QuantDev Brownbag - Cognitive psychometric approach for modeling cognitive aging and subtle cognitive changes across ambulatory assessments

Google map

Cognitive psychometric approach for modeling cognitive aging and subtle cognitive changes across ambulatory assessments

Abstract: 

Technological innovations (e.g., smartphone applications) make it possible to collect high-frequency repeated assessments about people's naturalistic, real-time cognitive functioning. Fitting cognitive psychometric models to such intensive longitudinal data can disentangle the underlying dynamic processes generating manifest cognitive performance. In this talk, we introduce a Bayesian exponential learning process model, which captures individual differences in cognitive performance in terms of four parameters: learning rate, retest gain, peak performance, and performance inconsistency. We cast the model in a multilevel framework which allows for exploring sources of individual differences by regressing these four parameters on predictors in a single-step analysis. First, we will demonstrate the utility of this model by analyzing data from the Einstein Aging Study. Results showed that participants with mild cognitive impairment had credibly lower peak performance, higher retest gain and more performance inconsistency across two cognitive domains. Second, we explore the utility of these parameters for predicting cognitive health status in older adults. These are promising results towards using the parameters of the proposed model as novel digital markers for risk of cognitive decline.