Design and sample
This study makes use of data from a school-based intervention, the HEalth In Adolescents (HEIA) study [
24]. The overall aim of the HEIA study was to develop and evaluate a multi-component intervention study aimed at healthy weight development through diet and physical activity. A total of 177 schools were invited, and 37 schools accepted the invitation. Schools were included in this study if they had a minimum of 40 enrolled pupils in the 6th grade. Schools were thus recruited from the largest towns/municipalities in seven counties from the Eastern part of Norway (Akershus, Buskerud, Hedmark, Oppland, Telemark, Vestfold, and Østfold). The schools were randomly assigned into 12 intervention and 25 control schools. All 6th graders in these schools and their parents/legal guardians were invited to participate in the baseline (BL) study which took place in September 2007.
The participants in the present study are students from the 25 control schools. Parental consent was obtained for 1014 children from these schools, with a rate of parental consent of 73%. A total of 975 students (96% of the 1014 returning parental informed consent) from these schools participated at baseline (BL). The first follow-up (T1) study took place in May 2008, and 934 (96%) of those participating at BL participated. The second follow-up (T2) was conducted in May 2009, and 885 (95%) of those participating at BL and T1 participated.
Ethical clearance was obtained from the Regional Committees for Medical Research and the Norwegian Social Science Data Service.
Measures
Four questions with pre-coded answer categories were asked to assess usual TV/DVD use and use of computer/electronic games: How many hours do you usually watch TV and/or DVD on a normal weekday? The same question was asked for a normal weekend day. The answer categories were (recoding in brackets): half hour [0.5], one hour [
1], two hours [
2], three hours [
3], four hours [
4], five hours or more [
5]. The two questions on computer/electronic game use were formulated in the same way as for TV/DVD, but the answer categories were: no playing [0], half hour or less [0.5], one hour [1], two hours [2], three hours [3], four hours or more [4]. Separate weekly scores for TV/DVD and computer/electronic games were calculated by summing hours reported for an average weekday (multiplied by five) and average weekend day (multiplied by 2), and then summed to create a total screen time (TST) variable.
Moderate test-retest correlation coefficients (0.66-0.73) for the weekly outcome measures were obtained from a separate test-retest study conducted at 10-14 days apart among 111 6th graders prior to the main data collection.
Independent variables
Perceived parental regulation of TV/DVD use and perceived parental regulation of computer/electronic games were assessed by a 4-item scale (e.g. My mother and father try to make sure that I do not watch too much TV; my mother and father try to make sure that I do not use the computer and play games too much) [
25]. The items in these measures were rated on a 5-point Likert-type scale coded 1 (lowest) to 5 (highest) and phrased "totally disagree" to "totally agree" with a neutral midpoint. Since the outcome measure for the regression analyses was TST, all items in both measures were used to form a parental TST regulation measure. Reliability analysis was done for this measure, and a high internal consistency was obtained: Cronbach's α = 0.80 (BL), 0.85 (T2).
Self-efficacy related to barriers for PA was assessed by a 5-item scale, adapted and modified from Motl et al. 2006 [
26] and Lytle 2009 [
27] (e.g. I can be physically active during my free time on most days even if I have the choice to watch TV or play video games instead). Items of the scale were rated on a 5-point Likert-type scale. The scale had high internal consistency with Cronbach's α = 0.75 (BL) and 0.77 (T2).
BMI was calculated as weight/height
2. Weight and height were objectively measured. The age and gender specific BMI cut-off values proposed by the International Obesity Task Force [
28] were used in order to categorize the adolescents into non-overweight and overweight/obese.
The pubertal development scale is based on the pubertal category scores defined by Carskadon and Acebo [
29]. The categories were: prepubertal, early pubertal, mid-pubertal, late pubertal or post-pubertal (re-categorization into pre-pubertal/early pubertal/mid, late and post-pubertal was done for the analyses).
Participants were divided into either ethnic Norwegian or ethnic minority. Ethnic minorities are defined as those having both parents born in a country other than Norway [
30], and constituted 6.5% of the total sample.
Living status of children was divided into two categories: those living with married or cohabitating parents constituted the first group; those living with their father or mother alone, equally with their mother or father, grandparents or another adult were grouped together in the second group.
Parental education was gathered as part of the parental informed consent for the adolescent. It was categorized into: low (12 years or less), medium (between 13 and 16 years) and high (more than 16 years). Educational status of the parent with the longest education or else the one available was used in the analyses.
Statistical analysis
Since schools were the unit of measurement in this study, we checked for clustering effect through the Linear Mixed Model procedure. Only 1.4-3.2% of the unexplained variance in the variables investigated was at the school level. Hence, adjustment for clustering effect was not conducted.
Mean values (standard deviations) were calculated for the outcome measures at baseline, T1 and T2, as these are approximately normally distributed. Gender-specific tertiles were computed from the average weekly TV/DVD use, computer/electronic game use and TST. The TST was also categorized into two, representing those with a weekly TST of less than 14 hours and those with a weekly TST of more than 14 hours. This cut-off was chosen based on recommendations for a maximum total daily electronic media use time of 2 hours or less in some countries [
31,
32].
Analysis of variance (ANOVA) was conducted to assess differences between males and females in mean levels of weekly hours spent watching TV/DVD, weekly hours spent using computer/electronic games, and mean hours of weekly TST at the three time points, and one-way repeated measures ANOVA was used to compare time spent on these SB over the three time points. ANOVA was also used to evaluate associations between TST and different characteristics of participants at BL and T2.
Tracking was examined using Generalized Estimating Equations (GEE), one of the latest statistical methods used for this purpose. GEE analysis has several advantages: it allows for all longitudinal data to be used to calculate a single stability coefficient, taking into account that repeated measurements within an individual are correlated; it also allows for adjustment for both time dependent and time independent covariates [
18]. In this method of analysis, the value of the initial measurement of the outcome variable at time 1 is regressed on the longitudinal development of the outcome variable from time 2 to time m (where m = number of measurements), adjusting for covariates. This allows for the assessment of the relationship between the first measurement and subsequent measurements, yielding one single regression coefficient, the standardized value of which can be interpreted as a longitudinal correlation coefficient [
18]. In this study, adjustment for ethnicity, parental education, family status and weight status was done. Correlation coefficients of < 0.30 were classified as low, 0.30 to 0.60 as moderate, and > 0.60 as moderately high [
33]. Then, to measure level of agreement between tertile membership at BL and T2, Cohen's weighted kappa was computed. According to Landis and Koch [
34], values of 0.01 to 0.2 indicate slight agreement, 0.21 to 0.40 fair agreement, 0.41 to 0.60 moderate agreement, 0.61 to 0.80 substantial agreement and 0.81 to 1.00 almost perfect agreement. Direct weighted Kappa calculation is not possible on SPSS, thus syntax from the IBM website [
35] was used. Data for imputation into the syntax was obtained from cross-tabulation.
Linear regression analysis was conducted to assess factors associated with change in TST between BL and T2, after correcting for baseline TST. The BL variables included in the regression analysis were perceived parental regulation of screen time, self-efficacy related to barriers for PA, BMI category, pubertal stage, ethnicity, living status and parental education.
Multinomial logistic regression analysis was then used to assess whether these BL characteristics of participants could be predictors of tracking high (vs low) TST. Subgroup analysis was done for this purpose including those in the lower and upper tertile at BL (low and high users). Then, three tracking patterns were identified: tracking high TST (those in the upper tertile for TST at BL who remained there at T2, or high-risk group), tracking low TST (those in the lower tertile for TST at BL who remained there at T2, or low-risk group) and no tracking of TST (increase or decrease).
Both univariate and multivariate analyses were conducted. Selection of variables for inclusion in the multivariate models was based on the results of univariate analysis (p ≤ 0.25). Results of the final adjusted models (unadjusted regression coefficients or odds ratios and 95% CI) are presented in text.
TST was used as an outcome variable instead of TV/DVD and computer/game use separately in the regression analyses. A review study has indicated that correlates of separate screen time behaviours are similar to those of overall screen viewing in younger children, with few exceptions [
36], and separate analysis (data not shown) showed that it was the case in this study as well. In addition, interventions are most likely to target overall screen time behaviours and not separate ones.
Attrition analysis was performed using independent samples t-test and chi-squared test of proportions to determine differences in baseline characteristics between participants and drop-outs (n = 90).
All statistical analyses were performed by IBM® SPSS® Statistics, version 18.0 (IBM Corp., Somers, New York, USA).