Geographic variation tells us less than we think

On Monday, J. Michael McWilliams and a team of colleagues, including me, published an article in Annals of Internal Medicine that compares geographic variation in cancer-related imaging in the VA to Medicare. The article was editorialized by Michaela Dinan and Kevin Schulman for AnnalsKate Madden Yee wrote a good, brief summary for

Rather than write my own, I’m going to suggest you go read Yee’s. She quotes Michael with the one sentence takeaway:

While geographic comparisons can be useful for understanding trends in provider behavior, our study demonstrates that geographic variation is not necessarily a reliable indicator of the extent of overuse in a healthcare system.

Our finding that greater average efficiency isn’t associated with less variation is consistent with that of other studies across a variety of settings, which we cite: VA vs. Medicare in medication use (Gellad et al.), HMOs vs. unmanaged care (Baker et al.), Medicare Advantage vs. traditional Medicare (Matlock et al.). If it’s not a reliable signal of efficiency, perhaps it’s time to move beyond studies that simply demonstrate geographic variation.

UPDATE: In light of some emails, I thought it worth clarifying that our article doesn’t rebut the idea that there is more waste in higher spending areas. That there is variation in the VA system and in Medicare means there could be more waste in some areas than others, as we and others have found. But as the IOM report notes, targeting areas for policy doesn’t make sense even if that is so. It would be better to target provider organizations/facilities within a system and measure performance at an organizational or facility level rather than at an area level. What our article does rebut is the idea that knowing how much variation there is tells you how much waste there is. Greater variation could exist at a lower level of waste or vice versa.


Hidden information below


Email Address*