Atul Gawande has a great piece in a recent New Yorker [the abstract is here]. It talks about programs aimed at reducing costs for the patients that have the highest medical expenses. Megan McArdle has a great discussion of the piece on her blog.
The thing that struck me the most is the big reduction in cost some programs that offered support services to patients whose medical expenses reached millions of dollar a year were able to achieve. I wasn't the reading the article with my economics thinking hat on, so I hadn't thought about a control group. As Atul pointed out and my econ thinking cap came on even without a program we would likely expect that some expensive patients would have better years the next year (not get hit by a car or fall down) that is they would move back toward the mean medical expenses for an average person. So even without a program their expenses might fall. It highlights the importance of having a non-treated comparison group.
Related an economist friend said the best way to show regression to the mean is to took look at batting averages for baseball players up until the All Star Break and then their final average.