Top Ten Tech Trends 2018: Health System Transformation Requires Rethinking Data Infrastructure
Editor’s Note: Throughout the next week, in our annual Top Ten Tech Trends package, we will share with you, our readers, stories on how we gauge the U.S. healthcare system’s forward evolution into the future.
Once health systems got most of their data into electronic and structured format, the next logical step was the development of data warehouses and analytics platforms, and capabilities to learn from all that data. Today those efforts are still in their adolescence in many organizations, but investments in warehousing and analytics continues, with a goal of identifying and solving institutional inefficiencies and clinical variance.
Handling all this new data also requires CIOs to rethink their data infrastructure strategy with an eye on shrinking data centers, hybrid cloud and public cloud solutions.
Seeing all these analytics needs coming down the pike, Minneapolis-based Allina Health entered into a tight partnership with vendor Health Catalyst in 2015, in which Allina took an investment stake in the company and some Allina employees became Health Catalyst employees. “It allowed us to start to accelerate our capabilities in the data space, in terms of warehouse, analytics, toolsets and talent to maximize our potential,” says Jonathan Shoemaker, Allina’s senior vice president and chief information officer.
Among other things, he says, Health Catalyst developed data transformation tools on the back end that can take raw data sets from more than 70 different sources and bring them into the data warehouse in an efficient manner. “The cycle time on that used to be fairly long,” Shoemaker recalls.
Now the health system is working to get better at analytics and change management. “You are only as good as your ability to execute on what the data says,” Shoemaker says. “We are thinking about how we can mature to the next level on our use of data at the physician level. We are starting to be able to put the power of data into the hands of individual providers.”
Jonathan Shoemaker
Besides the traditional enterprise data warehouse, some health systems are starting to create data “lakes,” which may contain new data types not found in traditional hospital clinical and financial data sets. “With a data warehouse you have to know ahead of time and plan out what the data structure is going to be so you can insert it,” explains Michael Gil, chief technology officer with Temecula, Calif.-based consulting firm Axene Health Partners. “With a data lake, you drop the data in in whatever file format it is in. It might be environmental or geographical data. You can start to pull that together without a pre-defined schema.”
Vijay Venkatesan, former senior vice president and chief data officer at Seattle-based Providence St. Joseph Health, told Healthcare Informatics in June 2018, when he was still working at the health system, that Providence created a data lake where it invited its other data asset owners to contribute their data into the lake, in exchange for data they want or don’t have access to. “We created a culture of convergence, to bring data in one place and share each other’s information so that the collective organizations benefit.”
Vijay Venkatesan
The second step, he says, involves creating a harmonizing data layer. “Think about it as your iTunes data catalog, where your albums in iTunes are categorized by rock, pop, alternative, etc. It’s the same idea. Now that we have data in one place, how do we create albums from the data?”
Shoemaker says that bringing in social determinant data is a key priority for Allina this year. “Consumer insights is another one we have put energy into,” he explains. “Hospital-centric data is necessary, but thinking of it from a retail experience level and understanding your role in the market you are trying to serve is relatively new to healthcare. We are trying to build that insights practice.” Another strong focus is on patient-reported outcomes, he adds.
Cynthia Burghard, research director of IDC Health Insights in Framingham, Mass., says CIOs tend to have a more short-term focus and are looking for analytics solutions that solve specific problems. If a CIO in a health system with tiny margins asks to invest $5 million for the next five years to build a data lake, they might be out of a job, she says. “But if that same CIO hears that the health system is getting penalized heavily for its readmission rates and can find something to help solve that problem analytically, that is a much more compelling argument.” In other words, CIOs are less focused on pristine infrastructure to do analytics than on getting help to solve today’s problems.
Of course, there is a personnel and talent management side to analytics as well. During a recent webinar talk, Shakeeb Akhter, director of the enterprise data warehouse at Northwestern University’s Feinberg School of Medicine, described what has worked well there. “When you’re first starting down the data warehousing and analytics path, it is important to find people who have a blend of experience with clinical operations, with the EHR workflows, and some business intelligence background,” he says. “We’ve hired people who have some sort of clinical background or have been application analysts. We made them data architects and started training them to think about things on a much larger scale from a data warehousing perspective.”
The proliferation of data, including large data sets involving genomics and research, is forcing CIOs to reconsider the traditional data center model and focus more on hybrid cloud and public cloud offerings from vendors such as Amazon and Microsoft. A recent blog post by David Cappuccio, a Gartner research vice president, was titled “The Data Center Is Dead.” He wrote that “the role of the traditional data center is being relegated to that of a legacy holding area, dedicated to very specific services than cannot be supported elsewhere, or supporting those systems that are most economically efficient on-premises.” He wrote that Gartner estimates that by 2025, 80 percent of enterprises will have shut down their traditional data center, versus 10 percent today.
In healthcare, however, one challenge is that many departmental systems, such as lab systems, were developed in older architectures. “You can’t just press a button and throw it up into the cloud,” IDC Health Insights Burghard says. “The application vendor has to modernize the software. Ripping and replacing a pharmacy system is not top of mind for CIOs these days. If they can limp along with it, they probably
are going to.”
Allina serves as a good example. It built a data center six years ago, which made business sense at the time, Shoemaker says. When it needed a second new data center, it chose a co-location design. “So far, we haven’t moved away from data center management, but we are always looking at when cloud-based management of our data stores makes more sense than on-premises solutions.” It requires regular re-assessment, he adds, “because the world does change every 18 months or so.”