A Big Idea—and Bigger Challenges

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon

Every once in a while, a big idea floats, like a sweet vapor, across the popular consciousness, invading every corner of US life, from science to commerce to entertainment. Curr-ently, our society (and business, in particular) is smitten with big data, and to be accurate, its reach is global, even unto health care. The lure is easy to understand. Take the case of biologist Pek Lum, PhD, working on a cure for cancer at a pharmaceutical company with a dataset that was a dozen years old. Lum found a correlation that she and fellow researchers had never seen before using a topology software that compresses relationships found in complex datasets into shapes more accessible to researchers than an ocean of numbers in columns would be.¹ She discovered that groups of patients thought to have molecularly similar cancer were not as alike as she thought and that others were not as dissimilar—and she went to work for the software developer. We spend a good deal of acreage in this issue exploring how using big data is different from using data in traditional ways, how radiology is using data to run departments and practices, and how data can be used to improve clinical practice. To date, however, the application of data analytics in radiology would be characterized as smaller than big and the challenges before it as bigger than small. Data Liberation The first challenge is one shared by all would-be users of big data in health care: A primary source of the data you deal with—personally identifiable health information—is highly regulated by the federal government and, therefore, fiercely guarded by each provider organization that collects the data. Consequently, these data are not freely shared. This is an access problem with implications for how data can be used. A second challenge is one that plagues any health-care data project: the ability to pull data from many disparate information systems that don’t speak the same language. In the language of health IT, this is interoperability. In the world of big data, however—for which a defining characteristic is the sheer variety of sources from which information is pooled—the answer is data governance. On March 4, 2013, at the annual Healthcare Information and Management Systems Society meeting in New Orleans, Louisiana, Kathryn A. Whitmore, MS, president of STS Consulting Group, spoke on this topic. She explains, “Data governance is not a committee; it is not a back-office function: We are talking about a coordinated set of processes that includes people, procedures, and processing—a top-down, cascading model that allows us to transform data thought to be owned by one individual into a corporate asset that is shared.” A properly instituted data-governance plan is not intended to lock up data, but to cleanse them and ensure that the people who can use the data have access to the data that they need to accelerate change. Yet another challenge is this: All leaders—but especially physicians steeped in the scientific method—will struggle to overcome their causation bias and embrace the ideas of correlation and relationship visualized in big-data projects. In their new book on big data, Mayer-Schonberger and Cukier write, “Society will need to shed some of its obsession for causality in exchange for simple correlations: not knowing why but only what. This overturns centuries of established practices and challenges our most basic understanding of how to make decisions and comprehend reality.”² An adjacent issue for organization leaders in all industries is the cultural shift inherent in replacing what wags call HiPPO, the highest-paid person’s opinion, with the willingness to defer to what the data indicate. An Evolutionary Path To deal with the access challenge, radiology departments and practices will need data ambassadors: highly skilled communicators who are able to articulate the need for data sharing across departmental and organizational borders—and to instill trust in partners. To address the data-governance challenge, health-care systems, departments, and practices will need a policy that is at once inclusive and disciplined. They will need data czars with absolute power. For physician leaders, in particular, it might be easier to adapt to correlation and pattern recognition if they are viewed as a coping mechanism—an evolutionary step in dealing with petabytes (rather than megabytes) of data. Take a moment to look beyond local information systems and imagine the variety of data