Big Data comes to the Practicing Physician
Practicing physicians today are principally concerned with caring for persons who present to their practices. The process of evaluation, clinical decision making, care planning and execution as recorded in the medical record was developed centuries ago and has evolved little since. Thus our medical record is a primary source of data reflecting that care process.
Consider, for example, some of the earliest extant medical records— Forman and Napier’s casebooks from 17th century England. These astrologer-physicians recorded their conversations, observations, judgments, plans (and charges) in much the same way we continue to collect and record information during the care process today. What we may know about any one patient or the populations of patients these early practitioners served depends critically on how accurately their records reflect what actually happened at the point of care.
Whether applied in the context of individual patient visits or to populations of patients our ability to understand the care process is largely derived from the medical record. Until recent times all of the medical record content was, similar to Forman and Napier, manually collected and recorded. Over the past 50 years or so the evolution of electronic health data management has transformed both data collection and documentation processes as well as medical records themselves, culminating in today’s Electronic Medical (Health) Record. While the content derived from our processes of conversation, observation, and judgment remains much the same, automation of data collection, automated workflow support, and automation of ancillary processes and data has resulted in the accumulation of substantially more documentation in today’s electronic medical record compared with its paper predecessor.
Why automation of medical records and those processes that generate it? For centuries the paper chart supported one physician seeing one patient at a time but the shortcomings of paper were legion. Information sharing and communication across providers was limited, clinical decision support was rudimentary, information extraction to inform population management was difficult, and analytics was essentially non-existent. The requirement to improve value in our healthcare system—to ensure better clinical care and outcomes at more affordable costs—was simply not achievable with legacy paper-based processes and was not much improved by our early attempts to automate them.
The parallel evolution of data processing capability and its application in healthcare has enabled us to evolve integrated delivery systems and more recently to create transformational care models such as Medical Home and Accountable Care Organizations. The wide-spread application of electronic health records has, however, occurred only since the 2009 HITECH (“meaningful use”) Act. While a recognized opportunity years earlier prior to HITECH perhaps 15 percent of physicians and a similar number of hospitals had basic electronic systems. Today more than half of the nation’s eligible providers and about 80 percent of the nation’s eligible hospitals are participating in the meaningful use program—a remarkable increase in only five years.
“In the era of “Big Data” I suggest it is not how “big” but rather we should concentrate on what is missing, inaccurate, or irrelevant”
With widespread adoption and use of electronic health records we may infer that our goals—better care for more people at a more affordable cost—can be achieved. Along with enhanced documentation and process support, an electronic health record should enable clinical decision support, pervasive across-provider communication, and informed, engaged patients. We understand that we are on the verge of being able to leverage “Big Data” to help achieve these goals. But do the exponentially growing and dynamically accumulated data really represent what is happening in our patients' lives and health—“big” yes but are these the “right” data?
What we know about our patients today is largely derived from our documentation of their current care processes most generally from observations made several times each year during visits they have with their care providers. With automation we are now able to electronically extract data (“quality metrics”) reflecting our patients' state of health or disease. But do these electronically extracted terms accurately reflect their true state or rather do they reflect wide variation in the electronic documentation of their care? We have more experience with the extraction of data reflective of care in an acute setting. In spite of years of electronic documentation evolution most of us still resort to manual chart abstraction to ensure that our “core measures” reflect the actual care process. While similar efforts in physicians' offices are less mature emerging data suggest there is significant variation and data gaps between what is extracted electronically and what actually happened.
Further the use of electronic ambulatory data to help providers consistently do better (“clinical decision support”) has been disappointing. We have successfully improved process but evidence of improved clinical outcomes, clinical or operational efficiencies, or cost savings is hard to find.
Why should this be and why has our electronic documentation and process support fallen short of our goals? In the era of “Big Data” I suggest it is not how “big” but rather we should concentrate on what is missing, inaccurate, or irrelevant. The current care paradigm wedded to the medical record must be expanded to include near continuous access to information regarding patients' health or disease. Pervasive access to smart phone and “wearable” technology provides clues about how this might be accomplished. Lastly our automation and derivative data must more usefully and accurately mirror how clinicians think and act.
Will the pervasive adoption and use of electronic health records and resultant “big Data” truly enable better care and lower costs? There is reason to be hopeful.