Investigating Epic’s AI & an unhealthy digital divide
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.
A glimpse behind Epic’s AI firewall
Epic Systems has mounted an aggressive push in recent years to spur the adoption of algorithms to predict everything from the onset of sepsis to who won’t show up for medical appointments. But in a new investigation, Casey found that health systems are finding these products often provide irrelevant or inaccurate information on seriously ill patients. In the case of sepsis, some hospitals reported favorable outcomes after a lengthy effort to customize the algorithm. But others said the company’s user guides omit crucial information needed to assess reliability and fairness, and that an Epic program that ties financial rewards in part to algorithm adoption is creating the wrong incentives in a context of largely unproven technology. Read the full story here.
Health systems looking for more information on algorithms developed by Epic and other companies or researchers may run into a gaping hole in the scientific literature. In a recent review of more than 4,000 studies of clinical decision support systems, researchers found only 12 had attempted to replicate a previous experiment. “We basically don’t replicate,” Enrico Coiera, director of the Center for Health Informatics at Macquarie University in Sydney and author of the review, told Katie. “Three in 1,000 papers is extremely low, even by the benchmarks of other disciplines that said they were in crisis,” like psychology. “So we have a problem.”
All FHIRed Up
Following on Amazon’s announcement of its newly-repackaged cloud services for health businesses, Google has introduced a new interoperability platform built off its Google Cloud Healthcare API that it’s sharing with limited private partners. The Healthcare Data Engine can transform and structure data from electronic health records using the Fast Healthcare Interoperability Resources (FHIR) format and help providers marry that data with other information, like claims data and clinical trial results. And lest we forget: Microsoft’s Azure cloud is in on health data sharing, too, last week joining its competitors in supporting NIH’s Science and Technology Research Infrastructure for Discovery, Experimentation, and Sustainability Initiative.
The digital divide is physical
The Agency for Healthcare Research and Quality released a new tool that makes it easy to see why Internet access is considered a social determinant of health. The data, which spans 2014 to 2018, highlights how access is unequally distributed across the United States — not just in terms of whether broadband is available, but how commonly households have access to Internet-connected devices. In the lightest blue counties, between 22% and 62% of homes have no computing devices.
AlphaFold goes deep and broad
Last year, Alphabet-owned AI outfit DeepMind shocked structural biologists by cracking open the problem of protein folding with its deep learning model, AlphaFold2. Seven months later, the company released the methods and code behind its model — and 350,000 of its protein structure predictions, which are now publicly available. That’s nearly double the number of experimentally-determined structures in the Protein Data Bank, and includes nearly every protein expressed in the human body, as the company describes in Nature. Read more in Katie’s story.
- The patient data aggregator Komodo Health said it will acquire Breakaway Partners, which collects and analyzes pharmaceutical sales and marketing data for life sciences companies. Komodo said the deal will help it create a comprehensive view of real-world outcomes and costs.
- Bon Secours Mercy Healthis investing in the AI startup Trilliant Health as part of a broader push into predictive analytics. Trilliant uses machine learning to help providers forecast growth opportunities. Bon Secours’ chief analytics officer, Deepesh Chandra, will now sit on its board of directors.
- Data infrastructure and interoperability don’t usually put people on the edge of their seats. But the VC firm Rock Health said investors are taking a keen interest in startups promising to improve those services, pumping in $836 million so far this year. That’s nearly triple the $293 million invested in all of 2021.