Assessing the Longitudinal Associations Between Decision-Making Processes and Attention Problems in Early Adolescence

Abstract

Cognitive functions and psychopathology develop in parallel in childhood and adolescence, but the temporal dynamics of their associations are poorly understood. The present study sought to elucidate the intertwined development of decision-making processes and attention problems using longitudinal data from late childhood (9-10 years) to mid-adolescence (11-13 years) from the Adolescent Brain Cognitive Development (ABCD) Study (n = 8918). We utilised hierarchical drift-diffusion modelling of behavioural data from the stop-signal task, parent-reported attention problems from the Child Behavior Checklist (CBCL), and multigroup univariate and bivariate latent change score models. The results showed faster drift rate was associated with lower levels of inattention at baseline, as well as a greater reduction of inattention over time. Moreover, baseline drift rate negatively predicted change in attention problems in females, and baseline attention problems negatively predicted change in drift rate. Neither response caution (decision threshold) nor encoding- and responding processes (non-decision time) were significantly associated with attention problems. There were no significant sex differences in the associations between decision-making processes and attention problems. The study supports previous findings of reduced evidence accumulation in attention problems and additionally shows that development of this aspect of decision-making plays a role in developmental changes in attention problems in youth.

Keywords: Attention problems; Decision-making; Development; Drift-diffusion modelling; Latent change score modelling; Longitudinal.

Fig. 1

Illustration of the stop-signal task. Note ITI = inter-trial interval, RT = reaction time, SSD = stop signal delay, SS = stop-signal, ms = milliseconds. From Casey et al. (2018)

Fig. 2

Flow chart of data pre-processing steps

Fig. 3

Latent change score models. Note: (A) Univariate latent change score model illustrating the average change in drift-diffusion parameter (Δ DDM0), variance in change (σ²Δ DDM0), and change depending on the initial measurement (red ρ). (B) Bivariate latent change score model illustrating correlation between DDM and attention problems (ATT) at baseline (purple φ), regression between DDM and change in ATT (blue γ1), regression between ATT and change in DDM (blue γ2), and co-occurrence of changes in DDM and ATT (yellow ρ). 0 = baseline, 2 = 2-year follow-up. The centred covariates were introduced at the baseline level. Adapted from Kievit et al. (2018)

Fig. 4

Graph showing the multigroup BLCS models for drift rate (v) and attention problems (att_prob). Note. The estimates are unstandardised

Fig. 5

Graph showing the multigroup BLCS models for decision threshold (a) and attention problems (att_prob). Note. The estimates are unstandardised

Fig. 6

Graph showing the multigroup BLCS models for non-decision time (t) and attention problems (att_prob). Note. The estimates are unstandardised