A quarter-century of research on geographic variation in Medicare costs has failed to find any positive association between high spending and better health outcomes. We conducted this study using a 5 percent random sample of Medicare beneficiaries with diabetes or heart failure in 2006 and 2007 to see whether there was any correlation between geographic variation in Part D spending and good medication-taking behavior-and, if so, whether that correlation resulted in reduced Medicare Parts A and B spending on diabetes and heart failure treatments. We found that beneficiaries residing in areas characterized by higher adjusted drug spending had significantly more "therapy days"-days with recommended medications on hand-than did beneficiaries in lower-spending areas. However, we did not find that this factor translated into short-term savings in Medicare treatment costs for these two diseases. This result might not be surprising, since returns from medication adherence can take years to manifest. At the same time, discovering which regional factors are responsible for differences in drug spending and medication practices should be a high priority. If the observed differences are related to poor physician communication or lack of good care coordination, then appropriately designed policy tools-including accountable care organizations, medical homes, and provider quality reporting initiatives-might help address them.