Accurately interpreting change in cognitive functioning is an essential aspect of clinical care for older adults. Several approaches to identifying ‘true’ cognitive change in a single cognitive measure are available (e.g., reliable change methods, regression-based norms); however, neuropsychologists in clinical settings often rely on simple score differences rather than advanced analytical procedures especially since they examine multiple test performances. This study sought to establish quick-reference normative criteria to help neuropsychologists identify how frequently significant change occurs across multiple cognitive measures in cognitively normal older adults.

Data were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Participants were 401 older adults who were classified as cognitively normal at baseline and at 24-month follow-up. In ADNI, these clinical classifications are made separately from the assessment of cognitive performance, including cognitive change. The sample was 50.1% female, 93.5% non-Hispanic White, 4.0% non-Hispanic Black, 1.5% Asian American, and 1.0% other race/ethnicity, with a mean age of 76.0 years (SD = 4.9). Mean education was 16.4 years (SD = 2.7). The cognitive battery included: Boston Naming Test, Category Fluency Test, Trails A & B, Clock Drawing Test, and Auditory Verbal Learning Test, Trial 1-5 Total and Delayed Recall. Change scores between baseline performance and 24-month follow-up were calculated for each measure. The natural distribution of change scores was examined for each measure and cut points representing the 5th and 10th percentile were applied to each distribution to classify participants who exhibited substantial declines in performance on a given measure. We then examined the multivariate frequency of statistically rare change scores for each individual.

As expected in a normal sample, overall cognitive performance was generally stable between baseline and 24-month followup. Across cognitive measures, 43.6% of participants had at least one change score fall below the 10th percentile in the distribution of change scores, and 21.9% had at least one score below the 5th percentile. 13.0% of participants had two or more change scores that fell below the 10th percentile, in comparison to 4.5% with two or more below the 5th percentile. 3.2% of participants had three or more change scores below the 10th percentile, versus 0.5% of participants who had three change scores below the 5th percentile.

Among cognitively normal older adults assessed twice at a 24-month interval with a battery of seven measures, it was not uncommon for an individual to have at least one score fall below the 10th percentile (43% of the sample) or even the 5th percentile (21%) in the natural distribution of change scores. However, only 3.2% of normals had more than two declines in test performance below the 10th percentile, and less than 1% of the sample at more than one change score at the 5th percentile. This suggests that individuals who exhibit more multivariate changes in performance than these standards are likely experiencing an abnormal rate of cognitive decline. Our findings provide a preliminary quick-reference approach to identifying clinically significant cognitive change. Future studies will explore additional batteries and examine multivariate frequencies of change in clinical populations.