Ethics approval was obtained from the Deakin University Human Research Ethics Committee (HEAG-H 04_2023). Written informed parental consent and verbal child assent were obtained before data collection. Household members were asked to provide verbal permission. Data were collected during May 2023 and June 2023 and analyzed between April 2025 and June 2025. Data collected are stored on a restricted shared drive at Deakin University, are presented in aggregate form, and no video footage is published, to protect the identity of participants. Participants received an Aus $10 (US $7.08) supermarket voucher per child participating in the study as a token of appreciation for their time and participation.

Data for this pilot observational study were drawn from a larger study that aimed to examine children’s physical activity and sedentary behavior in home environments using a combination of wearable devices. Participants wore 3 movement sensors (activPAL [PAL Technologies Ltd], ActiGraph GT3X+, and ActiGraph GT9X Link) and a wearable video camera (Vimel ultrahigh-definition body camera; 1296p; 36 MP) during waking hours. While all devices contributed to the broader assessment of children’s movement behaviors, this pilot study focused specifically on the screen use data captured via the wearable video cameras.

Participants included parents and caregivers (hereafter referred to as “parents”) and their children aged 2 to 5 years from Melbourne, Victoria, Australia. To be eligible, parents needed to be able to read English and be aged >18 years. Recruitment took place through convenience and snowball sampling, with participants identified through the research team members’ (KD, KDH, and SJJMV) known networks; institute-wide emails; and a public post on the institute Facebook page (October 30, 2024), which was also shared with the research team’s networks. Potential participants were emailed a link to the online study recruitment page (via Qualtrics [Qualtrics International Inc]), where they could view and download an information sheet, undergo eligibility screening, and provide consent should they agree to take part in the study. During the consent process, parents were asked to report their children’s date of birth; sex; and residential address, including postcode. Socioeconomic status was classified using the Australian Bureau of Statistics Index of Relative Socioeconomic Advantage and Disadvantage, which ranks postcodes in Australia according to neighborhood socioeconomic conditions from 1 (most disadvantaged) to 10 (most advantaged) [19].

Analyses were conducted in Stata (version 18.0; StataCorp LLC). Camera wear time was calculated as the total number of minutes during which video footage was recorded. Descriptive statistics (means and proportions) were used to describe the sample and calculate time spent in each screen use context. Independent-sample 2-tailed t tests were performed to examine differences between mothers and fathers in terms of coviewing session duration and total coviewing duration. Interrater reliability for the double-coded footage was calculated using the κ statistic and interpreted as follows: κ=0.00 to 0.20 indicated slight agreement, κ=0.21 to 0.40 indicated fair agreement, κ=0.41 to 0.60 indicated moderate agreement, κ=0.61 to 0.80 indicated substantial agreement, and κ=0.81 to 1.00 indicated almost perfect agreement [22].

Coder agreement was high across most screen use contexts. Perfect reliability was achieved for device type, multiscreen activity, and setting (κ=1.00), whereas screen content (κ=0.95), streaming service (κ=0.94), social interaction (κ=0.87), screen activity (κ=0.84), and coviewing partners (κ=0.82) exhibited very strong agreement. In the initial coding round, the concurrent behavior categories of play and fidgeting were differentiated using time-based criteria, whereby shorter durations were coded as “fidgeting” and longer durations were coded as “playing.” However, this interpretation was somewhat subjective, and the coding protocol was subsequently revised to distinguish between “object-based” and “non–object-based” play and fidgeting, which achieved substantial agreement (κ=0.64).

A total of 11 parents of preschool-aged children consented to take part in the study; however, 2 (18.2%) of the participants refused to wear the video camera. The final sample comprised 9 children, who recorded a total of 174,290 seconds (48.4 hours) of video camera footage over the observation period. Mean wear time was 5.4 (SD 2.9; range 0.3-10.1) hours. A total of 37,944 seconds (10.5 hours) of video camera footage from 8 children were coded as screen use, equating to 21.8% (37,944/174,290) of total camera wear time (range 0.3%-74.0%). Children were, on average, aged 3.8 (SD 1.1; range 2-5) years, 2 were girls, and 6 were boys. The child who recorded no screen use in their 0.3 hours (n=1245 seconds) of recording was a girl aged 3 years. Children were predominantly from advantaged backgrounds, with most residing in areas classified within Index of Relative Socioeconomic Advantage and Disadvantage deciles 8 to 10; 1 child was from an area classified as decile 6.

Table 1 shows the duration of screen use by device type, screen activity, setting, and social interaction. Overall, screen use was quite homogeneous, with program viewing (37,461/37,944, 98.7% of seconds), televisions (34,192/37,944, 90.1% of seconds), and lounge room use (33,710/37,944, 88.8% of seconds) dominating screen activity, device type, and setting, respectively. Program viewing was the sole screen activity when using televisions (34,192/34,192, 100% of seconds) and tablets (2577/2593, 99.4% of seconds). In contrast, screen activity during smartphone use was more heterogeneous, including program viewing (692/1159, 59.7% of seconds), communication (345/1159, 29.8% of seconds), and browsing (122/1159, 10.5% of seconds). However, smartphone use accounted for only 3.1% (1159/37,944) of total screen use. Coviewing occurred in approximately one-third (11,010/37,944, 29% of seconds) of all screen use but differed by device type (television: 8053/34,192, 23.6% of seconds; smartphone: 567/1159, 48.9% of seconds; tablet: 2390/2593, 92.2% of seconds). Concurrent behaviors occurred during 38.6% (14,632/37,944) of screen use time, with eating snacks (5751/37,944, 15.2% of seconds) and non–object-based fidgeting (5847/37,944, 15.4% of seconds) being the most common, and these were relatively consistent across device types.

A total of 44 coviewing sessions were captured. Among these 44 coviewing sessions, after excluding 3 (6.8%) involving both parents (to avoid double counting), 6 (13.6%) involving another child, and 2 (4.5%) where the coviewing partner could not be determined, there were 19 (43.2%) sessions involving fathers (with or without other family members) and 14 (31.8%) involving mothers (with or without other family members). Coviewing sessions involving fathers lasted, on average, 4.9 (SD 9.2; median 0.85, IQR 0.2-5.0; range 0.1-38.6) minutes, whereas sessions involving mothers lasted, on average, 1.1 (SD 1.2; median 0.75, IQR 0.3-1.5; range 0.2-4.1) minutes; however, this difference was not statistically significant either before (t₃₁=–1.50; P=.14) or after (t₃₀=–1.55; P=.13) removing 1 outlier involving a coviewing session with a father lasting 38.6 minutes.

Approximately one-third (11,010/37,944, 29%) of screen use involved coviewing. Table 2 shows the coviewing partners of preschool-aged children. Approximately two-thirds (7396/11,010, 67.2% of seconds) of coviewing instances involved at least one parent, mostly the father (5558/11,010, 50.5% of seconds) but occasionally both parents (882/11,010, 8% of seconds) or the mother (956/11,010, 8.7% of seconds). Coviewing with an older child was also common (3556/11,010, 32.3% of seconds). After excluding coviewing sessions involving both mothers and fathers to avoid double counting, fathers spent longer coviewing with their children (mean 1111.6, SD 923.07 seconds) compared with mothers (mean 159.33, SD 112.46 seconds), and this difference was statistically significant (t₉=–2.5; P=.03). However, after removing the outlier described above, this difference did not remain significant (coviewing involving fathers: mean 648.4, SD 680.42 seconds; coviewing involving mothers: mean 159.33, SD 112.46 seconds; t₉=–1.75; P=.11).

There was very little multiscreen activity (46/37,944, 0.1% of all screen use), with a single recorded instance. This instance occurred for a brief duration (46 seconds) and involved watching sports on television while simultaneously watching a song video on a smartphone.

Children viewed a range of content across various streaming services, including YouTube Kids (9/20, 45% of programs), Netflix (6/20, 30% of programs), SVT Barn (2/20, 10% of programs), Spotify (1/20, 5% of programs), ABC iview (1/20, 5% of programs), and Foxtel (1/20, 5% of programs). For 8.8% (3345/37,944) of seconds of video footage (6 occurrences involving 3 participants), the streaming service could not be determined due to the screen content being inconclusive.

Content classifications were available for 45% (9/20) of the programs, representing 61.8% (23,441/37,944) of screen use footage. Programs scored approximately 3.0 out of 5 stars for positive messaging (mean score 3.0, SD 1.00; range 1-4 stars) and positive role models (mean score 3.1, SD 1.17; range 1-5 stars) and approximately 1 star for violence and scariness (mean 1.3, SD 1.22; range 0-3 stars) and educational value (mean 1.7, SD 1.41; range 0-4 stars). Approximately one-third (3/9, 33.3%) of programs were classified as appropriate for an older age than those of the children in this sample (ie, for the ages of ≥6, ≥8, or ≥10 years).

Table 3 shows the mean content classification scores for coviewing vs solo viewing. Mean content classification scores differed according to social interaction. Specifically, mean scores were lower for educational value and positive messaging and higher for violence and scariness and positive role models when programs were coviewed than when programs were solo viewed.

This study appears to be the first, worldwide, to objectively assess preschool-aged children’s free-living screen use contexts using wearable video cameras. Our analysis yielded 4 main findings. First, screen use contexts among preschool-aged children were relatively homogeneous. Second, screen content was generally classified as low in terms of educational value, and one-third of content (3/9, 33.3% programs) was considered age inappropriate. Third, coviewing occurred in approximately one-third of screen use (11,010/37,944, 29%). Fourth, wearable video cameras provided a feasible approach for assessing preschool-aged children’s screen use contexts, albeit with some caveats. While a large amount of screen use data were captured (n=37,944 seconds), our findings should be considered alongside limitations associated with our small convenience sample (N=8).

This study showed that screen use contexts among preschool-aged children were relatively homogeneous. In our sample, almost all screen use involved program viewing of streamed or on-demand content on televisions in the lounge room. There was very little smartphone use or engagement in other screen activities such as communication or browsing. While the lack of variability in screen use may reflect our small, homogeneous sample, all from middle to high socioeconomic backgrounds, our focus on screen use in the home environment may have also influenced the results. Greater engagement with mobile devices and a wider range of screen activities might have been observed had we captured screen use outside of the home environment. For example, during interviews in a previous study, mothers (n=26) of preschool-aged children (aged 2-4 years) from the United Kingdom reported that they often allowed their children to use mobile devices outside of the home for a range of screen activities due to their multifunctionality, including playing games and taking and looking at photos, as well as program viewing [23]. Nonetheless, our finding that the television remains the primary source of screen use (34,192/37,944, 90.1%, of screen use in our sample) for preschool-aged children is consistent with results from previous studies [2,24,25] and suggests that older evidence from before the emergence of mobile devices is likely still relevant for this age group.

We found that the programs that preschool-aged children viewed were generally classified as having little educational content, and one-third of the programs (3/9, 33.3%) were classified as age inappropriate according to Common Sense Media ratings. Similarly, previous studies have shown that most content viewed by preschool-aged children is for entertainment rather than educational purposes [24,25]. It is possible that this finding reflects the nature of children’s television programs, with a 2019 examination of 88 popular children’s television shows in the United States indicating that, overall, shows aimed at children aged 2 to 5 years contained little educational content, scoring a mean of 2.44 out of 5 stars based on Common Sense Media classifications [21]. This is of concern given that substantial evidence demonstrates that high-quality educational television, such as Sesame Street, may benefit children’s cognitive, literacy, and social outcomes [26-29] and age-inappropriate content can negatively impact psychosocial outcomes [7]. An alternative explanation for the low educational value of the screen content observed in this study may be that, although screen guidelines from pediatric societies emphasize high-quality programming [8-10], this messaging may not be reaching parents. Thus, interventions aiming to support healthier screen use behaviors in preschool-aged children might focus on equipping parents with practical strategies to identify genuine educational content and recognize and avoid content that may be age inappropriate. Efforts could also involve engaging with industry to formally and scientifically evaluate content before making educational claims [10].

Coviewing in our sample occurred in approximately one-third of screen use (11,010/37,944, 29%). In contrast, Thomas et al [17] reported that just 16% of screen use involved coviewing among adolescents, with <1% involving interactive coviewing (n=7 images involving a laptop computer). However, the authors did not describe how this was coded from the camera images, making it difficult to ascertain how comparable this finding is to our data. In our study, we could not distinguish between interactive coviewing (eg, where another person asks questions or explains content) and passive coviewing (eg, where another person is present but not interacting with the child). This distinction is important as research suggests that interactive rather than passive coviewing can enhance language and cognitive development (eg, attention and memory) [24,30,31]. While the inclusion of audio data alongside video footage could be used to assess the quality of interactions, this approach may present additional ethical and feasibility challenges given the sensitive nature of recording in home environments. Future work may consider the inclusion of ecological momentary assessment to capture interaction during coviewing.

Wearable video cameras offer a promising and feasible approach to assessing preschool-aged children’s screen use in real-world contexts. Using these devices, we were able to objectively quantify a range of nuanced contextual information on preschool-aged children’s screen use. While wearable cameras reduce recall bias and offer a richer, more detailed account of children’s interactions with screens compared with self-report measures, several caveats must be considered. The success of this approach relies heavily on both parent consent and children’s willingness to wear the device consistently. A total of 11 parents of preschool-aged children consented to take part in this study, and most children (9/11, 81.8%) were willing to wear the device (mean wear time 5.4 hours), indicating that our protocol was feasible. Additionally, establishing strong interrater reliability is essential to ensure the reliability and replicability of the results. Our coding protocol achieved a high level of interrater reliability, supporting the robustness of our findings. Lastly, the substantial volume of video data generated from wearable video cameras makes coding time and labor intensive, underscoring the need to account for these factors in the study planning phase. As such, while wearable video cameras offer valuable insights into preschool-aged children’s screen use, their implementation requires careful ethical, practical, and methodological considerations.

There are a number of study limitations that should be acknowledged. While a large amount of screen use data were captured (10.5 hours), the small convenience sample and relatively homogeneous demographic characteristics limit generalizability to the wider preschool population. As such, future studies using larger samples are needed to overcome this limitation. Moreover, video footage data captured screen use exclusively in the home environment. As screen use outside the home environment may yield different patterns, future wearable camera studies examining screen use contexts both inside and outside the home environment are warranted. A previous wearable camera study in adolescents recorded camera footage both inside and outside the home environment, including on public transport, public food outlets, and community venues, suggesting that this approach is feasible and acceptable [17].

Among a small sample, we showed that preschool-aged children’s screen use is relatively homogeneous, predominately characterized by program viewing of streamed content on televisions in the lounge room. While there was some positive messaging and role models, the content viewed by children was generally considered of little educational value and at times age inappropriate. Coviewing occurred during approximately one-third of screen use and more often with fathers than mothers. We showed that wearable video cameras offer a feasible and acceptable approach to collect more accurate and detailed data on preschool-aged children’s screen use contexts compared with traditional self-report measures. This study provides some useful insights for intervention development. However, being the first study of its kind in this population, further research on preschool-aged children’s screen use contexts both inside and outside the home using wearable cameras in larger samples is warranted to confirm our findings.