The ability of clusters of psychometric profiles to predict cognitive
decline 3 to 5 years later was examined. Using Ward's method, research
ers conducted cluster analysis of age- and education-corrected Mayo Co
gnitive Factor Scale (MCFS) scores obtained at first evaluation for 37
6 normal elderly. Results revealed 4 of 16 cluster profiles were predi
cted to represent at-risk status for future cognitive decline; 258 of
the original sample completed psychometric reassessment. Membership in
an at-risk cluster did not powerfully predict subsequent medical eval
uation of cognitive or psychiatric disorder or status as lost to follo
w-up. Repeated measures analyses of variance of MCFS scores showed no
differential decline in learning or memory between normal and at-risk
clusters. Thus, in a nonclinical sample, relative cognitive impairment
s are not powerfully predictive of future cognitive decline.