Mealtime TV Use Is Associated with Higher Discretionary Food Intakes in Young Australian Children: A Two-Year Prospective Study

Background: Mealtime television use has been cross-sectionally associated with suboptimal diets in children. This study aimed to assess the two-year prospective association between baseline mealtime television use and subsequent diets in young children, and identify socioeconomic differences. Methods: Parents reported their child’s television use at meals, and fruit, vegetable, and discretionary food intakes. Multivariable linear and logistic regression analyses assessed the association between baseline mealtime television use and follow-up diet outcomes. Differences were assessed by socioeconomic position. Results: Participants were 352 Australian parents of children aged six months to six years. Daily mealtime television use (average frequency/day) was associated with higher daily frequency of discretionary food intakes (β 0.2, 95% confidence interval (CI) 0.07–0.67) at the 2-year follow-up. Individually, television use during breakfast and dinner (1–2 days/week compared to never) predicted higher daily intake frequency of discretionary food, β 0.36 (95% CI 0.12–0.60) and β 0.19 (95% CI 0.00–0.39), respectively. Similarly, 3–7 days/week of television use during breakfast and lunch predicted higher frequency of discretionary food intake, β 0.18 (95% CI 0.02–0.37) and β 0.31 (95% CI 0.07–0.55), respectively. Associations were not socioeconomically patterned. Conclusions: Investigating mealtime television use motivators across the socioeconomic spectrum could inform interventions targeting the high consumption of discretionary foods in children.


Introduction
One of the major contributing factors to poor health outcomes in high-income countries is sub-optimal nutrition, beginning in early childhood. In Australia, children consume well above the recommended intakes for discretionary foods [1,2], well below recommendations for vegetables [3], and an increasing proportion consume insufficient fruit with increasing child age [4]. These behaviours are contributing to high rates of childhood obesity [5] making child nutrition a public health priority. Early childhood is particularly important because nutrition during infancy and the preschool years influences growth and health outcomes across the entire life course [6,7]. During the early years, the majority of young children's eating occurs within the home [8], and the home environment contributes substantially to children's diets, eating behaviours, and learned food preferences [9,10].
A family's socioeconomic position (SEP) is also an important correlate of children's diets and obesity risk [11,12]. Children from disadvantaged backgrounds are more likely to have less healthy diets and poorer health outcomes when compared to those of higher SEP [2,13,14]. In particular, inadequate intakes of fruits and vegetables, and high intakes

Mealtime TV Use
At baseline, mealtime TV use was assessed by asking 'how often does (child's name) watch television while eating (each of breakfast/lunch/dinner/snacks)?' [41]. Five response options were available: never or <1 day per week, 1-2 days/week, 3-4 days/week, 5-6 days/week, and 7 days/week. These items and the response scale were informed by previous research [42,43]. Given the data distribution, where there was a high proportion of 'never' responses (38-66%), the five response options of mealtime TV use were collapsed into three categories (never, 1-2 days/week, and 3-7 days/week) for subsequent analyses.
To calculate average daily mealtime TV use, as conducted previously [19], a cumulative measure was first created out of a possible 28 occasions per week (four possible occasions per day: breakfast, lunch, dinner, and snack time). Where the response option was a range, the middle value was imputed in the cumulative measure (for example, 1-2 days was converted to 1.5 days). The weekly score out of 28 was divided by seven to create a daily average score out of four, which was treated as a continuous variable for analysis. The survey also assessed the use of other electronic media during mealtimes, using a separate question with the same format as TV use. However, very few (2-8%) children engaged in other electronic media at each mealtime; therefore, there was insufficient variability in the data to include in analysis (results not presented).

Intakes of Discretionary Foods, Fruits, and Vegetables
All dietary measures have been previously validated [44][45][46] and were considered the most relevant for this study at the time of data collection [47]. At both baseline and follow-up, we asked 'In the past month, about how often has (your child) had the following foods?'. Eight categories covering a range of discretionary items were included, with examples specific to Australia: fried potato products, savoury biscuits, sweet biscuits, takeaway food, baked goods, confectionary, and sugar-sweetened beverages [44]. Six response options were available: never or <once/month, 1-3 times/month, once/week, 2-4 times/week, 5-6 times/week, and 7 times/week. Intakes of discretionary foods and beverages are henceforth referred to as 'discretionary foods'. Appendix B includes a list of discretionary food items assessed. Intake frequency of each discretionary food item over the past month was converted to daily intakes using daily equivalent frequencies (never = 0; 1-3/month = 0.067; 1/week = 0.143; 2-4/week = 0.429; 5-6/week = 0.786; 1/day = 1.0; 2-3/day = 2.5; 4-5/day = 4.5; ≥6/day = 6.0) [48]. The daily frequency of each item was then summed to create average daily discretionary food frequency, analysed as a continuous variable (frequency per day).
Fruit and vegetable intakes were assessed by one item each, asking about 'how many serves of vegetables does (child's name) usually eat per day?' (with examples), and included five response options: my child does not eat these, <1 serve/day, 1 serve/day, 2 serves/day, 3 serves or more/day. Informed by the data distributions, intakes of fruit and vegetable were dichotomised for analysis as <2 and ≥2 serves per day, respectively.

Covariates
The baseline questionnaire also collected child age, child gender, location of meals, and family meal frequency. These variables have been considered as potentially influencing the association between diet and shared mealtimes [17,49], and were, therefore, included as covariates. Data collection methods for these measures have been described elsewhere [19]. The location of meals was dichotomised into an optimal category: sitting at a table/bench (highchair or chair), and a sub-optimal category: in the car, moving around the house, and sitting on the floor/couch. Family mealtime frequency for each meal type was defined as the responding parent and child eating together, and was categorised into 'never', occasionally (1-2 days), and more frequently (3-7 days) for the current analysis.

Socioeconomic Position
The education level of the responding parent at baseline was used to define SEP for this study. Previous research has found maternal education to be the strongest and most consistent socioeconomic predictor of children's dietary intakes [2,11,39,50], and the majority of responding parents (97%) in this study were mothers. Therefore, parental education was used as a proxy for SEP.
Due to the large number of responding parents who were university educated in this sample (>70%), education was dichotomised to 'university educated' and 'non-university educated'; comprising ≤ 12 year or equivalent, trade/apprenticeship (e.g., hairdresser, chef), and certificate/diploma (e.g., childcare, technician).

Survey Reliability
A test-retest study was conducted at baseline to measure the reliability of survey questions. A subsample of 54 participants completed the same survey again, two weeks after initial completion. Intraclass correlations (ICCs) were performed for each ordered categorical variable, and Cohen's Kappa was performed for each non-ordered categorical variable.

Statistical Analysis
Descriptive analyses were conducted to explore the sample characteristics. Multivariable logistic or linear regressions were used to examine the prospective associations between mealtime TV use at baseline and intakes of fruit and vegetables (<2 and ≥2 serves per day), and discretionary foods (intake frequency per day), at the 2-year follow up, respectively. Separate models were conducted for each meal type (breakfast, lunch, dinner, and snacks), as well as average daily mealtime TV use. Crude models were conducted first with mealtime TV use as the exposure, and dietary intakes at follow-up as the outcome with adjustment for baseline dietary intakes.
Adjusted models included the following covariates: child age, child gender, family mealtime frequency, SEP, and mealtime location. Pearson and Spearman correlations were conducted to test for potential multicollinearity between mealtime TV use and covariates. No significant correlation was found, so all covariates were included simultaneously in the adjusted model. Analyses to detect possible interaction between mealtime TV use and covariates were also performed by creating interaction terms (the product of two variables). The interaction between mealtime TV use and mealtime location was significantly associated with dietary intake; therefore, that interaction term was also included as a covariate in the adjusted model.
To test if the association between mealtime TV use at baseline and dietary intakes at follow-up differed by SEP, we conducted stratified analyses (crude and adjusted) by SEP, and the likelihood ratio test was used to test for differences between stratum-specific regression coefficients. For adjusted analysis, the model adjusted for all aforementioned covariates, except for SEP. All analyses were conducted in StataIC (version 1.0, StataCorp., College Station, TX, USA) with statistical significance set at p < 0.05.

Participant Demographics
Of 992 participants at baseline, 369 participants with complete data on mealtime TV use at baseline and dietary intakes at both baseline and follow-up were included in the crude model. Seventeen participants had missing data on covariates, resulting in a final sample of 352 included in the adjusted model ( Figure 1). A comparison of demographics between respondents and non-respondents at follow-up showed no significant differences between the groups (Appendix A Table A1).
sample of 352 included in the adjusted model ( Figure 1). A comparison of demogra between respondents and non-respondents at follow-up showed no significant ences between the groups (Appendix A Table A1).  Table 1 summarises the baseline characteristics of participants. A large proport respondents were university educated (72%). Child age distribution at baseline wa aged 6 months to <1.5 years, 29% aged 1.5-<3 years, and 40% aged 3-<6 years, w almost equal number of males and females.
At baseline, most children (74-86%) had optimal eating locations for main (breakfast, lunch, and dinner). In contrast, most children (71%) had suboptimal eati cations for snacks. A large proportion of parents ate each meal type with their child 7 days per week (74-90%). Around one third of children ate breakfast, lunch, and/o ner while watching TV at least once per week, and a similar number ate snacks watching TV on 3-7 occasions per week.

Dietary Outcomes
At baseline, 71% and 54% of children consumed ≥2 serves of fruits and ≥2 serve etables, respectively. At follow-up, 82% of children consumed ≥2 serves of fruits, an consumed ≥2 serves of vegetables. The mean daily frequency of discretionary food 0.96 (standard deviation (SD) 0.88) at baseline, and 1.30 (SD 0.85) at follow-up.

Baseline Television Use and Prospective Dietary Outcomes
Crude analysis revealed that overall daily mealtime TV use at baseline was a ated with a higher frequency of consuming discretionary foods at 2-years follow-0.16; p < 0.01) ( Table 2). No evidence of association was found for total daily mealtim use at baseline and prospective intakes of fruit or vegetables. Mealtime TV use (≥1 d week versus never) during breakfast was associated with a higher frequency of co ing discretionary foods (p = 0.03). Significant associations were also found be mealtime TV use during lunch (3-7 days per week), dinner (1-2 days per week), and  Table 1 summarises the baseline characteristics of participants. A large proportion of respondents were university educated (72%). Child age distribution at baseline was 31% aged 6 months to <1.5 years, 29% aged 1.5-<3 years, and 40% aged 3-<6 years, with an almost equal number of males and females.
At baseline, most children (74-86%) had optimal eating locations for main meals (breakfast, lunch, and dinner). In contrast, most children (71%) had suboptimal eating locations for snacks. A large proportion of parents ate each meal type with their children 3-7 days per week (74-90%). Around one third of children ate breakfast, lunch, and/or dinner while watching TV at least once per week, and a similar number ate snacks while watching TV on 3-7 occasions per week.

Dietary Outcomes
At baseline, 71% and 54% of children consumed ≥ 2 serves of fruits and ≥2 serves vegetables, respectively. At follow-up, 82% of children consumed ≥ 2 serves of fruits, and 69% consumed ≥ 2 serves of vegetables. The mean daily frequency of discretionary foods was 0.96 (standard deviation (SD) 0.88) at baseline, and 1.30 (SD 0.85) at follow-up.

Baseline Television Use and Prospective Dietary Outcomes
Crude analysis revealed that overall daily mealtime TV use at baseline was associated with a higher frequency of consuming discretionary foods at 2-years follow-up (β: 0.16; p < 0.01) ( Table 2). No evidence of association was found for total daily mealtime TV use at baseline and prospective intakes of fruit or vegetables. Mealtime TV use (≥1 day per week versus never) during breakfast was associated with a higher frequency of consuming discretionary foods (p = 0.03). Significant associations were also found between mealtime TV use during lunch (3-7 days per week), dinner (1-2 days per week), and snack times (3-7 days per week), and a higher frequency of discretionary foods. No significant associations were found between baseline mealtime TV use and vegetable intakes at follow-up, although dinner TV use 3-7 times per week showed a trend towards association with lower odds of consuming ≥ 2 serves of vegetables/day (odds ratio (OR): 0.56; p = 0.07). For fruit, mealtime TV use at lunch (1-2 days/week) and dinner (1-2 days/week) was associated with lower odds of consuming ≥ 2 serves of fruit per day, with odds ratios of 0.47 (p = 0.04) and 0.48 (p = 0.03), respectively. Table 2. Crude model. Association between children's mealtime TV use at baseline and consuming two or more serves of fruit and vegetables per day, and consuming non-core discretionary foods on more than one occasion per day at the follow-up, adjusting for baseline intakes. In the adjusted model (Table 3), the significant association between baseline mealtime TV use and discretionary food frequency at follow-up remained for most meals, except snacks. The significant mealtime TV predictors of fruit intake observed in the crude model were attenuated, and results related to follow-up vegetable intakes were similar. Table 3. Adjusted models. Association between children's mealtime TV use at baseline and consuming two or more serves of fruit and vegetables per day, and consuming non-core discretionary foods on more than one occasion per day at the follow-up, adjusting for baseline intakes and covariates (footnote).

Socioeconomic Position
In the multivariate analyses, parent education was not significantly associated with baseline mealtime TV use at any of the assessed mealtimes and prospective intakes of fruit, vegetables, and discretionary foods (not reported in tables; p = 0.15-0.82). Stratified analyses by two SEP groups are presented in Appendix A Table A2. There were no significant associations observed between mealtime TV use and diet for children whose mothers were non-university-educated. In children whose mothers were university-educated, higher mealtime TV use frequency 1-2 days per week for breakfast (β: 0.39; p = 0.01) and dinner (β: 0.24; p = 0.05), and 3-7 days per week for breakfast (β: 0.23; p = 0.03) and lunch (β: 0.39; p = 0.01), were significantly associated with discretionary food frequency. Additionally, watching TV during dinner 1-2 days per week was associated with lower odds of consuming ≥ 2 serves of fruit per day in this group (OR 0.41; p = 0.05). In both SEP groups, mealtime TV use was not associated with vegetable intakes (p < 0.05).

Discussion
In a cohort of Australian children between 6 months and 6 years, a higher frequency of mealtime TV use was prospectively associated with higher discretionary food intakes at the 2-year follow up. These results were broadly consistent for breakfast, lunch, and dinner times, as well as total mealtime TV use across the day.
Previous cross-sectional research has shown a consistent association between young children's mealtime TV use and sub-optimal diets, particularly lower intakes of fruit and vegetables, and higher discretionary food intakes [20,24,[26][27][28][29][31][32][33][34][35][36][37][38]. The present study is the first to evaluate the prospective association between mealtime TV use and dietary intake in young children, and thus, provides insights on the direction of influence. The significant prospective association between higher mealtime TV use and higher frequency of discretionary food intakes observed emphasises the importance of understanding the mechanisms by which this association plays out. Previous literature has highlighted that in adults, TV use during meals can attenuate feelings of fullness, leading to a higher consumption of food in one eating occasion, and contributing to an increased overall energy consumption [52]. In addition, regularly eating while watching TV may, in turn, result in TV triggering snacking, even when not hungry [53]. Further, watching TV is highly likely to expose children and families to discretionary food ('junk food') advertising [29], which is known to influence targeted food consumption. There are currently no regulations on discretionary food advertising in Australia. These factors, and others, might represent a cluster of sub-optimal eating behaviours among families more likely to engage in mealtime TV use.
Watching TV during breakfast showed the strongest prospective associations with discretionary food consumption. This is interesting because sugary breakfast cereals were not included in our list of discretionary items assessed, highlighting some important considerations. Among these is whether the association between TV use at breakfast and discretionary food intake stems from the food eaten at breakfast, or could it be that a clustering of less than optimal dietary and mealtime TV behaviours might play a role in determining dietary intake? For example, families who engage in breakfast TV use might also tend to purchase and consume more discretionary foods throughout the day. Acknowledging clustering in mealtime research is important [16,23]. Regardless, the prospective nature of this study suggests that mealtime TV use, particularly at breakfast, may be a determinant of diet outcomes in children.
Interestingly, adding location as a covariate reduced the strength of associations between TV use during snack times and the consumption of discretionary foods, fruits, and vegetables (not reported in tables), and the association with discretionary food intakes was no longer significant. Eating in a location other than at a table/bench (such as in the lounge room or the bedroom) has been found to be cross-sectionally associated with reduced vegetable intake [54], poorer overall diets in young children [55], and higher weight status in slightly older children [27]. Little research has focussed on younger children's mealtime location. It is possible that the consumption location may influence the types of foods parents offer their children. For example, one possibility could be that parents consider that meals eaten on the couch require foods which are easier to hold, can be eaten with fewer utensils, or that produce less mess. Alternatively, parents might intentionally offer lesspreferred foods (such as vegetables) while children are distracted by screens in an attempt to increase their intake. Furthermore, families of lower SEP may have fewer resources which enable them to eat in optimal locations, such as space for, or ownership of, a dining table [55], as well as the possible clustering of sub-optimal mealtime behaviours in some families [56]. Understanding potential interactions between mealtime TV use, location, and discretionary food intakes and linked contributors requires exploration. In particular, considering strategies which might address the behaviours in combination and whether these strategies are feasible and acceptable for parents of young children to implement is crucial.
This study highlights the importance of understanding young children's TV use during snacking, in the context of mealtimes. Given that snacks contribute substantially to young children's overall diet intake and diet quality [57], but are little studied in the context of mealtimes [17,20], the inclusion of snacks in this research is an important contribution. In this study, snacks were found to be least frequently eaten at a table or bench, compared to other meal types. Additionally, young children's mealtime TV use appears to be the highest during snack time [19], as well as the meal least-shared with parents [19]. This highlights that snack times may offer the greatest opportunity for improvements in young children's mealtime behaviours.
Although this study found no significant associations between mealtime TV use and prospective intakes of fruit and vegetables, all results were in the expected direction, with intakes lower as TV use increased. Previous cross-sectional studies supporting this association have used measures such as four-day [27] or seven-day dietary recall [26,37], TV time diaries [35], and Ecological Momentary Assessments [28], all of which are more detailed than measures in the present study and may have enabled the detection of smaller differences. It is noteworthy that in this study, there was minimal change in fruit and vegetable intakes from baseline to follow-up, which is consistent with existing evidence that young children's fruit and vegetable intakes tend to change little between the ages of nine months and five years [2]. This could contribute to explaining the minimal associations found between mealtime TV use and fruit and vegetable intakes in the present study. Alternatively, the prospective association between mealtime TV use and diet may well be more closely aligned with discretionary intakes than that of fruits and vegetables. This may provide a rationale for focusing on mealtime TV behaviours in nutrition promotion interventions which focus on limiting discretionary food intakes in young children.
This study did not find a differential effect of SEP on associations between mealtime TV use and diet. The strengths of the associations are similar, suggestive of the stratified analysis by SEP groups resulting in the low SEP group being too small to detect true differences. The results suggest that mealtime TV use is prospectively associated with discretionary food intakes, irrespective of SEP. This supports some cross-sectional research that mealtime components, such as TV, may not be influenced by SEP [23]. Although this study found no statistical differences, other evidence suggests that low SEP families are more likely to watch TV during meals [19,36,58], and, therefore, may be more susceptible to poorer dietary outcomes. The exploration of mealtime TV use in specific socioeconomic groups will help to inform targeted nutrition promotion initiatives.
To our knowledge, this is the first study to assess the prospective association between mealtime TV and intakes of fruit, vegetables, and discretionary foods. Given that the majority of mealtime research is cross-sectional, providing clarity of the temporal order of the outcomes measured is an important strength of this study. It is also the first study to assess a range of eating occasions across the day, including snacks. A potential limitation of this study may be the dietary measures, given their brevity, and that in the current study, child fruit and vegetable intakes were reportedly higher than the Australian average [3]. Though this may reflect variations in the dietary measures used between studies, it is possible that it reflects that higher child dietary quality is associated with a high level of parental education [59]. In the current study, 73% of parents were university educated versus ≈ 49% of Australian women aged 25-34, from national data [60], highlighting a need for future research focussing on populations of lower education. Moreover, although validated tools were used for intakes of all dietary items, future analysis would benefit from using a more detailed and robust measure of dietary intake. Additionally, snacks were conservatively included only once per day in calculating daily mealtime TV use. Future research could assess the proportion of children's meals eaten while using screens, accounting for how many meals an individual child eats in a day. Other limitations to this study include the potential influence of other unmeasured confounders, such as child ethnicity, body mass index (BMI), and other dietary outcomes, as well as potential residual confounding.

Conclusions
This study identified that a higher frequency of mealtime TV use in young children was prospectively associated with a higher frequency of discretionary food intakes two years later, irrespective of SEP. Reducing mealtime TV use could be a promising intervention strategy in the pursuit of reducing young children's discretionary food intakes, the promotion of optimal diet quality, and improving health outcomes for young children and future generations.  Table A2. Stratified analysis by parental education-Associations between children's mealtime TV use at baseline and consuming two or more serves of fruit and vegetables per day, and consuming non-core discretionary foods on more than one occasion per day at the follow-up, adjusting for baseline intakes and covariates. β: coefficient; OR: Odds Ratio; 95% CI: 95% Confidence Interval; covariates: baseline intakes, child gender, child age, location of meals, and family meal frequency in children from non-university-educated and universityeducated parents. Numbers in bold denote significant outcomes.