Objective: To establish quick-reference criteria regarding the frequency of statistically rare changes in seven neuropsychological measures administered to older adults.
Method: Data from 935 older adults examined over a two-year interval were obtained from the Alzheimer's Disease Neuroimaging Initiative. The sample included 401 cognitively normal older adults whose scores were used to determine the natural distribution of change scores for seven cognitive measures and to set change score thresholds corresponding to the 5th percentile. The number of test scores that exceeded these thresholds were counted for the cognitively normal group, as well as 381 individuals with mild cognitive impairment (MCI) and 153 individuals with dementia. Regression analyses examined whether the number of change scores predicted diagnostic group membership beyond demographic covariates.
Results: Only 4.2% of cognitively normal participants obtained two or more change scores that fell below the 5th percentile of change scores, compared to 10.6% of the stable MCI participants and 38.6% of those who converted to dementia. After adjusting for age, gender, race/ethnicity, and premorbid estimates, the number of change scores below the 5th percentile significantly predicted diagnostic group membership.
Conclusions: It was uncommon for older adults to have two or more change scores fall below the 5th percentile thresholds in a seven-test battery. Higher change counts may identify those showing atypical cognitive decline.
Keywords: Alzheimer’s disease; Cognitive aging; cognitive dysfunction; longitudinal studies; memory; neuropsychological tests.