I am perusing the usual morning email and run across "The value of Machine Learning in Value-Based Care" by Mary Hardy, Vice President of Healthcare for Ayasdi. Just like in every other case, the 'Machine Learning' they are shouting about is all corporate smoke and mirrors. I actually read the article (and did some research honestly) hoping to find a new and better way to accomplish things that need to be done. Maybe there was a big breakthrough in neural pathways or Inductive Logic Programming.

As it turns out, what the big corporate idiots are really talking about is a query on a database using aggregates. Any first year database developer could have taken the data given and answered the question asked. In fact, we at Sentia have the tools in place to ask that question without even the programmer, but I digress. The question was "What do patients who have had a total knee replacement who have the shortest length of stay have in common?" We could write that query in about 35 seconds, but that doesn't mean that anyone has ever written it before. Basically what they want is good outcomes, in this case short length of stay, correlated with non-obvious controllable factors. What they found is that patients who were given pregabalin, a drug used to mitigate the effects of shingles, obtained this better outcome.

While this has nothing to do with 'Machine Learning' it does sound like a bona fide, dyed in the wool medical breakthrough. …until you hear that there were four physicians who actually read the documentation that comes with pregabalin and believed that administering the drug prior to surgery would inhibit postoperative pain. It did. The point is that Mary's analytics (let's call a spade a spade and admit that there is no breakthrough here (unless it is by those four doctors who actually read the documentation (and probably love nested parentheticals)), much less Isaac Asimov turning at 6000 rpm in his casket) predicted nothing and actually had no value,in this case. If she had predicted the outcome before the pregabalin was administered and the surgery done, she might have had something. As it is that is kind of like me reading tea leaves and poking around in chicken innards in June of 1969 and then 'predicting' man would walk on the moon in our lifetimes. The smart kids were already doing the work necessary to get the outcomes they want. What Mary did was read some tea leaves and shuffle some tarot cards and have a junior developer tell her something the smart kids already knew.

What is the lesson here? Once again, big corporate entities are really good at telling you things you already know and making you think they have reinvented the wheel. As we've said before real innovation comes from small teams of dedicated people who work hard and achieve great things. Mary's thinking is certainly in the right place, but without the technical background to actually do the work, she is hopelessly outclassed and doesn't even know that Miss Cleo (RIP) is lighting candles and shuffling cards and generally putting on a show (well, not anymore) to make people believe that they've accomplished 'Machine Learning.' Yes, we know this is a tough, touchy subject. You, dear reader, should take away the fact that most of what you read on this subject is complete crap and you should do the deep dive and trust people who don't misuse technical terms and try to sell you on their "innovation" like Mary Hardy just did. We provide