Can accountable care drive precision analytics?

What has and will likely continue to keep providers focused on data-driven clinical interventions is “the spirit of accountable care,” which is to avoid overusing resources or using them inappropriately.

Approximately one-third of the healthcare provided in this country is wasteful, Richard G. Popiel, MD, executive vice president and corporate chief medical officer of Portland, Ore.-based Cambia Health Solutions, a nonprofit health insurance corporation, told attendees at America's Health Insurance Plans (AHIP) Institute & Expo 2017, in Austin, Texas.

Still, said Popiel, if providers are going to rely on evidence-based medicine-“which really can make a difference for patients”-they need better quality, cleaner data.

In his June 8 presentation, “Precision Analytics: How to Power Clinical Interventions,” Popiel pointed out that the U.S. healthcare industry is constantly trying to figure out how to work with healthcare providers to ensure that the care delivered to members and patients is safe and effective, but that’s no easy feat.

Predictive models that use historical data to “look in the rear-view mirror” can help determine who in the future might consume healthcare resources in order to prevent patients from getting sicker. But the challenge with these models-which he described as “blunt instruments”-is that they’re not very predictive, and they don’t accomplish the goal of focusing healthcare resources on the right individuals and optimizing the ability to provide safe and affordable care.

This isn’t easy work, he noted. While today’s predictive models are directionally correct, “they are far less optimal than we need,” he said. "That’s largely because this fragmented data comes from a variety of sources. For example, there’s lab data in one place and then there’s [electronic health record] data and claims data [in another place].”

It’s not merely a matter of “hiring someone who’s uber-smart,” he added. “We actually have 15 really smart NASA-like data scientists who are working on this. You have to build a way to fetch the data so that you can ingest it. You have to clean the data and store it in a way that’s consumable.”

Popiel referred to the concept of a data lake-typically understood as a repository that uses a flat architecture to store vast amounts of data in their original formats-as more of a “swamp,” in that the data found in such repositories is often “murky.” One of the most difficult aspects of predictive analytics is converting data into a format where it’s consumable by frontline staff who can apply the insight to improving patient care, he insisted.

Easily-understandable dashboards at care providers’ fingertips can help, he told Managed Healthcare Executive (MHE). For example, actionable data can be used by nurses in their treatment of diabetes patients.

In the past, a nurse would typically check in on her patient once a month to ensure that he was adhering to his treatment plan. With real-time, actionable insight into her patient, that nurse can determine whether her patient may be at greater risk for high blood pressure or coronary disease; the nurse also has access to claims-based data so she can learn when there are spikes in her patient’s claims, which can also improve the quality of care she provides.

Data-driven insight is particularly valuable in the treatment of Medicare patients, Popiel told MHE. For example, elderly, frail patients are more prone to falls and injuries, but it’s impossible to focus attention on all patients. With predictive analytics, providers can get very precise about which patients to focus on to help prevent falls.