Occasionally there are estimates that are specific to a particular intervention. In terms of a general estimate, you have just reminded me that there have been one or two studies out of the U.S. that have looked at this and tried to generalize it. Generally, though, they attracted a great deal of criticism from academics for poor methodology.
You're right. It's very difficult to argue that non-compliance of X does not lead to additional costs on the system of Y. I would say that it's next to impossible.