This a three part blog post. Emphasis of Big Data acquisition and analysis is supposed to improve healthcare. We emphasize that healthcare collaboration is a way to deal with the massive amounts of data considered to be medical knowledge that has grown beyond mastery of anyone. That said, we also are concerned that attempts to improve quality and profitabilility remove focus from care and caring. This first post will describe a source of error. The second will show how this error can happen in clinical practice -- with healthcare collaboration or not. The third will suggest a solution. The whole of the three should bring fair warning to those who hear the sirens of Big Data which is done poorly, and look to solutions which are done well.
Here is the problem. Many companies are moving from providing big data services for government to big data services for healthcare.
These companies need to understand what they are getting into.
We, as health care providers, need to realize that we affect the outcome of big data as we provide care alone or as we are part of healthcare collaboration.
Since healthcare collaboration is about taking care of the problem at hand, together, and patient care is about helping another human being either get better, be comfortable, or die peacefully, data entry and codes have no relevance.
On the other hand, we providers are asked to be data input specialists even though we don’t see a benefit for our patients. We have the suspicion that we are working for government -- and thus a boss who regulates; or insurance, and thus a boss who underpays.
And as happened in the office last week, we are sorely tempted to underperform as data input specialists. Thus, all of you big companies who have migrated from the TSA, CIA and NSA with your Big Data tools should be forewarned. Any of you providers, who see no value in coding precision, should consider the consequences.
Let’s use rounding errors as an example. Then compare the rounding errors to what we sometimes are tempted to (or need to) do as clinicians.
Kees Vuik of the Delft Centre for Computational Science and Engineering clearly and understandably highlights some disasters caused by numerical errors. These range from a Patriot Missile failure which killed 28 soldiers after it missed intercepting a Iraqi Scud missile, to errors on conversion of the Euro. In essence, they all (and there are many other examples) have the same root cause: the magnification of minute inaccuracies of a number by repeated calculation.
The next post, Part 2, will describe in detail how simple rounding errors will bring Big Data, Bad Data...