Healthcare Analytics, Population Health Management, Healthcare Big Data

Tools & Strategies News

For Predictive Analytics, Not All Data is Created Equal

By Jennifer Bresnick

- All different types of data may ultimately boil down to a string of zeroes and ones, but health IT and clinical analytics systems must work with data in its more complex forms. With layers of code and unexpected proprietary quirks that make interoperability and large-scale number-crunching a difficult proposition, healthcare organizations have a long journey to make before interoperability becomes a reality.  While predictive analytics that can accurately and effectively warn providers of expensive or dangerous events is the goal of most healthcare systems, why does gathering and leveraging the right data at the right time continue to be a challenge?

“People will say they’re doing predictive analytics, but it’s important to know where those analytics are derived from,” says Justin Lanning, Senior Vice President and Managing Director of Analytics at Xerox Healthcare Provider Services.  “Many are actually using retrospective claims data to predict the future while others are using more recent data sources and others still are using real-time data.”

Each of these analytics strategies has its merits, and organizations often have to choose their methods based on the attached price tag.  But real-time data analytics, based on a constant feed of information from EHRs, lab tests, bedside monitors, and other sources, can provide the deepest insight into both a hospital’s processes as well as an individual patient’s status, risk, and outcomes.

“Through some of the new government regulatory reimbursement programs, hospitals today don’t know for about 18 months to two years past their performance how their current and past performance is truly going to impact their reimbursement,” Lanning explains.  “So even though someone could say, ‘Hey, we really improved this area of metrics, and we feel really good about it,’ it could be that every other hospital in the nation did the exact same thing, too.  So you really didn’t actually gain any ground on that, and possibly you even lost some ground.”

“So it helps you be more targeted month to month if you use real-time data as part of these analytics.  It’s much better than operating for a year, improving all your metrics, but having absolutely no idea where that’s going to land you.  That has major financial implications and quality implications.”

READ MORE: Google Using FHIR, Deep Learning for Healthcare Predictive Analytics

“And then to contrast that, there’s the ability to use real-time data to know when certain vitals have changed and correlate that with lab values and the history of a particular patient so we can see if that patient is highly likely to be heading towards a septic or other type of event, and alert the provider ahead of time,” he added.

Developing a stream of real-time data that works together to create these rich portraits of performance is not an easy task, however.  As the industry struggles to adopt data standards that encourage health information sharing and break down data silos, providers must contend with the fact that not all systems wish to talk to each other just yet.  Interoperability may be the zeitgeist of the modern health IT industry, but market forces and a fragmented patchwork of old and new technologies make real-time predictive analytics a significant challenge.

“We’ve found that there are a number of finicky characteristics of each of the EHRs and how they treat data,” Lanning says.  “It’s one thing to say, ‘Hey, I wish you were all interoperable.’  That’s a big thing right now.  But interoperability also needs to come with an understanding of how these different things are built.”

“If one EHR sends us a piece of data, we might know that the way they date and timestamp their data is different than this other EHR, and another one and another one.  Maybe it’s stamped differently based on when someone opens a window as opposed to when they click ‘save.’  When you’ve got time-driven analytics, you have to know all that stuff, too.  So it’s one thing to say that if we had real smooth interoperability, that all the data would come over, but it requires that attention to these little details you might not think about.”

Those details may be one reason why healthcare leaders are frustrated by the inability of analytics to evolve quickly enough to answer the pressing questions of the day.  Understanding these differences in underlying data structure is crucial for healthcare organizations looking to harmonize multiple systems to create the possibility of robust data analytics.  As interoperability projects move forward, however, and vendors continue to adopt wider standards that encourage data exchange, the health IT community may eventually come to a consensus about how to build predictive analytics into products and services provide real-time insights to the bedside.


Join 25,000 of your peers

Register for free to get access to all our articles, webcasts, white papers and exclusive interviews.

Our privacy policy

no, thanks

Continue to site...