Rush University Medical Center is applying data technology in several areas, including documentation improvement, value-based care, dashboard development and implementation, readmissions reduction, and cost of care.
At Rush University Medical Center, an effort to investigate low quality scores has blossomed into a highly ambitious data analytics and machine learning initiative.
In early 2015, the Chicago-based medical center had three stars in Medicare's Hospital Compare rating system, and the organization's new CMO, Omar Lateef, DO, wanted to know why, says Bala Hota, MD, MPH, vice president and chief analytics officer.
"We are both practicing physicians. We saw patients at Rush, and we knew the quality of care for a lot of the service lines was excellent, but we weren't seeing that reflected in the ranking systems," says Hota.
While systematically reviewing documentation data, Hota and his colleagues found clinicians were underreporting the severity of patient illness. "We were consistently seeing that our severity of illness ranked among the lowest—in the bottom decile," he says.
The need for documentation improvement became even clearer after a review of patient transfers from other facilities, Hota says. "Patients were getting transferred to us and they were getting a lower acuity than they had at the hospital they were transferred from, which made no sense."
Documentation improvement that has improved severity of illness reporting has helped raise Rush's Hospital Compare rating to five stars.
Documentation is the key ingredient in most ranking systems at the Centers for Medicare & Medicaid Services, he says. To help find opportunities to improve documentation, Hota's team developed customized algorithms and gathered documentation data.
"We do see variation between different provider groups and different providers. There is more opportunity with some, with others there's less. Just being armed with that information helps us to educate, guide, and change behavior," Hota says.
Rolling out dashboards
After using data analytics to delve into Rush's quality ratings, Hota's team continued to seek opportunities for data to drive change.
A primary effort has been harnessing key performance indicator [KPI] data and displaying the data on dashboards, often in real time. "We have a level of detail needed to drive behavior change," Hota says.
The dashboards track several KPIs:
- Care variation
- Cost of care
- Length of stay
- Patient safety indicators
- Quality domains
"We have about 200 key performance indicators in executive dashboards that we have rolled out over the past four months. These dashboards are shown on large monitors in the offices of our president, chief operating officer, and chief financial officer," he says.
Rush is conducting predictive modeling with its Epic electronic medical record and with a cloud-based Big Data capability.
With Epic, Hota's team has developed five predictive models for illnesses, including congestive heart failure.
With the cloud-based technology, data can be drawn from any IT system in the cloud's "environment" such as sensor data and clinical engineering. "The advantage of this approach is we are using machine learning. Deep learning models allow us to customize models to our unique situation," Hota says.
Predictive modeling is being used primarily in three areas at Rush:
- Cost of care to identify high-risk patients and help ensure they are on the best care pathways or protocols
- Emergency department throughput to drive reductions in the length of time patients are spending in the ED
- Readmission reduction to identify negative factors after discharge
Data drives value-based care
Data analytics are playing a crucial role in value-based care at Rush, which features four acute-care hospitals.
In the Medicare Shared Savings Program (MSSP), Rush is using data analytics to track access to care, patient residence, in-network versus out-of-network utilization, and total cost of care.
Documentation is a key to MSSP success, Hota says. "We are looking at how documentation relates to our severity of illness and our care. We are looking for opportunities to improve what we document."
Data analytics is helping Rush manage the impact of the Hospital Readmissions Reduction Program and the Hospital-Acquired Condition Reduction Program.
For readmissions, Rush uses predictive modeling and risk adjustment to anticipate and affect readmission rates, which have improved. "We have seen readmission declines over the past four years, which has switched Rush from a Medicare penalty to a bonus for our academic medical center," he says.
Focusing on outcomes data is important to limiting hospital-acquired conditions, Hota says. "We are measuring every case of a condition. We are looking for where we have outcomes that we need to fix."
Scaling paced to growth
Rush has contained the costs for its data analytics initiative by leveraging internal resources and scaling growth strategically.
"We have tried to use existing staff and existing resources. We started out by dipping our toes in the water, [to] see whether there was value, and scaling it as we moved along," Hota says. "There has not been a lot of upfront expense."
Utilizing cloud technology is financially advantageous, he says. "From the start, we planned an in-the-cloud implementation because that allows our costs to scale as the data scales. So, we do not have a large capital investment for infrastructure on-site."
The cost of the initiative has increased over time, Hota says. "We were able to keep the costs relatively low—almost at a pilot-project level of funding—for the first six months. Since we have seen some successes, we are scaling the budget as time goes by."
Over the next few years, he expects Rush's investments in data analytics capabilities to reach "millions of dollars, not tens of millions of dollars."
"We want to grow this program and make it the source of truth for data and analytics at the medical center and the Rush system," says Hota.
Christopher Cheney is the senior clinical care editor at HealthLeaders.