Performance of the Brain Health Test-7, Mini-Mental State Examination, and Montreal Cognitive Assessment for detecting subjects with mild cognitive impairment.

Objectives: The Brain Health Test-7 (BHT-7), Mini-Mental State Examination (MMSE), and Montreal Cognitive Assessment (MoCA) are valid dementia and mild cognitive impairment (MCI) screening tools. Relevant validation studies have usually used receiver operating characteristic (ROC) curve analysis or a fixed-threshold approach. In this study, we adopted stratum-specific likelihood ratio (SSLR) analysis to capture more information about their performance in detecting MCI.

Design: Cross-sectional multi-site study.

Setting: Hospitals in northern and southern Taiwan.

Participants: 1090 subjects aged 50 years or older were assigned to a cognitively normal group, an MCI group, or a dementia group.

Measurements: BHT-7, MMSE, and MoCA to differentiate cognitively normal subjects from those with MCI or dementia.

Results: The three cognitive assessment tools were valid for detecting subjects with MCI or dementia according to ROC analysis. The overall area under the ROC curve (AUC) of the BHT-7 was significantly higher than that of the MoCA and MMSE in differentiating MCI or dementia from controls. Five strata were generated by SSLR analysis for the BHT-7 and MoCA, while 4 for the MMSE. The five strata of the BHT-7 and MoCA well represented the different degrees of probabilities of having MCI. However, it was still difficult to rule out the presence of MCI even by a test score within the highest-score stratum of the MMSE.

Conclusions: The BHT-7 performed slightly better than MoCA and MMSE in detecting subjects with MCI. The strata generated from the SSLR analysis were more informative than single cutoff values.