Collaborative analytics effort helps Advocate bend cost curve

BOSTON—Big data is not the end but the means to the end, said Tina Esposito, vice president of Advocate Health Care’s Center for Health Information Services in Chicago, speaking at the Big Data Healthcare Analytics Forum on Nov. 20.

Advocate maintains the largest accountable care organization covering 609,000 lives in addition to having 12 acute care hospitals, 250 sites of care on the outpatient side and $5 billion in annual revenue. To manage it all, Esposito said the executive team made a decision in 2005 to get quantifiable objectives around health outcomes. “We know we’re doing good work—we want to see how good.”

The question, she said, is no longer what is the number but why is that the number. Looking ahead, she said the question is going to be can we predict our performance? For now, “can we understand the care provided across the continuum.”

That’s where the big data need came to be, she said. “We cut our teeth on claims data but the ability to leverage clinical data is absolutely vital to support the front line. It’s not enough to know a patient is diabetic. We need to track hemoglobin A1C. That’s what you need to transform clinical care.”

Advocate partnered with Cerner, its vendor for the majority of its inpatient systems, to improve population health management capabilities. A collaborative team was charged with innovating to identify/risk stratify patients at risk; facilitate appropriate and early interventions as much as possible; and guide care across the continuum.

The team had specific goals including a readmission model and falls risk and prevention for year 1; predictive models to improve care in ambulatory settings in year 2; and drive health and care of the population with predictive analytical models across care settings in year 3.

During the first year, the team spent time assessing resources, talking with vendors and getting data into the big data platform, Esposito said. “We were very clear this was not going to be an academic exercise. If we never got to the front lines, it would all be for nothing.”

There was a significant effort to understand the science of implementation and a keen focus on how to deploy models, she said. The team deployed within five months an off-the-shelf product for falls assessment and learned from that to deploy a readmission model.

The overall platform is called Healthe Intent and has 64 separate data sources loaded in including claims, pharmacy and EMR vendors. The different care settings all have different vendors so “getting all that data together and in is an enormous amount of work,” said Esposito. “What became obvious and clear is that we had to get serious about patient identification and the enterprise master patient index.”

When they started the process, there were 3.4 million potential leakages or duplicates, she said. At this point, it’s down to 40,000. The clean-up makes for a “foundational item,” she said. “Our goal is to use these data for advanced analytics. When done right, the big data platform is just that—a platform for other things.”

Advocate is leading the industry when it comes to readmissions. The organization also is gaining efficiency with about 3.5 FTE productivity savings across the system and an automated continuous calculation of risk in the EMR. They’ve been able to reduce readmission rates by 20 percent for high-risk patients that received interventions. “That’s proof we were better targeting our strategic resources to support the outcomes we were after,” said Esposito.

The second phase of work is transitions of care and making sure patients are discharged to the most appropriate level of care. “We had a sense we weren’t always doing that but we needed an objective process around that.”

Advocate’s transition of care model uses clinical data to profile patients and look for similar patient profiles and where they have been most successful. About 30 percent of the time, the model predicted a different disposition for the patient than where he or she actually ended up, she said. “That indicates we are potentially underutilizing home care and overutilizing skilled nursing facilities, rehab and long-term care. That equals about $200 million total cost of care savings. It’s really about returning the patient to the level of independence they should be at.”

Esposito said the population health pyramid has the most complex, costly patients at the top with 12 percent of the population accounting for 50 to 60 percent of the total spend. “We want to focus our efforts here with any and all resources placed on this 12 percent.”

The assumption, however, that all of these patients are not being well managed because they are high cost is not true, she said. “They just have a high-cost diagnosis.” Also, this is not a static state with patients moving throughout the various segments of the pyramid. “We’re trying to develop a new way of stratifying.”

Doing so will help them better understand patient populations than just by cost, she said. She hopes they will learn how to better target interventions to those most impactful. “Everything in population health often turns to measures of cost and utilization. This potentially could help define news ways of defining quality metrics around population health. That’s compelling to us.”