2.1 Sample

Cross-sectional data (East Midlands, UK; April to June 2016) were used from the final follow-up visit of a cluster randomised controlled trial (RCT) of a school-based physical activity intervention.¹³ As there were no group effects on the primary outcome of moderate-to-vigorous-intensity physical activity (MVPA) at the final (14 month) follow-up ¹³ both groups were combined to conduct a cohort analysis. Participants (secondary school year groups 8 to 10) were randomly selected from all eligible girls in their school at the start of the trial (aged 11 to 14). Questions on screen use were included at the final follow-up data collection only. Participants attending the final follow-up visit (n = 1361) were given these additional questions if they first completed all the main testing procedures (n = 933). Of these participants, those who did not complete the screen use questions were excluded (n = 117) leaving a final analytic sample of 816 participants. Ethics approval was obtained from the University of Leicester ethics sub-committee for Medicine and Biological Sciences. Written informed consent was obtained from the school principal and an opt out consent process was used with parents/guardians. Participants provided verbal assent before completion of any measurement.

2.2 Screen ownership and access

Participants were presented with a list of common screens. Screens were taken from the Screen-Based Media Use Scale (SBMU) ¹⁴ with other contemporary screens added for the purposes of this study. The original SBMU had test-retest reliability of 0.52.¹⁴ Participants selected which screens they owned and had access to from a list of nine screens (see Appendix S1).

2.3 Screen usage

Participants self-reported all types of screens they used at the same time ‘straight after school’, ‘evening in your free time’, ‘in bed at night’ and ‘at weekends’. Questions from the SBMU were used.¹⁴ Participants were asked: ‘Young people sometimes use more than one screen device at the same time (such as a TV on in the background and also browsing on a phone). We call this ‘multi-screen-viewing’. For each of the following times of the day please put a tick next to ALL the screens that you use at the same time.' Options included nine common screens (see Supplementary file 1). Options of ‘I don't use any screens at this time’ and ‘Another type of screen’ were also given.

Concurrent screen use was determined by a simple sum of the number of screens that participants answered that they used at the same time for each time period.

2.4 Dependent variables

Objectively measured physical activity and sedentary time: Participants wore a GENEActiv (Activinsights, Cambs, UK) accelerometer (100 Hz recording) on their non-dominant wrist continuously for seven days.¹⁰ The.bin files were processed with R-package GGIR version 1.2–2 (http://cran.r-project.org).¹⁵ The processing steps (including definition of wear time) are included on page 12 of the original study report ¹⁶ MVPA (≥200 milligravitational (mg) units¹⁷) and average acceleration (Euclidean Norm Minus One or ‘ENMO’), indicative of total volume of PA were calculated. Daytime sedentary time was calculated by subtracting sleep (see below) from total sedentary time. Sedentary time was classified as <40 mg.¹⁸ Variables were calculated for all recorded days and for weekdays, weekend days and for the hours once the participants left school (‘after school+evening’ times based on each school's end times but approximated 4 to 9 pm).

Objectively measured sleep variables: A number of sleep-related variables were calculated from the GENEActiv data. ‘Time in bed’ represented night time sedentary behaviour which included sleep and any periods awake during the night. ‘Actual sleep duration’ was total time in bed minus the awake periods and essentially is the sleep variable. Both were calculated using the nocturnal sleep detection algorithm in GGIR¹⁹ and were calculated on all recorded days and for weekdays and weekend days individually. Sleep efficiency was calculated as a measure of the time spent in bed asleep (actual sleep duration/time in bed*100).

Body mass index (BMI): Height (average from two measurements using a Tanita portable stadiometer) and weight (Tanita SC330S bioimpedance scale) were measured to the nearest 0.1 cm and 0.1 kg, respectively. BMI was calculated as weight(kg)/height(m)². BMI z-score was calculated using participant BMI and self-reported age to provide a standardised measure relevant to the UK population.

Self-perceptions: Items from the physical self-perception profile questionnaire assessed self-esteem, physical self-worth and body attractiveness (six statements each).²⁰ The original questionnaire included two opposing forced choice statements that were adapted to represent a simple statement that participants responded to on a five point Likert scale from ‘strongly disagree’ to ‘strongly agree’. Negative statements were reversed for analysis.

Health-related quality of life: Participants completed the Child Health Utility-9D, a validated paediatric generic health-related quality of life measure with nine dimensions.²¹ See Supplementary file 1 for the questions.

2.5 Covariates

Self-reported total screen time: Participants completed a modified version of the adolescent sedentary activity questionnaire ²² which includes questions on 11 sedentary activities on all days individually. Participants reported how long in hours and minutes they spend in these activities on a typical school day and on a typical weekend day. Only the activities related to leisure time screen use were totalled, that is ‘watching a programme/film on TV/DVD/tablet/laptop’, ‘using the computer/laptop/games console for fun’ and ‘using a mobile phone/tablet to play games/search the internet/message friends’. Time spent in these three activities were totalled and split into sample-specific quintiles (for weekdays, Q1: 0.00–3.0 h; Q2: 3.01–5.0 h; Q3: 5.01–6.90 h; Q4: 6.91–10 h; Q5: >10 h; for weekend days, Q1: 0.00–5.0 h; Q2: 5.01–7.0 h; Q3: 7.01–10.0 h; Q4: 10.01–14.23 h; Q5: >14.23 h).

Two independent sample t-tests were performed in order to check differences in the dependent variables between the two ethnic groups (white Europeans and non-white Europeans). Significant differences between the two groups were found in MVPA variables (all-day, after-school, weekdays and weekends), overall PA, all-day and weekday sleep duration, sedentary times (all days, weekdays and weekends) and body attractiveness. In view of this, ethnicity was included as a covariate in all subsequent models.

Other covariates were participant school year group representing age, self-reported ethnicity (white European and non-white European), randomisation arm and accelerometer wear days (wear time). The index of multiple deprivation (IMD) decile (the official measure of relative deprivation of neighbourhoods in England with a lower value meaning more deprived) was calculated²³ based on participant self-reported postcode.

2.6 Statistical analyses

Participant-level characteristics for the 816 adolescents are presented as means (standard deviation) or numbers (percentages). Missing data for any variable have been identified in the tables and have been removed from the relevant analyses and percentages of the non-missing sample are presented.

Dependent variables were MVPA, BMI z-score, sedentary time, durations of actual sleep and time in bed, self-esteem, self-worth, body attractiveness and health-related quality of life utility index. The dependent variables were checked for normality (via qqplots) and MVPA, sedentary time and weekday and weekend days screen duration deviated slightly. The independent variable of number of screens was collapsed into five (0, 1, 2, 3 and 4+ screens) categories. Linear mixed effects models examined differences in each dependent variable between 1, 2, 3 and 4+ screens versus no screens (0) for each time periods (i.e., after school, in the evening, before bed and at weekends).

All models accounted for school level clustering (school ID; random effects). The models also accounted for covariates listed above. Ethnicity and year group variables were entered as factors (in two and three levels, respectively). Missing data were excluded for all of the variables in each model, and hence, number of observations in each model differs. R-Project for Statistical Computing (v.3.6.1) was used.

3.1 Screen access and usage

Table 2 shows that mobile/smart phones (94.3%) were the most common screen for participants to own (i.e. 79.2% own plus 15.1% own and have access to) followed by tablets (77.3%) and then laptops (71.4%). The proportion of participants reporting the use of each individual screen during each time period is presented in Figure S1. A mobile/smart phone was the most used screen device across the day. A TV was second most used after school, evening and weekend time periods, with a tablet being the second most used before bed.

3.2 Concurrent screen use

Overall, 59%, 65%, 36% and 68% of participants used ≥2 screens concurrently after school, in the evening, before bed and at weekends, respectively. The combination of TV/phone/tablet was the main combination after school (16.5%), evening (24.1%), bed (7.4%) and weekend (36.5%).

3.3 MVPA variables

As seen in Table S1, compared to using no screens, those using one (B = −18.4 mins/day, CI −32.9 to −3.9), two (B = −19.6, CI −33.6 to −5.6), three (B = −22.1, CI −36.0 to −8.2) and four+ (B = −20.9, CI −34.8 to −7.1) screens at weekends had significantly lower MVPA (mins/day) and significantly lower overall PA (1 screen: B = −7.57 mg, CI −13.6 to −1.5; 2: B = −8.03, CI −13.9 to −2.2; 3: B = −8.82, CI −14.6 to −3.0; 4+: B = −8.53, CI −14.3 to −2.7). Further, compared to those using no screens, weekend screen use of those using two (B = −23.4 mins/day, CI −43.2 to −3.6), three (B = −25.0, CI −44.7 to −5.4) or four+ (B = −25.1, CI −44.6 to −5.5) screens and one (B = 8.0, CI −14.9 to −1.2) or two (B = −8.6, CI −15.8 to −1.4) bed screens had significantly lower weekend MVPA (mins/day). Those using one screen in the evening had lower (B = −7.1 mins/day, CI-13.6 to −0.5) MVPA in the after-school and evening period.

3.4 Sedentary time

Compared to using no screens, those using two (B = 1.39, CI 0.11 to 2.66), three (B = 1.47 h, CI 0.21 to 2.74) and ≥four (B = 1.38 h, CI 0.12 to 2.64) screens at the weekend had significantly higher weekend sedentary time. No difference was seen between using no screen and one screen (Table S1).

3.5 BMI

Categorical comparisons indicated that compared to using no screens, those using one (B = 0.80 kg/m², CI 0.14 to 1.47), two (B = 0.73 kg/m², CI 0.09 to 1.38), three (B = 0.73 kg/m², CI 0.06 to 1.4) and four+ (B = 0.86 kg/m², CI 0.17 to 1.56) screens after school had significantly higher BMI (Table S2).

3.6 Sleep variables

As seen in Table S2, compared to no screens, those using one (B = −0.44 h, CI −0.87 to −0.01), two (B = −0.54 h, CI −0.95 to −0.12) and three (B = −0.68 h, CI −1.12 to −0.25) after-school screens had significantly lower sleep duration (hours) on weekdays. No difference was seen between 0 and 4 screens. Also, compared to no screens, those using one (B = −0.45 h, CI −0.85 to −0.05), two (B = −0.45 h, CI −0.85 to −0.07), three (B = −0.55 h, CI −0.95 to −0.15) and four+ (B = −0.47 h, CI −0.88 to −0.05) after-school screens had significantly lower time in bed on all days and lower time in bed on weekdays (1 screen: B = −0.54 h, CI −0.99 to −0.08; 2: B = −0.55 h, CI −0.99 to −0.11; 3: B = −0.65 h, CI −1.10 to −0.19; 4+: B = −0.52 h, CI −0.99 to −0.05).

3.7 Psychosocial and health-related quality of life measures

There were no differences in 1,2 or 3 screens versus no screens for any of these self-reported measures. However, compared to using no screens in bed those using 4+ screens had lower self-esteem scores (B = −1.3, CI −2.3 to −0.3).