Big data? Big deal!
The health care industry’s understandable excitement surrounding big data (Wow! the volume, the variety, the velocity) is gradually transforming into a queasy discomfort (Ugh! the volume, the variety, the velocity). And it’s not difficult to understand why. From electronic health records, claims data, health information exchanges, digital imaging, research studies and prescription data to every physician note, test result and vital sign on desktops, iPads, and smartphones, our digital footprint is expanding at a daunting rate.
And now that the significant regulatory and financial incentives are finally rounding into shape, healthcare institutions are wasting little time in embracing the digital infrastructure to support big data in healthcare. We now see rapidly accelerating mobile device adoption by clinicians and a slew of ever increasing health 2.0 apps for patients.
We see the broad emergence of health information exchanges that enable the data to follow the patient. Progress indeed. Of course, we should all applaud the amazing strides taken over recent years to ensure that the digital foundations for a future healthcare system are rapidly being put in place. What’s missing? The deep insights that can streamline clinical workflows, connect all the dots at the point of care, strengthen doctor-patient relationships, cut costs, and improve outcomes.
That’s because, as an industry, we’re only scratching the surface of the potential of big data. When you have this much healthcare data available, albeit still in multiple silos, the challenge shifts from, “We don’t have the data to make that decision,” to “We have the data to make that decision… It’s somewhere in this growing mountain.”
But with the right analytics, we can start to mine for the precious ore that will allow us to see crucial trends and patterns at both a macro and a detailed level. We will be able to more easily identify and better understand those key factors that can we can routinely use to predict the likelihood of adverse events, such as avoidable readmissions or hospital acquired infections – events that have a huge negative impact on quality, cost, and outcomes.
What can really make big data a “big deal” is when we start to move analytics from a few retrospective PowerPoint slides at a monthly department meeting – how did we perform and how can we improve? – into the individual patient encounter in the treatment room – how will we perform to improve this patient’s outcome? Patient-centered analytics bring the global correlations down to specific, actionable insights that improve healthcare one patient at a time through earlier (and more accurate) clinical intervention.
Those individual successes come from narrowing our focus to the areas that matter most. But how do we achieve that and deliver on the clinical promise inherent in big data? It starts with identifying the factors – the predictive dimensions – that are more likely to have an impact on a patient’s health. Inherent in the nature of big data is that we can now access massive data sets that can integrate dimensions we could never feasibly consider previously.
Of course, that implies the need to sift through hundreds or thousands of variables in real time to come up with the right answer. Instead, a well-built analytical model – devised, trained, and tested by both clinical experts and data scientists – can often pinpoint the “critical few” variables that contribute to the majority of the predictive capacity required for clinical decision support. And that means support for real-time diagnostics and interventions. In this exciting model, we move away from the institution- and physician-centric care model of today toward an appropriate continuum of care focused on the individual needs of the patient, which also encompasses thoughtfully planned pre- and post-visit interventions. For instance, the hospitalist, case manager, and primary care physician can work from the same playbook to help avert readmissions by coordinating analytic-driven recommendations for high-risk congestive heart failure patients who’ve been identified as good candidates for highly targeted, individualized sequences of interventions and outreach during the days leading up to and following hospital discharge.
We also have the opportunity to apply this model to the financial side of the picture as well. For years, other industries such as consumer packaged goods and financial services have been refining and applying these kinds of big data analyses to optimize their performance and giving us a generous body of best practices to draw from in our industry as well. Imagine the ability to integrate a stronger clinical evidence-based foundation into reimbursement processes. This could make pay-for-performance a discipline with a demonstrably greater rational foundation.
Even as we start to get a sense of the possibilities, it’s still time for us to gain a little “20/20 vision” and sketch out what we want to be achieving at the end of this decade. Collectively, we need to start making the expanded investments and share best practices to make the gains we’re capable of. As patients become more digitally connected, more savvy, activated consumers of healthcare services and the volume of available data grows, we owe it not only to our patients but to ourselves and ultimately our country to ensure that all stakeholders are prepared to derive the maximum value that this opportunity provides.
Let’s light the blue touchpaper and watch this analytic rocket fly.
Graham Hughes, MD, is chief medical officer, SAS Center for Health Analytics and Insights.