Original ArticleCog-4 Has Limited Diagnostic Test Accuracy and Validity for Cognitive Assessment in Stroke Survivors

BackgroundGuidelines recommend cognitive screening for all stroke survivors but do not suggest a preferred tool. Certain elements (orientation, executive function, language, and inattention) of the impairment scale, National Institutes of Health Stroke Scale (NIHSS), have been suggested as a short cognitive screening test-Cog-4. We aimed to describe accuracy and validity of Cog-4 against a more detailed cognitive assessment (Montreal Cognitive Assessment [MoCA]).MethodsWe assessed consecutive acute stroke unit admissions in 2 hospitals over 3 months. Four independent blinded assessors performed NIHSS and MoCA between days 1 and 4 poststroke. We described test properties of Cog-4 for MoCA-defined cognitive impairment using usual thresholds (Cog-4 ≥ 1 and MoCA < 26 of 30) and described the correlations of individual Cog-4 components with broadly equivalent MoCA domains.ResultsWe assessed 173 participants; 166 had Cog-4 data and 148 MoCA. MoCA described 84% (n = 124) of assessed participants as having cognitive impairment and the Cog-4, 37% (n = 62). Cog-4 had a sensitivity of .36 (95% confidence interval [CI]: .28-.45) and a specificity of .96 (95% CI: .80-.99) (positive predictive value: .98, negative predictive value: .23) for MoCA-defined cognitive impairment. Individual Cog-4 items correlated with certain MoCA domains, but the strength of association was modest (r = −.44 orientation, −.37 language, −.19 for inattention, and no significant correlation for executive function, P = .72).ConclusionsOur data suggest that many stroke survivors with MoCA-defined cognitive problems would not be detected by Cog-4. Subtest correlations suggest that Cog-4 may not be a valid measure of the cognitive domains that it purports to describe. Other brief cognitive screening tests may be better suited to acute stroke.