The healthcare industry is relatively data-rich. Yet due to technological impediments, much of the data generated...
by the electronic health record (EHR) systems in hospitals isn't stored and even less is analyzed, leaving many healthcare providers incapable of extracting real value from the mountains of data they produce every day.
"We're still bogged down in an EHR world that isn't quite working for us," Alan Weiss, chief medical information officer for ambulatory services at Houston-based Memorial Hermann Healthcare System, said during a panel discussion on healthcare data analytics at the Big Data in Healthcare Summit in Boston last week.
It wasn't too long ago that hospitals lacked even basic electronic record keeping software. Then in 2009, the federal American Recovery and Reinvestment Act created an incentive program known as meaningful use to get providers to adopt electronic health records. But a lot of EHR systems were adopted in a rush -- eligibility for the incentives expired in 2014 -- kicking in the penalty phase of the program. On the other hand, they continue in many cases to be based on decades-old proprietary technology. Now these systems are generating enormous troves of data that could be analyzed to optimize care delivery; but getting the data out of them to do analytics is easier said than done.
Weiss and his team at Memorial Hermann are trying to work around that problem by launching a number of initiatives to make patient care more data-driven. For example, they created a model in the provider's EHR system that scores each patient's likelihood of being readmitted and then displays the score at the top of individual records for physicians to see. The team also developed a scoring tool to evaluate patients' risk of developing sepsis, a potentially deadly complication from infections, during their hospital stays.
But because EHRs store data in unique formats and use a variety of terminologies for medical diagnoses, embedding these kinds of features into the EHR has been impossible on a broader scale at Memorial Hermann, which is made up of a collection of hospitals and medical offices. As it happens, some of these facilities use different EHR systems. "If only this were easy," Weiss said. "If only every EHR stored data in the exact same way with the same fields."
Industry leader, analytics roadblock?
It's hard to talk about the difficulties of getting data out of an EHR system for doing analytics in healthcare organizations without mentioning Epic Systems Corp. The Verona, Wis.-based EHR vendor is by far the industry leader in hospital implementations. A recent survey showed that Epic has roughly twice as many implementations as its next closest competitor, Cerner Corp.
DuWayne Willettchief medical informatics officer, University of Texas Southwestern Medical Center
But Epic is simultaneously praised for its track record as an established, solid vendor and criticized for its lack of technical innovation. The main drawback of Epic's EHR system, critics say, is that it's built around the Massachusetts General Hospital Utility Multi-Programming System, or MUMPS. The programming language was developed in the 1960s and isn't widely taught to programmers today. As a result, hospitals wanting to develop a tool to manipulate data in the Epic system or build applications on top of it often have a hard time finding developers with the right skills.
Instead of learning the programming language, most developers simply get training in one specific module of the Epic EHR, said DuWayne Willett, chief medical informatics officer at the University of Texas Southwestern Medical Center in Dallas. The system is broken down into modules that serve specific areas, like emergency departments, radiology and intensive care. But, Willett said, since developers know modules instead of the underlying programming language, people who work with data often don't know how or where it's stored. That makes it hard to develop analytics applications that span all of a hospital's services or combine data from different departments.
Willett added he's currently looking for developers who have broader skills. He and his team are also seeking to do more with metadata, which is typically generated but not stored by the EHR system, as well as mobile and patient-generated data sources. Those data sources could be easier to manipulate than medical records as well as useful for predicting patient behavior and recommending targeted interventions.
"We need to keep up on the data side," Willett said. "To be in this game, we need to get a lot better at predictive analytics."
Beating back analytics complexity
If doing healthcare data analytics on information from one EHR system is difficult, it pales in comparison to trying to work with data from multiple systems. Hospitals and health systems are going through a period of consolidation, and different hospitals bring their own EHR systems when they come together through mergers and acquisitions. This adds an extra layer of complexity to analytics efforts because various EHR systems are notorious for storing data in different formats that aren't compatible with each other.
That's a problem Stephen Allegretto, vice president of financial planning at Yale New Haven Health System, has tried to manage in recent years. The health system is comprised of three hospitals and a network of medical offices in Connecticut. The various providers once used a collection of EHR systems, but the organization has since consolidated all of them onto an Epic installation.
Allegretto said that has made it much easier to develop analytics applications to do things like predict adverse events, understand the drivers of high treatment costs and see how various treatments affect patients' lengths of stay. There have been some wins, but he added that implementing the analytics programs across several hospitals and doctors' offices still presents a great deal of complexity that makes further progress challenging.
"We're striving to have a single source of truth," Allegretto said. "But when you have a complicated structure that comes together like this, we create that truth one at a time. I'm not too sure we're there yet."
How to develop a plan for effective healthcare data analytics
Analytics and quality improvement teams must get along to move ahead
Get creative when thinking about implementing healthcare analytics processes