Detection of Vascular Mild Cognitive Impairment in Southeast Asia Using the Visual Cognitive Assessment Test: Machine Learning Analysis From the BIOCIS (Biomarkers and Cognition Study, Singapore)

TL;DR

The study shows that the Visual Cognitive Assessment Test (VCAT) is highly effective at detecting vascular mild cognitive impairment in Southeast Asia, performing as well or better than the Montreal Cognitive Assessment (MoCA).

Contribution

The study introduces the VCAT as a language-neutral, highly accurate tool for detecting VMCI in multilingual populations.

Findings

The VCAT achieved near-perfect accuracy in distinguishing VMCI from cognitively unimpaired individuals.

Both VCAT and MoCA showed optimal performance in differentiating VMCI from non-VMCI at a cutoff of 25.

VMCI participants had significantly lower neuropsychological test scores and higher vascular risk factors.

Abstract

Vascular mild cognitive impairment (VMCI) is a significant global health concern, particularly in Asia. The visual cognitive assessment test (VCAT) has shown promise as a language-neutral screening tool for cognitive impairment.

This study aims to assess the effectiveness of the VCAT in detecting VMCI and compare its diagnostic performance with the widely used and validated Montreal Cognitive Assessment (MoCA).

Cross-sectional data from 524 community-dwelling participants were analyzed from the BIOCIS (Biomarkers and Cognition Study, Singapore) and classified into cognitively unimpaired, non-VMCI, and VMCI groups. The participants underwent neuropsychological assessments and 3-T magnetic resonance imaging. The random forest technique and multivariable logistic regression were applied to assess the discriminative properties of the tests.

Participants with VMCI exhibited significantly lower performance across various neuropsychological tests (P<.001) and higher rates of vascular risk factors (P<.001). At a cutoff of 27, the VCAT achieved near-perfect accuracy in discriminating the VMCI group from the cognitively unimpaired group (area under the receiver operating characteristic curve=1; sensitivity=1; specificity=0.991). For differentiating the VMCI group from the non-VMCI group, both the VCAT and the MoCA showed optimal performance at a cutoff of 25 (area under the receiver operating characteristic curve=1.00; sensitivity=1.00; specificity=1.00).

The VCAT could be a valuable tool for detecting VMCI, particularly in diverse, multilingual populations. Its comparable or even superior performance to the MoCA, combined with its language-neutral design, positions the VCAT as a strong addition to cognitive assessment toolkits for VMCI. However, the complex nature of cognitive processing in VMCI suggests that a multifaceted approach that integrates both visual and verbal assessments may ultimately offer the most comprehensive evaluation.

RR2-10.14283/jpad.2024.89