Defining mild cognitive impairment: impact of varying decision criteria on neuropsychological diagnostic frequencies and correlates

Objective: To examine the impact of varying decision criteria on neuropsychological diagnostic frequencies and on their correlates.

Design: Descriptive and correlational study.

Setting: Florida Alzheimer's Disease Research Center.

Participants: A sample of 373 individuals with comprehensive baseline analyses participating in a longitudinal study of cognitive decline and early Alzheimer disease.

Measurements: Mild cognitive impairment (MCI) diagnoses were made on the basis of four sets of decision criteria created by crossing two approaches: varying the number of impaired test results required for a diagnosis within any domain (1 test versus 2) and varying the performance level required to determine impairment (1.5 or 2 standard deviations [SDs] below the normative mean) for any test.

Results: Under each criteria set, single-domain amnestic MCI was the most frequent MCI diagnosis. MCI global and subtype diagnosis frequencies were inversely related to the stringency of the criteria. The single test-1.5 SD criterion identified the largest number of cases as qualifying for an MCI diagnosis, and the two test-2.0 SD cutoff identified the fewest. Across all sets of criteria, the authors found significant positive associations between neuropsychological diagnoses and Clinical Dementia Rating score categories. Significant relationships between diagnoses and both apolipoprotein E (APOE) genotype and magnetic resonance imaging ratings of medial temporal atrophy (MTA) application were found only for the two test-1.5 SD and two test-2.0 SD cutoffs.

Conclusion: MCI diagnosis frequencies are substantively affected by the stringency of the criteria, but the relative rankings of MCI subtype diagnoses are fairly consistent regardless of the stringency of the criteria. Significant associations of neuropsychological diagnoses with independent markers such as APOE genotype and MTA are only found with more stringent criteria, suggesting that a coherent network of associations reflecting cognitive decline occurs with more restrictive definitions for impairment.