‘Big data’ is already here

When Thomas Insel, MD, recently announced he was stepping down from his post as director of the National Institute of Mental Health, some were surprised that he was leaving to join Google. But in reality, the move makes sense. Insel himself described in a blog post that big data companies, such as Google, have a keen interest in the $3 trillion healthcare system.

Of course, the term “big data” gets thrown around a lot these days, and for behavioral health providers, it might seem like a distant strategy far off in the future. Many providers are still taking baby steps with “small data” elements in electronic health records, so any thought of distilling terabytes of data into their daily work might seem far-fetched. But big data is already here—even for behavioral health.

In general, big data is essentially a combination of information that might come from several sources in several different forms, and it’s typically used for specific and complex analyses. For example, an acute care hospital might run an analysis of years’ worth of emergency room data to find the common variables among patients with asthma who were admitted for an inpatient stay. That’s big data.

Additionally, many large healthcare systems are starting to form consortia to combine their clinical data sets and run even larger analytics projects to discover patterns that might be clues as to how they might improve care. The federal government is now funding the creation of networks to gather large amounts of healthcare data as a first step toward the development of what is known as “a learning health system,” in which providers and researchers routinely share data to learn more about prevention and treatment.

And this is just the tip of the iceberg in big data.

Substantial archive

One promising example of big data research in behavioral health is the Knowledge Network, a partnership led by the Centerstone Research Institute (CRI) that involves behavioral healthcare providers, researchers, policymakers and analytics experts who have created a national data warehouse. Tom Doub, PhD, CRI’s chief executive officer, says the network brings together community mental health providers interested in technology, research and policy work.

“We realized that we had this remarkable resource in terms of data available to us,” he says. “We developed data-sharing arrangements, deidentified the data to be compliant with HIPAA, and built a data warehouse with half a million patients, 4.5 million prescription records, 2 million diagnoses, and 23 million service records. It is a very substantial archive of mental health data.”

Doub says researchers are starting to probe the data to look at variations in care.

“For people presenting with the same diagnosis of schizophrenia or depression, the kind of care they get varies a great deal,” he says. “That is probably not a surprise to most people in the field, but we haven’t had good ways to talk about it. We’re trying to understand what that means.”

Combining information into big data projects has great value for research and policy purposes. Doub has been surprised to find, for example, radically different care approaches between two organizations in the same state, with the same payer, same regulatory requirements and same cultural environment.

“And there is so little transparency around that,” he says.

Behavioral healthcare is often described as a cottage industry, and part of the reason is there is no transparency about provider performance. Such transparency is already occurring in medical health disciplines and seems imminent for behavioral health’s future.

“There are going to be outcome standards and public report cards we have to be accountable for, and as we move toward pay-for-performance models, in order to survive, providers are going to have to do a good job of creating value and managing populations,” Doub says. ‘You can’t just deal with people coming in the door. You have to go out and find people who are at risk and make sure you keep them out of the emergency room.”

Starting small

But there is optimism. Doub says organizations that want to calculate an analysis of their own data on a smaller scale now have more options than a few years ago. He says that Centerstone, CRI’s parent company, has a large amount of internal data and resources to mine that information and has increasingly been able to capitalize on it.

“But when we started, there weren’t a lot of good roadmaps for how to do that,” he says. “There were roadmaps in other industries, and that’s where we looked for inspiration. But there were no out-of-the-box solutions to address the data questions we had.”

The best place to start, Doub says, is in establishing one data point to investigate through big data analysis, rather than trying to align a lot of misaligned information.

“Trying to reconcile why two versions of the same spreadsheet are different is a colossal waste of time, energy and money,” he adds.

Networks of behavioral providers are finding ways to pool their data to look for actionable insights, however. For instance, the Florida Council for Community Mental Health used a commercialized analytics tool developed at CRI to analyze aggregate, deidentified data from its provider members statewide and identified cost, productivity and outcomes findings.

Another positive change Doub has noticed is that more managed care companies are open to sharing data with behavioral health providers, such as hospitalization data that a treatment center might not have access to otherwise.

“We can use that data to develop models to predict which individuals are at greatest risk of going into the hospital in the next 30 days and deploy reports and tools to our clinicians so that they can better manage their case load,” he says.

National data projects

A relatively new national initiative, the Patient Centered Outcomes Research Institute (PCORI), is establishing networks that will gather data, and several of the participants are in the behavioral health space.

For instance, the aim of the $1 million Mood Patient-Powered Research Network hosted by Massachusetts General Hospital is to bring together at least 50,000 patients with mood disorders to participate in studies powered by data contained in electronic medical records and patient-recorded outcomes. PCORI’s aim is to better understand which treatments work better comparatively and for whom.

Another example of a PCORI-funded big data research project in behavioral health is the three-year, $1.6 million Community Partnered Participatory Research Network being set up in Los Angeles County and New Orleans. A behavioral health data infrastructure will support research to look more broadly at the social determinants of health as they relate to depression.

Bowen Chung, MD, an assistant professor in residence in the department of psychiatry at the David Geffen School of Medicine at the University of California Los Angeles, says previous research studies on depression in Los Angeles neighborhoods have shown that broad sociodemographic factors often play as significant a role in outcomes as medications and therapy. Research also indicates that providing resources through churches and barber shops can have positive health outcomes.

“In the county of Los Angeles, we are working to track 175,000 to 200,000 individuals across county services, which include the sheriff, jail, housing, income supplements, traditional health, mental health and substance use,” Chung says.

The research network will track major policy initiatives that impact issues such as homelessness for behavioral health clients across all county systems. It might be challenging to define which interventions are truly considered as healthcare, Chung says, but the project will be able to start describing the relationships between the social determinants and health.

“Big data is going to be valuable to help us think about approaches to incorporating other social determinants, because we don’t yet know how to conceptualize it in a very sophisticated way,” he says, “especially for behavioral health.” 

Big data could help state and local governments create synergies across all their budget categories to create a more rational system that saves money and provides better care, he adds.

Heading up the New Orleans site work is Benjamin Springgate, MD, MPH, associate professor of clinical medicine and public health at Louisiana State University Health Sciences Center. He says the project will cull large, deidentified data sets from health systems and the Medicaid program, and eventually from federally qualified health centers and public hospitals, which are important sources.

 “Many of the people with serious behavioral health diagnoses end up in a public hospital or federally qualified health center, or end up receiving their care in prisons for a number of years,” Springgate says.

He also co-leads a smaller PCORI initiative that facilitates discussions about behavioral health within a church community. It is a stepping stone to future patient-centered research.

“By working with smaller organizations, we can enroll people in research studies who would never have found their way into larger data sets,” he says. “They are not going to go the ‘Mayo Clinic equivalent’ and find their way into a large data set.”

The future is now

Springgate says that smaller behavioral health organizations might not believe big data is relevant to them. However, they might not realize that they are already participating in big data research indirectly.

“They are participating by the fact that they accept insurance,” he says. “Insurers, particularly managed care companies, are very good aggregators of data and are good at finding ways to translate the data they aggregate into care programs, quality initiatives or value programs.”

Local collaboratives might give more behavioral health organizations a voice, and some might find that the intelligence being discovered about patient care elsewhere is relevant and applicable to their own patients.

David Raths is a freelance writer based in Pennsylvania.