We found that for age 4, watching TV for 18.5 hours or more per week (approximately 2 hrs 40 mins daily) was significantly associated with more significant behavioural problems in conduct problems and hyperactivity/inattention, peer problems, and lower prosocial behaviour. For age 6, 16.5 hours or more (approximately 2 hrs 20 mins daily) of watching TV was only associated with conduct problems and hyperactivity/inattention. Even after further adjustment for the SDQ status at age 4, only behavioural problems were significantly associated with greater than 16.5 hours weekly of TV watching time.
Our findings were in line with previous research on children aged 6 years, demonstrating associations between the use of mobile devices of 60 minutes or more and conduct problems and hyperactivity/inattention by using SDQ . A previous longitudinal UK study using the UK Millennium Cohort Study (MCS), in which screen time of 3 hours or more daily predicted a slight increase in conduct problems, but not hyperactivity/inattention in children between 5 and 7 years of age . Another longitudinal research on children aged 18–30 months illustrated the detrimental effects of high screen time on hyperactivity/inattention but not on conduct problems . Although not all studies showed negative findings for both hyperactivity/inattention and conduct problems, these problems can be interpreted using the broader category of “externalising behaviours”. Some researchers suggested a three-factor model rather than the original five scales of the SDQ scale, namely an externalised scale combining conduct and hyperactivity subscales, an internalised scale consisting of emotional and peer problems subscales and the prosocial scale [28–30]. Therefore, with this binary category, high screen time was associated with externalising problems by allowing the SDQ to be compatible with other measurement tools [31–34].
Different screen time limits from other studies can be explained by differences in sample size, age group, outcome measures, and identified confounders. Compared to the MCS study in the UK, we did not include a larger number of participants and other predictors such as parenting, mother-child relationship, household chaos, and child characteristics in the analysis due to the limited availability of such data in the GUS study. Even so, our findings are meaningful as the GUS study solely focused on children in the Scottish environment, which makes our findings reflect more on the real state of circumstances of children and their families in Scotland.
The strengths of our study include the large sample size and a broad range of children’s developments such as emotional, behavioural and social domains. Few studies have examined children in the UK, especially in Scotland, which this study focused on. The effective sample sizes of both sweeps 4 and 6 were very close to the actual sample sizes, which means that the estimate can be close to the true population value with a good precision level (Table S6) . Also, given that our multivariate binary regression models presented odds ratios with a relatively narrow 95% of CIs, our findings may be generalisable for most children in Scotland.
Even so, this study has several limitations. A very low proportion of non-white children was included in this cohort (age 4: 2.9%, age 6: 2.7%). In 2004, the number of non-white children was very low in Scotland compared to 2011(age 0 to 15, White: Non-white = 94%: 6% in 2011) [36, 37]. Given that cultural backgrounds related to parenting can be associated with TV viewing time for children, findings might not be fully supported in the population of different backgrounds . Despite a deliberate random selection process to ensure accurate representation, fewer from the most deprived areas agreed to participate in the GUS study, and this group has had the highest attrition over time. Non-response was higher amongst low Socioeconomic status (SES) families, lone parents, younger mothers, mothers who had not breastfed, families living in the more deprived area, and caretakers who did not join any groups for parents and children . Hence, children aged 6 were biased slightly towards the less deprived, less troubled families. This tendency did not significantly affect the association analyses of this study but identifying the reasons for non-responses and reducing them would be needed to reflect the true population in any analysis of further studies. Although we adjusted for the SIMD, a risk of misclassification of individuals could exist. That is, deprived individuals could have been misclassified as "not deprived" or vice versa. Considering that deprivation of rural populations can be less homogeneous since rural data zones cover larger geographical areas than urban ones and comprise more socioeconomically dispersed populations than in urban areas, the SIMD reflecting deprived populations can be less accurate .
Furthermore, lower SES families could show more hostility and less responsibility in interactions with their children . Therefore, children from lower SES families are likely to have negative emotionality, and this problem can be worsened by higher screen time. In this regard, future research should incorporate more accurate and precise data on SES as a potential factor.
In conclusion, our study found that TV viewing for more than 16 to 18 hours per week is associated with a high risk of children's behavioural problems. Regardless of earlier developmental outcomes, there is an association between a higher amount of screen time and behavioural problems developing later. Further research is needed to incorporate more potential confounders and data over a longer followed-up period.