Astrata, UPMC's newest spinoff, puts NLP to work for quality improvement
UPMC Enterprises announced this week the launch of Astrata, the newest company incubated in UPMC Enterprises, the Pittsburgh health system's innovation and commercialization wing.
WHY IT MATTERS
Astrata uses advanced analytics and artificial intelligence – natural language processing, especially – to help health insurers to more efficiently analyze unstructured clinical notes.
UPMC says the goal of Astrata is to help the insurance industry better make use of measured quality improvement programs as the push toward value-based care continues.
Data scientists at the new company use cloud-based NLP to build tools that enable payers to gain better understanding from unstructured EHR data – paving the way toward more accurate assessments of quality and population health measurements against the Healthcare Effectiveness Data and Information Set.
For many key HEDIS measures, UPMC notes, healthcare organizations are only able to determine quality rates manually, and only once at the end of the year.
Astrata's technology – built and validated in partnership with the 3.9 million member UPMC Health Plan – can enable year-round monitoring and quality improvement efforts at scale.
"Traditionally, health insurers use claims information to evaluate health care quality against HEDIS measures," said Dr. Rebecca Jacobson, president of Astrata. A wealth of additional information is available in medical charts, but it goes uncaptured because the process to extract it is labor-intensive, expensive and unscalable to entire populations."
THE LARGER TREND
The new spinoff is also piloting a real-time NLP monitoring platform with UPMC and its health plan, officials say. Its initial focus will be on a specific HEDIS measure that identifies older women with bone fractures who've not yet received appropriate imaging and intervention for osteoporosis.
Identifying candidates for screenings such as these is often delayed significantly because it relies on claims data. But Astrata's NLP tools could read clinical notes and identify plan members in need of intervention much earlier.
A UPMC pilot that started before the pandemic showed UPMC Health Plan members signing up for screenings at higher rates after the implementation of Astrata's technology – possibly because it allowed plan representatives to contact members more quickly.
Astrata, which aims increase its workforce by 30% over the coming year, joins a wide array of other spinoffs incubated at UPMC over recent years, including telehealth startup Abridge, LTPAC-focused Curavi (which merged this past year with two other companies to form Arkos Health) and Via Oncology, which was acquired by Elsevier in 2018.
UPMC also continues its work with the Pittsburgh Health Data Alliance, building innovative new machine models for an array of clinical use cases.
ON THE RECORD
"Over the last two years, UPMC Health Plan abstractors found they can work up to 38 times faster with the implementation of Astrata's NLP-assisted tools," said Diane Holder, president and CEO of UPMC Health Plan. "This partnership facilitates a more rapid and accurate flow of thorough, meaningful data between our quality team and our providers."
"As a 40-hospital health system serving millions of patients across three states, UPMC is always looking for ways to work with our providers and UPMC Health Plan to elevate the quality of care our patients receive," added Tami Minnier, chief quality officer at UPMC. "Our work with Astrata to date is only the beginning of understanding how NLP can help us achieve this goal."
The Diverse Roles of AI & ML
Machine learning applications are proliferating and improving, and this month we'll explore the many ways they're being put to work.