There are surprisingly few longitudinal studies of the oldest old, but these studies are of high importance because the number of oldest old continues to increase in most countries and because of the uniqueness in this population. The aims of this study were to investigate how health, activities of daily living (ADL), and use of care change over time in the oldest old and to seek how differences in health and ADL affect survival of the oldest old. The study was longitudinal in design, and the participants were interviewed by trained nurses. A group of 300 persons was randomly selected from three age-groups; 86, 90, and 94. For the first phase, in 1999, 157 persons could and wanted to participate; from these 98 persons continued to participate in the second phase and 62 in the third. Repeated measures (general linear model=GLM) from the oldest old showed a decline in objective health and ADL with increasing age, but subjective health remained positive and stable. The use of formal help increased with age, and once the oldest old entered the old-age care system, it was rare that they returned to independent living. Analysis using a Cox regression model showed that health and ADL significantly predicted survival, but age did not.