Objective: To investigate the accuracy of cognitive tests and clinical dementia diagnosis in distinguishing between mildly demented and nondemented subjects.
Design: Three-year longitudinal follow-up of a community-based cohort sample. Using the Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised as criterion variable, subjects were classified as demented and nondemented based on the results of cognitive test performance at time 1. These subjects were then examined prospectively for development of dementia at time 2.
Participants: The sample consisted of 63 mildly demented (Mini-Mental State Examination score > 18 < 24) and 398 nondemented subjects at time 1. At follow-up, due to death and refusal of participation, the sample consisted of 108 demented and 217 nondemented subjects.
Results: The group of subjects who were considered to be demented according to the results of the cognitive tests and nondemented by the clinical diagnosis at time 1 had a higher incidence of dementia at time 2 than did the subjects who were considered to be nondemented according to both cognitive tests and clinical diagnosis. Among the incident dementia cases, subjects considered to be demented according to the cognitive tests but nondemented by the clinical diagnosis performed at a lower level on most cognitive tests were older, had less education, and consisted of more women than did the group of subjects who were considered to be demented by the clinical diagnosis and nondemented by the cognitive tests at time 1.
Conclusions: A combination of cognitive tests, especially tasks assessing episodic memory, can detect many subjects in a preclinical state of dementia who could be missed in the clinical diagnostic procedure due to subjects' relatively high age, low education, and female sex. However, cognitive tests also may miss a sizable proportion of prevalent dementia cases due to possible insensitivity to the same demographic variables. These findings indicate the need of integration between clinical and cognitive data to increase the accuracy in detecting dementia in an early phase.