- Email: [email protected]
Mediators of the Relationship Between Maternal Education and Children’s TV Viewing
Mediators of the Relationship Between Maternal Education and Children’s TV Viewing Kylie Hesketh, PhD, Kylie Ball, PhD, David Crawford, PhD, Karen Campbell, PhD, Jo Salmon, PhD Background: Maternal education is consistently found to be inversely related to children’s television viewing and is associated with aspects of the family television environment. This study investigates whether family television environment mediates the relationship between maternal education and children’s television viewing. Methods:
Parents of 1484 children reported maternal education, time their child spends watching television, and 21 aspects of the family television environment (potential mediators) during 2002 and 2003. Separate regression analyses were conducted in 2006 for each potential mediator that met two initial conditions for mediation (associated with both maternal education and children’s television viewing (p⬍0.10)), to assess whether inclusion reduced the association between maternal education and children’s television viewing. Multivariable regression assessed the combined impact of all mediators.
Results:
Twelve of 21 potential mediators met the initial conditions for mediation. Inclusion of each resulted in decreased  values (3.2% to 15.2%) for the association between maternal education and television viewing. Number and placement of televisions in the home appeared to have the greatest mediating effect, followed by frequency of eating dinner in front of the television with the child and rules about television viewing during mealtimes. Together, the 12 mediators accounted for more than one-third of the association between maternal education and children’s television viewing time.
Conclusions: This study suggests the strong inverse relationship between maternal education and children’s television viewing is partly mediated by aspects of the family television environment. (Am J Prev Med 2007;33(1):41– 47) © 2007 American Journal of Preventive Medicine
Introduction
C
hildren’s high levels of television viewing are associated with a range of negative social and health effects including aggressive behavior,1,2 poor school achievement,2 social problems,2 poor selfrated health,3 consumption of high-energy drinks and snacks,4 low fruit consumption,4 lack of participation in organized physical activity,4 and overweight.5 In an attempt to reduce such negative effects, it has been recommended that parents limit the amount of time their children spend watching television.6 – 8 However, some parents may be better equipped than others to act on such recommendations. For example, maternal education level is consistently found to be inversely associated with children’s television viewing.9 –11 Maternal education also appears to be related to certain aspects of the family television environment.11 From the School of Exercise & Nutrition Sciences, Deakin University, Victoria, Australia Address correspondence and reprint requests to: Kylie Hesketh, PhD, School of Exercise & Nutrition Sciences, Deakin University, 221 Burwood Hwy, Burwood, Victoria, 3125 Australia. E-mail: kylie. [email protected].
The social ecologic model proposed by Davison and Birch12 provides a useful framework for investigating influences on health-related behaviors such as television viewing. According to this model, the family television environment can include both the physical (number and location of televisions) and the social environments (rules and practices associated with television viewing) within the family home. Maternal education has been found to be inversely associated with a child having a television in his or her bedroom and to be positively associated with parental concern about the amount of time their child spends watching television.11 Further, aspects of the family television environment, including frequency of families watching television together,10 mealtime television viewing,10,13 parents’ television viewing habits,9 and having a television in the child’s bedroom9 have also been found to be associated with children’s viewing time. This study aims to develop a greater understanding of the relationships between maternal education, the family television environment, and the amount of time children spend watching television. Specifically, aspects of the family television environment that may mediate the relationship between maternal education and
Am J Prev Med 2007;33(1) © 2007 American Journal of Preventive Medicine • Published by Elsevier Inc.
0749-3797/07/$–see front matter doi:10.1016/j.amepre.2007.02.039
41
children’s television viewing time will be investigated. By developing an understanding of the potential mediators, we will be better able to tailor interventions targeting mothers with lower levels of education, which aim to reduce television viewing and its associated consequences.
Methods Sample The sample consists of 1484 families involved in the Health, Eating and Play Study, conducted from 2002 to 2003. This study employed a stratified random sampling technique to select schools for participation in the study. All government and Catholic elementary schools in suburban Melbourne, Australia with a total enrollment exceeding 200 students were classified into socioeconomic status (SES) quintiles using the Australian Bureau of Statistics Socio-Economic Indexes for Areas Index of Relative Socio-Economic Disadvantage.14 In order to ensure representation from a spread of SES groups, schools classified in the highest, middle, and lowest quintiles were eligible for sampling (schools in the second and fourth quintiles were excluded). Thirteen schools from each of the three included quintile groups were randomly selected with a probability proportional to school size. Twenty-four schools (nine high, seven middle, and eight low SES area schools) agreed to participate. All families of children in their first year of school (mean age 6 years; younger children) at all 24 schools and families of children in school year levels 5 and 6 (mean age 11 years; older children) at 17 of the 24 schools were invited to take part. (Study 2 was initially a study of school-entry age children only. Opportunistically, children in years 5 and 6 were also included during 2003 data collection. The 17 schools from which the older children were included were evenly distributed across the low, middle, and high SES groups.) The study received ethics approval from the Deakin University Human Ethics Committee, the Victorian Department of Education and Training, and the Catholic Education Office. Under existing ethical guidelines, it was necessary to seek active written consent from parents for each child’s participation, and no information could be accessed regarding characteristics of nonrespondents. Written parental consent was received for 1562 children (42% response). No area-level socioeconomic gradient was noted in response rates (41% response at high, 39% at middle, and 48% at low SES area schools). Due to incomplete independent and/or dependent variable data, 78 children were excluded from analyses for this paper.
Measures All data were provided by the child’s main caregiver who completed a questionnaire at home. Respondents provided information on their own and, where applicable, their partner/co-caregiver’s highest level of schooling. The six response options were condensed to three levels of education: less than secondary school 42
(never attended school, primary school, some high school), secondary school (completed high school, technical or trade school certificate/apprenticeship) and post-secondary (university or tertiary qualification). Based on reported gender of the respondent and co-caregiver, maternal (mother or female co-caregiver) education was derived. Respondents reported the amount of time their child spends watching television (including commercial, noncommercial, cable/pay TV, videos, and DVDs) on a usual school day and usual weekend day (scale ranging from 0 to 6 or more hours, in half hour segments). School-day estimates were multiplied by 5, and weekend day estimates were multiplied by 2; the totals were summed and divided by 7 to generate average viewing time (minutes per day), which was slightly positively skewed (skewness, 0.93). Intra-class correlations (ICCs) and percent agreement were used to assess test–retest reliability. For continuous variables, an ICC above 0.75 was considered as indicating excellent reliability, an ICC between 0.4 and 0.75 as fair to good reliability, and an ICC below 0.4 was considered as poor reliability.15 For categorical variables, percent agreement values above 66% were considered fair.16 The ICC showed this measure to be reliable in repeated administrations (ICC⫽0.78, 95% confidence interval [CI]⫽0.69 – 0.84, p⬍0.001).4 Respondents also answered 22 items (assessing 21 mediators) providing information on their family’s television viewing environment. These items were developed, based on a social ecologic framework,12 to incorporate factors such as family television viewing practices, family television rules, parental modelling, division of responsibility, and access to sedentary pursuits at home. Two identical items inquired about how much the respondent and co-caregiver are concerned about how much time they spend watching TV (not at all, a little bit, quite a bit, or very much). As with level of education, based on reported gender of the respondent and co-caregiver, maternal level of care about their own television viewing was derived. The remaining 20 items are detailed in Table 1. Some of these items were developed and reliability tested as part of a previous study10 and were modified for use here. All 22 items underwent test–retest reliability testing as part of this study. Due to the time required to complete the entire questionnaire, the original questionnaire was divided in two for the retest, and separate subsamples derived from the main study sample were asked to complete one half of the survey 2 weeks after they had completed the initial (entire) questionnaire. A random subsample of 137 study parents completed the first half of the original questionnaire a second time (encompassing the majority of items involved in this paper), and a random subsample of 39 study parents completed the second half of the original questionnaire a second time (encompassing 5 items used in this
American Journal of Preventive Medicine, Volume 33, Number 1
www.ajpm-online.net
Table 1. Items used to assess family television environment (potential mediators) Itema
Test-retest reliability
Response options
How often did your child do the following things? My child ate dinner in front of the television My child ate snacks while watching television
Never or rarely, less than once a week, once a week, about 2–3 times/week, about 4–6 times/week, every day How often did you and/or the co-caregiver do the following things together with your child? I/we watched TV, videos or DVDs together with the child Never or rarely, less than once a week, I/we ate dinner in front of the television together with the child once a week, about 2–3 times/week, I/we ate snacks together with the child while watching television about 4–6 times/week, every day Indicate how much you agree or disagree with the following statements: I have to be sure that my child does not watch too much TV Disagree, slightly disagree, neutral, I will switch off the TV if I think my child is watching too much slightly agree, agree I let my child watch TV in exchange for good behavior If I did not guide or regulate my child’s TV watching, (s)he would watch too much I let my child watch any television show (s)he chooses I restrict how much time my child spends watching TV During meal times, I do not allow the TV to be on How concerned are you that your child watches too much Not concerned, a little concerned, television? fairly concerned, very concerned How many TVs do you have in your home? Open response Does your child have a TV set in his/her bedroom? Yes, no Who is responsible for the following? Deciding how much TV your child should watchb Parent(s) (mainly me, mainly the coDeciding when your child can watch TVb caregiver, shared role combined), mainly my child Deciding whether the TV can be on when eating our evening mealb How much do you agree with the following statements about you and the co-caregiver? Disagree, slightly disagree, neutral, My child’s co-caregiver and I have the same views about how slightly agree, agree much TV our child should watchb My child’s co-caregiver supports the rules I make about when our child is able to watch TVb
ICC ⫽ 0.82 ICC ⫽ 0.63 ICC ⫽ 0.54 ICC ⫽ 0.75 ICC ⫽ 0.60 ICC ICC ICC ICC
⫽ ⫽ ⫽ ⫽
0.56 0.43 0.62 0.50
ICC ICC ICC ICC
⫽ ⫽ ⫽ ⫽
0.39 0.46 0.71 0.50
ICC ⫽ 0.95 95% agreement 94% agreement 91% agreement 91% agreement ICC ⫽ 0.59 ICC ⫽ 0.60
a
All items relating to frequency of behavior asked respondents to think about the past few months. Test-retest reliability of these items employed a different subsample to that for the above items. ICC, intra-class correlation.
b
paper). ICC or percent agreement statistics (for continuous and categorical variables, respectively) were calculated for each questionnaire item, and the results are presented in Table 1.
Analyses Children’s gender is consistently found to be unrelated to television viewing,9 whereas it appears that older children may watch more television than younger children.9 These associations were observed in the current sample (independent samples t test for gender, p⫽0.37; and for age group, p⬍0.001), therefore, data were analyzed for the sample as a whole and separately for younger (grade prep) and older (grade 5– 6) children. To account for the cluster-based sampling, all analyses included school as the cluster unit to give a robust variance estimate that adjusts for within-cluster correlation. Baron and Kenny’s17 approach to testing mediation was used in this study. Individual linear and logistic regression analyses were used, as appropriate, to assess July 2007
which potential mediating variables met the initial two conditions for mediation: (1) the independent variable (maternal education) being significantly associated with the potential mediator and (2) the potential mediator being significantly associated with the dependent variable (children’s television viewing).17 For these initial criteria, an inclusive significance level of p⬍0.10 was set. The final condition of mediation is that inclusion of the potential mediator reduces the association between the independent and dependent variables.17 To test this, separate linear regression analyses were conducted for each potential mediator including the potential mediator and maternal education as the independent variables. To test for collinearity, pair-wise correlations (r ⬎0.7)18 and variance inflation factors ([VIF]ⱖ10)19 for potential mediators were assessed. Finally, a multivariable regression analysis was conducted including maternal education and all potential mediators simultaneously. The decrease in the size of the regression () coefficient was used as an indicator of the magnitude of mediation, and R2 was used to Am J Prev Med 2007;33(1)
43
Table 2. Sample characteristics n (%)
Characteristic Boys Area-level SESa Low (⬍26th percentile) Middle (26–75th percentile) High (⬎75th percentile) Maternal education Less than secondary Secondary Post-secondary Daily TV viewing (mean SD) Meets guidelinesb (ⱕ2 hr/day)
All children (nⴝ1484)
Younger children (mean age 6 yrs) (nⴝ589)
Older children (mean age 11 yrs) (nⴝ895)
705 (48)
291 (49)
414 (46)
263 (18) 513 (35) 686 (47)
141 (24) 206 (35) 235 (40)
122 (14) 307 (35) 451 (51)
355 (24) 596 (40) 533 (36) 188 (93) min 4 (0.3)
136 (23) 239 (41) 214 (36) 166 (87) min 1 (0.2)
219 (24) 257 (40) 319 (37) 202 (93) min 3 (0.3)
a
Australian Bureau of Statistics postcode-derived “Socio-economic Indexes for Areas Index of Advantage-Disadvantage” (Australian Bureau of Statistics, 2003). Australian recommendation that children should not spend more than 2 hours per day using electronic media for entertainment (Australian Government Department of Health and Aging 2004). SD, standard deviation; SES, socioeconomic status.
b
provide an indication of the proportion of variance in children’s television viewing accounted for by each model. All analyses were conducted in 2006.
Results Characteristics of the 1484 children in the sample are presented in Table 2. In both age groups, the sample was distributed across area-level SES and maternal education categories, providing a socioeconomically diverse sample. Mean daily television viewing for the total sample exceeded 3 hours and was higher in older than younger children. Only four children in the entire sample spent an average of 2 hours or less per day watching television and thus potentially met the Australian guidelines for electronic media use.8 Maternal education was strongly inversely related to daily television viewing in the total sample (⫽ ⫺31.0, 95% CI⫽ ⫺38.7, ⫺23.2), and among younger (⫽ ⫺25.3, 95% CI⫽ ⫺35.9, ⫺14.8) and older children (⫽ ⫺34.2, 95% CI⫽ ⫺43.7, ⫺24.8).
Assessment of Whether Potential Mediators Meet Initial Conditions for Mediation Of the 21 potential mediators assessed, six were not related (p⬎0.10) to maternal education (first initial condition to be met for mediation), and a further two were not related to daily television viewing (second initial condition to be met for mediation) in either age group. Of the remaining 13 potential mediators, one (co-caregiver supports television rules) was excluded, as it was only related to both maternal education and daily television viewing in younger children, and its inclusion resulted in the loss from analyses of children from single-parent families. The final 12 potential mediators 44
(Table 3) were included in subsequent analyses; only children with complete data (independent variable, dependent variable, and all 12 potential mediators) were included (n⫽1348). The vast majority of pair-wise correlations among mediator variables did not exceed r⫽0.2. Only one pair of potential mediators was correlated at r ⬎0.7:18 frequency of the child eating dinner in front of the TV with frequency of parent(s) eating dinner in front of the TV with the child (r⫽0.80). The VIFs of all of the individual mediators were low (⬍10)19 (range 1.06 – 3.17). As the VIFs were low, and exclusion of either variable in the final multivariable model resulted in negligible change in  values, neither of the highly correlated variables was excluded from the analyses involving all the potential mediators concurrently.
Effect of Potential Mediators on the Relationship Between Maternal Education and Children’s Television Viewing Table 3 presents  values and reduction in  values, for the association between maternal education and children’s television viewing, for each potential mediating model. For each individual mediator, inclusion in the linear regression model resulted in a decrease in  values (between the range of 3.2% and 15.2%) for the association between maternal education and television viewing. However, the association between maternal education and children’s television viewing habits remained significant (p⬍0.001) in all models. This supports the hypothesis that these family television environment variables partly mediate the association between maternal education and children’s television viewing. The two variables describing the number and
American Journal of Preventive Medicine, Volume 33, Number 1
www.ajpm-online.net
July 2007 Table 3. Linear regression analyses assessing change in  values for association between maternal education and children’s TV viewing when (a) each potential mediator is included independently and (b) when all potential mediators are included concurrently Total sample (nⴝ1348)
Younger children (nⴝ543)
Older children (nⴝ805)
Am J Prev Med 2007;33(1)
Variables associated with children’s TV viewing
 (95% CI) for maternal education
Reduction in  (%)
R2
 (95% CI) for maternal education
Reduction in  (%)
R2
 (95% CI) for maternal education
Reduction in  (%)
R2
Maternal education only Maternal education plusa: Frequency of eating dinner in front of the TV Frequency of eating snacks while watching TV Frequency of watching TV with child Frequency of eating dinner in front of TV with child Frequency of eating snacks with child while watching TV Will switch TV off if thinks child is watching too much Lets child watch any TV show (s)he chooses Restricts how much time child spends watching TV Do not allow TV on during mealtimes Number of TVs in the home Child has TV in his/her bedroom Responsibility for deciding if TV can be on during dinner Maternal education plus all potential mediatorsb
⫺31.0 (⫺38.7, ⫺23.2)
–
0.06
⫺25.3 (⫺35.9, ⫺14.8)
–
0.05
⫺34.2 (⫺43.7, ⫺24.8)
–
0.08
⫺27.7 (⫺38.1, ⫺20.2)
10.6
0.13
n/a
⫺30.0 (⫺38.5, ⫺21.6)
12.3
0.14
⫺28.2 (⫺35.7, ⫺20.6)
9.0
0.16
⫺22.9 (⫺32.2, ⫺13.5)
⫺31.4 (⫺41.0, ⫺21.7)
8.2
0.17
⫺28.3 (⫺35.3, ⫺21.3)
8.7
0.13
n/a
⫺32.2 (⫺39.9, ⫺24.6)
5.8
0.12
⫺27.4 (⫺35.2, ⫺19.5)
11.6
0.11
⫺22.5 (⫺33.6, ⫺11.4)
11.1
0.09
⫺30.7 (⫺39.5, ⫺21.9)
10.2
0.11
⫺27.8 (⫺35.3, ⫺20.3)
10.3
0.11
⫺21.9 (⫺31.8, ⫺11.9)
13.4
0.11
⫺31.5 (⫺40.2, ⫺22.8)
7.9
0.11
⫺29.8 (⫺37.3, ⫺22.4)
3.9
0.08
⫺23.8 (⫺33.7, ⫺13.9)
5.9
0.07
n/a
⫺28.8 (⫺36.2, ⫺21.4)
7.1
0.10
n/a
⫺32.1 (⫺40.1, ⫺24.0)
6.1
0.13
⫺28.6 (⫺36.5, ⫺20.7)
7.7
0.10
⫺23.0 (⫺32.7, ⫺13.3)
9.1
0.09
⫺32.1 (⫺42.6, ⫺21.6)
6.1
0.11
⫺27.5 (⫺34.7, ⫺20.2)
11.3
0.10
⫺22.8 (⫺32.4, ⫺13.2)
9.9
0.11
⫺31.2 (⫺39.1, ⫺23.3)
8.9
0.09
⫺26.9 (⫺34.0, ⫺19.8) ⫺26.3 (⫺33.8, ⫺18.7) ⫺30.0 (⫺37.3, ⫺22.7)
13.2 15.2 3.2
0/10 0.11 0.09
⫺23.8 (⫺34.4, ⫺13.2) ⫺21.9 (⫺32.6, ⫺11.2) n/a
5.9 13.4
0.06 0.09
⫺29.8 (⫺39.4, ⫺20.2) ⫺30.0 (⫺40.0, ⫺19.9) n/a
12.9 12.3
0.11 0.10
⫺18.8 (⫺25.8, ⫺11.9)
39.4
0.30
⫺16.9 (⫺25.9, ⫺7.8)
33.2
0.23
⫺22.3 (⫺30.7, ⫺14.0)
34.8
0.29
9.5
All significant at p⬍0.001; n/a indicates item did not meet mediator criteria for this group of children and was therefore not assessed. a Each potential mediator included in a separate model (not cumulatively). b Twelve potential mediators included concurrently for total sample, 8 for younger children, 10 for older children. CI, confidence interval.
0.15
45
placement of televisions in the home appeared to have the greatest mediating effect. That is, the mediator models including each of these variables individually resulted in the greatest reduction in  coefficients for the association between maternal education and children’s television viewing. Although the impact of a television in the child’s bedroom was similar for older and younger children, inclusion of the number of televisions in the home resulted in a greater reduction in the  coefficient for the association between maternal education and children’s television viewing in older than younger children. Of the social family television environment variables, frequency of eating dinner in front of the television with the child and rules about television viewing during mealtimes appeared to have the greatest mediating effects. In most cases, the magnitude of reduction in  values for the association between maternal education and television viewing was similar for both age groups. However, frequency of eating snacks with the child while watching television and restricting how much time the child spends watching television had a greater magnitude of effect in younger than in older children (13.4% vs 7.9% reduction in ). Inclusion of all mediators concurrently resulted in a 39% reduction in the  value, suggesting these family television environment variables account for more than one-third of the association between maternal education and children’s television viewing. Similarly, large reductions in the  value were observed for the younger (33.2%) and older (34.8%) groups of children. Again, the association between maternal education and children’s television viewing remained significant in these models (p⬍0.001 for total sample and older children; p⫽0.001 for younger children).
Discussion Children in this study generally spent more time watching television than the 2 hours per day recommended by current guidelines.8 Consistent with previous research,9 –11 children whose mothers had attained lower levels of education tended to spend more time watching television than those whose mothers had higher levels of education. This study suggests that the strong inverse relationship between maternal education and children’s television viewing is partly mediated by certain aspects of the family television environment, although temporal relationships cannot be ascertained from this cross-sectional data. In particular, physical aspects of the family television environment (number of televisions in the home and presence of a television in the child’s bedroom) and family rules and practices associated with eating while watching television appeared to be most important. Together, the 12 family television environment mediators accounted for more 46
than one third of the association between maternal education and children’s television viewing time. The strongest mediators of the relationship between maternal education and children’s television viewing were those related to the physical aspects of the family television environment. Although separate studies have reported that the presence of a television in the child’s bedroom is inversely associated with maternal education11 and positively associated with children’s television viewing time,9 we were able to show that having a television in the child’s bedroom may also mediate the association between maternal education and children’s television viewing habits. The number of television sets in the home was also a mediator of the association between maternal education and children’s television viewing. It appears that the greater television viewing time observed for children with less-educated mothers is partially related to the different physical environments provided by these mothers. Thus providing mothers, particularly those with lower levels of education, with messages about the importance of limiting children’s access to television sets in the home may be a useful strategy to reduce children’s television viewing time. The social aspects of the family television environment that most strongly mediated the association between maternal education and children’s television viewing were those related to rules and practices within the family related to eating while watching television. An association between mealtime television viewing and children’s television viewing time has previously been reported,10,13 but we have demonstrated that these rules and practices may be inequitably distributed depending on maternal education level. Parents with lower levels of education are less likely to report having household rules related to television viewing.20 Lesseducated mothers may have more difficulty, or place lesser importance on limit setting with regard to their child’s television viewing, with higher levels of viewing the consequence. Equally, less-educated mothers may be raising their children in more challenging physical environments (e.g., smaller homes, smaller backyards, neighborhoods with greater safety concerns and fewer outdoor play spaces) and may use television as a babysitter or an alternative to fee-incurring after-school activities. Although this study achieved only a modest response rate, the participants spanned a broad range of socioeconomic backgrounds. In addition, the response rate is similar to that achieved in other health surveys, and there is evidence that response rates for most major studies have been falling over the past few decades and that it is not necessary to have high response rates to ensure a broad spectrum of respondents are represented.21 Although the test–retest reliability results for some of the potential mediator variables were modest, the variables that most strongly mediated the associa-
American Journal of Preventive Medicine, Volume 33, Number 1
www.ajpm-online.net
tion between maternal education and children’s television viewing demonstrated greater reliability on repeat administration. The large number of statistical tests performed means that it is possible that some of the associations detected were chance findings, however, a larger number of associations were detected than would be expected by chance. Like all associational analyses based on cross-sectional data, the ability to draw firm temporal or causal conclusions of mediating effects from these data are limited. However, maternal education could plausibly be assumed to be established prior to the family television environment and children’s television viewing practices. Alternative explanations for these findings should also be considered. It is possible that an additional unmeasured factor, such as overall family SES influences maternal education, the television environment, and television viewing time. Furthermore, this study did not assess the potentially important role of the father or co-caregiver, who may also have an influential role in shaping the physical environment. The strengths of this study include a large sample, allowing stratification by age group and the consideration of multiple potential mediators of the relationship between maternal education and children’s television viewing time.
Conclusion The findings of this study have important implications for the development of interventions or public health messages aimed at reducing television viewing time and its associated consequences, in children from low maternal-level education families. As maternal education is often considered a proxy for family socioeconomic status, these findings may be able to be generalized to families facing other forms of socioeconomic disadvantage (e.g., those where parents have low incomes or low occupational status). Specifically, it appears that messages focusing on removing televisions from children’s bedrooms, limiting the number of televisions in the home, turning the television off during mealtimes, and discouraging eating (meals or snacking) while watching television, may have the greatest potential to effect change in viewing patterns and particularly to reduce elevated viewing times among children whose mothers have lower levels of education. The study was funded by VicHealth. K.H. is supported by an NHMRC/National Heart Foundation Postdoctoral Research Fellowship, K.B. is supported by an NHMRC/National Heart
July 2007
Foundation Career Development Award, D.C. is supported by an NHMRC/National Heart Foundation Career Development Award, K.C. is supported by a National Heart Foundation Postdoctoral Research Fellowship, and J.S. is supported by a VicHealth Public Health Research Fellowship. No financial conflict of interest was reported by the authors of this paper.
References 1. Singer MI, Miller DB, Guo SS, Flannery DJ, Frierson D, Slovak K. Contributors to violent behaviour among elementary and middle school children. Pediatrics 1999;104:878 – 84. 2. Ozmert E, Toyran M, Yurdakok K. Behavioral correlates of television viewing in primary school children evaluated by the child behavior checklist. Arch Pediatr Adol Med 2002;156:910 –14. 3. Brodersen NH, Steptoe A, Williamson S, Wardle J. Sociodemographic, developmental, environmental, and psychological correlates of physical activity and sedentary behaviour at age 11 to 12. Ann Behav Med 2005; 29:2–11. 4. Salmon J, Campbell KJ, Crawford DA. Television viewing habits associated with obesity risk factors: a survey of Melbourne schoolchildren. Med J Aust 2006;184:64 –7. 5. Dietz WH, Gortmaker SL. Do we fatten our children at the television set? Obesity and television viewing in children and adolescents. Pediatrics 1985;75:807–12. 6. American Academy of Pediatrics Committee on Public Education. Children, adolescents and television. Pediatrics 2001;107:423– 6. 7. Bar-on ME. The effects of television on child health: implications and recommendations. Arch Dis Child 2000;83:289 –292. 8. Australian Government Department of Health and Ageing. Australia’s physical activity recommendations for 5–12 year olds. 2004, Commonwealth of Australia: Canberra. 9. Gorely T, Marshall SJ, Biddle SJH. Couch kids: correlates of television viewing among youth. Int J Behav Med 2004;11:152– 63. 10. Salmon J, Timperio A, Telford A, Carver A, Crawford D. Association of family environment with children’s television viewing and with low level of physical activity. Obes Res 2005; 13:1939 –51. 11. Christakis DA, Ebel BE, Rivara FP, Zimmerman FJ. Television, video, and computer game usage in children under 11 years of age. J Pediatr 2004; 145:652– 6. 12. Davison KK, Birch L.L. Childhood overweight: a contextual model and recommendations for future research. Obes Rev 2001;2:159 –71. 13. Saelens BE, Sallis JF, Nader PR, Broyles SR, Berry CC, Taras HL. Home environmental influences on children’s television watching from early to middle childhood. Dev Behav Pediatr 2002;23:127–32. 14. Australian Bureau of Statistics. 1996 Census of Population and Housing: Socio-Economic Indexes for Areas. ABS Catalogue No. 2039.0. 1998. Commonwealth of Australia: Canberra. 15. Fleiss JL. The design and analysis of clinical experiments. New York: Wiley & Sons; 1986. 16. Portney L, Watkins M. Foundations of clinical research: applications to practice, 2nd ed. New Jersey: Prentice-Hall; 2000. 17. Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Social Psychol 1986;51:1173– 82. 18. Tabachnick BG, Fidell LS. Using multivariate statistics, 4th ed. Boston: Allyn and Bacon; 2001. 19. Kleinbaum DG, Kupper LL, Muller KE. Applied regression analysis and other multivariable methods. Boston: PWS-Kent Publishing; 1998. 20. Roberts DF, Foehr UG, Rideout VJ, Brodie M. Kids and media at the new millennium: a comprehensive national analysis of children’s media use. Menlo Park, CA: Henry J Kaiser Family Foundation; 1999. 21. Krosnick JA. Survey research. Annu Rev Psychol 1999;50:537– 67.
Am J Prev Med 2007;33(1)
47
Parents of 1484 children reported maternal education, time their child spends watching television, and 21 aspects of the family television environment (potential mediators) during 2002 and 2003. Separate regression analyses were conducted in 2006 for each potential mediator that met two initial conditions for mediation (associated with both maternal education and children’s television viewing (p⬍0.10)), to assess whether inclusion reduced the association between maternal education and children’s television viewing. Multivariable regression assessed the combined impact of all mediators.
Results:
Twelve of 21 potential mediators met the initial conditions for mediation. Inclusion of each resulted in decreased  values (3.2% to 15.2%) for the association between maternal education and television viewing. Number and placement of televisions in the home appeared to have the greatest mediating effect, followed by frequency of eating dinner in front of the television with the child and rules about television viewing during mealtimes. Together, the 12 mediators accounted for more than one-third of the association between maternal education and children’s television viewing time.
Conclusions: This study suggests the strong inverse relationship between maternal education and children’s television viewing is partly mediated by aspects of the family television environment. (Am J Prev Med 2007;33(1):41– 47) © 2007 American Journal of Preventive Medicine
Introduction
C
hildren’s high levels of television viewing are associated with a range of negative social and health effects including aggressive behavior,1,2 poor school achievement,2 social problems,2 poor selfrated health,3 consumption of high-energy drinks and snacks,4 low fruit consumption,4 lack of participation in organized physical activity,4 and overweight.5 In an attempt to reduce such negative effects, it has been recommended that parents limit the amount of time their children spend watching television.6 – 8 However, some parents may be better equipped than others to act on such recommendations. For example, maternal education level is consistently found to be inversely associated with children’s television viewing.9 –11 Maternal education also appears to be related to certain aspects of the family television environment.11 From the School of Exercise & Nutrition Sciences, Deakin University, Victoria, Australia Address correspondence and reprint requests to: Kylie Hesketh, PhD, School of Exercise & Nutrition Sciences, Deakin University, 221 Burwood Hwy, Burwood, Victoria, 3125 Australia. E-mail: kylie. [email protected].
The social ecologic model proposed by Davison and Birch12 provides a useful framework for investigating influences on health-related behaviors such as television viewing. According to this model, the family television environment can include both the physical (number and location of televisions) and the social environments (rules and practices associated with television viewing) within the family home. Maternal education has been found to be inversely associated with a child having a television in his or her bedroom and to be positively associated with parental concern about the amount of time their child spends watching television.11 Further, aspects of the family television environment, including frequency of families watching television together,10 mealtime television viewing,10,13 parents’ television viewing habits,9 and having a television in the child’s bedroom9 have also been found to be associated with children’s viewing time. This study aims to develop a greater understanding of the relationships between maternal education, the family television environment, and the amount of time children spend watching television. Specifically, aspects of the family television environment that may mediate the relationship between maternal education and
Am J Prev Med 2007;33(1) © 2007 American Journal of Preventive Medicine • Published by Elsevier Inc.
0749-3797/07/$–see front matter doi:10.1016/j.amepre.2007.02.039
41
children’s television viewing time will be investigated. By developing an understanding of the potential mediators, we will be better able to tailor interventions targeting mothers with lower levels of education, which aim to reduce television viewing and its associated consequences.
Methods Sample The sample consists of 1484 families involved in the Health, Eating and Play Study, conducted from 2002 to 2003. This study employed a stratified random sampling technique to select schools for participation in the study. All government and Catholic elementary schools in suburban Melbourne, Australia with a total enrollment exceeding 200 students were classified into socioeconomic status (SES) quintiles using the Australian Bureau of Statistics Socio-Economic Indexes for Areas Index of Relative Socio-Economic Disadvantage.14 In order to ensure representation from a spread of SES groups, schools classified in the highest, middle, and lowest quintiles were eligible for sampling (schools in the second and fourth quintiles were excluded). Thirteen schools from each of the three included quintile groups were randomly selected with a probability proportional to school size. Twenty-four schools (nine high, seven middle, and eight low SES area schools) agreed to participate. All families of children in their first year of school (mean age 6 years; younger children) at all 24 schools and families of children in school year levels 5 and 6 (mean age 11 years; older children) at 17 of the 24 schools were invited to take part. (Study 2 was initially a study of school-entry age children only. Opportunistically, children in years 5 and 6 were also included during 2003 data collection. The 17 schools from which the older children were included were evenly distributed across the low, middle, and high SES groups.) The study received ethics approval from the Deakin University Human Ethics Committee, the Victorian Department of Education and Training, and the Catholic Education Office. Under existing ethical guidelines, it was necessary to seek active written consent from parents for each child’s participation, and no information could be accessed regarding characteristics of nonrespondents. Written parental consent was received for 1562 children (42% response). No area-level socioeconomic gradient was noted in response rates (41% response at high, 39% at middle, and 48% at low SES area schools). Due to incomplete independent and/or dependent variable data, 78 children were excluded from analyses for this paper.
Measures All data were provided by the child’s main caregiver who completed a questionnaire at home. Respondents provided information on their own and, where applicable, their partner/co-caregiver’s highest level of schooling. The six response options were condensed to three levels of education: less than secondary school 42
(never attended school, primary school, some high school), secondary school (completed high school, technical or trade school certificate/apprenticeship) and post-secondary (university or tertiary qualification). Based on reported gender of the respondent and co-caregiver, maternal (mother or female co-caregiver) education was derived. Respondents reported the amount of time their child spends watching television (including commercial, noncommercial, cable/pay TV, videos, and DVDs) on a usual school day and usual weekend day (scale ranging from 0 to 6 or more hours, in half hour segments). School-day estimates were multiplied by 5, and weekend day estimates were multiplied by 2; the totals were summed and divided by 7 to generate average viewing time (minutes per day), which was slightly positively skewed (skewness, 0.93). Intra-class correlations (ICCs) and percent agreement were used to assess test–retest reliability. For continuous variables, an ICC above 0.75 was considered as indicating excellent reliability, an ICC between 0.4 and 0.75 as fair to good reliability, and an ICC below 0.4 was considered as poor reliability.15 For categorical variables, percent agreement values above 66% were considered fair.16 The ICC showed this measure to be reliable in repeated administrations (ICC⫽0.78, 95% confidence interval [CI]⫽0.69 – 0.84, p⬍0.001).4 Respondents also answered 22 items (assessing 21 mediators) providing information on their family’s television viewing environment. These items were developed, based on a social ecologic framework,12 to incorporate factors such as family television viewing practices, family television rules, parental modelling, division of responsibility, and access to sedentary pursuits at home. Two identical items inquired about how much the respondent and co-caregiver are concerned about how much time they spend watching TV (not at all, a little bit, quite a bit, or very much). As with level of education, based on reported gender of the respondent and co-caregiver, maternal level of care about their own television viewing was derived. The remaining 20 items are detailed in Table 1. Some of these items were developed and reliability tested as part of a previous study10 and were modified for use here. All 22 items underwent test–retest reliability testing as part of this study. Due to the time required to complete the entire questionnaire, the original questionnaire was divided in two for the retest, and separate subsamples derived from the main study sample were asked to complete one half of the survey 2 weeks after they had completed the initial (entire) questionnaire. A random subsample of 137 study parents completed the first half of the original questionnaire a second time (encompassing the majority of items involved in this paper), and a random subsample of 39 study parents completed the second half of the original questionnaire a second time (encompassing 5 items used in this
American Journal of Preventive Medicine, Volume 33, Number 1
www.ajpm-online.net
Table 1. Items used to assess family television environment (potential mediators) Itema
Test-retest reliability
Response options
How often did your child do the following things? My child ate dinner in front of the television My child ate snacks while watching television
Never or rarely, less than once a week, once a week, about 2–3 times/week, about 4–6 times/week, every day How often did you and/or the co-caregiver do the following things together with your child? I/we watched TV, videos or DVDs together with the child Never or rarely, less than once a week, I/we ate dinner in front of the television together with the child once a week, about 2–3 times/week, I/we ate snacks together with the child while watching television about 4–6 times/week, every day Indicate how much you agree or disagree with the following statements: I have to be sure that my child does not watch too much TV Disagree, slightly disagree, neutral, I will switch off the TV if I think my child is watching too much slightly agree, agree I let my child watch TV in exchange for good behavior If I did not guide or regulate my child’s TV watching, (s)he would watch too much I let my child watch any television show (s)he chooses I restrict how much time my child spends watching TV During meal times, I do not allow the TV to be on How concerned are you that your child watches too much Not concerned, a little concerned, television? fairly concerned, very concerned How many TVs do you have in your home? Open response Does your child have a TV set in his/her bedroom? Yes, no Who is responsible for the following? Deciding how much TV your child should watchb Parent(s) (mainly me, mainly the coDeciding when your child can watch TVb caregiver, shared role combined), mainly my child Deciding whether the TV can be on when eating our evening mealb How much do you agree with the following statements about you and the co-caregiver? Disagree, slightly disagree, neutral, My child’s co-caregiver and I have the same views about how slightly agree, agree much TV our child should watchb My child’s co-caregiver supports the rules I make about when our child is able to watch TVb
ICC ⫽ 0.82 ICC ⫽ 0.63 ICC ⫽ 0.54 ICC ⫽ 0.75 ICC ⫽ 0.60 ICC ICC ICC ICC
⫽ ⫽ ⫽ ⫽
0.56 0.43 0.62 0.50
ICC ICC ICC ICC
⫽ ⫽ ⫽ ⫽
0.39 0.46 0.71 0.50
ICC ⫽ 0.95 95% agreement 94% agreement 91% agreement 91% agreement ICC ⫽ 0.59 ICC ⫽ 0.60
a
All items relating to frequency of behavior asked respondents to think about the past few months. Test-retest reliability of these items employed a different subsample to that for the above items. ICC, intra-class correlation.
b
paper). ICC or percent agreement statistics (for continuous and categorical variables, respectively) were calculated for each questionnaire item, and the results are presented in Table 1.
Analyses Children’s gender is consistently found to be unrelated to television viewing,9 whereas it appears that older children may watch more television than younger children.9 These associations were observed in the current sample (independent samples t test for gender, p⫽0.37; and for age group, p⬍0.001), therefore, data were analyzed for the sample as a whole and separately for younger (grade prep) and older (grade 5– 6) children. To account for the cluster-based sampling, all analyses included school as the cluster unit to give a robust variance estimate that adjusts for within-cluster correlation. Baron and Kenny’s17 approach to testing mediation was used in this study. Individual linear and logistic regression analyses were used, as appropriate, to assess July 2007
which potential mediating variables met the initial two conditions for mediation: (1) the independent variable (maternal education) being significantly associated with the potential mediator and (2) the potential mediator being significantly associated with the dependent variable (children’s television viewing).17 For these initial criteria, an inclusive significance level of p⬍0.10 was set. The final condition of mediation is that inclusion of the potential mediator reduces the association between the independent and dependent variables.17 To test this, separate linear regression analyses were conducted for each potential mediator including the potential mediator and maternal education as the independent variables. To test for collinearity, pair-wise correlations (r ⬎0.7)18 and variance inflation factors ([VIF]ⱖ10)19 for potential mediators were assessed. Finally, a multivariable regression analysis was conducted including maternal education and all potential mediators simultaneously. The decrease in the size of the regression () coefficient was used as an indicator of the magnitude of mediation, and R2 was used to Am J Prev Med 2007;33(1)
43
Table 2. Sample characteristics n (%)
Characteristic Boys Area-level SESa Low (⬍26th percentile) Middle (26–75th percentile) High (⬎75th percentile) Maternal education Less than secondary Secondary Post-secondary Daily TV viewing (mean SD) Meets guidelinesb (ⱕ2 hr/day)
All children (nⴝ1484)
Younger children (mean age 6 yrs) (nⴝ589)
Older children (mean age 11 yrs) (nⴝ895)
705 (48)
291 (49)
414 (46)
263 (18) 513 (35) 686 (47)
141 (24) 206 (35) 235 (40)
122 (14) 307 (35) 451 (51)
355 (24) 596 (40) 533 (36) 188 (93) min 4 (0.3)
136 (23) 239 (41) 214 (36) 166 (87) min 1 (0.2)
219 (24) 257 (40) 319 (37) 202 (93) min 3 (0.3)
a
Australian Bureau of Statistics postcode-derived “Socio-economic Indexes for Areas Index of Advantage-Disadvantage” (Australian Bureau of Statistics, 2003). Australian recommendation that children should not spend more than 2 hours per day using electronic media for entertainment (Australian Government Department of Health and Aging 2004). SD, standard deviation; SES, socioeconomic status.
b
provide an indication of the proportion of variance in children’s television viewing accounted for by each model. All analyses were conducted in 2006.
Results Characteristics of the 1484 children in the sample are presented in Table 2. In both age groups, the sample was distributed across area-level SES and maternal education categories, providing a socioeconomically diverse sample. Mean daily television viewing for the total sample exceeded 3 hours and was higher in older than younger children. Only four children in the entire sample spent an average of 2 hours or less per day watching television and thus potentially met the Australian guidelines for electronic media use.8 Maternal education was strongly inversely related to daily television viewing in the total sample (⫽ ⫺31.0, 95% CI⫽ ⫺38.7, ⫺23.2), and among younger (⫽ ⫺25.3, 95% CI⫽ ⫺35.9, ⫺14.8) and older children (⫽ ⫺34.2, 95% CI⫽ ⫺43.7, ⫺24.8).
Assessment of Whether Potential Mediators Meet Initial Conditions for Mediation Of the 21 potential mediators assessed, six were not related (p⬎0.10) to maternal education (first initial condition to be met for mediation), and a further two were not related to daily television viewing (second initial condition to be met for mediation) in either age group. Of the remaining 13 potential mediators, one (co-caregiver supports television rules) was excluded, as it was only related to both maternal education and daily television viewing in younger children, and its inclusion resulted in the loss from analyses of children from single-parent families. The final 12 potential mediators 44
(Table 3) were included in subsequent analyses; only children with complete data (independent variable, dependent variable, and all 12 potential mediators) were included (n⫽1348). The vast majority of pair-wise correlations among mediator variables did not exceed r⫽0.2. Only one pair of potential mediators was correlated at r ⬎0.7:18 frequency of the child eating dinner in front of the TV with frequency of parent(s) eating dinner in front of the TV with the child (r⫽0.80). The VIFs of all of the individual mediators were low (⬍10)19 (range 1.06 – 3.17). As the VIFs were low, and exclusion of either variable in the final multivariable model resulted in negligible change in  values, neither of the highly correlated variables was excluded from the analyses involving all the potential mediators concurrently.
Effect of Potential Mediators on the Relationship Between Maternal Education and Children’s Television Viewing Table 3 presents  values and reduction in  values, for the association between maternal education and children’s television viewing, for each potential mediating model. For each individual mediator, inclusion in the linear regression model resulted in a decrease in  values (between the range of 3.2% and 15.2%) for the association between maternal education and television viewing. However, the association between maternal education and children’s television viewing habits remained significant (p⬍0.001) in all models. This supports the hypothesis that these family television environment variables partly mediate the association between maternal education and children’s television viewing. The two variables describing the number and
American Journal of Preventive Medicine, Volume 33, Number 1
www.ajpm-online.net
July 2007 Table 3. Linear regression analyses assessing change in  values for association between maternal education and children’s TV viewing when (a) each potential mediator is included independently and (b) when all potential mediators are included concurrently Total sample (nⴝ1348)
Younger children (nⴝ543)
Older children (nⴝ805)
Am J Prev Med 2007;33(1)
Variables associated with children’s TV viewing
 (95% CI) for maternal education
Reduction in  (%)
R2
 (95% CI) for maternal education
Reduction in  (%)
R2
 (95% CI) for maternal education
Reduction in  (%)
R2
Maternal education only Maternal education plusa: Frequency of eating dinner in front of the TV Frequency of eating snacks while watching TV Frequency of watching TV with child Frequency of eating dinner in front of TV with child Frequency of eating snacks with child while watching TV Will switch TV off if thinks child is watching too much Lets child watch any TV show (s)he chooses Restricts how much time child spends watching TV Do not allow TV on during mealtimes Number of TVs in the home Child has TV in his/her bedroom Responsibility for deciding if TV can be on during dinner Maternal education plus all potential mediatorsb
⫺31.0 (⫺38.7, ⫺23.2)
–
0.06
⫺25.3 (⫺35.9, ⫺14.8)
–
0.05
⫺34.2 (⫺43.7, ⫺24.8)
–
0.08
⫺27.7 (⫺38.1, ⫺20.2)
10.6
0.13
n/a
⫺30.0 (⫺38.5, ⫺21.6)
12.3
0.14
⫺28.2 (⫺35.7, ⫺20.6)
9.0
0.16
⫺22.9 (⫺32.2, ⫺13.5)
⫺31.4 (⫺41.0, ⫺21.7)
8.2
0.17
⫺28.3 (⫺35.3, ⫺21.3)
8.7
0.13
n/a
⫺32.2 (⫺39.9, ⫺24.6)
5.8
0.12
⫺27.4 (⫺35.2, ⫺19.5)
11.6
0.11
⫺22.5 (⫺33.6, ⫺11.4)
11.1
0.09
⫺30.7 (⫺39.5, ⫺21.9)
10.2
0.11
⫺27.8 (⫺35.3, ⫺20.3)
10.3
0.11
⫺21.9 (⫺31.8, ⫺11.9)
13.4
0.11
⫺31.5 (⫺40.2, ⫺22.8)
7.9
0.11
⫺29.8 (⫺37.3, ⫺22.4)
3.9
0.08
⫺23.8 (⫺33.7, ⫺13.9)
5.9
0.07
n/a
⫺28.8 (⫺36.2, ⫺21.4)
7.1
0.10
n/a
⫺32.1 (⫺40.1, ⫺24.0)
6.1
0.13
⫺28.6 (⫺36.5, ⫺20.7)
7.7
0.10
⫺23.0 (⫺32.7, ⫺13.3)
9.1
0.09
⫺32.1 (⫺42.6, ⫺21.6)
6.1
0.11
⫺27.5 (⫺34.7, ⫺20.2)
11.3
0.10
⫺22.8 (⫺32.4, ⫺13.2)
9.9
0.11
⫺31.2 (⫺39.1, ⫺23.3)
8.9
0.09
⫺26.9 (⫺34.0, ⫺19.8) ⫺26.3 (⫺33.8, ⫺18.7) ⫺30.0 (⫺37.3, ⫺22.7)
13.2 15.2 3.2
0/10 0.11 0.09
⫺23.8 (⫺34.4, ⫺13.2) ⫺21.9 (⫺32.6, ⫺11.2) n/a
5.9 13.4
0.06 0.09
⫺29.8 (⫺39.4, ⫺20.2) ⫺30.0 (⫺40.0, ⫺19.9) n/a
12.9 12.3
0.11 0.10
⫺18.8 (⫺25.8, ⫺11.9)
39.4
0.30
⫺16.9 (⫺25.9, ⫺7.8)
33.2
0.23
⫺22.3 (⫺30.7, ⫺14.0)
34.8
0.29
9.5
All significant at p⬍0.001; n/a indicates item did not meet mediator criteria for this group of children and was therefore not assessed. a Each potential mediator included in a separate model (not cumulatively). b Twelve potential mediators included concurrently for total sample, 8 for younger children, 10 for older children. CI, confidence interval.
0.15
45
placement of televisions in the home appeared to have the greatest mediating effect. That is, the mediator models including each of these variables individually resulted in the greatest reduction in  coefficients for the association between maternal education and children’s television viewing. Although the impact of a television in the child’s bedroom was similar for older and younger children, inclusion of the number of televisions in the home resulted in a greater reduction in the  coefficient for the association between maternal education and children’s television viewing in older than younger children. Of the social family television environment variables, frequency of eating dinner in front of the television with the child and rules about television viewing during mealtimes appeared to have the greatest mediating effects. In most cases, the magnitude of reduction in  values for the association between maternal education and television viewing was similar for both age groups. However, frequency of eating snacks with the child while watching television and restricting how much time the child spends watching television had a greater magnitude of effect in younger than in older children (13.4% vs 7.9% reduction in ). Inclusion of all mediators concurrently resulted in a 39% reduction in the  value, suggesting these family television environment variables account for more than one-third of the association between maternal education and children’s television viewing. Similarly, large reductions in the  value were observed for the younger (33.2%) and older (34.8%) groups of children. Again, the association between maternal education and children’s television viewing remained significant in these models (p⬍0.001 for total sample and older children; p⫽0.001 for younger children).
Discussion Children in this study generally spent more time watching television than the 2 hours per day recommended by current guidelines.8 Consistent with previous research,9 –11 children whose mothers had attained lower levels of education tended to spend more time watching television than those whose mothers had higher levels of education. This study suggests that the strong inverse relationship between maternal education and children’s television viewing is partly mediated by certain aspects of the family television environment, although temporal relationships cannot be ascertained from this cross-sectional data. In particular, physical aspects of the family television environment (number of televisions in the home and presence of a television in the child’s bedroom) and family rules and practices associated with eating while watching television appeared to be most important. Together, the 12 family television environment mediators accounted for more 46
than one third of the association between maternal education and children’s television viewing time. The strongest mediators of the relationship between maternal education and children’s television viewing were those related to the physical aspects of the family television environment. Although separate studies have reported that the presence of a television in the child’s bedroom is inversely associated with maternal education11 and positively associated with children’s television viewing time,9 we were able to show that having a television in the child’s bedroom may also mediate the association between maternal education and children’s television viewing habits. The number of television sets in the home was also a mediator of the association between maternal education and children’s television viewing. It appears that the greater television viewing time observed for children with less-educated mothers is partially related to the different physical environments provided by these mothers. Thus providing mothers, particularly those with lower levels of education, with messages about the importance of limiting children’s access to television sets in the home may be a useful strategy to reduce children’s television viewing time. The social aspects of the family television environment that most strongly mediated the association between maternal education and children’s television viewing were those related to rules and practices within the family related to eating while watching television. An association between mealtime television viewing and children’s television viewing time has previously been reported,10,13 but we have demonstrated that these rules and practices may be inequitably distributed depending on maternal education level. Parents with lower levels of education are less likely to report having household rules related to television viewing.20 Lesseducated mothers may have more difficulty, or place lesser importance on limit setting with regard to their child’s television viewing, with higher levels of viewing the consequence. Equally, less-educated mothers may be raising their children in more challenging physical environments (e.g., smaller homes, smaller backyards, neighborhoods with greater safety concerns and fewer outdoor play spaces) and may use television as a babysitter or an alternative to fee-incurring after-school activities. Although this study achieved only a modest response rate, the participants spanned a broad range of socioeconomic backgrounds. In addition, the response rate is similar to that achieved in other health surveys, and there is evidence that response rates for most major studies have been falling over the past few decades and that it is not necessary to have high response rates to ensure a broad spectrum of respondents are represented.21 Although the test–retest reliability results for some of the potential mediator variables were modest, the variables that most strongly mediated the associa-
American Journal of Preventive Medicine, Volume 33, Number 1
www.ajpm-online.net
tion between maternal education and children’s television viewing demonstrated greater reliability on repeat administration. The large number of statistical tests performed means that it is possible that some of the associations detected were chance findings, however, a larger number of associations were detected than would be expected by chance. Like all associational analyses based on cross-sectional data, the ability to draw firm temporal or causal conclusions of mediating effects from these data are limited. However, maternal education could plausibly be assumed to be established prior to the family television environment and children’s television viewing practices. Alternative explanations for these findings should also be considered. It is possible that an additional unmeasured factor, such as overall family SES influences maternal education, the television environment, and television viewing time. Furthermore, this study did not assess the potentially important role of the father or co-caregiver, who may also have an influential role in shaping the physical environment. The strengths of this study include a large sample, allowing stratification by age group and the consideration of multiple potential mediators of the relationship between maternal education and children’s television viewing time.
Conclusion The findings of this study have important implications for the development of interventions or public health messages aimed at reducing television viewing time and its associated consequences, in children from low maternal-level education families. As maternal education is often considered a proxy for family socioeconomic status, these findings may be able to be generalized to families facing other forms of socioeconomic disadvantage (e.g., those where parents have low incomes or low occupational status). Specifically, it appears that messages focusing on removing televisions from children’s bedrooms, limiting the number of televisions in the home, turning the television off during mealtimes, and discouraging eating (meals or snacking) while watching television, may have the greatest potential to effect change in viewing patterns and particularly to reduce elevated viewing times among children whose mothers have lower levels of education. The study was funded by VicHealth. K.H. is supported by an NHMRC/National Heart Foundation Postdoctoral Research Fellowship, K.B. is supported by an NHMRC/National Heart
July 2007
Foundation Career Development Award, D.C. is supported by an NHMRC/National Heart Foundation Career Development Award, K.C. is supported by a National Heart Foundation Postdoctoral Research Fellowship, and J.S. is supported by a VicHealth Public Health Research Fellowship. No financial conflict of interest was reported by the authors of this paper.
References 1. Singer MI, Miller DB, Guo SS, Flannery DJ, Frierson D, Slovak K. Contributors to violent behaviour among elementary and middle school children. Pediatrics 1999;104:878 – 84. 2. Ozmert E, Toyran M, Yurdakok K. Behavioral correlates of television viewing in primary school children evaluated by the child behavior checklist. Arch Pediatr Adol Med 2002;156:910 –14. 3. Brodersen NH, Steptoe A, Williamson S, Wardle J. Sociodemographic, developmental, environmental, and psychological correlates of physical activity and sedentary behaviour at age 11 to 12. Ann Behav Med 2005; 29:2–11. 4. Salmon J, Campbell KJ, Crawford DA. Television viewing habits associated with obesity risk factors: a survey of Melbourne schoolchildren. Med J Aust 2006;184:64 –7. 5. Dietz WH, Gortmaker SL. Do we fatten our children at the television set? Obesity and television viewing in children and adolescents. Pediatrics 1985;75:807–12. 6. American Academy of Pediatrics Committee on Public Education. Children, adolescents and television. Pediatrics 2001;107:423– 6. 7. Bar-on ME. The effects of television on child health: implications and recommendations. Arch Dis Child 2000;83:289 –292. 8. Australian Government Department of Health and Ageing. Australia’s physical activity recommendations for 5–12 year olds. 2004, Commonwealth of Australia: Canberra. 9. Gorely T, Marshall SJ, Biddle SJH. Couch kids: correlates of television viewing among youth. Int J Behav Med 2004;11:152– 63. 10. Salmon J, Timperio A, Telford A, Carver A, Crawford D. Association of family environment with children’s television viewing and with low level of physical activity. Obes Res 2005; 13:1939 –51. 11. Christakis DA, Ebel BE, Rivara FP, Zimmerman FJ. Television, video, and computer game usage in children under 11 years of age. J Pediatr 2004; 145:652– 6. 12. Davison KK, Birch L.L. Childhood overweight: a contextual model and recommendations for future research. Obes Rev 2001;2:159 –71. 13. Saelens BE, Sallis JF, Nader PR, Broyles SR, Berry CC, Taras HL. Home environmental influences on children’s television watching from early to middle childhood. Dev Behav Pediatr 2002;23:127–32. 14. Australian Bureau of Statistics. 1996 Census of Population and Housing: Socio-Economic Indexes for Areas. ABS Catalogue No. 2039.0. 1998. Commonwealth of Australia: Canberra. 15. Fleiss JL. The design and analysis of clinical experiments. New York: Wiley & Sons; 1986. 16. Portney L, Watkins M. Foundations of clinical research: applications to practice, 2nd ed. New Jersey: Prentice-Hall; 2000. 17. Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Social Psychol 1986;51:1173– 82. 18. Tabachnick BG, Fidell LS. Using multivariate statistics, 4th ed. Boston: Allyn and Bacon; 2001. 19. Kleinbaum DG, Kupper LL, Muller KE. Applied regression analysis and other multivariable methods. Boston: PWS-Kent Publishing; 1998. 20. Roberts DF, Foehr UG, Rideout VJ, Brodie M. Kids and media at the new millennium: a comprehensive national analysis of children’s media use. Menlo Park, CA: Henry J Kaiser Family Foundation; 1999. 21. Krosnick JA. Survey research. Annu Rev Psychol 1999;50:537– 67.
Am J Prev Med 2007;33(1)
47