Analytics project slashes sepsis deaths
Penn Medicine puts predictive analytics to work
The folks at Penn Medicine know a little something about putting data analytics to work. After identifying three years ago that their sepsis mortality rates were higher than expected, they set out to do something about it by harnessing predictive analytics. And the results? They're impressive.
After setting its sights on leveraging clinical systems data and building out an early warning system to make it work, the three-hospital health system saw promising numbers, noted Christine VanZandbergen, associate chief information officer at Penn Medicine, who helped spearhead the analytics initiative.
"One of the reasons we chose sepsis was, in terms of altering the care pathway, there is strong evidence to suggest that if you intervene early and appropriately, you can actually decrease mortality rate," said VanZandbergen, who will be speaking at the HIMSS/Healthcare IT News Big Data & Healthcare Analytics Forum Nov. 20 in Boston.
[See also: Clinical data analytics next big thing.]
And decrease the mortality rate they did. After building an algorithm for the early warning system that identified patients likely to benefit from early intervention, Penn Medicine watched its sepsis mortality rates fall 4 percent overall, from 17 percent at baseline to 13 percent today.
Big improvements in clinical care measures certainly didn't hurt. The team saw notable progress with antibiotic administration rates, which increased 10 percent – from 13 percent at baseline to 23 percent post implementation; IV fluid bolus also increased from 15 percent at baseline to 26 percent today.
Septicemia is currently responsible for the deaths of 36,000 people each year, according to data from the Centers for Disease Control and Prevention, making it the No. 11 leading cause of death in the U.S. Officials estimate average sepsis mortality rates to be more than 16 percent nationwide. In addition to the human death toll, the disease also costs the industry a pretty penny financially. It persists as the No. 1 most expensive hospital condition, costing more than $20 billion annually.
[See also: Healthcare analytics enters new age.]
Penn Medicine's project targeting sepsis mortality rates, as VanZandbergen pointed out, had both its straightforward tasks and its difficulties. For one, "the availability of the data was also relatively easy over the trajectory because we have been up on our clinical EMR for a period of time that allowed us to ensure that our data was reliable," she added.
But, although the clinical data piece was, for the most part, uncomplicated, and the algorithm proved "pretty straightforward," she and her team's biggest challenge pertained to "designing the operational response." For this part, VanZandbergen, who will detail Penn Medicine's case study at the data analytics forum, said forging relationships with both clinical and operations leadership proved integral.
In terms of what's next, based on the success of this initiative, VanZandbergen sees more data analytics in her future – clinical EMR data, specifically. Working with data scientists and Penn Medicine's data analytics center, officials hope to further leverage EMR data as it relates to quality outcomes, far beyond sepsis.