How Three Hospitals Use Predictive Analytics to Reduce Readmissions

Nursing leaders at West Des Moines, Iowa-based UnityPoint Health, a network of hospitals, clinics, and home care services in Iowa, Illinois, and Wisconsin, wanted to know why patients were being readmitted within 30 days.

So they asked patients, “Why do you think you’re back?”

This approach allowed patients to speak openly, which gave nursing leaders insight into patients’ greatest challenges. Patients can struggle with a lack of access to follow-up appointments, food insecurity, or an inability to pay for medications, says Rhiannon Harms, executive director of strategic analytics at the health system.

Open-ended questions are important. It’s difficult to ask a question such as “Can you pay for your medications?” says Harms.

UnityPoint Health combines the patient narrative with retrospective data on readmissions to create a readmission risk score for the patient-and that information is communicated to the patient’s care team.

For example, Patricia Newland, MD, a family medicine physician, relied on the predictive analytics tool to prevent a readmission by one of her patients. During a follow-up appointment, Newland used the tool to determine that her patient was likely to experience an onset of symptoms within the next 13 to 18 days. She shared this information with the patient and told her to call the practice at the onset of symptoms, which could include shortness of breath, wheezing, and coughing.

Related article: Reduce Hospital Readmissions with Virtual Care Technology

The patient called the practice in that timeframe complaining of those symptoms. Newland got the patient in for a same-day appointment, consulted with her patient’s pulmonologist, and changed her patient’s medications-and thus prevented a hospital readmission.

Newland relies on her practice’s clinical care coordinator to highlight the patients who are at highest risk of being readmitted. It also helps that her practice keeps same-day appointment slots open.

Team huddles, which take place at her practice three or four times a week, allow the team to coordinate patient care. Newland, who also serves as a physician leader in the health system, is responsible for educating clinical care coordinators across the health system about the use of predictive analytics to prevent readmissions.

Once physicians realize that access to this information can help their patients, they embrace this tool, she says. In fact, Newland considers the readmission risk score as the “fifth vital sign” in her patients’ follow-up care.

Payers can support this work by paying for home-based health providers and devices to monitor a patient’s vital signs in the home, says Newland. “If we could do those things, it might keep a patient out of the hospital.”

UnityPoint Health has reduced all-cause readmissions by 40% within 18 months of using the predictive analytics tool. The health system’s home health team also used it to determine their most vulnerable patients when the community was hit by a blizzard. With this insight, home health providers could tailor their visit schedules to see patients who were in most need of care, says Harms.

Preventing readmissions for diabetes patients

In his quest to reduce readmissions at Kansas City-based University of Kansas Health System, David Wild, MD, vice president of lean promotion, discovered that patients with diabetes were more likely than other patients to be readmitted three times within 90 days at the nonprofit, academic medical center.

Wild used predictive analytics to comb through variables such as total length of stay, number of chronic conditions, whether the hospital admission was planned or unplanned, smoking history, the age of the patient, and payer type (public or private).

Inability to access follow-up care, patient discharge disposition (the presence of a family member in the home or a stay at a skilled nursing or rehabilitation facility), and the total number of chronic conditions were the biggest drivers of readmissions for diabetes patients, he says.

With this insight, all patients who have been readmitted to the hospital three times in 90 days-not just those with diabetes-are connected with a case manager who helps them secure follow-up care. On an average day, five to nine inpatients meet that criteria, says Wild.

In May, when Wild started this project, the readmission rate for diabetes patients was 25%.

Currently, the readmission rate for diabetes patients is 13.9%. Wild attributes University of Kansas Health System’s success to improved engagement with nurses on inpatient units and incorporating diabetes educators for daily information sessions with patients on medications.

He welcomes support from payers. For example, coverage of diabetes education in the outpatient setting could help prevent readmissions.

Physician engagement helps reduce readmissions

Flagler Hospital, a nonprofit healthcare facility in St. Augustine, Florida, has 400 physicians, according to Michael Sanders, MD, its chief medical information officer. With that number of physicians, wide variation in clinical practice is likely. That’s a pet peeve for Sanders, who also laments the impact of waste on healthcare spending in the United States. (Wasteful spending in healthcare exceeds $1 trillion each year, an amount that two experts writing in HealthAffairssay “could fund the entire Medicaid program twice over.”)

To address this, Sanders decided to reduce variation in care for patients with pneumonia. After looking at more than 16,000 patients over four years, he discovered that many patients with pneumonia were getting daily X-rays and blood tests, which increases the cost of care without improving outcomes.

He also learned that introducing nebulizer treatments for patients with pneumonia and chronic obstructive pulmonary disease early in their hospital stays improves clinical outcomes.

Related article: Nine Ways to Reduce Hospital Readmissions from Hospital Executives

Sanders had the proof in hand, but changing how physicians practice medicine is tough. To tackle this, he taps his longstanding relationships with physicians involved in Flagler Hospital’s EHR deployment, which started in 2012. These physicians act as “ambassadors” for adhering to the care path designed for pneumonia patients, says Sanders.

Flagler Hospital has reduced pneumonia-related readmissions from 2.9% to .4% since March. In addition, the hospital has saved $1,350 per patient and reduced the length of stay for pneumonia patients by two days. Going forward, Sanders expects the hospital to save $726,300 annually because of this work.

Payers should care about this, he says. They’ll benefit from the elimination of unnecessary tests and the reduction in cost.

“They should be looking at ‘same cause’ readmissions and rewarding hospital systems that can not only reduce cost, but also readmission and mortality as well. They must recognize that this effort costs hospital systems in dollars and work effort, and reward them accordingly.”

Insurance companies could also pass on those savings to members who use hospital systems that reduce cost while improving outcomes, adds Sanders.

Aine Cryts is a writer based in Boston.