Testing Bidirectional Associations Between Screen Time and Inattention/Hyperactivity Symptoms From Childhood to Adulthood in a Brazilian Cohort

Abstract Background Screen time has been linked with inattention/hyperactivity symptoms, but studies often do not distinguish within-from between-person associations over long developmental periods. Methods Data were drawn from the Brazilian High-Risk Cohort for Mental Health Conditions, a school-based cohort with assessments at baseline (2010) and follow-ups in 2013–2014 and 2018–2019, 76% retention. The sample included 2,511 children (mean age = 10.2 years at baseline, 54.7% males). Daily screen time (hours spent exposed to computer, television or video games) was obtained by asking the parent or the primary caregiver (94.9% mothers). Inattention/hyperactivity symptoms were assessed with the Strengths and Difficulties Questionnaire (SDQ). Analysis was carried out using random intercept cross-lagged panel models, adjusting for sample representativeness, attrition, and demographic covariates. Sensitivity analysis was carried out using the Attention Scale of the Child and Behavior Checklist (CBCL), harmonized with the Adult Behavior Checklist (ABCL). Results We found no within-person relationship between screen time and inattention/hyperactivity symptoms from childhood to early adulthood. Significant between-person associations were found in the SDQ but not in the adjusted SDQ models (respectively: β = 0.25; 95%CI = 0.02, 0.45; p = 0.029, and β = 0.22; 95%CI = -0.00, 0.45; p = 0.052). For instance, the history of primary caregiver’s psychiatric diagnoses was associated with higher average screen time (β = 0.16; 95% CI, 0.07, 0.25; p < 0.001) and inattention/hyperactivity symptoms (β = 0.27; 95% CI, 0.21, 0.34; p < 0.001). Findings were replicated in the CBCL-ABCL model. Conclusion We found no longitudinal association between increasing screen time exposure beyond an individual’s average and inattention/hyperactivity symptomatology. Furthermore, between-person associations were absent after covariate adjustment and results weren’t sensitive to the questionnaire.