DETECTION OF COGNITIVE IMPAIRMENT AFTER ACUTE ISCHEMIC STROKE: VALIDATION OF BRIEF AND MORE COMPREHENSIVE COGNITIVE SCREENING INSTRUMENTS AND IMPLICATIONS FOR BEST PRACTICE

Abstract

Cognitive impairments after cerebrovascular accident (CVA) are associated with greater disability, mortality, and rising dementia rates over time. When evaluating screening approaches, including Hachinski's 2006 NINDS-CSN harmonization standards (1) and a MoCA 5-Minute protocol (M5M) (2), we posit that best cognitive screening practice will be (a) feasible in busy clinical settings, (b) sensitive and specific for deficits predicting practical outcomes, and (c) valid despite CVA-related limitations (e.g., aphasias, visuomotor deficits). Our group has been developing such an approach. 222 patients admitted for CVA were administered all or part of the LASS-I (Lexington Acute Stroke Scale – Inpatient), a 20-minute cognitive screen and received Barthel Index (BI) and Modified Rankin Scale (MRS) ratings. M5M scores were extracted from the LASS-I. M5M scores predicted both BI (R2 = 0.16, F(1, 153) = 28.63, p < .001) and MRS (R2 = 0.16, F(1, 161) = 30.32, p < .001). LASS-I score added predictive value for the BI (R2 = 0.22, F(2, 152) = 21.37, p < .001) and some predictive value for the MRS (R2 = .18, F(2, 160) = 17.06, p < .001). After analyzing individual domain-based contributions, combining best domain-based performances from the M5M (Executive Functions/Language) and LASS-I (Visuospatial/Construction & Praxis) was comparable to administering either instrument in full when predicting BI (R2 = .20, F (2, 144) = 18.35, p < .001) and MRS (R2 = .18, F (2, 152) = 16.62, p < .001). Effect sizes for all regression models were medium by Cohen's f2. ROC curve analysis yielded areas under the curve for the M5M at 0.74 (MRS) and 0.73 (BI), for the LASS-I at 0.79 (MRS and BI), and for the combined approach at 0.77 (MRS) and 0.72 (BI).