Screen Media

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

ORIGINAL ARTICLE

Time-varying Effects of Screen Media Exposure in the


Relationship Between Socioeconomic Background and
Childhood Obesity
Joost Oude Groeniger,a,b Willem de Koster,b and Jeroen van der Waalb

Conclusions: Our findings are consistent with the hypothesis that


Background: We investigated to what extent social inequalities in
social inequalities in screen media exposure contribute substantially
childhood obesity could be reduced by eliminating differences in
to social inequalities in childhood obesity.
screen media exposure.
Keywords: Causal mediation analysis; Childhood obesity; Health
Methods: We used longitudinal data from the UK-wide Millennium
inequalities; Marginal structural model; Screen media exposure
Cohort Study (n = 11,413). The study measured mother’s educa-
tional level at child’s age 5. We calculated screen media exposure (Epidemiology 2020;31: 578–586)
as a combination of television viewing and computer use at ages 7
and 11. We derived obesity at age 14 from anthropometric measures.
We estimated a counterfactual disparity measure of the unmediated
association between mother’s education and obesity by fitting an in-
verse probability-weighted marginal structural model, adjusting for
T he prevalence of excess weight among children has risen
dramatically in the last 4 decades.1,2 Childhood obesity is
linked to a range of adverse outcomes across the life course,
mediator–outcome confounders.
Results: Compared with children of mothers with a university de- including greater risk of chronic diseases, more mental health
gree, children of mothers with education to age 16 were 1.9 (95% problems, and lower socioeconomic attainment.3 Especially
confidence interval [CI] = 1.5, 2.3) times as likely to be obese. Those alarming is the differential distribution of childhood obesity
whose mothers had no qualifications were 2.0 (95% CI = 1.5, 2.5) across socioeconomic groups.4 Socioeconomically disadvan-
times as likely to be obese. Compared with mothers with university taged children are at a considerably higher risk to develop obe-
qualifications, the estimated counterfactual disparity in obesity at age sity, and recent evidence from the United Kingdom suggests
14, if educational differences in screen media exposure at age 7 and that these inequalities will keep rising.4 Given the already dis-
11 were eliminated, was 1.8 (95% CI = 1.4, 2.2) for mothers with proportionate health disadvantage of children growing up in
education to age 16 and 1.8 (95% CI = 1.4, 2.4) for mothers with no lower socioeconomic environments, and the need to intervene
qualifications on the risk ratio scale. Hence, relative inequalities in early in life to prevent obesity before it is established, tack-
childhood obesity would reduce by 13% (95% CI = 1%, 26%) and ling social inequalities in childhood obesity is listed as a vital
17% (95% CI = 1%, 33%). Estimated reductions on the risk differ-
public health strategy.5 Particularly for children, who have
ence scale (absolute inequalities) were of similar magnitude.
little control over the circumstances affecting their health, po-
Submitted September 23, 2019; accepted May 4, 2020.
tentially avoidable health inequalities are considered unjust.6,7
From the aDepartment of Public Health, Erasmus University Medical Centre Reducing these inequalities, however, requires evidence on
Rotterdam, Rotterdam, The Netherlands; and bDepartment of Public Ad- the effect of intervening on modifiable mechanisms in the re-
ministration and Sociology, Erasmus University Rotterdam, Rotterdam,
The Netherlands.
lationship between socioeconomic background and childhood
The authors report no conflicts of interest. obesity.
Supplemental digital content is available through direct URL citations Screen media exposure is a major risk factor for child-
in the HTML and PDF versions of this article (www.epidem.com).
Replication of results: Due to data protection regulations, the data cannot be hood obesity and an increasingly common leisure activity of
made available by the authors. Interested researchers may obtain the data children.8–10 Many children spend hours per day behind tel-
via UK Data Archive. Annotated Stata code is provided in the eAppendix. evision or computer screens, which substantially increases
Correspondence: Joost Oude Groeniger, Department of Public Health,
Erasmus University Medical Centre, PO Box 2040, 3000 CA Rotterdam, their obesity risk.11 Screen media exposure may affect body
The Netherlands. E-mail: [email protected]. weight by increasing food consumption and exposure to food
Copyright © 2020 The Author(s). Published by Wolters Kluwer Health, Inc. and beverage advertisements, lowering energy expenditure,
This is an open-access article distributed under the terms of the Crea-
tive Commons Attribution-Non Commercial-No Derivatives License 4.0 and reducing sleep duration.9,12 Moreover, screen media ex-
(CCBY-NC-ND), where it is permissible to download and share the work posure is substantially higher among children from lower so-
provided it is properly cited. The work cannot be changed in any way or cioeconomic backgrounds than among children from higher
used commercially without permission from the journal.
ISSN: 1044-3983/20/3104-0578 socioeconomic backgrounds.13,14 Limited financial resources
DOI: 10.1097/EDE.0000000000001210 to engage in more expensive leisure activities are likely to be

578 | www.epidem.com Epidemiology  •  Volume 31, Number 4, July 2020


Epidemiology  •  Volume 31, Number 4, July 2020 Screen Media Exposure and Inequalities in Childhood Obesity

associated with increased screen media exposure among lower Measures


socioeconomic status families. Moreover, more disadvanta- We used maternal education because it is a frequently
geous neighborhood conditions may discourage playing out- used and stable measure of socioeconomic position, a strong
side.15 Differences in screen media habits may also result from predictor of children’s life chances, and less sensitive to
other social determinants and transmit to children via sociali- measurement error than, for example, income.35–37 We used
zation and social learning practices.16–19 First, norms in more- mother’s highest attained educational level at child’s age 5 as
educated social environments have shifted to disapproval and the main exposure to minimize the number of mothers still
stigmatization of sedentary activities, such as television, view- enrolled in school, while still allowing for temporal ordering
ing in favor of a more active lifestyle.20,21 Second, childrearing of the measures. We excluded mothers who were a student at
practices of more-educated parents are increasingly aimed at child’s age 5 from the analysis (n = 151; 1%); we categorized
improving children’s development, resulting in more extra- mothers who had obtained qualifications overseas as missing
curricular activities and less screen media exposure.22 Third, and subsequently imputed their education (described below;
greater cognitive abilities may result in a higher awareness of n = 312; 3%). Educational categories include (1) university
the negative health consequences of screen media exposure (education to age 20+); (2) education to age 18 (A-level equiv-
and a preference for other activities that require greater infor- alent); (3) education to age 16 (O-level equivalent); and (4) no
mation processing capacities.23 qualifications.
To examine to what extent screen media exposure con- Screen media exposure was measured combining tele-
tributes to social inequalities in childhood obesity, we used vision viewing and computer use. At child’s age 7, mothers
longitudinal data from the Millennium Cohort Study. We reported how many hours, on a normal weekday, their child
aimed to estimate to what extent social inequalities (meas- spent (1) watching television, videos, or DVDs and (2) using a
ured by mother’s educational level) in childhood obesity at computer or playing electronic games outside school lessons.
age 14 would be reduced if differences in screen media expo- Answer categories were as follows: (1) none; (2) <1 hour; (3)
sure (television viewing and computer use) at ages 7 and 11 1 to <3 hours; (4) 3 to <5 hours; (5) 5 to <7 hours; and (6) ≥7
were eliminated. To do so, we used mediation methods that hours. At child’s age 11, the same questions were asked with
are able to estimate the effect of time-varying mediators even an additional answer category, differentiating between “1 to
in the presence of (time-varying) confounders that are also on <2 hours” and “2 to <3 hours.” To calculate an overall score
the causal pathway from exposure to outcome.24–27 for screen media exposure, the answer categories from the 2
variables were first recoded to 0, 0.5, 2 (1.5 and 2.5 for age
METHODS 11), 4, 6, and 8 hours/day, and subsequently summed. Because
of skewness and outliers, the resulting screen media exposure
Data variables (hours/day at age 7 and hours/day at age 11) were
The Millennium Cohort Study (MCS) is a nationally recoded into 4 categories: (1) <1 hour, (2) 1 to <3 hours, (3) 3
representative, prospective cohort study of UK children born to <5 hours, and (4) ≥5 hours.
between September 2000 and January 2002.28 A stratified Trained interviewers took anthropometric measures.
clustered sampling design was used to adequately represent Body mass index (BMI) was calculated from weight—mea-
children from disadvantaged areas, ethnic minority groups, sured using Tanita BF-522W (Tanita Corporation, Tokyo,
and children living in Wales, Scotland, and Northern Ireland. Japan) scales and recorded to the nearest 0.1 kg—and height—
Families were invited to participate when eligible cohort chil- measured using a Leicester height measure stadiometer. The
dren were 9 months old (72% response). Interviews were car- primary outcome measure was obesity defined by the Interna-
ried out in the home with the main respondent (over 99% tional Obesity Task Force (IOTF) age- and sex-specific cut-
were biologic mothers, hereafter referred to as mothers) and, offs for BMI.38 We identified potential confounders a priori
if applicable, the partner. Information was collected on var- from existing literature. Potential time-fixed confounders in-
ious topics relating to the child and their family. Additional cluded sociodemographic characteristics at baseline: sex, age,
data were collected when cohort members were 3, 5, 7, 11, ethnicity (white, Indian/Pakistani/Bangladeshi, black or black
and 14 years old from parent(s), siblings, teachers, and cohort British, other), country, mother’s age at birth, and mother’s re-
members. Parents were given the opportunity to opt out, and ligion (none, Christian, Muslim, other). In addition, mother’s
consent was sought and obtained at each contact. The MCS cognitive ability was included as a time-fixed confounder.
received ethical approval from the South West and London This was only measured at child’s age 14 and included in
Multi-Centre Research Ethics Committees of the National the analysis assuming that this measure is a valid indicator
Health Service. This study was restricted to singletons (n = for cognitive ability in previous waves.39 Mother’s cognitive
11,564) who participated in the latest wave (age 14; 61% re- ability was measured with a word activity assessment (range:
sponse), but not necessarily in all previous waves, using data 0–20) derived from a shortened version of the Applied Psy-
from all 6 waves.29–34 Data were obtained from the UK Data chology Unit Vocabulary Test.40 Potential time-varying con-
Archive, University of Essex. founders (all measured at ages 7 and 11) included equivalized

© 2020 Wolters Kluwer Health, Inc. All rights reserved. www.epidem.com | 579


Oude Groeniger et al. Epidemiology  •  Volume 31, Number 4, July 2020

household income (log transformed), managing financially requires only one exchangeability assumption: no unmeasured
(alright, getting by, difficult), housing tenure (own, public mediator–outcome confounding.44,47 To fulfill this assumption,
renting, private renting, other), area deprivation (in deciles), we adjusted for a comprehensive set of potential confounders
maternal BMI (<18.5 kg/m2, 18.5 to <25 kg/m2, 25 to <30 kg/ of the relationship between screen media exposure and obesity
m2, ≥30 kg/m2), maternal psychological distress (a score of (described previously).
≥13 on the Kessler-6 scale),41 child attends club outside of Counterfactual disparity measures (similar to CDEs)
school (no, yes), number of parents/carers (1, 2), natural fa- can be estimated for each level of the mediator. In the pres-
ther in household (no, yes), number of siblings (none, 1, 2, ence of an interaction effect between exposure and mediator
≥3), parent(s) not in work (no, yes), not enough time to spend on the outcome, these separate counterfactual disparity meas-
with child (no, yes), child illness that limits activity (no, yes), ures may differ depending on the magnitude of the interac-
child BMI (normal weight, overweight, or obesity), and ma- tion effect. However, in the absence of an interaction effect, all
ternal fair/poor self-rated health (no, yes). counterfactual disparity measures will be equal. We examined
the presence of interaction effects by including cross-product
Statistical Analysis terms between mother’s education and screen media exposure
First, we calculated descriptive statistics of the par- (eAppendix 2; http://links.lww.com/EDE/B673), but due to a
ticipants stratified by mother’s educational level to describe lack of precision in the estimated models, we were unable to
group differences in the prevalence of the outcome and me- observe interaction on either the risk ratio or risk difference
diator.42 Second, we fitted generalized linear models on both scale (although at least one must be present if both exposure
the risk ratio and the risk difference scale (described below).26 and mediator have an effect on the outcome).48,49 Because test-
We used multiple imputation by chained equations to impute ing the null hypotheses of no interaction resulted in P values
missing data (M = 20). eAppendix 1 (http://links.lww.com/ of >0.9 on both the risk ratio and the risk difference scale,
EDE/B673) lists the percentage of missings (ranging from 0% including the interaction terms would likely limit precision in
for age, sex, and country to 13% for maternal psychological our models even more and hinder inference. We, therefore,
distress). We used survey weights (age 14, whole UK analy- decided to omit the cross-product terms from the final analysis
ses) provided by the Millennium Cohort Study to correct for and estimated only one counterfactual disparity measure (sim-
sampling design and attrition.43 We conducted analyses using ilar to, e.g., Nandi et al50). As a result, our analysis assumes that
Stata 15 (StataCorp, College Station, TX). intervening to eliminate differences in screen media exposure
To assess to what extent social inequalities in childhood between children from different socioeconomic backgrounds
obesity could be reduced by intervening on screen media ex- has the same effect regardless of the amount of screen media
posure, we estimated a counterfactual disparity measure.44,45 exposure that is imposed by the hypothetical intervention.
The counterfactual disparity measure in this study comparing Subsequently, we calculated a “percentage reduction”
exposure level a* to level a is defined on the risk difference by dividing the difference between the total disparity (TD) in
scale in equation 1 and on the risk ratio scale in equation 2: childhood obesity and the counterfactual disparity measure
(CDM) by the total disparity (i.e., [TD − CDM]/[TD − 1] on
E[Y (m(t )) | A = a ] − E[Y (m(t )) | A = a*] (1)
the risk ratio scale and [TD − CDM]/TD on the risk difference
E[Y (m(t )) | A = a ] / E[Y (m(t )) | A = a*] (2) scale).51 This percentage reduction indicates how much the
 disparity in childhood obesity would be reduced if differences
where m(t ) denotes the mediator trajectory (i.e., screen in screen media exposure were eliminated. We bootstrapped
media exposure at ages 7 and 11). This measure can be inter- the percentage reduction parameter (1,000 repetitions) to ob-
preted as the magnitude of the association between mother’s ed- tain 95% bias-corrected confidence intervals (CIs).52
ucation and childhood obesity that would remain if a particular To estimate the counterfactual disparity measure, we
trajectory of screen media exposure was fixed at a specific value fitted a marginal structural model (MSM) using inverse prob-
uniformly in the population. A main advantage of the counterfac- ability of treatment weighting.24,25,27 This method uses weight-
tual disparity measure is that it can still be identified even if there ing to adjust for (time-varying) confounding, which bypasses
are confounders of the mediator–outcome relationship that are the need to condition on confounders in the outcome model as
also on the causal pathway from exposure to outcome.24,26,27,46 is traditionally done in mediation analysis. To do so, we first
Because the effect of screen media exposure on obesity may be calculated stabilized inverse probability weights (IPWs) of the
confounded by factors that are itself affected by mother’s educa- probability that each participant received the level of screen
tion (e.g., income, neighborhood deprivation), we estimated the media exposure that he/she actually received, given exposure,
counterfactual disparity measure to adjust for these factors. In mediator, and confounder history. For each individual i in the
this regard, the counterfactual disparity measure is similar to the sample, the mediator weight at time t is calculated by:
more widely known controlled direct effect (CDE). However, P[ M (t) = mi (t)|ai , mi (t −1)]
whereas a CDE also assumes no unmeasured exposure–outcome wM i (t)=
confounding, identification of a counterfactual disparity measure P[M (t) = mi (t)|ai , mi (t −1), li (t −1), vi ]

580  |  www.epidem.com © 2020 Wolters Kluwer Health, Inc. All rights reserved.
Epidemiology  •  Volume 31, Number 4, July 2020 Screen Media Exposure and Inequalities in Childhood Obesity

where ai, mi(t), li(t), and vi are the actual values of deprivation), but not via screen media exposure, this effect is
the exposure, the mediator, the time-varying confounders, still captured in the estimated disparity measure. However, to
and the baseline confounders, respectively, for individual the extent that the effect of screen media exposure on obesity
i.26 Second, we fitted generalized linear regression models is confounded by the measured covariates, this confounding is
with robust standard errors as shown in equations 3 and 4, removed by applying the weights.
weighted by the product of the inverse probability and survey Several effective interventions to reduce screen media ex-
weights53: posure among children exist, with some replacing screen time
with other activities (e.g., sports or extracurricular activities)
E[Y | A = a,M (t )=m(t )]=γ 0 +γ 1a +γ 2 m(t =age7)+γ 3m(t =age11)
 (3) and others targeted at decreasing screen time without encour-
aging replacement activities (e.g., by educational programs or
Log( P[Y =1 | A = a, M (t )=m(t )])=θ 0 +θ 1a +θ 2 m(t =age7)+θ 3m(t =age111)
(4)
 automatic time locks).54 The hypothetical intervention consid-
The parameters of this weighted regression give valid ered in our study is best envisioned by putting an automatic
estimates of the counterfactual disparity measure (assum- time lock on the television and computer, limiting screen time
ing no model misspecification, selection bias, or measure- uniformly for all children. As previously discussed, by omitting
ment error).25,27 eAppendix 3 (http://links.lww.com/EDE/ interaction terms, we assume that this hypothetical intervention
B673) provides more information and annotated Stata code. will have the same effect on social inequalities in childhood
By applying weights in the final regression model, a pseudo- obesity regardless of the amount of screen time set by the au-
population is created where the distribution of measured con- tomatic lock. Furthermore, it is important to note that by not
founders is unrelated to the effect of interest (as illustrated in specifying replacement activities, our models assume that these
the causal diagram in the Figure). In other words, to the extent activities do not differ between children from different socioec-
that mother’s education is related to obesity of the child via onomic backgrounds (at least with regard to their effect on obe-
the (time-varying) confounders (e.g., income, neighborhood sity).55,56 If, for example, children from more-educated mothers

FIGURE.  Causal diagrams (directed


acyclic graphs) of the proposed medi-
ation analyses before and after apply-
ing the inverse probability weights. A,
Causal diagram of the proposed
mediation analysis: A = mother’s ed-
ucation (in this diagram shown as
if it were effectively randomized),
Mt = screen media exposure, V = time-
fixed (baseline) confounders,
Lt = time-varying confounders,
Y = childhood obesity, U = unmeas-
ured confounders. Lt is on the causal
pathway A→Y, but also a confounder
in the relationship Mt→Y, which pro-
hibits conventional adjustment for
Lt. B, Causal diagram of the scenario
encountered after applying the in-
verse probability weights: A = pa-
rental education, Mt = screen media
exposure (depicted with a box to in-
dicate that they are conditioned on in
the regression model), V = time-fixed
(baseline) confounders, Lt = time-
varying confounders, Y = childhood
obesity, U = unmeasured confound-
ers. The dashed lines depict the
counterfactual disparity measure. By
applying the weights, several arrows
are “erased” (i.e., the effect of V and
Lt on Mt) and it becomes possible to
estimate the combined magnitude of
the dashed lines.

© 2020 Wolters Kluwer Health, Inc. All rights reserved. www.epidem.com | 581


Oude Groeniger et al. Epidemiology  •  Volume 31, Number 4, July 2020

TABLE 1.  Descriptive Statistics of the Millennium Cohort Study Participants Stratified by Mother’s Educational Level (Percentages)
Mother’s Educational Levela
University Education to Education to No Qualifications
(n = 4,050) Age 18 (n = 1,554) Age 16 (n = 3,571) (n = 1,223)
Female, %b 48 46 47 49
Ethnicity, %b
 White 87 88 91 69
 Indian/Pakistani/Bangladeshi 4 5 4 16
 Black or British black 4 2 2 8
 Other 5 5 3 7
Country, %b
 England 82 78 84 84
 Wales 5 6 5 4
 Scotland 9 12 7 7
 Northern Ireland 4 5 4 5
Mother’s religion, %b
 None 37 47 55 53
 Christian 57 46 40 24
 Muslim 3 4 4 19
 Other 3 3 2 4
Mother’s age at birth,b mean (SD) 31 (5.5) 27 (5.9) 27 (5.7) 26 (5.7)
Mother’s cognitive ability,c mean (SD) 13 (4.5) 11 (3.9) 9.7 (3.2) 7.1 (2.9)
Area deprivation decile,a mean (SD) 6.7 (3.1) 5.4 (2.8) 4.7 (2.6) 3.1 (2.1)
Household equivalized income,a mean (SD) 490 (272) 348 (193) 285 (158) 196 (101)
Managing financially, %a
 Alright 73 60 55 40
 Getting by 20 30 33 41
 Difficult 7 9 12 19
Housing tenure, %a
 Own 83 67 49 26
 Public renting 8 21 35 58
 Private renting 6 8 12 13
 Other 3 5 4 3
Maternal BMI (kg/m2), %a
 18.5 to <25 53 44 43 35
 <18.5 16 19 21 29
 25 to <30 20 23 21 20
 ≥30 10 14 14 16
Maternal psychological distress, %a 1 3 5 8
Child attends club outside of school, %a 16 11 8 5
One parent/carer, %a 12 19 28 34
Natural father not in household, %a 14 25 34 41
No. siblings, %a
 None 15 20 17 13
 1 54 50 46 32
 2 24 21 24 25
 ≥3 7 9 13 31
Parent(s) in work, %a 73 62 50 21
Not enough time to spend with child, %a 38 32 28 15
Child illness that limits activity, %a 5 6 6 8
Maternal fair/poor self-rated health, %a 2 4 5 7
Screen media exposure (hours) (per day; age 7), %
 <1 23 15 12 10
 1 to <3 48 47 44 42
 3 to <5 21 27 31 29
 ≥5 8 12 13 18
Screen media exposure (hour) (per day; age 11), %
 <1 16 9 7 7
 1 to <3 59 58 54 50
 3 to <5 16 19 21 22
 ≥5 10 13 18 20
Obesity (age 14), %c,d 5 6 10 10
Descriptive statistics calculated on nonimputed data weighted by the survey weights. Descriptive statistics of the confounders only shown for the earliest measurement.
a
Derived at age 5.
b
Derived at baseline.
c
Derived at age 14.
d
Defined by the International Obesity Task Force age- and sex-specific cut-offs for BMI.

582  |  www.epidem.com © 2020 Wolters Kluwer Health, Inc. All rights reserved.
Epidemiology  •  Volume 31, Number 4, July 2020 Screen Media Exposure and Inequalities in Childhood Obesity

spend this time on physical activity, while children from less- Children from mothers with education to age 18 were 1.3 (1.0,
educated mothers do not, this assumption would be violated and 1.7) times as likely to be obese and had a 1.6 (−0.4, 3.6) per-
the estimated counterfactual disparity measure may be biased. centage-point higher risk of obesity. Because of the relatively
We conducted several sensitivity analyses to investigate small inequality in obesity between children from mothers
the robustness of the results (eAppendix 4; http://links.lww. with education to age 18 and mothers with university quali-
com/EDE/B673). First, analyses were repeated using the UK fications, we refrain from making inferences for this contrast.
1990 growth reference (UK90) BMI cut-offs.57 Whereas the Results from the inverse probability-weighted regres-
IOTF cut-off defines obesity as an age- and sex-specific cut- sion model showed that longer exposure to screen media is
off extrapolated from the adult BMI cut-off of 30 kg/m2, the associated with a higher risk of childhood obesity (Table 3).
UK90 cut-off defines obesity as those at or above the 95th Five hours or more of screen media per day at age 11 was as-
percentile based on age- and sex-specific reference charts. sociated with 1.7-fold (1.0, 2.8) increased risk of obesity or
Using the UK90 cut-off, the prevalence of childhood obesity 3.9 (0.4, 7.4) percentage-points, compared with <1 hour/day.
ranges from 15% for children from mothers with a university Compared with mothers with university qualifications,
degree to 25% for children from mothers with no qualifica- the estimated counterfactual disparity in obesity at age 14, if
tions. Second, we repeated analyses using the highest attained educational differences in screen media exposure at ages 7 and
educational level in the household (either from the mother 11 were eliminated, was 1.8 (1.4, 2.2) for mothers with educa-
or partner) and household income quartiles as the exposure tion to age 16 and 1.8 (1.4, 2.4) for mothers with no qualifica-
(while controlling for education). Third, we repeated analy- tions on the risk ratio scale. On the risk difference scale, the
ses using only television viewing and using only leisure-time same comparison was 4.3 (2.5, 6.1) for mothers with educa-
computer use as a mediator, instead of a combined measure tion to age 16 and 4.6 (2.0, 7.2) for mothers with no qualifica-
(while including the other measure as a confounder). Fourth, tions (Table 3). This corresponds to an estimated reduction in
we repeated analyses without using imputed data for exposure relative inequalities in childhood obesity of 13% (1%, 26%)
and outcome (n = 9,749). for mothers with education to age 16 and 17% (1%, 33%) for
those with no qualifications (Table 4), and an estimated reduc-
RESULTS tion in absolute inequalities of 15% (2%, 28%) for mothers
Among the children’s mothers included in the study, with education to age 16 and 18% (−1%, 37%) for those with
39% had a university degree, 15% had education to age 18, no qualifications (Table 5).
34% had education to age 16, and 12% had no educational Sensitivity analyses (eAppendix 4; http://links.lww.com/
qualifications. Table 1 shows that 8% of 7-year-old children EDE/B673) showed (again respectively contrasting children
and 10% of 11-year-old children from mothers with a uni- from mothers with education to age 16 and no qualifications
versity degree were exposed to ≥5 hours of screen media per against children from mothers with university qualifications)
weekday. This percentage increased steadily among children
from mothers with a lower educational level to 18% of 7-year-
TABLE 3.  Results From the Inverse Probability-weighted
old children and 20% of 11-year-old children from mothers Regression Model Regressing Obesity on Mother’s
with no educational qualifications. Educational Level and Screen Media Exposure
Children from mothers with no educational qualifica-
RR (95% CI) RD (95% CI)
tions were 2.0 (confidence interval = 1.5, 2.5) times as likely
to be obese and had a 5.6 (3.1, 8.1) percentage-point higher Mother’s educational level
risk of obesity than children from mothers with a univer-  University 1 0
sity degree (Table 2). Children from mothers with education  Education to age 18 1.2 (0.9, 1.7) 1.1 (−1.1, 3.3)
to age 16 were 1.9 (1.5, 2.3) times as likely to be obese and  Education to age 16 1.8 (1.4, 2.2) 4.3 (2.5, 6.1)
had a 5.1 (3.4, 6.7) percentage-point higher risk of obesity.  No qualifications 1.8 (1.4, 2.4) 4.6 (2.0, 7.2)
Screen media exposure per day (hour) (age 7)
 <1 1 0
TABLE 2.  Total Disparity in Childhood Obesity  1 to <3 1.3 (0.9, 1.9) 1.3 (−1.1, 3.7)
 3 to <5 1.3 (0.9, 1.9) 1.5 (−1.1, 4.0)
RR (95% CI) RD 95% CI)
 ≥5 1.2 (0.7, 1.8) 1.0 (−2.1, 4.1)
Mother’s educational level Screen media exposure per day (hour) (age 11)
 University 1 0  <1 1 0
 Education to age 18 1.3 (1.0, 1.7) 1.6 (−0.4, 3.6)  1 to <3 1.3 (0.8, 2.1) 1.8 (−0.8, 4.4)
 Education to age 16 1.9 (1.5, 2.3) 5.1 (3.4, 6.7)  3 to <5 1.6 (1.0, 2.7) 3.4 (0.1, 6.7)
 No qualifications 2.0 (1.5, 2.5) 5.6 (3.1, 8.1)  ≥5 1.7 (1.0, 2.8) 3.9 (0.4, 7.4)
RD indicates risk difference (in percentage-points); RR, risk ratio. RD indicates risk difference (in percentage-points); RR, risk ratio.

© 2020 Wolters Kluwer Health, Inc. All rights reserved. www.epidem.com | 583


Oude Groeniger et al. Epidemiology  •  Volume 31, Number 4, July 2020

TABLE 4.  Reduction in Relative Inequalities in Childhood Obesity if Educational Differences in Screen Media Exposure Were
Eliminated
Total Disparity Counterfactual Disparity Percentage Attenuated

RR (95% CI) RR (95% CI) Estimate (95% CI)

Mother’s educational level


 University 1 1
 Education to age 18 1.3 (1.0, 1.7) 1.2 (0.9, 1.7) 29% (−18%, 174%)
 Education to age 16 1.9 (1.5, 2.3) 1.8 (1.4, 2.2) 13% (1%, 26%)
 No qualifications 2.0 (1.5, 2.5) 1.8 (1.4, 2.4) 17% (1%, 33%)
RR indicates risk ratio.

TABLE 5.  Reduction in Absolute Inequalities in Childhood Obesity if Educational Differences in Screen Media Exposure Were
Eliminated
Total Disparity Counterfactual Disparity Percentage Attenuated

RD (95% CI) RD (95% CI) Estimate (95% CI)

Mother’s educational level


 University 0 0
 Education to age 18 1.6 (−0.4, 3.6) 1.1 (−1.1, 3.3) 30% (−27%, 206%)
 Education to age 16 5.1 (3.4, 6.7) 4.3 (2.5, 6.1) 15% (2%, 28%)
 No qualifications 5.6 (3.1, 8.1) 4.6 (2.0, 7.2) 18% (−1%, 37%)
RD indicates risk difference (in percentage-points).

that using the UK90 obesity cut-offs resulted an estimated mothers (8% at age 7 and 10% at age 11). We estimated that
reduction in social inequalities in childhood obesity of 11% up to 17% of relative and 18% of absolute inequalities in
and 11% for relative inequalities and 9% and 9% for abso- childhood obesity would be reduced if differences in screen
lute inequalities. Using highest parental educational level, the media exposure were eliminated.
estimated reduction was 8% and 9% for relative inequalities This study has several limitations. First, even though the
and 10% and 11% for absolute inequalities. Using house- sample consisted of 11,413 UK children, our estimates have
hold income quartiles, the estimated reduction was 19% and limited precision. Most notably, this obstructed inclusion of
17% for relative inequalities and 19% and 16% for absolute interaction terms in the models. Second, although obesity was
inequalities. Both television viewing and leisure-time com- derived from anthropometric measures, covariates were self-
puter use contributed independently to the estimated reduction reported, which risks higher measurement error. Third, although
in inequalities in childhood obesity, although including only great care was given to adjust for potential confounding, the ob-
television viewing as a mediator resulted in a slightly higher servational nature of the data implies that there is no guarantee
estimated reduction in inequalities in obesity (15% and 17% that we were able to fulfill the exchangeability assumption. Spe-
reduction in relative inequalities and 11% and 12% reduction cifically, the assumption of no unmeasured mediator–outcome
in absolute inequalities) than including only leisure-time com- confounding implies that we have to presume that the risk of
puter use (16% and 13% for relative inequalities and 9% and obesity among children who were exposed to, for example, ≥5
6% for absolute inequalities). Last, not imputing exposure and hours of screen media per day would be comparable—given
outcome resulted in an estimated reduction in inequalities in the measured confounders—to the risk of children who were
obesity of 16% and 21% for relative inequalities and 16% and exposed to <1 hour of screen media, if, counter to the fact, they
22% for absolute inequalities, respectively. were exposed to <1 hour of screen media per day themselves
(and vice versa). In other words, we assume that if we were able
DISCUSSION to reduce screen media exposure, these children would replace
Children of the least-educated mothers were almost watching television or playing on computers with (healthier)
twice as likely to be obese at age 14 than children of the most- activities comparable to those of the children in our cohort who
educated mothers. Similarly, children of the least-educated have less screen media exposure (instead of substituting screen
mothers had greater levels of screen media exposure (19% at media exposure for an activity with a similar or even higher
age 7 and 20% at age 11) than children of the most-educated risk of obesity). Violation of this assumption is perhaps most

584  |  www.epidem.com © 2020 Wolters Kluwer Health, Inc. All rights reserved.
Epidemiology  •  Volume 31, Number 4, July 2020 Screen Media Exposure and Inequalities in Childhood Obesity

likely for factors that affect screen media exposure and other 2. N. C. D. Risk Factor Collaboration. Worldwide trends in body-mass in-
dex, underweight, overweight, and obesity from 1975 to 2016: a pooled
lifestyle-related factors that may lead to a higher probability analysis of 2416 population-based measurement studies in 128.9 million
of obesity among children, such as habits and preferences re- children, adolescents, and adults. Lancet. 2017;390:2627–2642.
lated to food consumption and physical activity. In an attempt 3. Commission on Ending Childhood Obesity. Report of the Commission
on Ending Childhood Obesity. Geneva, Switzerland: World Health
to block these pathways, we adjusted for mother’s BMI because Organization; 2016.
the same factors would likely also lead to a higher BMI of the 4. Bann D, Johnson W, Li L, Kuh D, Hardy R. Socioeconomic inequalities
mothers. However, mother’s BMI may not fully account for in childhood and adolescent body-mass index, weight, and height from
1953 to 2015: an analysis of four longitudinal, observational, British birth
these confounding effects. Fourth, in addition to differences in cohort studies. Lancet Public Health. 2018;3:e194–e203.
the quantity of screen media exposure, children from different 5. Marmot M, Allen J, Goldblatt P, et al. The Marmot Review: Fair Society,
socioeconomic backgrounds may be exposed to different screen Healthy Lives. The Strategic Review of Health Inequalities in England
Post-2010. London: University College London; 2010.
media content (e.g., children from more-educated mothers may 6. Pearce A, Dundas R, Whitehead M, Taylor-Robinson D. Pathways to in-
more often consume media with less exposure to food adver- equalities in child health. Arch Dis Child. 2019;104:998–1003.
tisements). Because we had no data on screen media content, 7. Law C, Parkin C, Lewis H. Policies to tackle inequalities in child health:
why haven’t they worked (better)? Arch Dis Child. 2012;97:301–303.
we could not adjust for these differences. Fifth, whereas exten- 8. Epstein LH, Roemmich JN, Robinson JL, et al. A randomized trial of the
sive data were collected from mothers, less data were available effects of reducing television viewing and computer use on body mass
from their partners. Moreover, because a substantial number index in young children. Arch Pediatr Adolesc Med. 2008;162:239–245.
9. Robinson TN, Banda JA, Hale L, et al. Screen media exposure and obesity
of mothers had no partner, partner information could not be in children and adolescents. Pediatrics. 2017;140(suppl 2):S97–S101.
included in the models. To the extent that the level of screen 10. Zhang G, Wu L, Zhou L, Lu W, Mao C. Television watching and risk of
media exposure of children and their risk of obesity was af- childhood obesity: a meta-analysis. Eur J Public Health. 2016;26:13–18.
11. Rideout VJ, Foehr UG, Roberts DF. Generation M 2: Media in the Lives of
fected by partner’s factors independently of maternal factors, 8-to 18-Year-Olds. Menlo Park, CA: Henry J. Kaiser Family Foundation;
this may have affected our results. Sixth, mother’s cognitive 2010.
ability was measured as knowledge of vocabulary, which may 12. Jordan AB, Robinson TN. Children, television viewing, and weight sta-
tus: summary and recommendations from an expert panel meeting. Ann
not reflect the full spectrum of cognitive abilities. Am Acad Pol Soc Sci. 2008;615:119–132.
Previous studies have found indications that televi- 13. Tandon PS, Zhou C, Sallis JF, Cain KL, Frank LD, Saelens BE. Home
sion viewing tracks from childhood to adulthood, suggest- environment relationships with children’s physical activity, sedentary
time, and screen time by socioeconomic status. Int J Behav Nutr Phys
ing that intervening on screen media exposure in childhood Act. 2012;9:88.
may also affect screen media exposure in later life.58 More- 14. De Craemer M, Verloigne M, Ghekiere A, et al. Changes in children’s
over, because childhood obesity is a strong predictor of adult television and computer time according to parental education, parental
income and ethnicity: a 6-year longitudinal EYHS study. PLoS One.
obesity and other adverse health outcomes, intervening on 2018;13:e0203592.
the causes of childhood obesity will positively affect health 15. Ellaway A, Macintyre S, Kearns A. Perceptions of place and health in so-
chances throughout the life course. Future research that exam- cially contrasting neighbourhoods. Urban Stud. 2001;38:2299–2316.
16. Bandura A, Walters RH. Social Learning and Personality Development.
ines how screen media exposure can be effectively reduced in New York, NY: Holt, Rinehart, & Winston; 1963.
socioeconomically disadvantaged families is, therefore, war- 17. Bourdieu P. Distinction: A Social Critique of the Judgement of Taste.
ranted. Furthermore, given the fact that new forms of screen Cambridge, MA: Harvard University Press; 1984.
18. Singh-Manoux A, Marmot M. Role of socialization in explaining social
media emerge rapidly (e.g., smartphones, tablets, virtual re- inequalities in health. Soc Sci Med. 2005;60:2129–2133.
ality media) and are increasingly used by (young) children, 19. Stamatakis E, Hillsdon M, Mishra G, Hamer M, Marmot M. Television
screen media exposure may become an increasingly relevant viewing and other screen-based entertainment in relation to multiple so-
cioeconomic status indicators and area deprivation: the Scottish Health
determinant of childhood obesity in the next decade. To pre- Survey 2003. J Epidemiol Community Health. 2009;63:734–740.
vent a further surge of (social inequalities in) obesity, we need 20. Berry B. Disparities in free time inactivity in the United States: trends and
to carefully monitor how these technologic innovations affect explanations. Sociol Perspect. 2007;50:177–208.
21. Currid-Halkett E. The Sum of Small Things: A Theory of the Aspirational
our youth’s health across different social groups. Class. Princeton, NJ: Princeton University Press; 2017.
22. Lareau A. Unequal Childhoods: Class, Race, and Family Life. Berkeley,
‍ACKNOWLEDGMENTS CA: Univ of California Press; 2011.
23. Notten N, Lancee B, van de Werfhorst HG, Ganzeboom HB. Educational
We are grateful to the Centre for Longitudinal Studies stratification in cultural participation: cognitive competence or status mo-
(CLS), University College London Institute of Education, for tivation? J Cult Econ. 2015;39:177–203.
the use of these data and to the UK Data Service for making 24. Nandi A, Glymour MM, Kawachi I, VanderWeele TJ. Using marginal
structural models to estimate the direct effect of adverse childhood social
them available. However, neither CLS nor the UK Data Service conditions on onset of heart disease, diabetes, and stroke. Epidemiology.
bear any responsibility for the analysis or interpretation of 2012;23:223–232.
these data. Thanks to the anonymous reviewers and editor for 25. Robins JM, Hernán MA, Brumback B. Marginal structural models and
causal inference in epidemiology. Epidemiology. 2000;11:550–560.
their suggestions, many of which were very helpful. 26. VanderWeele T. Explanation in Causal Inference: Methods for Mediation
and Interaction. New York, NY: Oxford University Press; 2015.
REFERENCES 27. VanderWeele TJ. Marginal structural models for the estimation of direct
and indirect effects. Epidemiology. 2009;20:18–26.
1. GBD 2015 Obesity Collaborators. Health effects of overweight and obe-
28. Connelly R, Platt L. Cohort profile: UK Millennium Cohort Study

sity in 195 countries over 25 years. N Engl J Med. 2017;377:13–27.
(MCS). Int J Epidemiol. 2014;43:1719–1725.

© 2020 Wolters Kluwer Health, Inc. All rights reserved. www.epidem.com | 585


Oude Groeniger et al. Epidemiology  •  Volume 31, Number 4, July 2020

29. University of London, Institute of Education, Centre for Longitudinal 43. University of London, Institute of Education, Centre for Longitudinal
Studies. Millennium Cohort Study: First Survey, 2001-2003. [data Studies. Millennium Cohort Study: Longitudinal Family File, 2001-2015.
collection]. 12th ed. UK Data Service, 2017. Available at: http://doi. [data collection]. 2nd ed. UK Data Service, 2017. Available at: http://doi.
org/10.5255/UKDA-SN-4683-4. Accessed 13 May 2020. org/10.5255/UKDA-SN-8172-2. Accessed 13 May 2020.
30. University of London, Institute of Education, Centre for Longitudinal 44. Naimi AI, Schnitzer ME, Moodie EE, Bodnar LM. Mediation analysis for
Studies. Millennium Cohort Study: Second Survey, 2003-2005. [data health disparities research. Am J Epidemiol. 2016;184:315–324.
collection]. 9th ed. UK Data Service, 2017. Available at: http://doi. 45. Lin HS, Naimi AI, Brooks MM, Richardson GA, Burke JG, Bromberger
org/10.5255/UKDA-SN-5350-4. Accessed 13 May 2020. JT. Child maltreatment as a social determinant of midlife health-related
31. University of London, Institute of Education, Centre for Longitudinal quality of life in women: do psychosocial factors explain this association?
Studies. Millennium Cohort Study: Third Survey, 2006. [data collec- Qual Life Res. 2018;27:3243–3254.
tion]. 7th ed. UK Data Service, 2017. Available at: http://doi.org/10.5255/ 46. Valeri L, Vanderweele TJ. Mediation analysis allowing for exposure-
UKDA-SN-5795-4. Accessed 13 May 2020. mediator interactions and causal interpretation: theoretical assumptions
32. University of London, Institute of Education, Centre for Longitudinal and implementation with SAS and SPSS macros. Psychol Methods.
Studies. Millennium Cohort Study: Fourth Survey, 2018. [data collec- 2013;18:137–150.
tion]. 7th ed. UK Data Service, 2017. Available at: http://doi.org/10.5255/ 47. VanderWeele TJ, Robinson WR. On the causal interpretation of race
UKDA-SN-6411-7. Accessed 13 May 2020. in regressions adjusting for confounding and mediating variables.
33. University of London, Institute of Education, Centre for Longitudinal Epidemiology. 2014;25:473–484.
Studies. Millennium Cohort Study: Fifth Survey, 2012. [data collec- 48. Greenland S, Lash TL, Rothman KJ. Concepts of interaction. In:

tion]. 4th ed. UK Data Service, 2017. Available at: http://doi.org/10.5255/ Rothman KJ, Greenland S, Lash TL, eds. Modern Epidemiology. Vol. 3.
UKDA-SN-7464-4. Accessed 13 May 2020. Philadelphia, PA: Lippincott Williams & Wilkins; 2008.
34. University of London, Institute of Education, Centre for Longitudinal 49. VanderWeele TJ, Knol MJ. A tutorial on interaction. Epidemiologic

Studies. Millennium Cohort Study: Sixth Survey, 2015. [data collection]. Methods. 2014;3:33–72.
5th ed. UK Data Service, 2020. Available at: http://doi.org/10.5255/ 50. Nandi A, Glymour MM, Subramanian SV. Association among socioec-
UKDA-SN-8156-5. Accessed 13 May 2020. onomic status, health behaviors, and all-cause mortality in the United
35. Massion S, Wickham S, Pearce A, Barr B, Law C, Taylor-Robinson D. States. Epidemiology. 2014;25:170–177.
Exploring the impact of early life factors on inequalities in risk of over- 51.
VanderWeele TJ. Policy-relevant proportions for direct effects.
weight in UK children: findings from the UK Millennium Cohort Study. Epidemiology. 2013;24:175–176.
Arch Dis Child. 2016;101:724–730. 52. Mackinnon DP, Lockwood CM, Williams J. Confidence limits for the
36. Rougeaux E, Hope S, Law C, Pearce A. Have health inequalities changed indirect effect: distribution of the product and resampling methods.
during childhood in the new labour generation? Findings from the UK Multivariate Behav Res. 2004;39:99.
Millennium Cohort Study. BMJ Open. 2017;7:e012868. 53. Brumback BA, Bouldin ED, Zheng HW, Cannell MB, Andresen EM.
37. Galobardes B, Shaw M, Lawlor DA, Lynch JW, Davey Smith G.
Testing and estimating model-adjusted effect-measure modification using
Indicators of socioeconomic position (part 1). J Epidemiol Community marginal structural models and complex survey data. Am J Epidemiol.
Health. 2006;60:7–12. 2010;172:1085–1091.
38. Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard def- 54. Schmidt ME, Haines J, O’Brien A, et al. Systematic review of effective
inition for child overweight and obesity worldwide: international survey. strategies for reducing screen time among young children. Obesity (Silver
BMJ. 2000;320:1240–1243. Spring). 2012;20:1338–1354.
39. Hertzog C, Schaie KW. Stability and change in adult intelligence: 1. Analysis 55. VanderWeele TJ, Vansteelandt S. Conceptual issues concerning media-
of longitudinal covariance structures. Psychol Aging. 1986;1:159–171. tion, interventions and composition. Stat Interface. 2009;2:457–468.
40. Closs S, Hutchings M. APU Arithmetic Test. London: Hodder and
56. Cole SR, Frangakis CE. The consistency statement in causal inference: a
Stoughton; 1976. definition or an assumption? Epidemiology. 2009;20:3–5.
41. Kessler RC, Barker PR, Colpe LJ, et al. Screening for serious mental ill- 57. Cole TJ, Freeman JV, Preece MA. Body mass index reference curves for
ness in the general population. Arch Gen Psychiatry. 2003;60:184–189. the UK, 1990. Arch Dis Child. 1995;73:25–29.
42. Ward JB, Gartner DR, Keyes KM, Fliss MD, McClure ES, Robinson WR. 58. Smith L, Gardner B, Hamer M. Childhood correlates of adult TV viewing
How do we assess a racial disparity in health? Distribution, interaction, and time: a 32-year follow-up of the 1970 British Cohort Study. J Epidemiol
interpretation in epidemiological studies. Ann Epidemiol. 2019;29:1–7. Community Health. 2015;69:309–313.

586  |  www.epidem.com © 2020 Wolters Kluwer Health, Inc. All rights reserved.

You might also like