Recent studies suggest that delirium is associated with risk of dementia and also acceleration of decline in existing dementia. However, previous studies may have been confounded by incomplete ascertainment of cognitive status at baseline. Herein, we used a true population sample to determine if delirium is a risk factor for incident dementia and cognitive decline. We also examined the effect of delirium at the pathological level by determining associations between dementia and neuropathological markers of dementia in patients with and without a history of delirium. The Vantaa 85+ study examined 553 individuals (92% of those eligible) aged ≥85 years at baseline, 3, 5, 8 and 10 years. Brain autopsy was performed in 52%. Fixed and random-effects regression models were used to assess associations between (i) delirium and incident dementia and (ii) decline in Mini-Mental State Examination scores in the whole group. The relationship between dementia and common neuropathological markers (Alzheimer-type, infarcts and Lewy-body) was modelled, stratified by history of delirium. Delirium increased the risk of incident dementia (odds ratio 8.7, 95% confidence interval 2.1–35). Delirium was also associated with worsening dementia severity (odds ratio 3.1, 95% confidence interval 1.5–6.3) as well as deterioration in global function score (odds ratio 2.8, 95% confidence interval 1.4–5.5). In the whole study population, delirium was associated with loss of 1.0 more Mini-Mental State Examination points per year (95% confidence interval 0.11–1.89) than those with no history of delirium. In individuals with dementia and no history of delirium (n = 232), all pathologies were significantly associated with dementia. However, in individuals with delirium and dementia (n = 58), no relationship between dementia and these markers was found. For example, higher Braak stage was associated with dementia when no history of delirium (odds ratio 2.0, 95% confidence interval 1.1–3.5, P = 0.02), but in those with a history of delirium, there was no significant relationship (odds ratio 1.2, 95% confidence interval 0.2–6.7, P = 0.85). This trend for odds ratios to be closer to unity in the delirium and dementia group was observed for neuritic amyloid, apolipoprotein ε status, presence of infarcts, α-synucleinopathy and neuronal loss in substantia nigra. These findings are the first to demonstrate in a true population study that delirium is a strong risk factor for incident dementia and cognitive decline in the oldest-old. However, in this study, the relationship did not appear to be mediated by classical neuropathologies associated with dementia.