Purpose: Current measures of children’s screen use limit the ability to acurately assess children’s exposure to screen media for surveillance and research. Most studies rely on self- or parent-reports. Advances in image technology, such as facial verification and gaze tracking offer solutions to objectively measure children’s screen use. We report two steps in the development of an objective screen use monitoring system: FLASH (Family Level Assessment of Screen use in the Home).  

Methods: FLASH-TV uses machine learning algorithms to process videos based on convolutional neural networks (CNN) to 1) detect faces, 2) verify presence of target child’s face, and 3) estimate time child’s gaze is on screen. Here steps 2 and 3 are reported. A video dataset was collected from 12 families in an observational lab. The target child, parent, and sibling spent time watching and not watching a TV in different positions and lighting conditions. Trained staff coded video data for target child’s presence and gaze on TV (10% double coded) to serve as gold standard. Videos from 10 families were available to assess face verification. Face verification was stratified by gaze/no-gaze and reported as accuracy (true positives) and false positive rate (FPR). In a separate protocol, 5 other parent-sibling triads took part in observational protocols on two visits to assess face verification across days. A CNN-based gaze detection algorithm was trained on 5 families and, to date, tested on 4 families’ data.

Results: Kappa of staff double-coding gaze was 0.91 (SD 0.15). Face verification accuracy of target child was 94.7% (SD 4.8%) for gaze and 78.5% (SD 14.4%) for no gaze, with FPR of 1.96% and 15.3% respectively. Face verification accuracy for visit 1 was 96.6% and 96.4% for visit 2. FLASH gaze detection achieved accuracy of 84.0%.

Conclusions: Current face verification algorithms are robust during gaze. Face verification during no-gaze will not be necessary to calculate screen use estimates. Our short-term goal is to optimize FLASH gaze detection to accuracy of 95%. In parallel, we are developing FLASH-Mobile, which will be a background application on mobile devices.

Purpose: The home environment has a key influence on family screen behaviours (i.e. TV viewing, smartphone use etc.). Due to the shared environment, screen behaviours of parents and children may cluster, forming unique typologies of familial behaviours. Determining correlates of these typologies can inform intervention design and implementation. This study identified familial typologies of multiple screen behaviours within the home, and their modifiable correlates.

Methods: Parents (n=542, 40.7yrs, 94% female) in the Sitting in the Home (SIT) study self-reported their own, and proxy reported their child’s (11.2yrs, 46% female) participation in six screen behaviours (TV/videos/DVDs, computer/laptop [homework/work and leisure use separately], tablet/smart phone [homework/work and leisure use separately], and video game use), 6 intrapersonal (e.g. behaviour preference), 10 social (e.g. role-modelling), and 6 physical environment (e.g. screen availability) characteristics which were assessed as modifiable correlates.

Latent class analysis identified familial typologies of screen behaviours. Characteristics of typologies were compared using one-way ANOVAs and chi-square tests. Logistic regression identified correlates of typology membership.

Results: Three typologies were identified: 1)‘computer users’ (36%), characterised by high parent and child computer use for homework/work and leisure; 2)‘leisure screenies’ (25%), high parent and child TV/video/DVD and tablet/smart phone use for leisure; and 3)‘low-users’ (39%), low parent and child use of all screen types.

Compared to other typologies, ‘low-users’ reported the most sleep, physical activity and restrictive rules, and the lowest screen time preference, use of screens as a babysitter, peer screen time co-participation, and screen options in the bedroom and home. The requirement to use a tablet/laptop for homework was highest in the ‘computer users’ and lowest in the ‘leisure screenies’.

Four correlates significantly predicted typology membership (reference: low-users): sedentary behaviour preference was associated with increased likelihood of being a ‘computer user’ and ‘leisure screenie’. The frequency that child’s homework requires tablet/computer and parental sedentary behaviour encouragement and support was associated with increased likelihood of being a ‘computer user’. Parental physical activity was associated with a reduced likelihood of being a ‘leisure screenie’.

Conclusion The three typologies show similar patterns of screen behaviours of children and parents. Strategies need to target the whole family.

Purpose: To evaluate the relationship between media use and metabolic syndrome (MetS) and its components: waist circumference, blood pressure, dyslipidemia and homeostasis model assessment of insulin resistance (HOMA-IR) in children and adolescents and whether there is an interrelation between sedentary behaviour (SB), moderate to vigorous physical activity (MVPA) and media use.

Methods: 7263 children (2-16 years old) from 8 European countries of the I.Family study were included in the cross-sectional analysis. Self-reported media use (hours/day) included TV, computer, game console and internet use. Children were categorized as having MetS and high level of its components when the latter exceeded the 90^th percentile, based on reference curves. Using logistic regression models we examined the association between media use and MetS, adjusting for various covariates, including dietary behaviours and puberty status. In a subsample including accelerometer data (3640 children), SB, MVPA and media use were considered to check for interrelation in the association with MetS and its components. Based on WHO guidelines we examined whether meeting screen time recommendations(≤1h/day for children 2-5 years and ≤2h/day for children ≥5 years old) was associated with MetS.

Results: Children used media for 2.3 hours/day on average (SD=1.4). Only 49% of children met the guidelines for media use. There was a significant association of media use with high waist circumference (OR=1.16, 95%CI=1.09-1.23), dyslipidaemia (OR=1.06, 95%CI=1.008-1.13) and MetS (OR=1.22, 95%CI=1.07-1.40). The association remained after examining for interrelation between SB, MVPA and media use. Positive association was seen between media use and blood pressure (1.05, 95%CI=0.93-1.19) and HOMA-IR (OR=1.14, 95%CI=0.95-1.37). Children who met the media use guidelines, had lower odds for any of the metabolic outcomes, but significant only for having high waist circumference (0.75, 95%CI=0.67-0.84) and dyslipidaemia (OR=0.86, 95%=0.77-0.96), independently of covariates.

Conclusions: Media use was associated with metabolic syndrome, waist circumference and dyslipidaemia. Our novel approach suggests that this association is independent of dietary habits and physical activity patterns. However, meeting the media use guidelines seemed to have a protective role on the risk for MetS and its components. Longitudinal associations between media use and MetS will be investigated.

Purpose: There is limited evidence of the association between device-measured sitting time and health markers and how physical activity impacts this relationship among children. The aim of this cross-sectional study was to examine whether device-measured light-intensity physical activity (LIPA) and moderate-to-vigorous physical activity (MVPA) moderate the associations between sitting time and cardiometabolic health markers: waist circumference z-score (zWC), body mass index z-score (zBMI), blood pressure (BP), high-density lipoprotein (HDL), low-density lipoprotein (LDL), cholesterol, triglycerides, glucose, insulin, vitamin D and C-reactive protein (CRP) among children.

Methods: This study included data from 114 children (mean age 8.2±0.5), collected in 2010 in Melbourne, Australia. ActiGraph GT3X accelerometers were used to determine time spent in LIPA, low-LIPA and high-LIPA (i.e. split at mid-LIPA cutpoint), and MVPA. Average time spent sitting was obtained from activPAL inclinometers. Anthropometric measures and resting BP were assessed by trained staff using standard protocols. Fasting-blood samples were obtained at a commercial pathology laboratory using standard protocols.

Interactions between sitting and LIPA, low-LIPA, high-LIPA, and MVPA on the health markers were examined using linear regression, adjusting for age and sex, and moderation by LIPA, low-LIPA, high-LIPA and MVPA was examined by adding interaction terms. Significant interaction effects were probed by comparing associations at the mean and at one standard deviation below and above the mean.

Results: There was a positive association between sitting time and glucose (p=0.015) that remained significant after adjusting for specific physical activity variables. A significant negative association was found between sitting time and vitamin D, however, this was non-significant after adjusting for physical activity variables. No moderation effects were found for any of the physical activity variables in the relationship between sitting time and cardiometabolic health markers.

Conclusions: The association between sitting time and glucose appeared to be independent of time spent in physical activity, which is consistent with research amongst adults. However, other cardiometabolic health markers were not independently related to children’s sitting. Experimental evidence and more sophisticated analyses (e.g. compositional) are needed to further explore the moderation effects of the different physical activity intensities and relationships between sitting time and cardiometabolic health markers.

Purpose:There is a well-documented trend of physical activity disengagement that is disproportionately observed among girls in adolescence, compared to age-matched boys. Negative perceptions of weight have been linked with lower rates of participation during distinct phases in adolescence, however this data is limited to cross-sectional designs. Due to the declining rates of physical activity participation and potential changes in weight perceptions throughout adolescence, it is important to understand changes over time during this critical developmental period. As such, the present investigation aimed to assess the association between weight perceptions and the odds of participating in physical activity in girls throughout adolescence. 

Methods:The sample consisted of a 4-year linked (Years 2015-2018) sample ofn = 1066 adolescent girls in the COMPASS study who had completed yearly self-reported survey assessments (i.e., self-reported body mass index [BMI], weight perceptions, participation in school-based physical activity, school-based competitive sport, and organized sport outside of school). Generalized estimating equation models were conducted to test the effect of weight perception on physical activity over time, controlling for BMI, age, race/ethnicity, and geographic location. 

Results: Beyond the effects of BMI, girls who perceived their weight to be “about right” (OR = 0.88, 95% CI: 0.80-0.99) or “slightly overweight” (OR = 0.89, 95% CI: 0.81-0.99) were less likely to be disengaged from school-based physical activity, compared to girls who perceived being “very overweight”. However, only BMI, and not weight perceptions, were significant in predicting participation in school-based sport or competitive sport outside of school. 

Conclusions: Girls who perceive their weight to be higher than “socially acceptable” are more likely to disengage from or avoid behaviors that are weight-salient, and thus may be underrepresented from recreational physical activity contexts (e.g., intramurals, non-competitive clubs). Meanwhile, higher actual weight rather than perceived weight, may be associated with less competitive and performance-based sport engagement due to sport-related weight restrictions. Overall, the pervasive stigmatization and discrimination of higher-weight bodies in Western societies warrants further investigation as it may be associated with the decline in recreational physical activity observed in adolescent girls.

Purpose: Existing literature shows that there is an inverse association between socioeconomic position and screen time among youth. What is less known is the mechanism behind these differences. The study aimed to explore correlates of total screen time among adolescents and to assess their mediating role in the association between parental education and total screen time.

Methods: A cross-sectional study including 706 adolescents (mean age of 13.6 (SD=0.3)) was conducted in 2016. Data were collected at school through an online questionnaire.  Multiple regression analyses were used to explore factors associated with total screen time. Mediation analyses were conducted to assess whether these factors mediated the association between parental education and total screen time.

Results: Multiple linear regression analyses showed that parental modelling of TV and movie streaming (B= 4.44 (CI: 2.90, 5.98)), self-efficacy towards limiting TV and movie streaming (B= -0.62 (CI: -1.03, -0.21)) , self-efficacy towards limiting computer/electronic game use (B= -1.18 (CI: -1.54, -0.82)), TV/movie streaming during dinner (B= 1.10 (0.07, 2.13)) access to screens (B= 1.81 (CI: 0.42, 3.20)) and the perceived opportunities for physical activity in the neighborhood (B= -1.39 (CI: -2.47, -0.31)) were related to total screen time. All of these factors except self-efficacy towards limiting TV and movie streaming mediated the association between parental education and screen time.

Conclusions: The study identified several modifiable factors at the individual, interpersonal and perceived neighborhood environmental levels that can be targeted in interventions aimed at decreasing screen time among youth in general and among those with a low socioeconomic position in particular.