Abstract
The Montreal Cognitive Assessment (MoCA) is widely used to screen for cognitive impairment, yet commonly applied cutoff scores have been found to perform poorly among US Latinos. Prior studies relied on small samples, combining persons with mild cognitive impairment (MCI) and dementia into a single group, or failing to account for multiple intersecting demographic factors. We identified optimal MoCA cutoffs for MCI and dementia among US Latinos while addressing these limitations. We analyzed cross-sectional data from the National Alzheimer's Coordinating Center Unified Data Set. Participants included English- and Spanish-speaking Latinos who completed testing in their primary language. Research diagnostic groups consisted of cognitively unimpaired (CU), MCI, and dementia. Groups were further stratified by testing language, age, and level of education. Diagnostic accuracy and receiver operating characteristic (ROC) analyses were performed. The Youden Index was used to determine the optimal cutoff score. Of the 1,673 participants in the total sample, 46% completed the MoCA in Spanish and 54% in English, 54% were CU, and 28% had MCI and 19% had dementia. Area under the curve (AUC) values for CU vs. MCI were 0.70 for Spanish-tested participants and 0.79 for English-tested participants, while values for MCI vs. dementia were 0.85 and 0.89 for the Spanish and English tested participants, respectively. Overall AUC values were 0.76 for CU vs. MCI and 0.86 for mild cognitive impairment vs. dementia. Optimal cutoffs were consistently found to be lower among participants tested in Spanish, those older than age 75, and participants with the fewest years of education, with some optimal cutoffs shown to be substantially lower than the traditionally used standard cutoff. These findings provide cutoffs that better reflect differences amongst language and demographic groups. We also provide a scoring calculator for clinical and research use.