Predictors of mortality in 2,249 nonagenarians--the Danish 1905-Cohort Survey.

OBJECTIVES : To elucidate whether well-known predictions of mortality are reduced or even reversed, or whether mortality is a stochastic process in the oldest old. DESIGN : A multidimensional survey of the Danish 1905 cohort conducted in 1998 with follow-up of vital status after 15 months. SETTING : Denmark. PARTICIPANTS : All Danes born in 1905, irrespective of physical and mental status were approached. Two thousand two hundred sixty-two persons of 3,600 participated in this survey. MEASUREMENTS : Professional interviewers collected data concerning sociodemographic factors, smoking, alcohol consumption, body mass index, physical and cognitive performance, and health during a visit at the participant's residency. Cox regression models were used to evaluate predictors of mortality. RESULTS : Five hundred seventy-nine (25.7%) of the 2,249 participants eligible for the analysis died during the 15 months follow-up. Multivariate analyses showed that marital status, education, smoking, obesity, consumption of alcohol, and number of self-reported diseases were not associated with mortality. Disability and cognitive impairment were significant risk factors in men and women. In addition poor self-rated health was associated with an increase in mortality in women. CONCLUSION : In the oldest old, several known predictors of mortality, such as sociodemographic factors, smoking, and obesity, have lost their importance, but a high disability level, poor physical and cognitive performance, and self-rated health (women only), predict mortality, which shows that mortality in the oldest old is not a stochastic process.