Diagnostic accuracy of cognitive screening tools under different neuropsychological definitions for poststroke cognitive impairment

Abstract Objectives The accuracy of cognitive screening tools to detect poststroke cognitive impairment (PSCI) was investigated using various neuropsychological definitions. Methods Hospital‐based stroke patients underwent a comprehensive neuropsychological assessment. The rate of PSCI was estimated using thresholds of 1, 1.5, or 2 standard deviations below the normal control and memory impairment defined by a single or multiple tests. Meanwhile, the diagnostic accuracy of cognitive screening through face‐to‐face assessment using the Mini‐Mental State Examination (MMSE) and the Montreal Cognitive Assessment Scale (MoCA), and telephone assessment using a 5‐minute NINDS‐Canadian Stroke Network (NINDS‐CSN) scale and a six‐item screener (SIS), was both tested under different definitions, with the optimal cutoff selected based on the highest Youden index. Results In stroke patients, the rate of PSCI ranged from 46.3% to 76.3% upon different definitions. The face‐to‐face MoCA was more consistent with the comprehensive cognitive assessment compared to MMSE. The optimal cutoff of PSCI was MMSE ≤ 27 and MoCA ≤ 19. For the telephone tests, the 5‐minute NINDS‐CSN assessment was more reliable, and the optimal cutoff was ≤23, while for SIS ≤ 4. Conclusions Cognitive screening tools including the face‐to‐face MMSE and MoCA, together with the telephone assessment of NINDS‐CSN 5‐minute protocol and SIS, were simple and effective for detecting PSCI in stroke patients. The corresponding threshold values for PSCI were 27 points, 19 points, 23 points, and 4 points.