Abstract
Background: The expected growth in dementia prevalence mandates an efficient and cost-effective triage system for timely tertiary prevention. Fast and accurate cognitive screening assessments such as the Montreal Cognitive Assessment (MoCA) are essential tools in the system, although the effects of education posted some challenges. We investigated the potential use of item-response theory analysis to improve measurement precision of MoCA in a culturally homogeneous sample of older persons with diverse educational background. Methods: Data were available from 1,876 community-dwelling Cantonese-speaking older persons aged 60 years or above in Hong Kong. Participants were randomly sampled from public rental housing estates territory-wide. Bayesian information criterion was used for model selection and any formal education was used as a grouping variable to estimate multiple group item-response theory models. Results: A total of 852 (45%) of the participants had no formal education. For people without formal education, reliability of the MoCA sum score and ability estimates was 0.79 and 0.83, respectively; that for people with any formal education was 0.73 and 0.78, respectively. Ability estimates provide a closer match in percent classified for the 2nd, 7th, and 16th percentile than cutoff scores. Item functioning of Cube, Clock Number, and Clock Hand was superior in people without formal education. Conclusions: Scoring using ability estimates is more reliable than sum scores. Certain items contain higher item information in people with no formal education. Item response theory analysis should be incorporated in future IT applications of MoCA to improve measurement precision across populations with different education level.