I'm glad he didn't succeed. Because prescribing medication to patients (who might suffer from a serious illness) should be based on years of learning, carefully vetted information from medical peers, judicious thinking from the doctor and not some statistic models which can analyze dubious quality data.

What if your model has small errors? What if the data is wrong? What if the data is good but people you've payed cheaply to enter it in the database did some small mistakes? It's not like we are dealing with consumer goods, where small mistakes might not be so important. We are dealing with people lives and health, where any kind of mistake should not be tolerated.