The intersection of population health, DSS and ACOs

BOSTON--“ACOs want to do population health, but it’s not that easy to do,” said Hadi Kharrazi, MHI, MD, PhD,  Johns Hopkins University School of Public Health and School of Medicine, speaking at the Medical Informatics World Conference on April 28.

For ACOs to succeed, they need good population decision support that utilizes a swath of real-time data that are not just from an EHR. “Population health IT is on the frontier of health IT solutions,” he said. “It’s not just about collecting data in the EHR.”

Meaningful Use (MU) data is a good start, but not enough for predictive analytics, he said.

With data mapping, organizations can link MU and ACO measures to help streamline reporting. MU Stage 1 covers 35 of the 40 National Committee for Quality Assurance’s ACO measures, while MU Stage 1 covers 20 of the Centers for Medicare & Medicaid Services' 33 ACO measures. While such data can help calculate how well an ACO is performing, they fall short of the type of data necessary for predictive analytics, which is a major success factor of ACOs.

Meanwhile, one of the key processes to support population health within ACOs is cohort management, which requires robust clinical decision support, the ability to monitor individual patients and patient engagement, among other functions.

Data to drive effective analytics must derive from a variety of sources on the physician side and on the patient side. From the physician, it could include data from EHRs, claims data, national datasets and clinical decision support systems. Data from the patient side could include those from mHealth apps, social networks and social data.

“The easiest place to run decision support systems is through HIEs,” Kharrazi said, noting that maybe it’s the answer to their sustainability struggles.

Kharrazi illustrated how the Indiana HIE pulled in EHR, claims, imaging, laboratory, registry and personal health record data into its enterprise master patient index.

The HIE encompasses 22 hospital systems, with 70 hospitals, five large medical groups and clinics, laboratories, imaging centers and five payers, which handle 11 million unique patients. “We literally had billions of rows of data,” he said.

Utilizing these data, the organization was able to run automated, real-time, HIE-based population decision support systems.  

“You can look at the entire population of 70 hospitals in one diagram and you can see the gaps and differences. And it’s all automated and in real time,” he said.

At the Center for Population Health IT (CPHIT) at Johns Hopkins, Kharrazi is working to advance state-of-the-art health IT across public and private health organizations. This entails a focus on the application of EHRs, mobile health and other health IT tools targeted at communities and populations.

Kharrari said the center is working with Maryland’s HIE, the Chesapeake Regional Information System for our Patients (CRISP), to pull in data and conduct real-time analytics to predict readmissions.

“We are now calculating readmission risk based on zip code,” he said, and this information gets passed along to primary care providers.

CPHIT also is working with Maryland as it moves to a population health model. Unique to the state, Maryland negotiates rates directly with CMS on behalf of all hospitals. In January, the state announced a five-year plan outlining its move to value-based care, and capitated the all-payer per capita growth rate to 3.58 percent with the goal of $330 million savings.

As such, the incentive to maximize the potential of predictive analytics is greater than ever.