Digital Cognitive Assessment Tests for Older Adults: Systematic Literature Review

JMIR Mental Health

Cubillos, C., & Rienzo, A. (2023).

JMIR Mental Health, 10, e47487.

This systematic review investigates the psychometric properties of computerized cognitive tests for older adults.

Not stated

Not stated

Published primary studies

20

Computerized cognitive tests sufficiently correlated to their traditional paper-and-pencil counterparts for assessing cognition in older adults. The computerized tools that showed the highest correlation to traditional or specialized cognitive tests included: <br /> <ul> <li><span style="color: #333333;"><strong>the computerized assessment battery for cognition (C-ABC)</strong>: validity and correlation coefficient (VALID-CO)= 0.75 for individuals with dementia or mild cognitive impairment (MCI); </span></li> <li><span style="color: #333333;"><strong>the Self-Administered Gerocognitive Examination (eSAGE)</strong>: VALID-CO = 0.77 for individuals with MCI or mild dementia; </span></li> <li><span style="color: #333333;"><strong>the Brain Health Assessment Cognitive Score (BHA-CS)</strong>: VALID-CO = 0.77 for individuals with dementia or MCI; </span></li> <li><span style="color: #333333;"><strong>the Inbrain Cognitive Screening Test (Inbrain-CST)</strong>: VALID-CO = 0.85 for individuals with dementia or amnestic MCI ; and </span></li> <li><span style="color: #333333;"><strong>the Cognivue application</strong>: VALID-CO = 0.86 for individuals at risk of dementia or age-related cognitive decline. </span></li> </ul> <span style="color: #333333;">Limitations of this review include small sample sizes and heterogeneity between studies regarding the population under observation (e.g., dementia, MCI, age-related cognitive changes). Additional research is needed to investigate psychometrics for specific populations and to determine the effect of patient characteristics (e.g., education, gender, culture, technology literacy) on assessment validity.</span>

Across all tools, sensitivity of computerized cognitive tests ranged from 0.63 to 0.95 for older adults with cognitive impairments. The following tools demonstrated the highest sensitivity: <br /> <ul> <li><span style="color: #333333;"><strong>the BHA</strong><strong>-CS</strong>: sensitivity = 0.81-0.95, specificity = 0.71-0.85 for older adults with MCI or dementia; </span></li> <li><span style="color: #333333;"><strong>the BHA</strong>: sensitivity = 0.71-0.91, specificity = 0.85 for older adults with MCI or dementia; and </span></li> <li><span style="color: #333333;"><strong>the eSAGE</strong>: sensitivity = 0.63-0.95, specificity= 0.67-1.00 for individuals with MCI or mild dementia.</span></li> </ul> Limitations of this review include small sample sizes and heterogeneity between studies regarding the population under observation (e.g., dementia, mild cognitive impairment, age-related cognitive changes). Additional research is needed to investigate psychometrics for specific populations and to determine the effect of patient characteristics (e.g., education, gender, culture, technology literacy) on assessment validity.

Across all tools, specificity of computerized cognitvie tests ranged from 0.54 to 1.00 for older adults with cognitive impairments. The following tools demonstrated the highest specificity for identifying individuals with MCI or dementia: <br /> <ul> <li><span style="color: #333333;"><strong>the Trail-Making Test Part A (TMT-A)</strong>: sensitivity = 0.86, specificity = 1.00; and </span></li> <li><span style="color: #333333;"><strong>the TMT Part B (TMT-B)</strong>: sensitivity = 0.71, specificity = 0.98.</span></li> </ul> Limitations of this review include small sample sizes and heterogeneity between studies regarding the population under observation (e.g., dementia, mild cognitive impairment, age-related cognitive changes). Additional research is needed to investigate psychometrics for specific populations and to determine the effect of patient characteristics (e.g., education, gender, culture, technology literacy) on assessment validity.

Additional findings about the psychometric properties of computerized cognitive test for older adults with cognitive impairment included: <br /> <ul> <li><span style="color: #333333;"><strong>The Tower of London application </strong>showed the lowest error rate (0.38) when comparing the errors made among individuals with mild cognitive impairment. </span></li> <li><span style="color: #333333;"><strong>The University of California, San Francisco (UCSF) application</strong> showed the highest error rate (1.00) when comparing the errors made by individuals with dementia. </span></li> <li><span style="color: #333333;"><strong>The UCSF application</strong> and <strong>the BHA application</strong> presented higher receiver operating characteristic (ROC) curves for discriminating between individuals with and without a cognitive impairment (UCSF ROC curve = 0.94-0.99; BHA ROC curve = 0.94-0.95).</span></li> </ul> Limitations of this review include small sample sizes and heterogeneity between studies regarding the population under observation (e.g., dementia, mild cognitive impairment, age-related cognitive changes). Additional research is needed to investigate psychometrics for specific populations and to determine the effect of patient characteristics (e.g., education, gender, culture, technology literacy) on assessment validity.