OBJECTIVES: Compare transitions between mild, moderate and severe health states, and to death or institutionalization for Alzheimer disease (AD) and non-AD dementia patients. METHODS: The National Alzheimer Coordinating Center's Uniform Data Set (NACC-UDS) is a large, longitudinal dataset funded by the National Institute of Aging that includes AD and non-AD dementia patients, and non-demented controls. Using the mini-mental state examination (MMSE) we classified patients as mild (27-20), moderate (19-10) or severe (<10). We calculated the association between previous and current MMSE stage and the association between previous MMSE stage and institutionalization using multinomial logistic models with random effects to account for individual and center level correlation, controlling for demographic characteristics and time since last observation.  The coefficients from the regressions were used to calculate predicted probabilities using the population means for each of the covariates. RESULTS: Our analysis was limited to 3,418 patients with dementia (52.3% probable AD) and two or more observations with complete data for our covariates of interest.  Average baseline age was 76.72 years old. Average MMSE was 17.05. The majority were female, 50.4%. The regression coefficient for the previous MMSE stage was significant in all cases and suggests that being in a higher MMSE stage at the previous visit is associated with being in a higher MMSE stage or dead. Previous MMSE stage was found to be strongly associated with institutionalization. White, non-Hispanic, unmarried patients were more likely to be institutionalized, all else equal.  Patients with non-AD dementia were more likely to be institutionalized than patients with AD. CONCLUSIONS: Patients with AD in the NACC -UDS database transition more quickly to more severe stages of MMSE than non-AD dementia patients.  Non-AD dementia patients are more likely to transition to institutionalization and die than patients with AD.