everythingpossible - Fotolia
Healthcare providers have known for a while that applying predictive analytics to the problem of patient readmissions was going to be important to their financial future. Now that they've had a few years to think about the problem, they're finding ways to make meaningful improvements in their readmission rates.
New regulations stemming from the Affordable Care Act compel health systems to operate under reimbursement programs that reward quality over quantity, but this arrangement puts a premium on predictive analytics in healthcare. Knowing who your patients are, what conditions they have and how to keep them healthy and prevent them from needing expensive follow-up care are now top priorities.
Speaking at the Big Data and Healthcare Analytics Forum in Boston, Pamela Peele, chief analytics officer for the insurance division of Pittsburgh-based healthcare provider UPMC, said she and her team started working on this problem in 2008 when Medicare announced it would stop reimbursing for care related to readmissions for certain conditions, such as heart failure or pneumonia. But it's only been since 2013, when team members took a more systematic approach to healthcare analytics, that they saw a meaningful reduction. Today, the health system's readmission rate has dropped two percentage points to 13.5%.
Measuring the success of predictive analytics in healthcare
Measuring the success of specific interventions has been one of the most important factors in UPMC's readmission rate improvement. Early on, the analytics team started developing a score for every patient that predicted his or her likelihood of being readmitted to the hospital. Team members initially targeted follow-up care to patients in the highest risk brackets. But this failed to make a dent in the readmission rate.
Pamela Peelechief analytics officer, insurance division of UPMC
The team eventually realized the reason: The highest risk patients are also the ones with the most serious health problems, and no intervention is going to keep them out of the hospital. When analysts started targeting interventions more to people with moderately high risk of readmission, they found that care teams were able to keep those people from returning to the hospital.
They also found that patients who received a follow-up appointment at an outpatient clinic within five days of hospital discharge were significantly less likely to be readmitted than patients whose follow-up visit came later. As a result, UPMC changed physician bonus structures to encourage them to schedule follow-up appointments with patients sooner.
Aside from measuring the success and failure of various interventions, Peele said communicating effectively to executives is the key to the success of predictive analytics in healthcare. Leaders may understand that reducing readmissions is important, but they may not get why a certain change in procedure is necessary just by looking over a report or slide deck. Peele recommended only giving executives the most important information -- and in a clear format.
"We're awash in data," she said. "Every report generates five more report requests. We are raining information down on our decision makers, and when they look befuddled, we create more reports. After a while you just can't stand it anymore."
Keep it simple to get healthcare analytics buy-in
Doctors don't typically have a lot of time, so any tool that attempts to change the way they do things to reduce readmissions needs to be streamlined, said Trey La Charite, medical director for clinical integration at the University of Tennessee Medical Center in Knoxville, Tenn.
He said his team started down the road toward readmission reduction with a tool that required doctors to manually enter information about the patient and wait for it to return a measure of the patient's risk of returning to the hospital. But the doctors found the tool clunky to use. Adoption was low, and most clinicians fell back on their own intuition of which patients were at the greatest risk of being readmitted.
However, in September 2015, his hospital introduced a new tool that takes data from the hospital's electronic health record system and clusters patients into different risk levels. It also assesses historic data to determine care strategies that have worked in the past for certain kinds of patients. Case managers get these scores each morning. They then meet with clinicians who are treating patients getting discharged that day. Together, they go over the system's recommendations and develop a plan. La Charite said the hospital has seen a 10% reduction in readmissions since switching to the new approach.
He said one of the biggest reasons for the success of the tool is that doctors don't have to be too involved in its use. They get recommendations each morning and then are able to apply their clinical skills.
"I think this predictive analytics tool is bringing together quality health with financial stability and isolating it away from the providers, which free them to do what they do best, which is keep patients healthy," he said.
Analytics helps healthcare providers avoid penalties
Analytics in healthcare helps keep the data flowing
Healthcare analytics market set for huge growth