In 2021, 14 large healthcare organizations banded together to build a large-scale data platform company called Truveta. That membership has since grown to 25 health systems. Truveta executives spoke with Healthcare Innovation this week about the roll-out of Truveta Studio, an analytics solution built on top of an exabyte of health data.
The executives said this is the first solution to make massive streams of daily clinical data useful for analytics through the integration of AI-powered natural language processing and de-identification.
“Our intent has always been to structure, normalize, and de-identify this data and make insights possible on patient care and outcomes,” said Truveta CEO Terry Myerson. “There's nothing on the market to study patient care outcomes. There are endless products to study the cost of care, availability of care, utilization of your operating rooms, or to help understand which physicians are prescribing which drugs. We have data from 25 health systems responsible for 16 percent of all clinical care in the United States. It's very exciting to be transitioning from the mode of discussing what is needed to actually introducing it.”
In addition to the 25 health systems contributing data, life science firm Pfizer and medical device company Boston Scientific are early adopters.
One of the innovations is what the company calls Truveta Prose, which makes medical concepts computable, explained Michael Simonov, M.D., Truveta’s vice president of clinical informatics. These concepts can combine events from a patient’s longitudinal history, including diagnoses, labs, procedures, medications, vaccinations, devices, or any concept found within a clinical note. The company said that like Google searches the Internet, researchers can search Truveta data for any Prose-defined population within a few seconds.
Simonov discussed the value of the platform in the context of his career as an informaticist at Yale New Haven Health. “The major pain points I think Truveta is resolving were my pain points as a researcher,” he said. When the COVID pandemic emerged, if he didn’t know whether to intubate patients early or late, it would take three months just to get the data. “It's really like pulling teeth at health systems to get someone to provide you the data.” He added. “Then once I have the data, I have to clean it and that's another several months process. Then, on top of that, how am I going to define COVID over these data? How am I going to define heart failure over this data?”
He said Truveta has robust normalization processes across the EHR data coming in from across the country. Unstructured content with the medical record is mapped to clinical ontology standards, such as LOINC for lab tests and GUDID for medical devices.
“We can clean that data for the researcher,” Simonov said. “You can get started right away and you can be confident that the diabetes definition is the diabetes definition across the country. This is not just Providence’s definition of diabetes, but it's really normalized across all of our data.”
With Truveta Studio, he said, within seconds he can search across the entire United States and generate populations using concepts in Prose that are transparent and can be leveraged by researchers.
Truveta Studio includes an integrated Jupyter notebook atop a serverless SQL experience, pre-installed with medical statistics and visualization libraries, with full support for R and Python. The company said the integrated analytics make it hassle-free for distributed research and data science teams to study Truveta Data populations within Studio – and for the underlying statistics to be shared transparently.
Myerson said Truveta’s unique membership and governance structure is actually beneficial. “If you're a company that’s owned by a private equity organization or venture capital organization, ethics is just a word on a marketing website. It's about squeezing every penny out of every corner of the business,” he said. “I think it's actually vital to what we're doing to have the health system governance. It's not like this optional overhead; it's actually very intrinsic to what we're doing. It takes an investment of effort, but right now I think it's working incredibly well. We're being quite nimble.”
