Advocate Health Uses Big Data to Improve Value Based Care

Advocate Health Care is looking to big data and predictive analytics to improve patient care and better control costs, partnering with Cerner to develop analytical tools hosted on a cloud-based platform.

 Advocate Health Care is looking to big data and predictive analytics to improve patient care and better control costs.

The health system has partnered with Cerner, a vendor of electronic health-records systems and other healthcare products, to develop analytical tools hosted on Cerner’s cloud-based population-health management software platform, called HealtheIntent. 

Advocate has a lot at stake. It operates 10 hospitals in metropolitan Chicago and two hospitals in central Illinois, a home health business, and a multi-specialty medical group. It also operates Advocate Physician Partners (APP), a collaboration between Advocate’s hospitals, physicians employed by Advocate and independent but closely aligned physicians.  APP—which includes more than 4,900 physicians—oversees patient-care coordination among its members and managed-care contracts with governmental and commercial payers. 

ALSO SEE: Mercy's Big Data Project Aims To Boost Operations

And the revenue model for all healthcare providers is changing quickly as insurers, as well as Medicare and Medicaid, increasingly tie the amount they pay to the quality and efficiency of the health care services those systems provide to patients.

Advocate, however, has been operating for years as a value-based care provider.

In 2011, Advocate inked its first value-based contract with Blue Cross Blue Shield of Illinois, and in 2012 joined Medicare’s Shared Savings Program—in which doctors and hospitals earn incentive payments for working together to meet cost and quality targets. 

Advocate has signed many other value-based contracts since 2012, ranging from traditional fee-for-service models with incentive payments to arrangements in which Advocate assumes full financial responsibility for managing the care of a group of patients.

But as the organization began its transition to value-based patient care with that first contract in 2011, Advocate’s executives soon realized that they needed to provide physicians, nurses and other caregivers with comprehensive and timely information on which to base patient-care decisions.

They wanted clinical information on all of its patients’ encounters with the healthcare system—including test results, medications and procedures—integrated into a single, longitudinal record for each patient.

They wanted algorithms to predict medical events, such as patients at high risk for an unplanned hospital stay.

And they wanted access to real-time patient data for analysis as opposed to relying on claims data, which can be dated.  

“It was really about how we leverage data and information to understand the population that we have attributed to us—their needs” and how to “target the right patient to the right intervention,” Tina Esposito, vice president of Advocate’s Center for Health Information Services, says.

But Advocate’s rich store of real-time electronic patient data resided in a disparate group of proprietary EHR products, such as for inpatient and outpatient services, and ancillary clinical systems, such as for pathology and radiology. This silo approach made it difficult for Advocate to aggregate information about patients’ interactions across sites of care, such as physicians’ offices, emergency departments, and hospitals. 

“We knew we had to create a new data and analytics platform that would support that new type of thinking,” Esposito says.

Advocate, which already was a customer of Cerner’s cloud-based EHR product, Millennium, decided to deploy the vendor’s HealtheIntent, which is a Hadoop-based big data management and analytics system that facilitates distributed storage and processing of many types of data in its native format across clusters of commodity hardware.

Within the HealtheIntent platform, analytic jobs run in the Hadoop environment automatically and the results are pushed out to either secure web-based software services or, when possible, integrated directly into an EHR system, according to Nathan Beyer, Cerner’s principal architect for the HealtheIntent platform.

With HealtheIntent deployed, Advocate Health Care and Advocate Physician Partners then worked with Cerner to develop software products that the health system needed.

The tools that resulted from the partnership are a configurable suite of HealtheIntent products that facilitate tracking of patients based on automatically updated clinical data, such as blood glucose levels, and predicting outcomes, such as an unplanned hospital stay. 

Also, through the partnership, Cerner now provides Advocate with access to an enterprise data warehouse called HealtheEDW and HealtheAnalytics analytics software that allows the health system to produce custom reports using Tableau visualization.


At Advocate, the cloud-based data store is extensive.

The health system feeds data from more than 60 different clinical and transactional sources into the HealtheIntent platform, including numerous proprietary EHRs deployed in hospitals, doctor’s offices and Advocate’s home care operation.  It uses Cerner’s EHR product at 10 of 12 hospitals, but uses other vendors’ products at two other hospitals, its home care operation and physician’s offices.

Advocate also pulls in data from insurance claims for both medical services and pharmacy benefits and other ancillary clinical systems, such as for pathology and radiology. In the future, Esposito and her colleagues would like to add other types of data, such as from wearable fitness devices, air quality reports or Census data. 

An important step in the data-management process is Advocate’s enterprise master-patient index (EMPI)—a database that uses a unique identifier for patients to ensure that data, such as their name, address and medications, are consistent across various, disparate software systems, such as EHRs. Advocate uses IBM InfoSphere Master Patient Data Management to standardize and manage the data.

Once the data is fed into the Cerner platform, it is mapped and normalized to resolve issues with ontology and terminology—such as different diagnostic codes to describe the same medical problem or different brand names for a common medication, such as aspirin, according to Bharat Sutariya, M.D., vice president and chief medical officer of population health at Cerner.

Advocate completed the first iteration of its HealtheIntent data environment in 2013, but is “adding new sources of data all the time,” says Esposito.  As of August 2015, Advocate had aggregated information on nearly 17 million patients.

The aggregated data runs in the HealtheIntent environment via an application, called HealtheRecord, that’s comprised of a longitudinal view of each patient, including 10 years of clinical data, such as tests results and surgeries; claims data, such as information on prescriptions and payments; and administrative data, such as home addresses and insurance carriers.


From the beginning of their data partnership, Cerner and Advocate have been working to co-develop algorithms that predict patients’ level of risk for a given medical outcome, such as a hospital stay, by analyzing data in the Hadoop platform. 

The most recent result of the Advocate-Cerner collaboration is an algorithm that leverages the HealtheIntent data set to help physicians choose the next level of care—such as a skilled nursing home, long-term acute-care facility or in-home health services—for patients being discharged from a hospital. The goal is to send patients to a level of care that reduces their risk of returning to the hospital within 30 days of discharge.

Advocate and Cerner conducted a randomized controlled research trial to test the algorithm from October 2014 to January 2015.  The findings, which involved a total of 5,600 patients at one of the system’s largest hospitals, “look very positive. We have seen that when the model is followed, there is a lower readmission rate,” Esposito says. She declined to release specific numbers because she and her co-researchers would like to publish a peer-reviewed journal article. 

Advocate plans to roll out the algorithm, which is already integrated into its inpatient EHR at most of its hospitals, in the next six to nine months.  It’s designed to be an add-on product to Cerner’s Readmission Prediction Solution, which is a tool that the two organizations co-developed earlier in their partnership.

The Readmission Prevention Solution assesses hospital patients’ risk of readmission every two hours throughout their hospital stay based on about 30 variables, such as the primary medical issue, secondary diagnoses, and medications.

Advocate had developed another in-house algorithm prior to the partnership with Cerner, but was unhappy with it because it did not update risk scores throughout a hospital stay and required clinical staff to input information manually into the EHR. Through these shortcomings, staff members could easily override the algorithm’s scientific objectivity if they checked “off the boxes that they knew would get the risk score that they had in their head was true for that patient,” Esposito says.

The HealtheIntent readmission tool solves these problems. Not only does it automatically calculate a risk score, but the information is integrated directly into Cerner’s EHR, and as of July 2015, into a Meditech EHR solution deployed at Advocate BroMenn Medical Center in Normal, Ill., which is one of two Advocate hospitals that uses a non-Cerner EHR product. 

The information also isavailable via secure, web-based software hosted at Cerner, allowing physicians access to information from their offices, where they also use EHR solutions from vendors other than Cerner.

Physicians or other caregivers drill down to find out each patient’s risk score (low, medium or high), what factors led to the score, and what clinical interventions would be appropriate to lower the risk.

Standard interventions include intense patient education or physical therapy while in the hospital, or follow-up phone calls, and in-home visits after discharge.

Advocate saw results fairly quickly from the Readmission Prediction Solution.  The application was live at all Chicago-area hospitals by the end of 2013. By March 2014, the 30-day, all-cause readmission rate among high-risk patients had dropped 20 percent, as these were the ones who received most of the intensive resources.  

Advocate then decided to help develop the companion tool—which matches patients to the appropriate next level of care after discharge from a hospital based on 100 variables—because “there was definitely a consensus in the organization that we needed to be more objective in how we placed our patients,” Esposito says.

The Advocate-Cerner predictive-analytics team plans to move into the ambulatory setting next. For example, the team would like to develop an algorithm to determine which patients should be assigned to outpatient care managers, who coordinate healthcare encounters proactively and encourage patients to adhere to their care plans, which include instructions on medications and lifestyle.

As Esposito explains, “Not everyone needs a care manager. It wouldn’t be efficient to deploy a care manager to everyone.”

More for you

Loading data for hdm_tax_topic #better-outcomes...