Objective: To explore whether prevalence, survival, and incidence of dementia have changed from 1987-1994 to 2001-2008 in Stockholm, Sweden. Methods: This study is based on 2 cross-sectional surveys of people aged 75 years or over conducted in central Stockholm: the Kungsholmen Project (KP) (1987-1989, n = 1,700) and the Swedish National study on Aging and Care in Kungsholmen (SNAC-K) (2001-2004, n = 1,575). In both surveys we diagnosed dementia according to DSM-III-R criteria, following the identical diagnostic procedure. Death certificates were used to determine survival status of KP participants as of December 1994 and SNAC-K participants as of June 2008. We used logistic and Cox models to compare prevalence and survival, controlling for major confounders. We inferred incidence of dementia according to its relationship with prevalence and survival. Results: At baseline, 225 subjects in KP and 298 in SNAC-K were diagnosed with dementia. The age- and sex-standardized prevalence of dementia was 17.5% (12.8% in men; 19.2% in women) in KP and 17.9% (10.8% in men; 20.5% in women) in SNAC-K. The adjusted odds ratio of dementia in SNAC-K vs KP was 1.17 (95% confidence interval 0.95-1.46). The multiadjusted hazard ratio of death in SNAC-K vs KP was 0.71 (0.57-0.88) in subjects with dementia, 0.68 (0.59-0.79) in those without dementia, and 0.66 (0.59-0.74) in all participants. Conclusions: Prevalence of dementia was stable from the late 1980s to the early 2000s in central Stockholm, Sweden, whereas survival of patients with dementia increased. These results suggest that incidence of dementia may have decreased during this period.