On disentangling states versus traits: demonstration of a new technique using the Alzheimer's disease assessment scale.

Part of the challenge in research on degenerative neurologic disease relates to distinguishing those measurements that essentially describe patient characteristics stable across the course of illness (traits) from those that vary systematically within subjects (states), particularly those specifically related to stage or duration of illness. A components-of-variance approach was used to examine the state versus trait aspects of the Alzheimer's Disease Assessment Scale (ADAS) Cognitive and Noncognitive subscales, a clinical instrument frequently used in research on Alzheimer disease. Subjects were 190 patients with probable AD followed longitudinally. Stage of illness was indexed by mental status scores. Analysis of variance was used to partition total variance into that associated with subjects (trait), stages (state: stage), subjects x stages (state: other), and error. ADAS Cognitive scores were strongly related to stage of illness (83% of true variance). ADAS Noncognitive scores were modestly related to stage (approximately 21% of true variance) and moderately related to state: other (47%). We discuss how state-trait analyses can be helpful in focusing attention on those areas of assessment most likely to accomplish specific objectives.