An item in FierceHealthIT pointed me to an intriguing paper from the Institute of Medicine, Making the Case for Continuous Learning from Routinely Collected Data:
The availability and reliability of large volumes of relevant longitudinal digital data from a variety of clinical and nonclinical sources are core features of a system that learns from each care experience, a learning health system. Common clinical repositories include data from electronic health record (EHR) systems used to manage patient care and claims data necessary for billing purposes. In some cases, data sources can be linked, using either institution-specific identifiers or matching algorithms, to create disease-specific patient registries that enable research. Integration of large pools of disparate clinical data from EHRs and claims is a major function of health information exchanges, which will be increasingly important to ensure seamless management of health information across institutions. Nonclinical sources of patient information may also include data from retail sales of over-the-counter medications, dietary supplements, walking and running shoes, and personal preferences and behaviors.
A lot of what I've seen written about Big Data is misleading and faddish, an opinion I share with some folks at Gartner Group, among others, but the thesis of this paper has the ring of truth. I'm only partway through the paper, but it makes a lot of sense so far.
It does, however, bump up against another concern I have: the tension between wanting to see more data captured and made use of in primary care, and the critical requirement for direct, human patient-physician engagement in the clinical encounter. It's a theme I've blogged about before, in various forms (see the Forget Everything You Know link below). I'll be posting more about that in the near future, due to a couple other bits of reading I'm ploughing through at the moment.