[cc81]

But it does not have to be that extreme. Let us take a made up example where there are two categories of patients. One that are healthy enough that a positive outcome of surgery is almost guaranteed, almost all survive. If you operate on people like that you will have excellent stats.

However let us add someone that is really sick and will most likely die within 6 months if they don't get surgery. They are in bad shape so surgery has a 25% risk of killing them. Most patients would probably want to take that gamble but for the surgeon who operates on patients like that frequently that would mean that their stats would go from almost perfect to pretty bad. You don't want to surgery from a guy with where 10% of his patients die from heart surgery when you can get another guy where only 0.1% die.