Investigating the Measurement Precision of the Montreal Cognitive Assessment (MoCA) for Cognitive Screening in Parkinson's Disease Through Item Response Theory

Background: The Montreal Cognitive Assessment (MoCA) is widely used to evaluate global cognitive function in older Brazilian adults. However, concerns persist regarding its applicability in non-homogeneous socio-demographic groups. This study scrutinizes the Brazilian version of the MoCA, focusing on its measurement properties in a diverse cohort of patients with Parkinson's disease (PD). Purpose: This study examined the psychometric properties of the Brazilian Portuguese MoCA in a heterogeneous sample of PD patients using item response theory (IRT) methods. Material and Methods: In a multicenter cross-sectional study, 484 PD patients aged 26-90 years (mean ± SD, 59.9 ± 11.1 years), with disease durations ranging from 1 to 35 years (mean ± SD, 8.7 ± 5.4 years), underwent MoCA testing. IRT analyses, including the Graded Response Model, evaluated item parameters, such as difficulty (location) and discrimination. Differential item functioning was analyzed vis-à-vis age and education using multiple indicators multiple causes (MIMIC) modeling. Results: The MoCA exhibited essential unidimensionality and satisfactory model fit. Attention and naming demonstrated high discrimination. Orientation and naming items were less challenging. Multiple domains showed differential item functioning related to age and education, underscoring the necessity of considering background characteristics when interpreting total scores. Conclusion: This study enriches validity evidence for the MoCA in PD by providing a detailed analysis of its measurement properties and sources of score bias. Tailoring test content and norms based on education and establishing computerized scoring algorithms leveraging item parameters may optimize the tool’s reliability and fairness. Refinements to mitigate differential item functioning could enable precise cognitive screening across diverse socio-demographic backgrounds.