Servers and storage are a primary focus for one hospital’s support upgrades.
For Partners HealthCare’s Brent Richter, the growth and increased use of his organization’s Integrated Data Environment for Analytics (IDEA) — a data lake storage platform that allows researchers to request and access information in a cataloglike fashion — represents a new frontier for the Boston-based nonprofit healthcare system, and for the health industry overall.
The goal of IDEA, conceptualized in 2014 and supported by Dell EMC infrastructure technology, is to speed the path of internally developed research within Partners to clinically relevant settings. As the director of enterprise research infrastructure and services at Partners, Richter says flexibility will prove crucial to the platform’s continued evolution.
HealthTech Managing Editor Dan Bowman chatted with Richter about IDEA’s development, its impact on research efforts to date and plans to expand.
RICHTER: It’s a learning platform. We began conceptualizing it three years ago. For the first year, we effectively took that time to understand the architecture, the technologies, the availability of vendor solutions. We put that together in year two.
EMC was integrating technology and companies around data science — what it coined as “data lakes” — and had a first iteration with us. Since then, EMC has come out with a federation business data lake and an analytics insight module. For us, a lot of that was informed by what our users — investigators and clinical innovation folks — were bringing to the platform; pools they wanted to leverage; their use cases, particularly during both iterations. We wanted to create a flexible platform for these communities and their needs. It’s about providing a suite of tools and a set of frameworks in both development and analytics, with underlying storage to satisfy multiple needs for our internal groups.
RICHTER: We built the system from the get-go with security and privacy as key requirements. While some investigators on the basic research side might say that their data is not patient data or not identifiable, there’s always a risk that it could be de-identified. And while those researchers might not even work with patient data right now, next year that might not be the case.
RICHTER: Overall, there’s a lot of excitement. Some groups are doing a lot of development and are now in the midst of transforming their clinical practices and including the decision support. Others use it as a sandbox to determine how it could be useful.
There are several very large projects or initiatives getting started in the hospitals that will leverage data science. One area that’s received a lot of funding is cardiology at Brigham and Women’s Hospital, where they’re trying to leverage data science and techniques to treat heart disease much earlier in a patient’s life. Researchers want to identify patients at age 25 so they can start pretreating them, hopefully preventing them from showing up in the hospital 35 years later.
RICHTER: For the next year, we’re concentrating on training and providing intense sessions where people can bring their data and work with a couple of data scientists from our partners. We continue to evolve the platform in terms of the tools needed, and algorithms and libraries, as users take advantage of the system and tell us what they need.
We are building this learning environment to provide a baseline and platform for a new frontier. As the frontier evolves, so will the platform. That’s our ultimate mission.
RICHTER: The greatest issue, and it’s ongoing, is training for our community around the tools being used: the data ingest tools, manipulation tools, some of the analytics tools.
Another training issue has been understanding how to do data science versus statistics, which a lot of this research has been steeped in for many years. Some statistics are involved, but our thinking is more along the lines of a complete shift out of statistics and into predictive analytics, which is a different mindset. Reducing the learning curve for very smart, technical people, helping them to think more like a developer rather than a statistician, has been a great hurdle.
With these new technologies, it requires a mind shift. It requires new knowledge.
RITCHER: We’re working right now with 22 individual groups that span the Partners HealthCare system, which includes our academic institutions and our community hospitals. Studies occur worldwide, and there are a lot of consortiums with other academic medical centers across the country to collect data, to collect patients for studies. Some groups work internationally as well, in Rwanda, Spain, Haiti and elsewhere.