At the intersection of big data and cloud computing: Analytics

The provider has been making big use of the cloud to manage its huge and ever-growing piles of data
By Mike Miliard
10:11 AM
Share

Two of today’s most prominent trends in healthcare – big data and cloud computing – are fated to intertwine in ways that will bolster health data analytics.

Indeed, it’s already happening at UPMC. The provider has been making big use of the cloud to manage its huge and ever-growing piles of data as part of a long-term partnership with IBM to completely revamp its storage and analytics of huge volumes of patient and research data, all leading up to the medical center’s $100 million personalized care initiative, announced Oct. 1.

UPMC has a deployed a dynamic virtualized infrastructure that shortened information backup times by 20 percent and recovery times by half, officials said. IBM helped with storage and server virtualization, enabling the network to be more flexible to accommodate exponential data growth at UPMC’s 20 hospitals and 400 outpatient locations. A key part of this is a private cloud that supports mission-critical applications.

Back when UPMC made the first moves toward virtualization in 2005, “We probably had a little more than a million unique patient records in our EHR and lab systems,” said Christian Carmody, vice president of the Information Services Division at UPMC. “Where we sit today, we’re at 7.3 million unique patient records.”

[This article was drawn from "Navigating the Cloud," a free e-supplement containing this and other must-read information about cloud computing in healthcare.]

Big data, to be certain. But it’s not just about volume. “The richness of that data has ballooned as well,” Carmody added. “We’ve experienced a tremendous amount of organic growth as we’ve merged with different hospitals over that timeframe.”

And the need for data storage “just keeps going up,” he said. “We’re at probably four petabytes of data today, and we anticipate that at the end of this agreement with IBM, this next four years, that we’re going to grow to 19 petabytes.”

The good news? “We’ve turned the corner as far as trying to get physicians to adopt technology,” Carmody said. The challenge? “They are now reliant upon it, and they want more.” And that means more data.

Luckily, that’s something UPMC has been anticipating for nigh on a decade. We had the “foresight and vision ... to virtualize our environment,” Carmody said. “We were dealing with issues of organic growth and running out of datacenter space back when we were doing the first deal with IBM.

“Now we’re on the cusp of this concept of big data, which to us is this 10-plus years of electronic health record data, the financial data, and the upcoming use and inclusion of genomic data into our environment,” Carmody added. “The combination of all those different elements, being run through a superior analytics program, is what we’re challenged with as we move forward.”

Big data and cloud considerations

Most of the cloud solutions being adopted in healthcare are private clouds, but hybrid clouds are also becoming more accepted, depending on the application and usage pattern, said Greg Caressi, senior vice president, healthcare and life sciences, at Frost & Sullivan. Public cloud solutions are deployed in healthcare, but are more rare, he added.

Infrastructure-as-a-service (IaaS) technology “will become increasingly important as the data explosion in healthcare continues to stretch the capabilities and budgets of organizations to manage these petabytes of data,” Caressi explained, “and as we need to collaborate and share access to healthcare data among a wider range of stakeholders.”

Dennis Schmuland, MD, chief health strategy officer at Microsoft’s U.S. health and life sciences division says too many hospitals “sell themselves short by focusing only on the cost-savings” of the cloud. Instead, they should be looking at “how the cloud can focus on what they do best, rather than keeping the lights on.”

That could mean collaborative analytics for big data, he said – pointing specifically to “three areas that are exciting.”

One is improving patient safety. “A lot of those data streams can be aggregated and analyzed to signal early warnings from oximeters, ventilators, blood pressure, body temperature, heart rate,” Schmuland explained. “Those early warnings can really predict adverse events well in advance of the condition being serious: detect infections earlier, detect cardiac arrest or a blood clot before it occurs.”

A second area is reducing practice variability. “Many physicians do not know how their practice compares with others. Big data could quickly enable every physician to see how they compare with best practices – and even help provide them with feedback over time, moving toward best practices and reducing variability and improving their adherence to guidelines,” he said.

The third has to do with comparative effectiveness research. “The challenge with CER is that it takes so much time to do comparative studies: You compare one drug to another, and report and publish results, Schmuland said. “Big data could actually do a lot of comparative effectiveness research based on data that already exists. A lot of this data is currently lost in the system.”

Three fine examples, sure. But ultimately there’s not much limit to what can be accomplished with the affordability and collaborative opportunities enabled by the cloud.

Indeed, it seems we find ourselves at a pivotal moment. Taken together the cloud and big data – those two purported buzzwords – are setting the stage for huge changes in the quality and cost- effectiveness of care. Which is precisely the situation in which UPMC finds itself.

Back at UPMC

There are more challenges – albeit fun and exciting ones – ahead.

“Then you hear about things like ‘The Internet of Things’” – discrete physical objects, when represented in a virtual Web-like structure – “where all these different devices will be connecting to the Internet, and we’ll have this additional abundance of data coming in,” UPMC’s Carmody said. “We have a unique opportunity ahead of us as far as how we effectively manage, store and enable that data to be transformed into information and knowledge for our clinicians.”

Eight years ago, even with nearly one-eighth as many patient records to store as there are now, there was already an “explosion of physical servers just taking over our datacenter floor space,” he said. “We were looking at having to build or lease floor space at another datacenter just to deal with that organic growth.”

The decision to virtualize UPMC’s infrastructure was appealing, but the process was anything but easy. “It definitely took a lot of effort, a lot of time – three or four years to go through the transformation to where we were fully up and running on our own little private cloud,” Carmody said.

The biggest obstacle was standardization. “We needed to consolidate the varieties of platforms and solutions we’d previously been supporting for all the different application vendors, and really work with them and reach out to them to push us toward where we could reduce the number of operating systems and databases to a manageable few that we could have confidence in as we moved forward to a virtualized environment.”

Calling that “a tremendous challenge and a lot of work,” Carmody added that the benefits “have been overwhelming.” Once threatened with having to spillover into another datacenter, “We’ve reduced our physical footprint, there’s actually room to grow,” Carmody said, “which positions us well for this next expansion into big data.”

Today, with the data deluge gathering strength, “We are well-positioned to be flexible, to be scalable, and to grow with the computing and storage needs that the enterprise analytics program is going to place on the infrastructure,” he said. “As our organization grows and introduces and integrates the genomics data with the electronic health record data with the financial and operational data – and then crank through those different algorithms and infuse that into the clinician’s workflow, the infrastructure is going to be heavily relied-upon. We’ve very confident and feel very capable.”

UPMC’s leading-edge datacenter, all in the service of a transformative approach to personalized medicine, is a great example of what a world-class institution with huge financial wherewithal can do with the cloud.

But providers of every size should be taking a careful look at cloud technology as data volumes grow inexorably.

Indeed, far from being intimidating to small hospitals, cloud technology should be looked upon as an opportunity as it will “likely be an equalizer between large hospitals with sophisticated IT capabilities and big budgets and smaller facilities that will not be able to keep up without ‘HIT as a service’ solutions,” Frost & Sullivan’s Caressi said. “Both will use cloud solutions, but smaller organizations will be dependent on these solutions for their survival and to keep up with the big hospital systems."