The University of Iowa Hospitals and Clinics has seen a 74 percent drop in infections through a predictive modeling program targeted at surgical sites.

John Cromwell, MD, clinical associate professor of surgery at Iowa, said his team used a combination of process, technology and the human factor to leverage live surgical data to inform their decisions.

"We basically developed a data warehouse to use at the back end for all of our modeling. That was the first thing we did right," Cromwell said.

See all of our HIMSS16 previews

"Going beyond that is a prescriptive analytics for how to manage wounds at high risk of developing SSIs."

Cromwell will share insights gleaned from Iowa’s work during the “Real-Time Data, Predictive Analytics Can Reduce Infections,” session at HIMSS16. "We started very small, writing R code and running models off of desktop computers.”

As the project grew, Iowa built out infrastructure to include enterprise tools like Dell’s Statistica, which also demanded new investments in personnel and IT resources.

Indeed, predictive analytics tools are fairly mature at this point, Cromwell said, but access to data science practitioners is a barrier. “We have lots of informaticists who are doing research, but they're not practitioners from the standpoint of being able to just bring them in and say get going on this project,” he said.

Having the right team in place for these sorts of detailed analytics projects is key, he said.

[Also: 11 essential quotes from notable HIMSS keynotes]

"You really need someone who's a content expert – a physician or provider who understands the clinical problem quite well,” Cromwell said. “Someone who understands the whole infrastructure on the back end, and a data scientist in the middle who can translate those two.”

Asked for some tips that might be applicable to other providers looking to try similar projects, Cromwell said one important first step is to "pick a high-visibility problem, but start on a small scope. Don't try to change your entire organization' infrastructure in order to begin using predictive analytics. We wanted to incrementally increase infrastructure to demonstrate the ROI of doing this."

So far, that ROI has been hard to argue with: Begun back in 2012, the first two years of the project saw SSIs reduced by 58 percent; after the third year, they'd been reduced by 74 percent.

[Like Healthcare IT News on Facebook]

"We still don't have all the surgical departments even close to all using this, we chose to focus on divisions where the opportunity was the greatest," Cromwell explained. "We chose the top two or three areas, and it's had a huge impact just in those. As far as I know, this is the first systematic approach to really make clinical decisions. It's gone beyond research now, and this is our clinical practice."

Surgical site infections, after all, are "probably the most costly complication, the most common healthcare associated infection, the leading source of readmission for patients to the hospital," said Cromwell. "So tackling this is a big goal for us."

Twitter: @MikeMiliardHITN

This story is part of our ongoing coverage of the HIMSS16 conference. Follow our live blog for real-time updates, and visit Destination HIMSS16 for a full rundown of our reporting from the show. For a selection of some of the best social media posts of the show, visit our Trending at #HIMSS16 hub.