Computational Phenotyping of Cognitive Decline With Retest Learning

Area of Science:

• Computational neuroscience
• Cognitive psychology
• Gerontology

Background:

• Cognitive change involves performance gains and decline, necessitating methods to track subtle changes.
• Identifying risk factors for Alzheimer's Disease and Related Dementias (ADRD) requires disentangling these complex dynamics.
• Ultra-brief, high-frequency cognitive assessments via smartphones offer novel data collection methods.

Purpose of the Study:

• To apply a computational cognitive model to analyze learning and forgetting dynamics in cognitive assessment data.
• To identify cognitive markers that are sensitive to preclinical cognitive change and mild cognitive impairment (MCI).
• To assess the utility of computational markers in distinguishing MCI from normal aging, independent of age effects.

Main Methods:

• Utilized data from the Einstein Aging Study (EAS; n=316) including smartphone-based cognitive assessments.
• Applied a computational model of learning and forgetting to response time data from visual short-term working memory (Color Shapes) and processing speed (Symbol Search) tasks.
• Extracted markers including intraindividual variability, within-burst learning, asymptotic performance, and forgetting of retest gains.

Main Results:

• Asymptotic performance correlated with mild cognitive impairment (MCI) and age, with evidence of asymptotic slowing over time.
• Long-term forgetting, learning rate, and within-person variability were unique indicators of MCI, independent of age.
• Computational cognitive markers demonstrated sensitivity and specificity in identifying MCI.

Conclusions:

• Computational cognitive markers show potential as sensitive and specific indicators of preclinical cognitive change.
• These markers can aid in the early identification of individuals at risk for ADRD.
• The findings support the use of computational markers for targeted interventions and improved cognitive health monitoring.