Objectives. To evaluate a wide range of sociodemographic, neurological and clinical variables as to whether they predict mortality in a representative sample of demented elderly.
Design. A three-stage community survey was conducted, based on a total of 3721 elderly patients whose cognitive status was assessed by their general practitioners (stage I). A strati®ed random sample of patients underwent a standardized research interview, including cognitive testing and the assessment of mental status, physical illness, sensory impairment and motor disability (stage II). After a mean interval of 28 months, all patients were recontacted. For deceased patients a close reference person was interviewed and the exact date of death was recorded (stage III). The in¯uence of the predictor variables on mortality was determined by using the Cox proportional hazards model.
Subjects. A strati®ed random sample of 117 patients in primary care (mean age 82 years) suering from mild, moderate or severe dementia (Alzheimer type, vascular or mixed dementia).
Materials. Hierarchical Dementia Scale (HDS), a modi®ed version of the Hamilton Depression Scale, other clinical rating scales and CAMDEX criteria for clinical diagnosis and a degree of severity of dementia.
Results. Fifty-three of the 117 demented patients had died during the follow-up interval. The mortality risk increased steeply with the degree of severity of dementia. By controlling for this variable, only age and motor disability contributed signi®cantly to the prediction of mortality, whereas gender, social class, type of dementia, extrapyramidal signs and other clinical features showed no or only a weak eect on the outcome.
Conclusion. The remaining life expectancy of the demented elderly depends primarily on the severity of the dementia, the patients' age and their general physical health. The in¯uence of other clinical features which often have been hypothesized as indicators of speci®c subgroups of dementia was mainly due to their relationship to the disease severity.