OBJECTIVES: To examine differences in the prevalence of other diseases for patients with and without Alzheimer's Disease (AD), and the independent effect of AD on cost, beyond the expected cost of treating other medical conditions. METHODS: We used MarketScan Medicare Supplemental and Coordination of Benefits claims data, 2003-2004, for over-age-65 individuals with Medicare and employer-sponsored plans including drug benefits. We identified AD patients by an AD diagnosis or an exclusively prescribed AD medication in 2003, and selected a demographically-matched, non-demented Control (3:1 ratio to AD). Applying Diagnostic Cost Groups (DCGs), a comprehensive disease classification and prediction system, we calculated a prospective relative risk score (RRS) that predicts 2004 costs from non-AD illnesses in 2003; the mean RRS of all individuals enrolled for all of 2003 and =1 month in 2004 was set to 1.00. We used regression to estimate AD's independent effect on cost overall and among patients with selected non-AD conditions. RESULTS: The AD Cohort (n = 25,109) is sicker (mean RRS 1.23 vs. 1.04) than the Controls (n = 75,327); they have more comorbidities (mean of 8.1 unique medical conditions vs. 6.5) and 34% higher costs ($13,936 vs. $10,369). However, excess annual costs attributable to AD are estimated to be only $2307, with outpatient pharmacy being the key driver ($1711 in excess costs). The AD Cohort has more diabetes, heart, mental health, injuries, vascular, and urinary problems. For patients with certain comorbidities (such as, anxiety disorders), excess costs attributable to AD reach $6000. CONCLUSION: AD patients are sicker than demographically similar patients, yet cost even more than is accounted for by their excess morbidity. Much of the additional cost is due to greater use of outpatient pharmaceuticals.