Nobels: Will the quest for profits quash collaboration?

Artificial intelligence was all over this year's Nobel Prize roster. John Hopfield and Geoffrey Hinton, who laid the groundwork for machine learning, won the physics prize; the next day, a group of researchers whose work using AI to predict the structure of proteins built on Hopfield and Hinton's work took home the chemistry award.

The chemistry prize, the scientific community agrees, is a reflection of the power of collective research shared freely, write my colleagues Katie Palmer and Brittany Trang. But the spirit of open collaboration could be in jeopardy as companies race to commercialize research.

"The deep learning stuff we're doing, and the deep learning stuff that the DeepMind people are doing, really is building on the life work of tens of thousands of scientists," the University of Michigan's David Baker, who won the award along with Alphabet-owned DeepMind's Demis Hassabis and John Jumper, told Katie and Brittany. Read more about how open science contributed to their findings, and why commercialization discourages some scientists from sharing, here.

medical records

Why nurses liked an AI bot more than doctors did

An interesting research letter out in JAMA Network Open suggests attitudes toward large language models in health care could vary based on license and professional role. In an assessment of 9 clinics and 166 users spanning from nurses, medical assistants, and advanced practice clinicians to physicians, the majority of nurses — about 92 percent — felt an AI chatbot that drafted responses to patients' portal requests "helped improve efficiency, empathy, and tone," whereas other health care professionals were "less favorable," authors wrote.

Nurses, they found, were more likely than other health care workers to agree that the chatbot reduced the need to forward patient messages to physicians and advance practice clinicians; it's possible then that the messages that were passed on to physicians and APCs were more complex, and therefore harder for the chatbot to process, leading to less utility for those groups. These differences, and a chatbot use rate of about 12 percent, suggests that "LLMs may need to be tuned to recognize who will receive the message (MA, nurse, or physician or APC) and create a reply accordingly," authors wrote.

business forecasts

Payers, employers to boost digital health spend

The Peterson Health Technology Institute — a relatively new non-profit that aims to independently evaluate digital health — polled a few hundred digital health decision makers across payers, health systems and employers about current buying and future plans. And it's largely good news for health technology purveyors: The majority of decision makers plan to ramp up digital health spending, and are interested in risk-based contracts.

Despite ramping up spending over the past few years, buyers appear to be increasingly hawk-eyed about their contracts and ensuring that the digital health services they pay for demonstrate concrete benefits, like making employees measurably healthier or saving money. Almost 60 percent of buyers said their digital health contracts are structured for two years or less. And while employers are more likely to evaluate contracts based on their employees' engagement with digital health offerings, insurers and providers are more concerned about clinical outcomes.

startup studio

Catching up with health admin AI upstart Anterior

We've written at length about health systems' hesitation to deploy AI in clinical settings. But earlier this month, I chatted with physician Abdel Mahmoud, who heads a company taking aim instead at administrative workflow, namely prior authorization. New York-based Anterior (previously known as Co:Helm) recently raised a $20 million Series A round led by New Enterprise Associates. The company sells to payers, but its generative AI tech is primarily used by the nurses and doctors doing medical reviews for health insurance approval, Mahmoud told me. The tool, he said, can highlight relevant parts of medical records and policy documents.

At a time when insurers are coming under fire for using AI to determine health care decisions, Mahmoud emphasized that Anterior doesn't automate decisions. Instead, the tool helps clinician reviewers determine whether certain treatments are medically necessary. "If you want to really dumb down what our product does, it's a better PDF reader," he said.

The young company said it could share some of its health plan customers publicly in coming months. In the meantime, I was intrigued to learn that more than half of its few-dozen staff are clinicians, spanning from AI engineers to utilization management and quality assurance professionals.

And, as far as patients are concerned, Anterior's technology hasn't meaningfully changed the rate of approvals or denials, he said. "If we do our job right it should be something that happens in the background and you're not even aware of it."

More around STAT

Check out more exclusive coverage with a STAT+ subscription

Read premium in-depth biotech, pharma, policy, and life science coverage and analysis with all of our STAT+ articles.

What we're reading

How the wearable neurotech market is shaping up, TechCrunch
MITRE RISE has a new health tech accelerator for founders of color, People of Color in Tech
Broad Institute announces layoffs in IT, software engineering, STAT
Abridge raising $250 million, The Information

Thanks for reading! More on Thursday - Mohana

Mohana Ravindranath is a Bay Area correspondent covering health tech at STAT and has made it her mission to separate out hype from reality in health care.