These days, virtually all hospitals and health systems are looking at ways to manage population health. Most of their approaches assume that it’s a matter of identifying the right big data tools and crunching the numbers, using the data already in-house. Doing this may be costly and time-consuming, but it can be done using existing databases, integration engines and the appropriate business analytics tools, or so the conventional wisdom holds.
However, at least one health IT leader disagrees. Adrian Zai, MD, clinical director of population informatics at Massachusetts General Hospital, argues that current tools designed to enable population health management can’t do the job effectively. “All of the health IT tools companies call population health today will be irrelevant because the data they look at can only see what goes through hospital, which is far too narrow in scope.”
Zai points out that most healthcare organizations attempt to leverage claims data in doing population health management analyses. But that approach is far from ideal, he told Healthcare IT News. Claims data, he points out, is typically one to two months old, which significantly limits the value healthcare providers can generate from the data. Also, most hospitals’ claims data only covers about 20% to 30% of the area’s population, he notes.
Instead, organizations need to study real-time data drawn from a significantly broader population if they hope to achieve population health management goals, Zai argues. For example, it’s important to look at the Medicaid population, whose members may get most of their care through community health centers. It’s also important to collect data from other consumer touch points. (Zai doesn’t specify which touch points he means, but mobile health and remote patient monitoring data come to mind immediately.)
I think Zai make some excellent points here. In particular, while achieving true real-time analysis is probably well the future for most healthcare organizations, the fresher data you can use the better. Certainly, analyzing archival data has a purpose, but to have a major impact on outcomes, it’s important to foster behavior change in the present.
However, I’d argue that few providers are ready to roll ahead with this approach. After all, to achieve his goals means establishing some new definitions as to what data should be included in population health analysis. And that’s not as simple as it sounds. (For a recent look at how providers look at population health, check out this survey from last summer.)
First, providers need to take a fresh look at how they define the term “population,” and develop a definition that takes in a more comprehensive view of patient data. Certainly, claims data analysis is start, but that by definition is limited to insured patients seen at the hospital. Zai recommends that population health management efforts embrace all patients seen at the hospital, insured or not. In other words, he’s recommending hospitals address the community in which they are physically located, not just the community of patients for whom they have provided care.
Just as importantly, hospitals and health systems need to consider how to collect, incorporate and analyze the exponentially-growing field of digital health data. While some middleware solutions offer to serve as a gateway for such data, it seems likely that providers will still need to do a lot of hands-on work to make use of these data sources.
Finally, providers need to continually improve the algorithms they use to pinpoint problems in a given population, as well as the ways in which they create actionable subsets of the population. For example, it may be appropriate to target patients by disease state today, but other ways of improving outcomes might arise, and providers’ IT solutions need to be flexible enough to evolve with the times.
Over time, the industry will evolve best practices for population health management, and definedthe IT tools best suited to accomplish reasons. And while some existing tools may work, I’d be surprised if most survive the transition.