OBJECTIVE: To examine if method of case ascertainment affects estimates of the prevalence and severity of dementia in Medicare nursing home residents. METHODS: A nationally-representative sample of Medicare beneficiaries residing in skilled-nursing facilities (N=1100) was identified from the Medicare Current Beneficiary Survey (MCBS) for 2001. The MCBS contains detailed information from medical records and personal interviews on socio-demographics, health and medical conditions, and health care use. Survey information can be linked to Medicare claims, drug administration records, and Minimum Data Set (MDS). Dementia prevalence rates were determined using following four sources of diagnosis information, alone and in combinations: survey, MDS, Medicare claims, and drugs for dementia treatment. Concordance between sources for dementia diagnosis was measured as percent agreement and with kappa statistics. The severity of dementia cases from each source was determined using cognitive, physical, and behavioral functioning limitation measures from MDS. Chi-square tests were performed to identify statistically significant differences at p<0.05. RESULTS: Among four measures considered singly, the lowest dementia rates were obtained using drugs (12.4%) and the highest using claims (57.3%). Rates were higher when sources were combined and reached 68% using all four sources together. As for concordance, the percent agreement ranged from a low of 19.2% between claims and drugs to a high of 97.5% between drugs and claims with survey or MDS. Kappa statistics were the lowest between drugs and survey or MDS with claims (kappa 0.12), and were highest between survey and MDS (kappa 0.70). Sources were similar in the severity of dementia cases. CONCLUSIONS: Although there was a wide variation in prevalence and concordance of dementia cases by diagnostic source, there was no systematic bias based on disease severity. A combination of all four sources presents the most inclusive measure of disease prevalence available to researchers working on dementia in long-term care.