Factors associated with inconsistent diagnosis of dementia between physicians and neuropsychologists

OBJECTIVE: To explore reasons for discrepancy in the diagnosis of cognitive impairment between physicians and neuropsychologists. DESIGN: Retrospective analysis of national survey data. SETTING: Canadian Study of Health and Aging Phase 1, a national survey of community-dwelling and institutionalized older Canadians. PARTICIPANTS: 1879 subjects who completed all components of a clinical examination. MEASUREMENTS: Data available to both disciplines (demographic data, functional status, Modified Mini-Mental State (3MS), schedule H of the Cambridge Mental Disorders of the Elderly Examination (CAMDEX)), results of the physician's history and physical examination, and results of a psychometric test battery. Subjects were classified as No Cognitive Impairment, Cognitive Impairment Not Dementia (CIND), and Dementia, the latter according to the criteria of the Diagnostic and Statistical Manual, 3rd Ed., Revised. Preliminary diagnoses by physicians and neuropsychologists were compared. RESULTS: In univariate modeling, higher education increased consistency. Lower scores on the 3MS, depression reported in the CAMDEX, focal neurological signs, and all neuropsychological variables decreased agreement. In multivariate modeling, higher education and identification of long-term memory impairment by the neuropsychologist increased agreement; lower scores on the 3MS, depression reported on the CAMDEX, and identification of short-term memory impairment or constructional impairment led to disagreement. When the category of CIND was removed, kappa for agreement increased from 0.51 to 0.92. CONCLUSIONS: Physicians and neuropsychologists have different, complementary approaches to the diagnosis of dementia, and a consensus approach should be used. The category of CIND requires elucidation. Identification of dementia in subjects with depression or low education is difficult, and new strategies are required.