30-Day Readmissions

Opportunity

In an effort to reduce hospital readmissions, OSF HealthCare implemented a BOOST-based navigator inside of EPIC. This required nurses to assess patients on multiple criteria in an effort to identify which are at the highest risk of readmission. However, the approach was found to take a significant amount of nurse time over the course of an inpatient stay, resulting in more than $3 million in salary and benefits in nursing time each year to assess patient risk. It also produced a large volume of inefficient downstream work through false-positive generation.

Solution

Healthcare Analytics built a 30 Day Readmission Risk Model that helps clinicians identify patients most at risk for readmissions, driving work processes and helping better align patients with existing interventions such as case management.

Impact

Over the course of a year, this resulted in about 425 fewer readmissions than expected in our medium-high and high-risk patients. The team also found it was able to reduce about 67 percent of nursing assessment activities and decrease the flow into case management by about 44 percent. These staff time reductions translate to a little more than $2 million per year that we can put back into direct patient care.

"The 30 Day Readmission Model is currently being used to direct workflow activity in a variety of operational areas such as; inpatient case management, ambulatory care management, post discharge follow-up phone calls, outpatient palliative care and home care reporting/monitoring. Through the use of this model, we're able to efficiently focus resources to those patients who are in the most need."
- Chris Franciskovich, manager of Advanced Analytics, Healthcare Analytics at OSF

Download PDF

Predictive Model Reduces Readmission Rates Among Most Vulnerable Patients

Like many hospital systems around the U.S., OSF HealthCare is continually working to reduce its hospital readmission rate. In one of many efforts to do this, OSF implemented a BOOST-based navigator inside of EPIC, our Electronic Health Record. This required nurses to assess patients on multiple criteria in an effort to identify which are at the highest risk of readmission.

However, the approach was found to take a significant amount of nurse time over the course of an inpatient stay, resulting in more than $3 million in salary and benefits in nursing time each year to assess patient risk. It also produced a large volume of inefficient downstream work through false-positive generation, meaning up to 85 percent of all in-patients were getting referred into case management even though they didn't all warrant it.

Building the Model

The Healthcare Analytics team came up with a more efficient way to proactively identify patients in need of risk mitigation. The group explored hundreds of variables and their interactions, but ultimately built a predictive model that includes around 50 variables and automatically identifies at-risk patients in four levels.

"The predictive model uses many variables from data within the electronic health record (EHR) to assign a risk level to each patient so that OSF clinicians can take proactive steps to improve care coordination instead of spending time trying to identify which patients may be at high risk for a readmission."
- Mark Hohulin, senior vice president of Healthcare Analytics at OSF

When we first started, the model was implemented in our Enterprise Data Warehouse (EDW) and its output shared with clinicians via a daily report. The report grouped patients by unit and risk level and then sorted by probability to readmit, but was delivered outside of the clinicians' normal EPIC workflows. As the utility and effectiveness of the model was proven, Healthcare Analytics partnered with Information Technology to develop a communication pathway between the EDW and EPIC via a process called DataLink. This allowed the model's output to be incorporated directly into clinicians' daily workflows, reducing a potential barrier to use.

Results

Over the course of a year, this resulted in about 425 fewer readmissions than expected in our medium-high and high-risk patients. The team also found it was able to reduce about 67 percent of nursing assessment activities and decrease the flow into case management by about 44 percent. These staff time reductions translate to a little more than $2 million per year that we can put back into direct patient care.

The model has been in active use for more than three years. While it started as a way to help direct case management activities inside the hospital, the use of the model has expanded to provide work direction assistance to : inpatient case management, ambulatory care management, follow-up phone calls, outpatient palliative care and homecare reporting/monitoring.

"By leveraging predictive readmission risk scores, we are able to focus Case/Care Management resources on clinical care and post-hospitalization for the appropriate patients" said Hoa Cooper, Vice President of OSF HealthCare Care Management.

Overall, the unadjusted readmission rate has remained steady, but we did see a change in the distribution of our population with an increase in medium-high patients and a decrease in those within our low risk category.

Comparing the historical performance by category against the current performance, we get the following:

• Low and Medium Low are not showing significant differences.
• Medium High and High are showing significant differences.

We can't yet say which interventions are driving changes, but we can say there is an indication that targeting these activities is having an impact.

"Using proactive, predictive modeling to enhance and enrich traditional EHR data will be key in the future for OSF clinicians to continue to improve patient care and improve outcomes for all of our patients" said Hohulin.