The Perils of Metrics

Got blood?  Here's an elegant example of the positive power of data-informed decision-making.

Transfusions save lives, but they are also expensive and not without risks.  Each decision whether or not to transfuse is the output of some decision-making process, presenting the opportunity to reach a better balance by improving the algorithm.  It may one day be possible to run every lab test that might be relevant at a reasonable cost and feed the results into an extremely accurate ML model developed at a reasonable cost.  But we aren't there yet.

So these researchers built a system to compare readily available lab results to previously-developed guidelines, then inform the doctor whether or not a transfusion was indicated.  The data did not drive the decision, and the primary goal of great patient outcomes was not altered.

The result:  "By simply asking doctors to think twice about transfusions, the hospital had not only reduced costs, but also improved patient outcomes."

http://www.nature.com/news/evidence-based-medicine-save-blood-save-lives-1.17224