Differential item functioning analysis of Montreal Cognitive Assessment Scale on educational level

Abstract

Objective To explore how much impact that the educational level may exert on the Montreal Cognitive Assessment Scale (MoCA) for screening mild cognitive impairment (MCI) among the old people. Methods 3 230 participants aged 60+ years in 6 districts of Xiamen were selected using multi-stage stratified randomized sampling. One-way ANOVA and Scheffe were performed using SPSS 23.0 to compare the MoCA scores among different educational level groups. Binary logistic regression was used to analyze differential item functioning (DIF). If the P<0.000 3 125 (Bonferroni correction), the corresponding item is considered to have DIF. Results The educational level of the elderly in Xiamen is generally lower, with 39.63% (1 133/2 858) illiterate and 28.04% (801/2 858) educated only at primary school, respectively. MoCA scores of elderly with different educational backgrounds were quite different with statistical significance(χ2=413.73, P<0.01), but Scheffe showed that there was no significant difference in MoCA scores between high school educated and undergraduate. Age, gender and MoCA scores of the elderly were significantly correlated (P<0.01) . The items that had DIF included: trail making test, copy cube,numbers of clock, hands of clock, abstraction 1 and 2, and word 1,2,3, and 5 of delayed recall, adding up to 11 items. The items of trail making test and copy cube only have DIF in the people above junior high school education level, while the 2 items of abstraction dimension only show DIF in the people above senior high school education level. Conclusion Most items of MoCA scale are suitable for screening people with all educational levels except when applying trail making test, copy cube and the 2 items of abstract dimension. Key words: Aged; Differential item functioning analysis; Montreal Cognitive Assessment Scale; Educational