Kathleen Hayden, Oksana Makeeva, Ashley Dunham, Valentina Markova, Zara Melikyan, Irina Zhukova, Natalia Zhukova, Larisa Minaycheva, Stepan Buikin, Marina Abushaeva, Svetlana Burlutskaya, Ekaterina Karas, Elena Starinskaya, Yuka Maruyama, Heather Romero, Cassandra Germain, Brenda Plassman, Kathleen Welsh-Bohmer: Construct validity of MoCA subscales across community-based registries, from North Carolina to Tomsk, Russia // Alzheimer's & Dementia: The Journal of the Alzheimer's Association. Volume 9, Issue 4, Supplement, Pages P849–P850, July 2013
The Montreal Cognitive Assessment (MoCA) has been used internationally for the evaluation of cognitive performance in pre-clinical samples. As the field moves toward larger, multinational clinical trials in Alzheimer's disease prevention, it is important to evaluate MoCA performance as a phenotyping tool within the global context.
Using data from three community studies of normal aging (Durham, NC, Kannapolis and Cabarrus County, NC, and Tomsk, Russia), MoCA subscales for memory, executive function, attention, language, orientation, and visuospatial ability were derived. Measures and scales were normalized and standardized based on pooled sample means; results are presented as standard deviation units (SDU). Subscales were compared to established measures that are recognized as being sensitive to early cognitive changes: the CERAD Word List Memory Test delayed recall (WLM), Trail Making Test Part B (Trails B), and the self-report ADCS Mail-In Cognitive Function Screening Instrument (MCFSI). Multilevel modeling was used to evaluate associations between each of the measures separately with MoCA subscales, adjusting for age, sex, and education level.
Data from 2343 participants were analyzed. All were living independently, without diagnosed cognitive disorders in Durham, NC n=1073, Kannapolis/Cabarrus, NC n=469, and Tomsk, Russia n=801. The sample was 70.8% female, mean age 68.5 years, and most had a college education or higher (58.9%). Mean total raw scores differed significantly across sites: MoCA (Durham, 26.9; Kannapolis, 24.8; Tomsk, 23.2), WLM (Durham, 7.0; Kannapolis, 6.5; Tomsk, 6.3), Trails B (Durham, 92.3 sec.; Kannapolis, 108.0 sec.; Tomsk, 161.8 sec.), MCFSI (Durham, 2.5; Kannapolis 3.0; Tomsk, 4.9). The MoCA memory subscale most strongly predicted performance on WLM relative to other subscales, however there were significant differences in performance between Durham and Tomsk (p =.01). The executive subscale best predicted Trails B performance although the association was stronger for Durham compared to Tomsk (p =.02) The MCFSI was broadly associated with memory, executive, and orientation subscales with some variation across sites on orientation.
Overall, the general form of the associations between subscales, WLM, Trails B, and MCFSI demonstrate construct validity of the MoCA across sites. However, significant variations in performance suggest the need for norms that better represent the individual countries.