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.
Sitting down with Collective Health’s Ali Diab
I visited Collective Health‘s San Mateo headquarters this week looking for insight into large employers’ projections and digital health buying patterns. Employer customers use Collective Health to choose the providers and benefits they want to offer their workers. CEO Ali Diab told me his company is seeing customers scrutinize their spending more than ever.
The pressure to cut costs has led to growing frustration among employers that large health insurance companies aren’t letting them experiment or mix-and-match providers or benefits, he said. “For an employer, they feel they’re being constrained further and further.”
At the moment, Collective charges employers a fixed, flat fee per month calculated based on the number of employees. But Diab also said the company is experimenting with pay-for-performance, if they can drive down costs while also making beneficiaries healthier — including by nudging them to get preventive or behavioral care.
Still, that kind of model can be technically complicated, Diab explained: Collective has to be able assess each members’ specific health risks and then recommend next steps, like getting a colonoscopy or mammogram. The company draws from beneficiaries’ lab results, their interactions with the site, and claims, among other data sources. And even if they’ve successfully assessed beneficiaries’ risk, “it’s another thing to be actually able to engage them” so they follow up on the recommendations, he said.
How health care should shift its thinking on AI
Top health tech leaders mused on the opportunities for and limitations of artificial intelligence — and in particular, generative AI systems like ChatGPT — in the delivery of health care at STAT’s Breakthrough Summit last week. On stage, Vanna Health‘s Tom Insel — formerly of Mindstrong and Verily — told me he’d been playing around with the newest large language models and exploring their potential in health. “Its ability to give you good clinical decision support is scary it’s so good.”
“AI is terrific at creating the kind of documentation you need in the public health field so you can get paid,” he said, explaining that for behavioral health, some automation could “increase our workforce by 30 to 40% tomorrow just with the tools we have now.”
Still, misconceptions about AI abound, and experts urged the industry to be aware that applications reach far beyond ChatGPT, my colleague Brittany Trang writes.
“One of the things that I find extremely jarring is the extreme zoom laser focus on ChatGPT, per se,” said Suchi Saria, who heads the Machine Learning and Healthcare Lab at Johns Hopkins University. “People are thinking of ChatGPT as like a box. The progress in the field has not been this box,” she said. The progress is the technology in side that box, which can do more if trained on other health care data streams. Read more here.
Tracking FDA’s progress on health tech
Speaking of artificial intelligence, FDA commissioner Robert Califf gave a rundown on AI, digital health and patient engagement at a National Health Council symposium on Monday. The crux? Health tech is rapidly advancing, and the FDA needs to become far more nimble to both regulate it, and better draw patients into the conversation. “I think we’re behind, and it’s going to be really hard to catch up,” Califf said.
As the agency’s framework for testing devices that use AI and mobile health apps remains in flux, so does its approach to drawing in patients’ perspectives, my colleague Lizzy Lawrence writes.
AI devices learn and grow from real-world use. So it’s more difficult for the FDA, and patients, to understand their full risk/benefit profiles at the start of the development cycle. Read Lizzy’s dispatch from D.C. here. (A version of this item also ran in STAT’s D.C. Diagnosis newsletter. Sign up here.)
UC San Diego’s plan for an AI mission control center
A $22 million donation to UC San Diego Health will support a new center housing its efforts to use artificial intelligence to guide clinical care — and leaders are well under way in planning, The San Diego Union-Tribune reports from a symposium there late last week. Among its recent efforts: Developing an AI system to predict bowel obstructions, and allowing AI to recommend responses to patient emails. Another system predicting sepsis has already reached patients, UC San Diego Health’s chief medical and digital officer Christopher Longhurst told the Union-Tribune.
“We implemented this algorithm six months ago, and our emergency department, in the last six months, we’ve had the lowest observed (versus) expected mortality and sepsis that we’ve ever seen at UC San Diego Health,” he said. Still, it’ll take years for the center to be fully functional, Longhurst said.
The latest on Carbon’s insurance dustup
Last week I reported on Carbon Health‘s rare public dispute with Anthem Blue Cross of California, which declined to raise the primary care tech companies’ payment rate and subsequently let its contract lapse. (At the time, Carbon said, Anthem Blue Cross was refusing to reimburse out-of-network claims for patients who continued to visit Carbon, and more recently had stopped processing the claims altogether — though Anthem denied the latter charge. )