Detection of dementia of the Alzheimer type in a population-based sample: neuropsychological test performance.

The ability to detect dementia of the Alzheimer type (DAT) in a community-dwelling sample of elderly individuals on the basis of neuropsychological test performance was examined. Three hundred sixty community-dwelling individuals were identified by neurological examination as having probable or possible Alzheimer's disease, being at risk for Alzheimer's disease, or having no cognitive impairment. A logistic model comprised of tests of verbal and nonverbal memory, mental flexibility, and confrontation naming correctly classified 82% of DAT subjects and 98% of normal elderly subjects. The logistic model classified 77% of subjects who were diagnosed as at risk for Alzheimer's disease as being cognitively normal. A cross-validation with a clinically based sample of subjects correctly classified 89% of DAT patients and 100% of normal control subjects. The results suggest that psychometric discrimination of dementia may be less accurate in community-dwelling populations than in clinically based samples.