Health tech companies are facing a wave of layoffs – STAT

Biotech & Pharma
Health Tech
Business & Policy
Health & Science
Exclusive analysis of biotech, pharma, and the life sciences
Follow Us
By Casey Ross and Katie Palmer June 14, 2022
You’re reading the web edition of STAT Health Tech, our guide to how tech is transforming the life sciences. Sign up to get this newsletter delivered in your inbox every Tuesday and Thursday. 
Digital health is tightening its belt
Digital health is facing a wave of layoffs like it hasn’t seen since the beginning of the pandemic. Economic pressure is hitting every industry, and traditional and tech-forward companies alike are using headcount as a release valve. But the whiplash is especially severe for health tech, which has seen unprecedented growth and investment over the past two years. Noom raised $540 million in a single round last May; now it’s laying off 495 employees, many of them wellness coaches. Carbon Health went on an acquisition spree and is losing 250. Online pharmacy DivvyDose promised to hire for hundreds of positions, and is instead cutting 62.
The list goes on: Thirty Madison cut 24 employees after merging with Nurx, and Truepill said it laid off 15% of its workforce. Those losses don’t necessarily reflect the promise of young digital health companies — but some of the contraction is a direct response to recently-exposed cracks in their foundations. Cerebral is facing down an undisclosed number of layoffs on July 1 in the wake of the federal inquiry into its ADHD medication prescribing practices, and companies in the space, including ADHD-focused Ahead, have simply closed up shop. Whether it’s a temporary protective move or a response to a fundamental flaw, the belt-tightening is likely far from over.
FDA clears a Parkinson’s app for Apple Watch
The Food and Drug Administration has cleared software to track Parkinson’s disease symptoms using sensors from the Apple WatchMario reports. The StrivePD system from neurological data startup Rune Labs keeps track of patient’s tremors and dyskinetic symptoms using Apple’s Movement Disorder API, and allows them to be displayed to clinicians — some of whom have been using the system at the University of California San Francisco and Mount Sinai. Despite Apple’s own work demonstrating symptom tracking for Parkinson’s, it has said it doesn’t plan to market movement disorder monitoring. But the clearance could nonetheless turn the watch into a more regular part of Parkinson’s care — if only because it makes it possible for clinicians to use certain billing codes when reviewing data from the device.
Will Oracle be a pioneer or an also-ran?
Reactions to Oracle chairman Larry Ellison’s promise to create a “national health records database” following its Cerner acquisitions have ranged from eye rolls to cautious optimism. Following the data aggregator’s playbook, Ellison emphasized the most altruistic of goals — improved patient care and better public health research — in pledging to build a highly-detailed repository of patient information. However, such information compiled by other data miners are often focused on driving private profits instead of broader public benefits. It also remains to be seen whether Oracle will get the regulatory backing and cooperation from clients and other health records vendors to launch such a database. Mohana has the full story.
When the AI solution is worse than the problem
It’s a common problem in health outcomes prediction: If the dataset to train a model is heavily weighted toward people with the condition the AI is trying to predict, it may have trouble accurately identifying patients who don’t have that condition, and vice versa. But a new study published in JAMIA finds that the solutions being offered — which involve adding more examples of the minority outcome — are often making matters worse  The authors found that sampling fixes for logistic regression models resulted in “strong overestimation” of probability for patients in the minority class, translating to lower accuracy and greater miscalibration. The authors call for further study of sampling fixes applied to other types of AI prediction models, such as random forests and neural networks.
Dollars, data, & deals
Tapping new talent
What we’re reading
National Technology Correspondent
Casey covers the use of artificial intelligence in medicine and its underlying questions of safety, fairness, and privacy. He is the co-author of the newsletter STAT Health Tech.
Health Tech Correspondent
Katie Palmer is a health tech correspondent at STAT.

This name will appear with your comment
There was an error saving your display name. Please check and try again.
By Mario Aguilar
By Jonathan Wosen
By Matthew Herper
By Damian Garde
By Tara Bannow
Reporting from the frontiers of health and medicine


Leave a Reply

Your email address will not be published.