Purpose: Our objective in this study was to compare Medicare costs of treating older adults with and without dementia in nursing home settings. Design and Methods: An expert panel established the dementia status of a stratified random sample of newly admitted residents in 59 Maryland nursing homes between 1992 and 1995. Medicare expenditures per-person month (PPM) were compared for 640 residents diagnosed with dementia and 636 with no dementia for 1 year preadmission and 2 years postadmission. Multivariate analysis with generalized estimating equations was used to identify the source of Medicare cost differentials between the two groups. Results: Medicare expenditures peaked in the month immediately preceding admission and dropped to preadmission levels by the third month in a nursing home. Adjusted PPM costs postadmission for the dementia group as a whole were 79% (p < .001) of the Medicare costs of treating residents without dementia. For the subgroup of residents admitted without a Medicare qualified stay (MQS), those with dementia had Medicare costs of just 63% (p < .001) of those without dementia. Overall Medicare costs PPM were insignificantly different between the two groups admitted with a MQS. Implications: Whether nursing home residents are admitted with a MQS is the single most important factor in assessing treatment cost differentials between residents admitted with and without dementia. Failure to consider this factor may lead researchers and policy makers to misdirect their attention from the true source of the differential-dementia patients admitted without a qualifying stay.