Study population
Study participants were panel members of the Australian Health and Social Science project, recruited through annual population surveys conducted by the Population Research Laboratory at Central Queensland University. Between October and November 2013, 3901 randomly selected panel members were invited to participate in an online survey. Of these, 2034 respondents (52.1 %) across all States and Territories of Australia completed the survey [12]. Participants provided informed consent and the Human Ethics Committee at Central Queensland University approved the study (H13/09-163).
Measures
Socio-demographic variables
Socio-demographic variables measured included age, sex, parental status (parent of children 0–12 years, parent of children ≥ 13 years, non-parent) and level of education (high school, trade certificate/diploma, university degree).
Adult screen time behaviour
Adult screen time behaviour was assessed using two items from the Workforce Sitting Questionnaire developed by Chau et al. [13]. The items measure discretionary screen time during leisure (excluding screen time at work), and have shown acceptable test-retest reliability (ICC = 0.56–0.91) and criterion validity against accelerometry (r = 0.23–0.34) [13]. Adults were asked to report their time spent sitting while 1) watching TV and 2) using a computer at home for leisure activities (e.g., email, games, information, chatting) on a work day and a non-work day in the last 7 days. Total screen time on work days and non-work days, respectively, was the sum of time adults spent watching TV and using the computer on a work day/non-work day. The variables ‘total screen time on a work day’ and ‘total screen time on a non-work day’ were dichotomised into ‘≤2 h/day’ and ‘>2 h/day’ to assess whether adults themselves would exceed screen time limits recommended for children [9].
Adult views on screen time restrictions for children
Adult views on screen time restrictions for children were assessed using two items. Adults were asked ‘What is the maximum time children aged between 5–12 years should be allowed to spend watching TV?’ and ‘What is the maximum time children aged between 5–12 years should be allowed to spend using a computer at home for fun (e.g. games, emails, chatting, surfing the internet)?’ The questions were asked separately for a usual school day and a usual weekend day. The age group 5–12 years was chosen because children’s screen time habits tend to develop during this age [14]. Response options for both questions were no more than 15 min, no more than 30 min, no more than 1 h, no more than 2 h, no more than 3 h, no more than 4 h and 5 h or more. Based on distributions, responses for maximum screen time were collapsed into the categories no more than 30 min (≤30 min), no more than 1 h (≤1 h), no more than 2 h (≤2 h), and 2 h or more (>2 h). Total screen time adults would allow children on school days and weekend days, respectively, was the sum of maximum time adults would allow children for watching TV and using the computer on a usual school day/weekend day.
Statistical analyses
Chi-square and independent t-tests were performed to assess differences in socio-demographic variables, screen time behaviour and views on screen time restrictions for children between men and women, and included and excluded participants. All variables were checked for normal distribution by examining descriptive statistics, boxplots and histograms. Ordinal logistic regression was used to assess associations between adults’ screen time behaviour and their views on screen time restrictions for children. Predictor variables were adults’ ‘total screen time on a work day’ and ‘total screen time on a non-work day’. Outcome variables were ‘screen time adults would allow children on school days’ and ‘screen time adults would allow children on weekend days’ (reference category: > 2 h a day). Four ordered logit models were run by combining one of the two predictor variables with one of the two outcome variables. An interaction term was entered in the ordered logit models to test for differences in associations by gender and parental status. Analyses were adjusted for adult age, sex, parental status and level of education (covariates). Variance inflation factors, R2-square values and parameter estimates were inspected to ensure there was no multicollinearity amongst predictor variables and covariates. Associations are presented using proportional odds ratios (ORs), 95 % confidence intervals (CIs) and p-values with significance levels set at p < 0.05. Participants with missing data across all variables were excluded from analyses (n = 498). Analyses were performed in IBM SPSS Statistics (version 22.0).