Watching the watchers: assessing the nature and extent of children’s screen time using wearable cameras

Children use a variety of screens in their daily lives, including mobile devices, computers, tablets and televisions. Such use may present both risks and benefits for their health and development.¹ Evidence from systematic reviews suggests that higher time spent on screens (all types combined) is associated with obesity, unhealthy diets, depressive symptoms, shorter and poorer quality sleep and poor cognitive performance.^2–6 More recently, the “fear of missing out” on things including social media access has emerged as a key driver of problematic screen use among adolescents, which (in turn) may have consequences for their mental health and wellbeing.^7–9 In addition, children’s exposure to bullying on social media is of substantial concern.^10–12 Potential benefits of screen use may arise from opportunities to socialise and access to age-appropriate educational content,^13,14 although evidence of positive health impacts from systematic reviews has been inconsistent.² Screen use increased rapidly during the COVID-19 pandemic,¹⁵ highlighting the need for contemporary methods to keep pace with technological developments and changing patterns of children’s screen use.

Owing to the health risks associated with screen time, several countries and health organisations have issued guidelines on children’s screen use. However, the contents of these guidelines vary. The World Health Organization (WHO) recommends screen time restrictions for children under age 5, but currently has no guidelines for older children and adolescents.¹⁶ Some countries, including New Zealand, recommend that children and adolescents (outside school time) spend no more than 2 hours per day on screens.^17,18 Guidelines from other countries have offered more general advice rather than time limits, including recommendations to consider screen types and activities, and children’s age and stage of development.^19–21

To help inform policy to promote healthy screen use behaviour, researchers need reliable and accurate measures of screen activity. A weakness in the screen time literature has been a lack of data on non-television media (e.g., computers, smartphones and tablets)² and reliance on self-report methods or parent proxies to measure screen use. For example, Scharkow²² found that, among 3,401 people aged 14–80 from individual United States households, self-report measures have poor accuracy for determining internet use compared with recorded logs of online activity. While Scharkow’s study participants kept a log record of their screen use, there are limitations associated with recorded logs, owing to high participant burden and the possibility that brief or reflexive uses are missed.²³ Multi-screen use—that is, the use of two or more media devices simultaneously, such as a TV and laptop or a handled device—is a growing phenomenon that may carry additional health risks than single-screen activity (e.g., poorer sleep quality), ^24,25 yet few studies have evaluated multi-screen activity.

Wearable cameras offer a valuable opportunity to explore screen time behaviours. These devices capture images of the wearer’s surroundings at fixed intervals (typically several images per minute). A pilot study of 15 adolescents from New Zealand aged 13–17 found that wearable cameras provide a feasible, acceptable method of measuring pre-bedtime screen behaviour, including multi-screen activity.²⁶ Given this background, we aimed to use wearable cameras to examine the extent (duration and frequency of use) and nature (types of screens, activities and when used) of children’s screen time during the after-school period, using data collected in the 2014/2015 Kids’Cam project.²⁷ Kids’Cam was a cross-sectional observational study that recruited 168 randomly selected children, aged 11–13 years from 16 randomly selected schools in the Wellington Region of New Zealand.²⁷

Methods

The Kids’Cam project

The study was conducted over a 12-month period (July 2014 to June 2015) to account for seasonal differences in the participants’ environments and activities. Sampling was stratified by school decile and child ethnicity to enable equal explanatory power for socio-economic and ethnic subgroups. Each child was provided with a wearable camera (Autographer) and a GPS device (Qstarz BT-Q1300ST Sports Recorder). Children were instructed to wear the devices for 4 consecutive days (2 school and 2 weekend days) on lanyards around their necks. Children were asked to wear the devices for all waking hours, but to remove the camera in situations where privacy could be expected, if they felt uncomfortable, when swimming or playing vigorous sport, or if requested by others.²⁷ Ethical approval was obtained to study all aspects of children’s lives relevant to public health from the University of Otago Human Ethics Committee (Health) (13/220). Further methodological details are published elsewhere.²⁷

In this ancillary study of children’s screen time, we included 108 Kids’Cam participants (64.3% of total sample) who captured at least 30 minutes of image data on Thursday afternoons after school. The after-school period was selected because it accounts for the largest proportion of children’s weekday recreational time. Of the 2 weekdays on which data were collected—Thursday and Friday—Thursdays were chosen as being the most like usual weekdays; after-school behaviours often differ on Fridays, being the end of the school week.

Coding for screen time

A coding protocol was developed to guide the coding of children's screen time (Appendix 1). Screen time was defined as the duration of time spent engaged with a screen. The coding process differentiated between screen mediums (i.e., type of screen) and screen activities, as detailed below. Codes were “tagged” to each image using customised software. Prior to coding, a reliability test was conducted using a test dataset of five participants (n=4,279 outside school images), on which three coders (one of whom coded all the data) achieved 90% or more agreement.

Screen mediums included televisions, computers, tablets and mobile devices (full definitions are available in Appendix 1). Multiple screen use was defined as the use of any two or more screen mediums in an image, e.g., watching television while playing on a tablet. Screen activities included programmes, games, social activities (e.g., social media), internet, background, “other” and undetermined (Appendix 1). Background activity included situations where a screen was present in a child’s vicinity, but the child did not appear to be fully engaged with it (e.g., they were facing away or doing something else). This generally applied to television, where children could still be influenced by the screen (e.g., through hearing advertising). “Other” was defined as any other type of screen-based activity, such as listening to music through a screen device or using productivity software such as Microsoft Word. Activities were coded as undetermined in situations where it was clear that the child was engaging with a screen, but the coder was unclear what was occurring on the screen; for example, due to obstruction of the screen in the image (e.g., food), interference of light or other image quality issues.

Statistical analysis

Statistical analyses were performed in Stata IC/16. Rates of screen time/hour (presented as means with 95% CIs) were calculated with negative binomial regression, using counts of screen time images as the numerator and total images captured as the denominator. Images were specified as contributing 7 seconds of recording time (this being the median interval between images). Analyses accounted for the stratified sampling design using Stata’s svy command and associated weighting options, to better reflect the target population. Subgroup differences in screen time were examined with rate ratios (from the negative binomial models), mutually adjusting for: ethnicity, gender and socio-economic deprivation (New Zealand Individual Deprivation Index [NZiDep])²⁸ simplified to lower deprivation (NZiDep groups 1, 2 and 3) and higher deprivation (NZiDep groups 4 and 5) and body weight status according to Cole cut-offs: overweight/obese (BMI >25.0) and non-overweight (BMI <24.9).²⁹ Weight status was included given the evidence demonstrating an association between screen use and increased risk of unhealthy weight gain owing to greater sedentary behaviour/reduced physical activity, passive overconsumption and exposure to the marketing of unhealthy food.^30,31 Participants with unknown weight status (n=4) and socio-economic deprivation (n=3) (Table 1) were excluded from these comparisons.

Results

Sample characteristics

The characteristics of the 108 children are shown in Table 1. Just over half (56%) were female and 44% were overweight/obese children, which reflects the national statistics for children of this age at the time of the study. The ethnic distribution was 43% NZ European, 35% Māori and 22% Pacific (reflecting the stratified sampling design). There were more than twice as many children in the lower socio-economic deprivation group (70%) than the higher socio-economic deprivation group (28%).

Children captured a median of 2.0 hours’ (interquartile range [IQR]: 1.4, 2.9) worth of images over the observation period, of which 95.8% were codable for screen activities. There was some variation in image capture across groups (Table 1), with children of higher socio-economic deprivation capturing fewer images than children of lower socio-economic deprivation.

View Tables 1-4, Figure 1.

Screen time

Children’s mean rate of screen time was 23.1 minutes/hour, which included 2.3 mins/hour of multi-screen use (10.0% of total). Televisions accounted for the highest proportion of screen time (9.8 mins/hour; 42.4% of total), followed by computers (7.4 mins/hour; 32.0% of total), mobile devices (3.0 mins/hour; 13.0% of total) and tablets (2.9 mins/hour; 12.6% of total) (Table 2). Image examples of screen types and screen activities are shown in Figure 1.

Differences by key demographic groups are presented in Table 2. Females spent just over half as much time on screens (total screen time) (rate ratio [RR]=0.58, 95% CI 0.37–0.93) and a fifth of the time on computers (RR=0.19, 95% CI 0.04–0.85) than males. Total screen time was similar for Māori, NZ European and Pacific children (Table 2), though there were some differences by ethnicity in television viewing (relative to NZ European: RR for Pacific=2.10, 95% CI 1.14–3.87; RR for Māori=1.38, 95% CI 0.95–2.00). There were some patterns of screen time according to deprivation. Although total screen time was similar by deprivation, there was evidence that high deprivation children spent less screen time on computers (RR=0.17, 95% CI 0.05–0.54) and mobile devices (RR=0.33, 95% CI 0.14–0.75) relative to those of low deprivation. There was no strong evidence for patterning of screen time use according to overweight status (total screen time RR=0.76, 95% CI 0.46–1.23 for overweight/obese compared to not overweight group).

Screen activities

Of the screen activity categories (Appendix Table 1), watching programmes accounted for the highest proportion of total screen time (6.3 mins/hour; 27.0% of total), followed by games (5.6 mins/hour; 23.9% of total), other (3.3 mins/hour; 14.0% of total), background (3.0 mins/hour; 12.8% of total), social activities (1.8 mins/hour; 7.8% of total) and internet (1.6 mins/hour; 6.9% of total). On average, 1.3 minutes of screen activities were coded as “unknown” (7.7% of all screen time). 10 times lower rates of screen use for games were observed among girls (relative to boys) (RR=0.10, 95% CI 0.03–0.30) and games were used more than half as often by overweight children (relative to non-overweight children) (RR=0.31, 95% CI 0.10–1.00). Children of higher deprivation spent less time engaged in “other” screen activities than children of lower deprivation (RR=0.16, 95% CI 0.04–0.57).

Rates of screen use were highest in the late evening period (after 8 pm, mean of 37.7 mins/hour) than in the early evening period (5:30 pm–8 pm, mean of 24.6 mins/hour) and early afternoon period (3 pm–5:30 pm, mean of 20.6 mins/hour) (Table 4). Higher rates of screen time closer to bedtime was predominantly explained by television use (26.3 mins/hour in the late evening; 69.7% of screen use), compared with 11.6 min/hour (46.9% of screen use) in the late evening and 6.6 mins/hour (32.1% of screen use) in the early afternoon (Table 4).

Discussion

Children in this study used screens, on average, for over one third of the after-school period, including over half the time after 8 pm. Television accounted for the highest proportion of screen time, which is consistent with previous studies,³² although it is possible that screen use patterns have changed since this data was collected in 2014/2015. The high rate of screen activity raises health concerns as it likely displaced other activities such as active play and sleep.³³ In addition, it is particularly problematic given the risk of exposure to cyberbullying.^10–12 The incidence of bullying on social media is particularly high among New Zealand children, with more than one in four parents reporting that their child had experienced cyberbullying.¹⁰ High rates of screen time after 8 pm raised particular concerns for children’s sleep hygiene; that is, practising behaviours that facilitate sleep and avoiding behaviours that interfere with sleep, given that national and international evidence demonstrate pre-bedtime screen use is associated with poor sleep outcomes.^5,6 Furthermore, the most popular screen activities (programmes and gaming) may have limited the opportunities for learning or development relative to other activities the children could have engaged in.

We found that children engaged in multi-screen activity 10% of the time while using screens, which is higher than 5% reported among a pilot study of adolescents aged 13–17.²⁶ Qualitative research suggests that children may use multiple screens for several reasons, including tempering impatience while a device is loading, filtering out unwanted advertising and because it is enjoyable.²⁴ A recent review found limited research on multiple screen use in the literature,² although there is some evidence that multiple screen use is associated with poorer sleep quality than single screen use.²⁵

While we found no associations by bodyweight, we found several patterns in screen use by other socio-demographic characteristics, which are largely consistent with previous studies. These include: higher rates of total, computer and gaming screen time among boys than girls;³⁵ lower computer use among children of higher deprivation, consistent with their lower access to computers;³⁶ and higher rates of television use among Pacific⁶ and Māori children than NZ European children.³⁷ The differences by ethnicity and socio-economic deprivation add to previous concerns about “digital divides”, characterised by differences in the nature of digital screen access by deprivation.³⁸ A surprising finding was more screen time on tablets among children of high deprivation than those of low deprivation, which may be explained by the lower cost of these devices compared to computers.

Our study identifies some strengths of wearable cameras for assessing screen time, which echo some of Smith et al.’s pilot study findings.²⁶ The method enabled the recording of children’s screen use as they went about their day, potentially making this one of the first studies to do so. Differentiating between screen activities is important given evidence that the type of activity affects health outcomes.¹ The passive method of data collection also minimizes participant burden. This is particularly important for capturing mobile device use, which often occurs for brief periods of time and is likely under-reported in previous research. It also enables the study of any screen device that is in front of the child. However, cameras cannot determine where children are directing their attention. This presents a challenge for identifying children’s engagement with “background” screens (e.g., televisions). Correctly identifying these activities may therefore require wearable cameras to be used alongside other methods, e.g., self-report or activity logs. The coding of images is also time intensive. While automated image recognition could expedite coding of some visual elements, this is less feasible for the variable nature of screen activities.

As well as the strengths of wearable cameras identified above, a key strength of this study was the high rate of image capture. Cameras worn in the Kids’Cam project captured images of children’s surroundings approximately every 7 seconds, which was more than twice as frequent as previous research.²⁶ This likely yields a more accurate measure of brief bouts of screen activity (e.g., mobile phone use). Further, the sample size of 108 was considerably larger than previous research,²⁶ helping to identify the utility of this methodology on a larger scale.

The study has some limitations. It is possible that the 2014/2015 dataset may not accurately reflect current trends in screen type usage and screen activities, particularly since the COVID-19 pandemic. Our sample was limited to children of Māori, Pacific or NZ European ethnicity. To gather more comprehensive information, future studies should be designed to include New Zealand’s other ethnic groups. As the cameras captured a median of 2 hours after school, we only recorded approximately a quarter of children’s after-school time. Also, because we excluded 60 children with fewer than 30 minutes of image data, we do not know their use. Nevertheless, for the majority of children in the study, it is possible to see the nature of their screen use and determine that screens play a dominant role in the children’s lives.

Conclusions

In this study, wearable cameras were used to explore the nature and extent of children’s screen time. The approach enabled an objective and reliable assessment of screen activity across all types of screens, including multi-screen activity. Children in the study spent over one third of their after-school time using screens, with higher rates of screen time in the late evening period (after 8 pm). Most screen use involved watching programmes and gaming. The high rate of recreational screen time, including pre-bedtime, reinforces the need for consistent guidelines to promote healthy screen time behaviour among children. Further research is needed to monitor the impact of screens on children’s wellbeing, including any socio-demographic differences, and for innovation in protecting children from harm in the online space.

View Appendix.