Objective To examine the effect of dementia on longer term survival after hospital admission, and to assess whether dementia is an independent predictor of mortality. This information is vital for the provision of appropriate care. Methods A prospective cohort study, in a large urban acute general hospital, of 616 people (70 years and older) with unplanned medical admission. The principal exposure was DSM‐IV dementia and main outcome mortality risk. Dementia severity was analysed by using the Functional Assessment Staging scale. We examined a range of modifying variables: acute physiological disturbance (Acute Physiology and Chronic Health Evaluation), chronic comorbidity (Charlson Comorbidity Index, CCI) and pressure sore risk (Waterlow score). Results A total 42.4% of the cohort had dementia. Nearly half (48.3%) had died 12 months after admission (median survival time 1.1 years compared with 2.7 years in people without dementia). Unadjusted hazard ratios for mortality in people with dementia was 1.66 (95% CI 1.35–2.04) and for people with moderately severe/severe dementia 2.01 (95% CI 1.57–2.57). After sequential adjustment (age, gender, Acute Physiology and Chronic Health Evaluation score, Charlson Comorbidity Index and Waterlow score), patients with dementia had a mortality risk of 1.24 (95% CI 0.95–1.60) and those with moderately severe/severe dementia 1.33 (0.97–1.84). Conclusions People with dementia had half the survival time of those without dementia. The effect of dementia on mortality was reduced after adjustment, particularly by the Waterlow score, a marker of frailty. The median survival of 1 year suggests clinicians should consider adopting a supportive approach to the care of older people with moderate/severe dementia who have an emergency hospital admission. Copyright © 2012 John Wiley & Sons, Ltd.