Big data in healthcare splashed onto the front page of the Wall Street Journal earlier this summer, heralding its arrival as a new and important topic for mainstream media to follow.
In reality, we’ve had big data in healthcare for as long as I’ve been around. What we haven’t had are the tools for effective and actionable analytics across our data.
But we do now.
That's why I think this new era of big data in healthcare should more appropriately be dubbed "big analytics." With more information available to us and better software and hardware to digest and synthesize it, our real challenge is to shape a vision for big analytics that will improve all aspects of healthcare delivery without drowning us in meaningless metrics.
Our first steps to this vision require us to abandon the "silo think" we have innocently and comfortably adopted. We have grown to view healthcare IT as supporting clinical, financial and operational decisions. Are these decisions separate and independent? I argue not.
To use an example from my administration days, a seemingly simple clinical problem — the occasional failure to order indicated imaging studies out of the emergency department — was actually rooted in complex clinical, financial and operational causes.
In our analysis of the issue, transporter staffing levels during day shifts turned out to be the root cause. Financial data had led us to tightly manage transporter hours to reduce costs. Operationally, we reinforced this behavior by rewarding managers to limit overtime.
When there were more than the expected call-offs for a shift, patient transports slowed, and scheduled imaging for admitted patients backlogged into the evening (just when the emergency department imaging utilization was peaking). This led to marked delays in treatment in the emergency department.
Of course, such delays were unacceptable because this angered patients and subverted a separate operational goal of improving patient satisfaction in the ED. Mindful of our patient satisfaction metrics, some clinicians, we discovered, were simply skipping indicated imaging when their clinical suspicions were low.
What does this have to do with big analytics?
Big data encompasses disparate sources, structured and unstructured, clinical, financial and operational, as well as extramural sources. When properly applied to these data, big analytics should unveil relationships previously unknown to the decision-maker. Often these are subtle but significant relationships — like that between transporter staffing levels and clinical decision-making.
When we analytically examine the data, we can provide the decision-maker with all the consequences (financial, patient satisfaction, employee engagement, clinical quality) of being short two transporters on the 7 – 3 shift. This is predictive analytics.
When big analytics reaches maturity, we should expect it to provide weighted solutions for the decision-maker. For example, we can show lost margins from missed MRIs balanced against the costs associated with overtime or increased staffing levels to recommend a course of action to decision-makers. This is prescriptive analytics.
Big analytics should also ultimately head off problems. Are staffing threats aligning? Who’s on vacation? Are we close to a holiday weekend? When critical components converge, we should expect big analytics to sense and actively warn the decision-maker. This is preventive analytics.
The data and analytic tools for this new era are already available. It’s our vision that will ultimately determine how we transform healthcare with big analytics.