OBJECTIVES: The impact on mortality is a major driver in determining the cost-effectiveness of diagnostics and treatment of dementia disorders. We examined the relationship between the degree of dementia severity and mortality in a longitudinal, population-based study in Sweden. METHODS: From a total sample of 1810 subjects aged 75 years or older, 211 were identified as having a clinical diagnosis of dementia at baseline and were included in the study. Disease severity was assessed with the Mini-Mental State Examination (MMSE) as well as the Clinical Dementia Rating (CDR), administered at baseline and again at two follow-up visits after approximately 40 and 80 months respectively. Mortality data was obtained from the national death statistics, 10 years post baseline. Survival analysis was conducted using Weibull regression with baseline as well as time-varying covariates. Age and gender were also included as covariates in addition to dementia severity. RESULTS: A total of 198 deaths were observed during the observation period, and the time to death was 935 days on median. Annual mortality rates in females were estimated to 12%  for mild dementia (MMSE 21-26), 15% for moderate dementia (MMSE 10-20) and 19% for severe dementia (MMSE 0-9) at baseline. The corresponding estimates for males were 19%, 24% and 31% respectively. Each point lower result on the MMSE scale was associated with a decrease in survival by 2.5%. There was no statistically significant relationship between baseline CDR scores and mortality, though a trend was seen towards increasing mortality in more severe CDR states. Similar results were observed in an analysis incorporating changes in disease severity over time. CONCLUSIONS: Mortality in subjects with dementia increase with the severity of dementia, as measured by the MMSE. Incorporating differential survival by disease severity has important implications for the long-term cost-effectiveness of diagnosis and therapies for dementia disorders.