Lilly offers biotechs its AI models

Eli Lilly today announced it's making its artificial intelligence models that predict the behavior of experimental compounds available to some smaller biotechs at no cost. Notable names like Schrodinger have signed up to use the models already. And Insitro will be contributing its own model to the program.

The new offering, called TuneLab, could be a boon to smaller biotechs that, unlike Lilly, don't have billions of dollars and decades of data to develop their own models. The only catch is that companies will need to contribute their data to help train Lilly's models to participate. In her new story, STAT's Brittany Trang explains how the models can help, and importantly, how Lilly's new program differs from other collective AI development efforts.

Health data

Epic must face monopoly claims, judge rules

A judge on Friday ruled that Epic Systems must face a lawsuit over claims it unfairly used its market power to hurt Particle Health, an upstart company that serves both insurers and health care providers. In an antitrust suit filed last year Particle alleged that Epic used its control over patient data to block Particle from business opportunities serving "payer platforms," a type of software that allows health insurers to retrieve and analyze large amounts of patient data. Epic tried to have the case dismissed, but Particle succeeded in convincing a judge to let it move forward.

As Casey Ross reports, the case is being watched closely for its implications for Epic as well as the broader landscape of businesses arrayed around increasingly valuable data held in electronic health records.

personnel file

FDA digital health leader latest to leave government for industry

The parade of government health IT talent to industry continues. Sonja Fulmer, who until last month was acting director of the Food and Drug Administration's Digital Health Center of Excellence has a new job working on health policy at Mayo Clinic. You'll recall that Micky Tripathi, who was the top health IT regulator during the Biden administration, also headed to Mayo this year. Read my whole story for more context and a rundown of notable departures from government to Oracle.

Also last week, I reported that the FDA's Digital Health Advisory Committee will meet in November. During the committee's inaugural meeting last year it tackled challenges around regulating generative artificial intelligence. After all the shakeups at FDA there was some question as to whether the committee would meet again. The full announcement of the committee meeting is expected this week. Read more here.

Artificial Intelligence

Duke's experiment with AI for ARIA

The FDA in August said it would recommend people taking Alzheimer's drug Leqembi undergo an additional MRI screening for ARIAs, or amyloid-related imaging abnormalities, that are associated with dangerous brain swelling and fluid build-up. Screenings have been recommended since the drug was first approved, and in a sign of how tech might be used to help, researchers at Duke University recently conducted a quality control initiative using Icobrain Aria, AI software from Icometrix that flags potential trouble spots in scans for clinician review. The FDA-cleared software has been previously studied, but its use on scans from patients in the real world is notable because it gives us a sense for the promise of technology in less tightly controlled conditions.

The researchers used Icobrain Aria on 66 scans from 23 patients. In that small sample, the software identified seven cases of ARIA, three of which had not been identified otherwise. On the upside, the researchers reported that negative findings from the AI were reliable, but the software also identified ARIAs when there were none. The researchers note that the technology may help with the growing need for screening as use of anti-amyloid treatments increases, but there are risks associated with adopting AI, for example, that clinicians will default to the judgement of the technology.

fundraising

Strive raises $550 million for tech-powered kidney care

Strive Health announced it has raised $550 million to expand its care model that uses predictive analytics to improve outcomes and lower costs for people with chronic kidney disease.

The details: The funding round includes $300 million of equity financing led by NEA with participation from CVS Health Ventures, CapitalG, Echo Health Ventures, Town Hall Ventures and Redpoint. The company also raised $250 million in debt financing led by Hercules Capital. CEO Chris Riopelle told me the deal values Strive at $1.8 billion.

How it works: Most of Strive's patients are on Medicare. Strive contracts with Medicare Advantage plans and risk-bearing provider organizations to take on chronic kidney disease populations. By analyzing population data, the company identifies people it thinks may need care. It shares in cost savings with customers.

The numbers: Riopelle told me the company currently cares for 145,000 people. He didn't share revenue numbers but said that the company has $5 billion worth of annual care under contract.

How does it save money? For example, Riopelle touted an "optimal starts" program. He said the company can predict with "greater than 95% accuracy" the month someone's kidneys will fail. This allows the patient and company to delay that date if possible through interventions and to start making preparations for the logistics and lifestyle changes associated with dialysis treatment. This also helps avoid a costly and traumatic hospitalization where a patient "wakes up in the ICU and someone tells him to pick a dialysis center," said Riopelle.

What's next? More investment in technology, seeing more patients, and exploring care offerings for conditions Strive frequently sees in kidney disease patients, like congestive heart failure.

More around STAT