University of Iowa Hospitals and Clinics Transforms the O.R.
Accessible, predictive analytics for far fewer infections
In the United States, roughly one in 20 patients admitted to a hospital develops an infection. According to the US Centers for Disease Control and Prevention, surgical site infections are the most common, accounting for more than 30 percent of occurrences, and putting patients at risk for illness, prolonged hospitalization, and death. In addition, the total cost of hospital-acquired infections is estimated at $10 billion per year. Therefore, reducing infections has major implications for overall patient health as well as hospital cost savings.
Surgeons at the University of Iowa Hospitals and Clinics, one of the most highly regarded medical facilities in the US, wanted to know when patients were susceptible to surgical infections so they could make critical treatment decisions in the operating room.
Dr. John Cromwell, director of gastrointestinal, minimally invasive, and bariatric surgery, believed that use of predictive analytics could prevent a high percentage of surgical site infections and decrease healthcare costs. He had used predictive analytics previously, however, and realized that the division's desktop analytics environment could not handle large distributed data volumes. The team faced a major roadblock in the sheer multiplicity of sources. Yet connecting to them is a prerequisite for valuable analytics, as was a robust way of deploying analytics to frontline staff.
To improve quality of care, the team needed a flexible, enterprise-grade, advanced analytics platform that encompassed the entire analytics lifecycle—from data aggregation and preparation, to model development, deployment, and monitoring.
Jose Maria Monestina, senior application developer at UI Hospitals and Clinics, was charged with implementing TIBCO Statistica™ technology that would predict outcomes for Cromwell's team.
"Dr. Cromwell has a unique approach to predictive analytics because he uses patient healthcare data, historical data to create predictive models, and real-time data. This combination has allowed us to deliver predictive analytics in a real-time environment to improve healthcare."
Broad Data Aggregation for Individualized, Data-Driven Decisions
The first step was to connect to the disparate enterprise systems and bring the data into a common data set with embedded analytical tools, moving from desktop to enterprise analytics, and turning prediction theory into a life-enhancing reality in the operating room. "We have complex workflows, and different data sources," explains Mr. Monestina.
"We take information from electronic medical records and other enterprise sources, including real-time data from the operating room, to determine whether patients are likely to get a surgical site infection,” says Dr. Cromwell. "This allows us to modify and individualize the type of preventive care we're delivering, in real time."
Real-Time Preventive Care, Drastically Reduced Infection
"Predictive analytics is allowing us to manage the ever-increasing types of data that healthcare institutions are tasked with," says Dr. Cromwell. "The number of sources continues to increase rapidly, and the Statistica tools allow us to keep track of the various models we need for that type of exercise."
Running predictive models, the team can anticipate which patients are at risk for surgical site infections before they occur. Predictions can be stored for future analysis to aid in follow-up patient care.
"Using these tools and other methods, we've been able to reduce surgical site infections by about 58 percent. "That's a revolutionary concept in gastrointestinal surgery."
Accessible, Flexible Analytics Deployment
"A small group of people can now deliver predictive analytics," says Mr. Monestina. "Statistica simplifies deployment, execution, sharing of models, and analysis of the data—all in one package. You can store the data model in a server and then reuse it. You can share the data models across the research group. You are not bound to a specific PC or a server. You can run those models using a mobile application or a web browser and access the results."
"Big data and predictive analytics are transforming outcomes at virtually every point in patient care," says Cromwell. "We see so many other areas where this could be useful, including drug delivery, population health, managing patient flow, and every other aspect of medicine that allows us to deliver high-quality healthcare."