Reliability and validity of an algorithm for the diagnosis of normal cognition, mild cognitive impairment, and...

The traditional consensus diagnosis (ConsDx) of normal cognition, mild cognitive impairment (MCI), and dementia relies on the reconciliation of an informant-based report of cognitive and functional impairment by a physician diagnosis (PhyDx), and a neuropsychological diagnosis (NPDx). As this procedure may be labor intensive and influenced by the philosophy and biases of a clinician, the diagnostic algorithm (AlgDx) was developed to identify individuals as cognitively normal, with MCI, or dementia.

The AlgDx combines the PhyDx with the NPDx, using a diagnostic algorithm that provides cognitive diagnoses, as defined by the National Alzheimer Coordinating Center/Uniform Data Set nomenclature. Reliability of the AlgDx was assessed in 532 community-dwelling elderly subjects by its concordance with the ConsDx and association with two biomarkers, medial temporal atrophy (MTA) scores of brain magnetic resonance imaging scans, and Apolipoprotein E (ApoE)-epsilon4 genotype.

A high degree of concordance was observed between ConsDx and AlgDx with a weighted Cohen's kappa of 0.84. Concordance of the AlgDx to the same ConsDx categories ranged from 85% to 92%. Excellent discriminative validity was observed using AlgDx, MTA scores, and ApoE-epsilon4 allele frequencies, each of which distinguished subjects with amnestic MCI and dementia from normal subjects.

The AlgDx of normal cognition, MCI, and dementia is a valid alternative that reduces time, effort, and biases associated with the ConsDx. The inherent reliability of a fixed algorithm, together with its efficiency and avoidance of individual bias, suggests the AlgDx may be used in longitudinal, multisite clinical trials, and population studies of MCI and dementia.