Diagnostic accuracy of the Montreal Cognitive Assessment in screening for cognitive impairment in initially hospitalized COVID-19 patients: Findings from the prospective multicenter NeNeSCo study

Abstract

Objective: This study aimed to investigate the prevalence and nature of cognitive impairment among severely ill COVID-19 patients and the effectiveness of the Montreal Cognitive Assessment (MoCA) in detecting it. Method: We evaluated cognition in COVID-19 patients hospitalized during the first wave (March to June 2020) from six Dutch hospitals, nine months post-discharge, using a comprehensive multi-domain neuropsychological test battery. Test performance was corrected for sex, age, and education differences and transformed into z-scores. Scores within each cognitive domain were averaged and categorized as average and above (z-score ≥ -0.84), low average (z-score -1.28 to -0.84), below average (z-score -1.65 to -1.28), and exceptionally low (z-score < -1.65). Patients were classified with cognitive impairment if at least one domain's z-score fell below -1.65. We assessed the MoCA's accuracy using both the original cutoff (<26) and an optimal cutoff determined by Youden's index. Results: Cognitive impairment was found in 12.1% (24/199) of patients, with verbal memory and mental speed most affected (6.5% and 7% below -1.65, respectively). The MoCA had an area under the curve of 0.84. The original cutoff showed sensitivity of 83% and specificity of 66%. Using the identified optimal cutoff of <24, maintained sensitivity while improving specificity to 81%. Conclusions: Cognitive impairment prevalence in initially hospitalized COVID-19 patients is lower than initially expected. Verbal memory and processing speed are primarily affected. The MoCA is a valuable screening tool for these impairments and lowering the MoCA cutoff to <24 improves specificity.