Control for area differences in population health (casemix adjustment) is necessary to measure geographic variations in medical spending. Studies use various casemix adjustment methods, resulting in very different geographic variation estimates. We study casemix adjustment methodological issues and evaluate alternative approaches using claims from 1.6 million Medicare beneficiaries in 60 representative communities. Two key casemix adjustment methods—controlling for patient conditions obtained from diagnoses on claims and expenditures of those at the end of life—were evaluated. We failed to find evidence of bias in the former approach attributable to area differences in physician diagnostic patterns, as others have found, and found that the assumption underpinning the latter approach—that persons close to death are equally sick across areas—cannot be supported. Diagnosis-based approaches are more appropriate when current rather than prior year diagnoses are used. Population health likely explains more than 75% to 85% of cost variations across fixed sets of areas.