Paradox

Everyone, including this blog writer who has been touting the virtues of the great discoveries of data already or will soon be in the electronic health record (EHR), are available, which heralds the learning health [1, 2]. It is sometimes unbridled enthusiasm that the data collected in the clinical systems, perhaps combined with the research data, such as gene sequencing, will supply us easily knowledge about what works in health care and as new treatments [3, 4] are developed. The data is unstructured? No problem, just use natural language processing [5].
I honestly must share in this enthusiasm, but I also know that it can be alleviated, or at least at a dose of reality. In particular, we must remember that our great and data analysis algorithms only get us so far. If we have poor underlying data, the analysis may end up misleading us. We must be careful to problems of data incompleteness and inaccuracy.
There are all sorts of reasons for a lack of data in EHR systems. Probably the most important is that those data, ie, physicians and other clinicians, you usually give it for reasons other than the data analysis. I've often said that the clinical documentation may be what stands between a busy clinician and home for dinner, so he or she charting terminate before the end of the work day.
I also know many doctors, whose enthusiasm for the entry of correct and complete data is tempered by their view of the input data as a black hole. That is, they give the data in but never derive its benefits. I like to think that most doctors would be an opportunity to look at the aggregate view of their patients in their practice and / or patient, the outliers in a measure or another to enjoy identify. But a common complaint I hear from doctors is that the data collection priorities are more attempt by the hospital or clinic to maximize their refund than to support clinicians in better patient care driven ..