[Skip to Navigation]
August 1994

Screening for Cognitive Impairment in Older Individuals: Validation Study of a Computer-Based Test

Author Affiliations

From the Neurobehavioral Program, Department of Neurology, Emory University School of Medicine, the Wesley Woods Center (Drs R. Green and J. Green and Ms Harrison); and the Division of Biostatistics, Emory University School of Public Health, Atlanta, Ga (Dr Kutner).

Arch Neurol. 1994;51(8):779-786. doi:10.1001/archneur.1994.00540200055017

Objective:  This study examined the validity of a computer-based cognitive test that was recently designed to screen the elderly for cognitive impairment.

Design:  Criterion-related validity was examined by comparing test scores of impaired patients and normal control subjects. Construct-related validity was computed through correlations between computer-based subtests and related conventional neuropsychological subtests.

Setting:  University center for memory disorders.

Participants:  Fifty-two patients with mild cognitive impairment by strict clinical criteria and 50 unimpaired, age- and education-matched control subjects. Control subjects were rigorously screened by neurological, neuropsychological, imaging, and electrophysiological criteria to identify and exclude individuals with occult abnormalities.

Results:  Using a cut-off total score of 126, this computer-based instrument had a sensitivity of 0.83 and a specificity of 0.96. Using a prevalence estimate of 10%, predictive values, positive and negative, were 0.70 and 0.96, respectively. Computer-based subtests correlated significantly with conventional neuropsychological tests measuring similar cognitive domains. Thirteen (17.8%) of 73 volunteers with normal medical histories were excluded from the control group, with unsuspected abnormalities on standard neuropsychological tests, electroencephalograms, or magnetic resonance imaging scans.

Conclusions:  Computer-based testing is a valid screening methodology for the detection of mild cognitive impairment in the elderly, although this particular test has important limitations. Broader applications of computer-based testing will require extensive population-based validation. Future studies should recognize that normal control subjects without a history of disease who are typically used in validation studies may have a high incidence of unsuspected abnormalities on neurodiagnostic studies.

Add or change institution