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.
Ransomware is plunging rural hospitals into crisis
Ransomware is no longer a threat reserved for only the largest health institutions. Small and rural providers are also getting hit with a wave of attacks, in some cases forcing them to resort to pen-and-paper record keeping to continue serving patients. “We were woefully unprepared,” said John Gaede, director of information services at Sky Lakes Medical Center in rural Oregon. The health system was hit with an attack in October 2020, just as it was responding to its first local surge of Covid cases, making a tough situation nearly impossible to manage.
Are voice biomarkers ready for prime time?
Using AI to find hints of disease in voice samples remains a new field of inquiry, but the effort appears to be advancing rapidly. An increasing array of apps promise to detect mental health conditions such as anxiety and depression by analyzing voice samples. And researchers from Mayo Clinic and the University of Tel Aviv just published a study showing a strong association between a specific voice biomarker and health events tied to coronary artery disease. The research is extremely promising, but also preliminary. Asking whether AI can do something in a lab, and asking whether it can do it safely and reliably in the real world are very different questions.
It ain’t all about your AUC
In initial testing, an AI algorithm to detect a specific kind of hip fracture outperformed radiologists and registered a nearly perfect AUC. But a closer examination of the model by researchers at the Australian Institute for Machine Learning revealed unexpected problems. In validation on an outside dataset from Stanford University, the algorithm’s AUC remained sky high, but it also registered a sharp rise in false positives at its prespecified calibration. Sure, that can be fixed with careful local training.
But a deeper audit of its performance also turned up inhuman mistakes, such as misdiagnosing severe fractures in certain instances. Those kinds of issues can lead to unacceptable errors in clinical settings, and they are easy to miss if algorithm developers and regulators aren’t looking carefully.
Virtual care by the numbers…
Year-end operating stats shared with us by UnitedHealth Group offer some insights on virtual care worth pondering:
- 28 million virtual care visits were recorded in 2021 — a sustained 2,500% increase from pre-pandemic levels and an indication of remote-care’s staying power.
- 14 million of those virtual visits, or fully half, addressed behavioral health needs, which tells you why behavioral health startups are continuing to rake in cash.
- 64% of virtual care users were women, up 2% from 2020. The growth from year-to-year shows that dynamic may remain durable over the long run.
Dollars, data, & drones
- Health data brokers Komodo Health and Datavant are expanding their partnership to allow life sciences companies to connect proprietary and third-party clinical trial data to Komodo’s software platform. The software allows users to link de-identified data from electronic medical records and other sources to gain a more complete picture of patients’ care and outcomes.
- Alphabet-owned Fitbit has received FDA clearance for an algorithm to detect atrial fibrillation. Plenty of other wearables detect A-fib, but Fitbit is emphasizing the ability of its product to flag the condition while you’re asleep, potentially allowing for more comprehensive monitoring.
- Another Alphabet-owned company, Wing, has launched a project using drones to deliver over-the-counter medicines and other products ordered by Walgreens customers in Little Elm, TX, near Fort Worth. Items are attached to lines dropped by the drones and then reeled in for delivery.
- Diligent Robotics, the maker of a robot that transports supplies for frontline caregivers, raised $30 million in a Series B round led by Tiger Global. The company is pitching the product as a way to combat the national nursing shortage.
- Viz.ai, a company known for using AI to accelerate stroke detection, raised $100 million in a Series D round led by Tiger Global and Insight Partners. The company said its technology is now in use in more than 1,000 hospitals.
What we’re reading