivalent to Andreessen Horowitz and the OpenAI Startup Fund.
- Digital well being firm Eko is launching a brand new cardiac illness detection platform known as Sensora, which works to determine coronary heart murmurs.
- Philips executives are waiving their bonuses — other than former CEO Frans van Houten, who netted about $222,000 in his bonus — after a disastrous yr resulting from its ventilator remembers.
- The FDA has permitted a brand new affected person monitor from ache administration firm Medasense by means of the de novo pathway. The monitor makes use of physiological alerts to measure ache in anesthetized sufferers.
- Axena Well being, a medical machine firm specializing in ladies’s pelvic well being, secured $25 million in a Sequence A funding spherical.