The longitudinal structure of cognition in the ABCD study and associations with neural function and psychopathology: A Bayesian probabilistic principal components analysis.

Longitudinal studies of cognition are necessary to estimate individual developmental trajectories and to link these trajectories with neurodevelopment. Short task batteries, if psychometrically sound, may have benefits over lengthy assessments. This paper focuses on the assessment of cognition in the Adolescent Brain Cognitive Development™ (ABCD) Study. Previously, we performed a Bayesian Probabilistic Principal Components Analysis (BPPCA) on ABCD's baseline data, extracting components representing general ability, executive function and learning/memory. Because ABCD's cognitive assessments changed over time, a strategy is needed to evaluate the longitudinal stability of these component scores based on the modified batteries, changes in scores over time, and associations with external variables. The current analyses address these gaps using available data from the ABCD baseline, year 2 and year 4 assessments. We extracted components from the modified task batteries, aligning them with those described by Thompson et al. (2019) at baseline. When the battery is reduced from 9 to 7 variables (baseline to year 2) and from 7 to 5 variables (year 2 to year 4), a three-component structure remains a good fit with task loadings similar to those observed for the full baseline battery. There are strong intercorrelations among component scores across 5, 7, and 9 subtest solutions. Using components derived from the 5-subtest solution, age- and practice-related changes are longitudinally evident; longitudinal component scores also differentially associate with psychopathology indices and working-memory task-related MRI imaging activation. Researchers may wish to leverage this approach to assess how cognition relates to neurodevelopment and other sources of individual variation.