STAT Health Tech is our new weekly guide to how tech is transforming health care and the life sciences. Sign up here to receive it in your inbox every Wednesday.
Good morning. We’re thrilled to bring you the first edition of STAT Health Tech!
Every Wednesday, we’ll steer you through the crowded intersection of health care and technology. You’ll get the latest on how tech giants like Amazon and Alphabet and startups like Akili and Insitro are trying to disrupt the delivery of care and the development of drugs. And we’ll give you our take on the underlying financial, ethical, and privacy issues.
Why did we decide to launch a newsletter on this topic? Because advanced artificial intelligence and novel software tools are challenging the status quo in every corner of the industry. Our goal is not only to update you on the news, but also to separate hype from real progress, and tell you how changes in regulation and the broader business environment will affect uptake of new technologies within health care.
A little background on us: Casey is STAT’s national technology correspondent, based in Cleveland, and Rebecca leads our West Coast coverage of health tech and biotech from San Francisco. Combining our efforts will allow us to cover more ground and tackle this beat from different angles, examining the impact on the lives of doctors and patients as well as the bottom lines of entrepreneurs and established companies.
We’ll show when the interests of those parties align, and when they do not. We want to cover the great promise of tech innovation and the many challenges of deploying it in an industry where human life is at stake, not to mention a fifth of the U.S. economy and the fortunes of the nation’s biggest companies.
With that out of the way, let’s get to this week’s news.
The talk of health tech
Aidoc announced today that it received FDA clearance for software that uses AI to identify warning signs of pulmonary embolism (a potentially fatal blood clot) in chest CT scans. The Israel-based company said it has eight more imaging products in active clinical trials. The news came two days after another Israeli company, Zebra Medical Vision, said it got the agency’s sign-off for AI software that can identify a pneumothorax (collapsed lung) from a chest X-ray. Two larger takeaways here: Israel, specifically Tel Aviv, is becoming a hotbed for digital health innovation. Second, the pace of FDA approval of AI products is accelerating rapidly: By one count, 14 had been approved/cleared as of January. Since then, the tally has doubled.
A key study supporting a fertility app called Daysy just got retracted. The journal, Reproductive Health, pulled the paper because of “fundamental flaws in [the study’s] methodology,” as Buzzfeed reported yesterday. Apps that claim to help or hinder conception by predicting when a person is ovulating have been around for years — and Daysy is far from the only one that has claimed to be very effective. One app called Dot says its studies show the app is 95 percent effective at preventing pregnancy; the only FDA-cleared app for contraception, Natural Cycles, claims a 93 percent effectiveness rate with typical use.
It can be easy, amid Elon Musk’s many and varied antics, to forget that he founded a very quiet, very ambitious startup called Neuralink. The San Francisco company, which is trying to develop a system to link the human brain to computers, is raising money. It’s already brought in $39 million of an anticipated $51 million funding round, the company said in a new filing with the SEC. Next up: Watch for the company to announce “something notable” in the coming weeks, if you trust a recent tweet from Musk.
AI has drawn interest as a promising tool in the quest to prevent suicide, spurring experiments everywhere from Facebook to the Canadian government. Now, the Trevor Project, a nonprofit that works to try to prevent suicide in LGBTQ youth, is giving it a try, too. Google will give the organization $1.5 million — plus access to its AI technology — to use machine learning and natural language processing to try to improve its crisis services, NBC News reports. But keep in mind: There’s not yet evidence linking the use of AI to fewer suicides.
Remember Jawbone, the fallen consumer electronics unicorn that burned through $1 billion in a decade before liquidating in 2017? A health tech company born in its ashes just raised $65 million, according to a new SEC filing from Jawbone Health. Hosain Rahman, the entrepreneur behind both companies, has said that Jawbone Health will use monitoring devices and machine intelligence to try to prevent disease. News of the fundraising, first reported by TechCrunch, prompted some serious skepticism — particularly about the judgement of Rahman’s “very forgiving & optimistic investors,” as digital health evangelist Matthew Holt put it.
Predicting patient decline from a suburban hospital bunker
A visit to the Cleveland Clinic’s Central Monitoring Unit, or CMU for short, gave us an inside view of how the clinic is using AI and real-time data feeds to speed up responses to patient emergencies. Technicians in the CMU, situated in a leafy suburban office park outside Cleveland, use a home-grown algorithm to identify and remotely monitor the health system’s highest-risk patients based on an analysis of cardiac telemetry data and other clinical factors.
Dr. Daniel Cantillon, a cardiologist and medical director of the unit, said the next step is to use machine learning to begin to predict cardiac emergencies as much as an hour before they happen. He said the clinic is working with unnamed commercial partners to identify digital biomarkers of deterioration and build a product to increase warning times: “We want a system that is of the highest reliability for our own patients and also for future use cases for other health systems,” Cantillon said. It may be some time before such software is deployed, but other major medical centers, including Yale New Haven Hospital and Johns Hopkins, are also exploring AI-based approaches to predicting patient emergencies. Stay tuned.
The latest from academia
As algorithms inundate medicine, universities are scrambling to keep up. The latest sign of that came this week, with the launch of two new academic centers in New York. At the Icahn School of Medicine at Mount Sinai, the Center for Computational Immunology will try to bring algorithms to bear on the development of cancer therapies that harness the immune system. And at Stonybrook University, the Institute for AI-Driven Discovery and Innovation will support research in bioinformatics, among other applications.
A new paper by Johns Hopkins researchers in Nature calls for a “digital health scorecard” to assess the value of tech products in medicine. Our inboxes are a testament to the need for an objective assessment mechanism. The pitches are constant, but it is difficult to cut through the puffery to identify products likely to deliver real benefits to doctors and patients, and returns for investors. The paper’s authors propose a scorecard based on four criteria: technical accuracy, clinical value, usability, and cost. “For digital health solutions to have greater impact, quality and value must be easier to distinguish,” they wrote. It is easy to identify instances in medicine where those criteria were not fully considered. Three words: electronic health records.
Having trouble keeping up with the unrelenting stream of new research papers on AI in medicine? We are, too — which is why we were heartened to see the launch yesterday of Doctor Penguin, a project to curate by email the top papers of the week. It comes by way of a team of AI researchers from Stanford and Scripps Research.
Hype watch: a small machine-learning study with a big headline
A bold headline out of a European cardiology conference caught our eye this week: “Machine learning overtakes humans in predicting death or heart attacks.” The headline was attached to a study detailing the performance of an algorithm in predicting patients’ risk of heart attack or cardiovascular death. The study involved 950 patients; 24 suffered heart attacks and 49 died. After repeatedly analyzing 85 variables in imaging data and clinical histories, the system identified the variables that predicted those outcomes with a 95 percent accuracy rate. Some experts were unimpressed, however, noting that there were too few patients in the study to proclaim the algorithm was ready for clinical use. Eric Topol, an AI expert and author of the book “Deep Medicine,” tweeted that the headline, which rocketed around the internet this week, was an “exemplar of AI in medicine hype.”