Abstract
The object of the study was the assessment of the mortality risk for persons in a representative two-wave community sample assessed longitudinally. In the first cross-section a total of 358 (89.1%) subjects of Munich, Germany, aged 85 years and above were interviewed by research physicians. One year later 263 (73.5%) persons were reexamined. Death certificate diagnoses were obtained after an interval of 4 years 8 months. 58% of the total sample were deceased. Sociodemographic factors, mental disorders, subjective health status and need for care were analysed in relation to mortality by Cox regression. The probability of death was increased in those diagnosed as having a dementia or depressive disorder, in those of increasing age, living in institutions, being in need of care and of bad health status. In the multi-variate Cox regression model the influence of these different factors was examined and evaluated. Need for care was the most powerful predictor of mortality.