How Two Health Systems Use Predictive Analytics to Reduce Readmissions

Reducing hospital readmissions requires just the right mix of clinical expertise and predictive analytics, according to industry experts. Here are two health systems that are hitting their stride.

1. Atrium Health, formerly Carolinas HealthCare System

James Hunter, MD, chief medical officer at Atrium Health, formerly Carolinas HealthCare System, says the large healthcare system with hospitals, emergency departments, urgent care centers, and medical practices in North Carolina and South Carolina began its journey to reduce readmissions in 2010. At that time, it focused on providing the best care to patients in the acute-care setting.

About three years into that journey, it became clear that focusing solely on the acute-care setting wasn’t enough. That’s why, in 2013, the health system began using predictive analytics to learn the types of supports patients needed outside the hospital or emergency department.

Today, its EHR leverages 42 different variables-which include clinical and socioeconomic information-to determine if a patient falls into one of three categories: low, medium, or high risk of readmission.

Much of this information is captured in patient interviews by case workers. Case workers are able to determine, for example, if patients are able to afford their medications, are experiencing food insecurity, have support from family members at home, and access to transportation for follow-up appointments.

“More and more, these are the things impacting readmissions-it’s not about whether a doctor put a congestive heart failure patient on a statin,” says Hunter.

One program that has grown out of Atrium Health's predictive analytics program is its “Heart Success” program. High-risk patients are seen in a multidisciplinary cardiology transition clinic for 30 days by cardiologists with experience treating patients with congestive heart failure, in addition to nurses, social workers, and pharmacists.

Patients who live within a 45-minute drive to the cardiology transition clinic in Charlotte, NC, receive care in person. Patients further away receive virtual visits facilitated by nurses who come to their homes.

After being seen at the cardiology transition clinic for 30 days, patients are returned to the care of their primary care physicians, who are briefed on their patients’ status through the EHR, says Hunter.

At that point, patients have a lot more knowledge about their disease, the signs and symptoms to look for, and people to call if there’s a problem, he adds.

As a result of this and other programs, Atrium Health has been able to reduce its readmission rate for congestive heart failure patients from 19% to 15%. Hunter notes that this is 10% below the national average of 25%.

Next: University of Kansas Health System

2.  University of Kansas Health System

About 960 miles away in Kansas City, KS, David Wild, MD, vice president of lean promotion at the University of Kansas Health System, says the academic medical center was driven to improve its readmission rate when it learned that other hospitals within the Vizient network-of which his facility is a member-were achieving better outcomes. Vizient is a member-driven performance improvement company.

While the healthcare system’s readmission results were always better than the national average, Wild says the healthy competition led him and his team to seek improvements.

They decided to use predictive analytics because of their practical knowledge that many patients were being readmitted because of socioeconomic and other factors. For example, often patients lacked certain resources, such as housing and family members who could take care of them after they were released.

The health system uses a variety of data-from laboratory results to a “random smattering,” such as the day of the week the patient is admitted and insurance plan information.

One of the “biggest clinical changes” the hospital has made as a result of this analysis is to improve care transitions by providing more detailed discharge information to patients earlier.

For example, congestive heart failure patients now learn earlier in their hospital stay that having too much fluid in their system is a problem-and that they need to be very careful about the amount of sodium they consume. As patients get closer to discharge, they meet with nurses and a pharmacist to learn that they’ll need to weigh themselves regularly to assess weight gain, which could be a sign that they’re retaining or losing water and indicate that the heart isn’t functioning properly.

High-risk congestive heart failure patients learn upon discharge that they will have follow-up appointments within three days and another follow-up appointment within seven days, says Wild. These patients receive a phone call from their care team within 24 hours of discharge, and subsequent calls from a nurse or a social worker to transition them to the home environment.

The team’s use of predictive analytics also showed that patients taking no medications are more likely to be readmitted than those taking many medications. That surprised Wild and his team.

While they can’t prove it yet, they believe patients who are prescribed and taking a variety of medications may be more aware of their condition and understand more fully why they are taking medications. Patients without medications may not know as much about their conditions and are ripe for education. Wild hopes to validate this theory by testing it in his predictive analytics models.

As a result of these and other efforts, the University of Kansas Health System has reduced its readmission rate for congestive heart failure patients from 21% to 14%, with isolated incidents when the health system performed better, he says.

Aine Cryts is a writer based in Boston.