Vanderbilt using biostatistics software to fight readmissions

Vanderbilt University Hospital (VUH) is testing a new computer program designed to estimate patients' risk of returning within 30 days.

The program, Cornelius, calculates two risk scores for each newly arrived patient—one estimating the patient’s risk of developing pressure ulcers during the current hospital stay, and the other estimating the patient’s risk of returning for readmission within 30 days of discharge.

“I realized we had to close the gap between publishing these models and making them part of the clinical workflow, and that’s what Cornelius is about,” said biostatistician Dan Byrne, MS, leader of the Cornelius team and director of quality improvement and program evaluation with the department of biostatistics.

Cornelius’s risk models use a handful of clinical and demographic factors that are typically documented in the EMR within 24 hours of hospital admission.

The models rest on a statistical analysis of some 30,000 VUH patient records. In that analysis, Byrne and Hank Domenico, MS, biostatistician, tested more than 400 clinical and demographic variables for correlation with pressure ulcers and readmission.

In a randomized controlled trial of Cornelius’s usefulness, for half of all VUH patients the two risk scores are completely suppressed, hidden from the care team and from researchers. For the other half of patients, the scores are taken into account by the care team as they try to discern which patients may benefit from which preventive measures. All arriving VUH patients are randomly assigned to one of the two groups.

The study began in 2012 and is expected to continue at least another year.

“With the research component of Cornelius, we would like to show the country, if we can, how you can carry translational science all the way into the clinical space in a seamless, efficient way,” said Gordon Bernard, MD, associate vice chancellor for clinical and translational research.