The validity of brief screening cognitive instruments in the diagnosis of cognitive impairment and dementia after first-ever stroke

Abstract

INTRODUCTION: Brief cognitive tests such as the Mini-mental State Examination (MMSE) and the Informant Questionnaire for Cognitive Decline in the Elderly (IQCODE) have been used to detect cognitive impairment and dementia in studies of stroke patients. However, there are few data on their validity for such use. We have evaluated their validity in detecting cognitive impairment not dementia (CIND) and dementia in a community-based sample of first-ever stroke patients. METHODS: The standardized MMSE (S-MMSE) and the 16-item IQCODE were administered to 79 patients 1 year after a first-ever stroke. CIND and dementia were diagnosed independently using a comprehensive cognitive battery. The performances of the two tests were evaluated using receiver operating characteristic (ROC) analyses. Combined performance was evaluated when their scores were used in parallel (the "or rule"), in series (the "and rule") or as a weighted sum (the "weighted sum rule"). RESULTS: Both tests were extremely poor at detecting CIND individually and in combination. For dementia, at traditional cut-points, the S-MMSE (< or =23) was insensitive (0.50, 95% CI 0.16-0.84) and the IQCODE (> or =3.30) nonspecific (0.63, 95% CI 0.51-0.75). An acceptable balance between sensitivity and specificity was achieved for dementia using the "or rule" combination, but with only modest positive predictive value. CONCLUSIONS: The S-MMSE and the IQCODE were individually poor at detecting CIND and dementia after a nonaphasic first-ever stroke. The combination was useful in detecting dementia but it does not replace the need for detailed neuropsychological tests.