There is no magic with big data and machine learning techniques. If the data have an efficient information, then forming a reasonable conclusion is possible.
The simple answer to this question as Mr. Sarsam has explained earlier would be getting enough information about patient demographics, symptoms and medical history to determine the correct diagnosis. While data can help with this. We are moving ahead one step where computing power and large databases help make relevant treatment suggestions.
A few examples of this kind of technologies would be IBM Watson Health and Flatiron products which help in cancer analytics. I believe you will find interesting answers in their white papers.