Prediction of post-stroke cognitive impairment by Montreal Cognitive Assessment (MoCA) performances in acute stroke: comparison of three normative datasets

Abstract

Background: Cognitive assessment in acute stroke is relevant for identifying patients at risk of persistent post-stroke cognitive impairment (PSCI). Despite preliminary evidence on MoCA accuracy, there is no consensus on its optimal score in the acute stroke setting to predict PSCI. Aims: (1) To explore whether the application of different normative datasets to MoCA scores obtained in the acute stroke setting results in variable frequency of patients defined as cognitively impaired; (2) to assess whether the normality cut-offs provided by three normative datasets predict PSCI at 6–9 months; (3) to calculate alternative MoCA cut-offs able to predict PSCI. Methods: Consecutive stroke patients were reassessed at 6–9 months with extensive neuropsychological and functional batteries for PSCI determination. Results: Out of 207 enrolled patients, 118 (57%) were followed-up (mean 7.4 ± 1.7 months), and 77 of them (65%) received a PSCI diagnosis. The application of the normality thresholds provided by the 3 normative datasets yielded to variable (from 28.5% to 41%) rates of patients having an impaired MoCA performance, and to an inadequate accuracy in predicting PSCI, maximizing specificity instead of sensitivity. In ROC analyses, a MoCA score of 22.82, adjusted according to the most recent normative dataset, achieved a good diagnostic accuracy in predicting PSCI. Conclusions: The classification of acute stroke patients as normal/impaired based on MoCA thresholds proposed by general population normative datasets underestimated patients at risk of persistent PSCI. We calculated a new adjusted MoCA score predictive of PSCI in acute stroke patients to be further tested in larger studies.