OBJECTIVES: To characterize people of advanced old age in their last year of life and compare those dying in their late 80s with those dying aged 90 and older to inform policy and planning. DESIGN: Retrospective analysis of prospectively collected population-based data from the Cambridge City over-75s Cohort (CC75C) Study, United Kingdom. PARTICIPANTS: Men and women aged 85 and older at death who died less than 1 year after taking part in any CC75C survey (N=321). MEASUREMENTS: Physical health, functional disability, self-rated health, cognitive status. RESULTS: Functional and cognitive impairments were markedly higher for those who died aged 90 and older- predominantly women-than for those who died aged 85 to 89. At least half (49.4-93.6%) of subjects aged 90 and older needed maximum assistance in virtually every daily activity; those aged 85 to 89 needed this only for shopping and laundry. Disability in basic and instrumental activities rose from 59.1% before to 85.4% after the age of 90 and cognitive impairment (Mini-Mental State Examination score < or =21) from 41.7% to 69.4%. Despite this and proximity to death, 60.5% and 67.0%, respectively, rated their health positively. Only one in five reported needing more help. CONCLUSION: This study provides new data identifying high levels of physical and cognitive disability in very old people in the year before death. As the very old population rises, so will support needs for people dying in extreme old age. The mismatch between health perceptions and functional limitations suggests that these vulnerable older adults may not seek help from which they could benefit. These findings have major policy and planning implications for end-of-life care for the oldest old.