Background
The amount and nature of time spent using screen-based devices such as televisions, computers, and mobile phones has changed over recent years. A report in 2017 suggested that British children aged 5–15 years spent 1.5 more hours per week online than watching TV which is in contrast to their findings in 2007 when they spent roughly 5 h more per week watching TV than online [
1]. Patterns of screen use also differ depending on time of the week, with more time spent using screens on weekends than weekdays [
2]. The report found that screen-based products were commonly used by children and adolescents, with 79% of 12–15 year olds owning their own smart phone, and 48% of 5–15 year olds having a TV in their bedroom in 2016 [
1]. Alongside increases in screen time there has been an increase in the recorded incidence of common mental health disorders in children and adolescents [
3], leading us to question whether they are related.
Teychenne and colleagues [
4] recently systematically reviewed the literature on the association between sedentary behaviour and anxiety; they included studies that specifically examined screen time. Of the four studies in the review that explored the association between increased screen time and anxiety, two found positive associations [
5,
6]. However, like many in this field, these studies were cross-sectional and could not assess the temporal direction of association. The two remaining studies either found no association [
7] or an inverse association [
8] (in a cross-sectional study and prospective cohort, respectively). Of the four studies, only one [
5] was assessed as having strong methodological quality. Other reviews of the literature concluded that there was insufficient or inconclusive evidence for an association between screen time and anxiety [
9,
10].
There is more consistent evidence for an association between screen time and depression [
11‐
13]. However the evidence base is still limited, with research conclusions restricted by methodological limitations such as cross-sectional designs and broad age ranges (including both children and adults) [
4]. What evidence there is indicates that associations between screen time and depression may operate in both directions [
12,
14].
We therefore examined the association between screen time and both anxiety and depression during adolescence using prospectively collected longitudinal data from the Avon Longitudinal Study of Parents and Children (ALSPAC) [
15,
16]. Building on previous research, ours is the first study to assess the association between screen time (and different types of screen time) in a prospective UK cohort. Importantly, we also attempted to adjust for a range of other activities in order to identify what other activities are sacrificed for screen time. Such measures include time spent outside, time spent socialising, and time spent alone. We also separately investigated the associations with weekday and weekend screen use.
Discussion
Our results indicate that there is a small positive association between computer use at age 16 and both anxiety and depression two years later. Although the increase in the risk of developing anxiety and depression is small, given the high prevalence of screen use in young people, effects of small magnitude may still result in a substantial population burden and could therefore be clinically significant. Increased time spent alone attenuated the associations, particularly for anxiety.
The existing evidence regarding the association between screen use and anxiety is limited, whereas the evidence for an association between depression and screen time is more consistent [
4,
10,
13]. However, these studies cannot tell us whether any associations are likely to be causal. Several studies have found evidence for an association between anxiety and screen time [
5,
6], but none adjusted for time spent alone, which attenuated the association in our study. In addition, none of these studies were longitudinal, so it was not possible to establish the temporality of the association. Our findings for depression are in line with previous research suggesting there is an association with screen time when time spent alone is not adjusted for [
12].
There are different possible explanations for the results relating to time spent alone. It is possible that the measure of time alone used in this study may be measuring variance in anxiety or depression rather than confounding the relationship. Alternatively, time alone and screen time could be common markers of underlying causes of depression such as family circumstances or peer relationships. Our research highlights that time spent alone is an important factor (potentially as a confounder or as a marker of depression or anxiety) in this association that until now has been overlooked.
Besides time spent alone, various other mechanisms could explain the associations found between screen time and both anxiety and depression. Screen time allows for social comparisons with both fictional characters and real people who are perceivably higher up the social ladder than the viewer. In support of this theory, negative social comparisons on social networking sites are related to higher levels of depression and anxiety [
29]. Cyber bullying (whereby individuals are bullied via social media and texting) could also partly explain this association; victims report feeling depressed and worried as a consequence of cyber bullying [
30]. Alternatively, the sedentary nature of the screen time measured in this study may be the mechanism by which screen time and anxiety and depression are associated, as sedentary behaviour has been shown to be associated with both [
4,
31].
A common criticism in the reviews of the literature is the widespread use of cross-sectional rather than longitudinal data [
10]. An important strength of our study was the use of data from a longitudinal study and, in particular, our ability to adjust for previously identified anxiety and depression. Another strength of our study was the ability to adjust for a wide range of potential confounders. Nonetheless, there may be important confounders that were not measured in ALSPAC, or that were measured imperfectly. As a result, there is potential for residual confounding.
Another limitation is the extent of missing data; the proportion of individuals with complete data was low, which could have resulted in bias. We found evidence that individuals with anxiety and depression at age 7 years were more likely to have missing anxiety and depression data at age 18. This suggests (but cannot establish for certain) that the outcome data – depression and anxiety – were MNAR conditional on the covariates included in the analysis model – that is, the probability that depression and anxiety data were missing depended on their (unknown) missing values, even after taking account of the observed variables. If an outcome measure is MNAR then both a complete case analysis and MI will generally produce biased estimates of exposure-outcome associations, although there are exceptions to this if the outcome is binary [
32]. However, since we had four earlier measures of anxiety and depression (that were more complete than the measures at 18 years), we were able to include these as auxiliary variables in the MI models, thus giving a better approximation to MAR and hence reducing the likelihood of bias [
33]. The results for depression were generally weaker in the MI models, indicating some bias may have been present in the complete case analysis, although we cannot determine whether we have eliminated bias by using MI. We carried out sensitivity analyses making the assumption that imputed values of anxiety and depression were underestimates; although this weakened the results, the general conclusions remained the same.
A final limitation of our study is that screen use patterns have changed over time [
34], and the data for the current study were gathered between 2007 and 2009. This predates the wide availability of smart phones, smart watches and tablets that allow for use of screens (and particularly social media which was not assessed in this study) at times and in situations where screen use may have previously been limited. It is difficult to ascertain whether the findings of this study would apply to young people and screen use today, and evidence actually suggests that increased screen time using more recent technology may have positive effects on social capital [
35]. Additionally, screen time no longer necessarily means sedentary behaviour - some screen-based games, such as Pokemon Go, even encourage physical activity [
36]. There is clearly a need to capture different aspects of screen time including the context and amount of time spent using screen-based devices and types of devices as well as types and range of different activities being undertaken in order to fully investigate how screen time affects mental health in young people. Recent research by Przybylski and Weinstein [
37] suggested that moderate screen use may be beneficial. In their study, between one and four hours (depending on the type of activity) was found to be beneficial but was negatively associated with wellbeing above this threshold. They also found that any beneficial effects depended on whether use was on weekdays or weekends with negative effects on wellbeing seen at lower thresholds of use on weekdays. We found some differences between the effects of weekday and weekend exposure. However, the highest category of screen time measured in our study was three or more hours; as such, we could not differentiate between moderately high and very high use – and, as a consequence, could not assess whether there was a stronger association with very high levels of screen time. Furthermore, the categorisation of the screen time measure used in this analysis may not have been sensitive enough to detect moderate use between 2 and 3 h. As is the nature of secondary data, we were unable to create a category for 2 to 3 h due to the wording of the answer options provided in the questionnaire.
Different types of screen time may have different effects, both in terms of wellbeing and in terms of poor mental health. In their study, Przybylski and Weinstein found that different types of screen use had different effects on wellbeing [
37]. For example, there was a negative linear trend for smart phone use on weekends in relation to wellbeing whereas there were positive trends for TV, computer or video game use below the pivot point for beneficial vs non-beneficial use. We also found differences between type of screen use, where only time spent on the computer was clearly associated with an increase in anxiety and depression whereas there was little evidence of associations with time spent texting or watching television. Evidently, the association between screen use and mental health is complex and there is not a linear association between simply any type of screen use and mental health. This difference could be due to the use of social networking sites, which were primarily accessed through computers at the time of the study, whereby negative social comparisons may be the mechanism of the association found. Alternatively, texting could be associated with social behaviour whereas computer use could be associated with exam and work-related stress. Another theory to explain the difference is that some screen types may induce effects at lower levels of exposure than others, perhaps due to perceived level of immersion; young people may be more likely to multi-task when watching TV, and texting is intermittent whereas computer use may be more focussed and continuous. The pattern of association between computer use and anxiety also differs from the association between computer use and depression. Where the effects for anxiety seem to be consistent across the time of the week, the association for depression is much stronger with computer use on weekends than weekdays suggesting the mechanisms underlying these effects may be different for anxiety and depression. This highlights the need for on-going research in the area to assess the effect of specific types of activity on mental health in young adults in order to provide up-to-date, accurate advice for screen use.
Acknowledgements
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses.