Drawing on 10 years of EHR data

The project "started just by thinking outside the box with respect to how is it we can use analytics," Metcalf said. "Renown Health has been sitting on 10 years of EHR data, and they had not looked at it in total. They did not have the luxury of being able to look backward and see if there were any trends or things that could be identified that might improve outcomes."

Jim Metcalf

Researchers are in the early stages of using SAS Analytics to identify trends that could lead to improving population health. "Breakthroughs in statistical and machine learning methods, and even artificial intelligence, are making possible what could only be imagined in healthcare and the life sciences five or 10 years ago," said Saurabh Gupta, director of advanced analytics and artificial intelligence product management at SAS.

Using the SAS platform, Gupta said, Renown IHI "is able to leverage these advances along with the powerful data management and visualization capabilities to examine vast amounts of clinical, environmental and socioeconomic data, in combination. Better understanding the complex interplay between these factors and related effects on population health holds the key to better diagnostics, treatments and potentially even cures."

Tackling 'huge data munging project'

I can use [SAS Analytics] to pretty much do anything I want to do with data, and this is a huge data munging project.

Metcalf calls SAS his Swiss army knife. "I can use it to pretty much do anything I want to do with data, and this is a huge data munging project," he said. "The electronic healthcare records alone are 4 TB in size, 18,000 tables, and the SAS platform is really the only platform that I'm aware of that is able to slice and dice and apply advanced analytics to data that size."

A pilot project, which involved collecting DNA samples from 10,000 Nevadans, served as proof of concept "to ensure we could actually do this, and we have the genetic results and we're beginning to do genome-wide association studies with the data."

A second phase, which began in March, will add an additional 40,000 Nevadans to the project, making a total of 50,000. "That," Metcalf said, "is the size of study that we need to be able to get the statistical power that we require in order to look at trends."

To embark on this project, Renown IHI had to comply with HIPAA, which has provisions for affiliates of providers to become a business associate. As an affiliate of Renown Health, "we went through the business associate process, which requires that you go through a risk assessment," Metcalf said.

Complying with HIPAA regulations

The vendor risk assessment (VRA) process includes 1,800 questions, Metcalf said, "requiring proof that you have adequate physical protections in place, adequate logging and auditing mechanisms in place, that your IT shop has a robust ability to respond, deal with the things that go along with keeping a perimeter secure."

After completing the VRA process, Metcalf said, "we constructed a HIPAA moat, if you will, which is essentially a physical containment area for the data we are using. It's sealed off from the outside web. We have two-factor authentication, the servers sit behind four different layers of physical access. This is something we're incredibly careful about. It is one of our topmost concerns -- to ensure that as a business associate we are the gold standard of how these data are treated."

Metcalf said Renown IHI has a dedicated staff "whose role is to ensure the integrity and security of what we've constructed." Other staff members are data scientists who work with the project's complex data. "The schema is unlike anything any of us has seen," Metcalf said, "and because of that complexity, we meet with the hospital once a week to ensure we're understanding the schema in the same way they're understanding the schema."

Metcalf said it's difficult to find people who can be statisticians, as well as data scientists. "So we have some very specific requirements of the people who interact with these data. ... A lot of statisticians want their data to be brought to them. Data scientists, in my view, are statisticians willing to get their hands dirty, who can manipulate data and get it to where it needs to be for them to apply the algorithm that will yield the insight they are seeking."

Improving population health through the project model

Metcalf believes a program such as the Healthy Nevada Project, in which "a local provider uses an external group to help them with their analytics," can help other healthcare organizations identify trends that might lead to improving population health. "Our vision is that we can cookie cutter this model and use it in hospitals in other areas."

"We are learning an awful lot," Metcalf said. "Ultimately, we'd like to be able to take these learnings and incorporate them into some kind of a best practices organization where we can approach other healthcare systems and say 'Hey, look, we've been through this, we know how to do it, we've learned through the crucible of pain everything we know not to do the next time.'"