To demonstrate that artificial intelligence can have a better outcome for patients than real doctors, they worked with a database of 6700 patients with chronic physical disorders like diabetes, hypertension and cardiovascular disease.
They had 30 to 35 percent increase in patient outcomes, with a cost of treatment divided by more that two, compared to what doctors prescribed.

Though this is not one study that is going to change the art of prescribing of doctors, able to exercise their intuition in a ocean of uncertainties, it shows that big data increases in importance. Like in other fields of expertise, big data becomes easily handed. A few years ago it was unthinkable to treat a high quantity of diverse data using other methods than shortcuts and intuition.
Deep blue chess computer proved that it is possible to beat Kasparov, the best chess player ever, with a computer. More recently, Watson computer proved that it is possible to beat best Jeopardy! players. The day will come when best doctors will be challenged by artificial intelligence.

This is great, sure. But even if we have dozens of super computers giving the best diagnosis and prescriptions, the ultimate decision will always belong to a human being with medical knowledge. For thousands of reasons.
One reason is the liability of medical devices manufacturers. This kind of computer is a medical device. As such, the intended use claimed by the manufacturer will never be to decide the diagnosis or treatment. It will be sold as an aid for diagnosis or treatment.
One can argue that we already rely on computers to keep us alive, when we travel in a plane with electric flight commands or in a automatic subway. True. These machines give great services but they don't take decisions. It's the pilot in the plane or the operator in the subway command station who take decisions.

Super-computers analyze, doctors decide.