With all due respect to Jimmy Buffet, the real opportunity for progress on the use of data to predict, steer, or bemoan our ongoing care and cost conundrum lies not in the latitudes but in the longitudes.
By way of brief background, for quite sometime the bedrock on which most data analytics programs were based was (and in many cases, is) claims data. Paid claims data, no matter how ugly, can at least be a predictor of future cost. Moreover, claims data reflects the soul of how we historically have delivered health care services, one ICD 9 code at a time, and how we diagnosed patients based upon their physiologic conditions at the moments in time when they were standing mostly naked in front of their doctors.
Skeptics about using claims data for health analytics purposes rightfully assert issues around the three “A’s”- Age, Aggregation and Accuracy of the data. To which I would add three “I’s” Integration, Information and Institutionalization. Claims data is not always timely available, not always fully complete (particularly in migratory populations such as Medicaid patients who may move in and out of health coverage), and may not always be accurate due to billing errors, coding opportunities and other challenges. We certainly receive plenty of “clean claims” which appear to have covered mammography for males, as an example. (Of course, I defer in cases when this is appropriate.)
My concerns also revolve around the fact that claims are only one part of a patients’ story, and quite frankly the least important. To be fair, claims can include pharmacy data, and some evidence of quality standards met or missed. But even then, if the central object of this exercise is to answer the fundamental question, ‘How healthy is this patient?” claims data alone leaves us wanting more.
Moreover, claims data is simply a reflection of the institutional practice of medicine – it’s a bill after all. I would suggest there are times when claims data is no more reflective of a patient’s health than a lawyers’ bill is reflective of the likelihood of successful outcome in a case. Now obviously, claims data are far more useful than legal bills, but you get my point.
What is needed is the opportunity to augment claims data, which in the best circumstances are helpful, along with other longitudinal data – labs certainly – but also physiologic data collected on a daily or at least regular interval.
We stand at a moment – smartphone in hand, iPad at the ready – when many of our patients are mobile, multimedia, and now monitored, measured and compared (or at least capable of being compared) with other patients. Some would suggest this is the beginning of personalized health care. (Which it is, though, in subsequent blogs, I will argue that personalized health care may still be a decade away.) But for the moment consider this, health care can and is using phone, video conferencing, email, social networking web sites (Patients Like Me or Diabetes Mine for instance), GIS, tweets, on-body/in-body sensors, digestibles, implantables, wearables, and now medical body area networks.
Each of these communication tools collect ongoing, longitudinal patient data. In this evolving “there-is-an-app-for-that medicine,” investments by HHS through the Office of the National Coordinator are pushing providers toward a digital health enterprise containing electronic health records, patient-controlled records and health information exchanges.
We are on the verge of creating zetabytes of medical data. But this explosion in data for the moment runs ahead of our legacy systems’ claims-based analytic architecture. (By the way, it also runs ahead of our legacy-based, episodic care medical architecture.)
So what is needed to turn longitudinal data into information, knowledge and wisdom?
First we need integrated technology platforms that can absorb active or passive collected data from diagnostic/monitoring devices, and then combine that information with more traditional sources of health, pharmaceutical and lab data.
Second, we need a new suite of data mining and integrative analytics and network solutions.
Third, we need new performance and outcomes analysis metrics. That is, we need to turn our focus to measuring outcomes rather than process.
Fourth, we need the output of this analysis to be translated to bedside practice, producing more targeted care and increasingly efficient use of resources.
Which, of course and interestingly enough is exactly what this new forum is all about.
Carl Taylor is a Partner at the Fraser Institute for Health Research, Princeton, and Assistant Dean in the University of Southern Alabama’s College of Medicine.