Do Not “Let Them Eat Cake”: Correlation of Food-Consumption Patterns among Rural Primary School Children from Welfare and Non-Welfare Households

Physical and financial access impacts food choice and consumption, while educational attainment, employment, income, gender, and socioeconomic status are also influential. Within this context, the aim of the paper is to examine the association between various foods consumed and eating patterns of children between low and higher income households. A paper-based survey was completed by parents/carers of children in 41 primary schools in rural and regional areas of Victoria. Data collected included demographics and the consumption of fruit, vegetable, and other foods including drinks. Ordinal data were analysed using Spearman’s rank-order correlation. The main findings were that children who consumed more fruit and vegetables tended to have a higher intake of healthy drinks (plain milk and water) as well as a lower intake of unhealthy snacks and drinks (sugar sweetened drinks). Those who perceived that fruit and vegetables cost too much reported greater consumption of unhealthy snacks and sugar-sweetened beverages, which was more prominent in low-income households. Changing food consumption behaviours requires a complex systems-based approach that addresses more than just individual issues variables. A participatory approach that works with local communities and seeks to build an understanding of unique challenges within sub-groups has potential for embedding long-lasting and meaningful change in eating behaviours.


Introduction
It has been established that low consumption of nutrient-rich foods, such as fruits and vegetables, has short-and long-term negative health implications [1]. Poor nutrition is responsible for greater morbidity and mortality than alcohol, tobacco, and low-level physical activity combined [2]. A number of dietary risk factors have been implicated in chronic conditions, such as cardiovascular disease, type 2 diabetes, and some cancers [3]. Low consumption of fruits and vegetables is more prevalent in low-socioeconomic populations [4]. Physical and financial access to food has been clearly shown to affect food consumption [5]. Food literacy, or knowledge of healthy eating, may also affect food choices and methods of preparation and consumption [5]. Other social determinants, such as educational

Research Instrument
The "survey on eating patterns of children" was designed by a research team of dietitians, health promotion workers, a local government community wellbeing officer, and an academic. Development of the tool and detail of the tool has been described earlier [9]. The survey examined demographics of child and parent/carer, fruit and vegetable consumption of the child and additional factors that influence eating patterns, which has been published in detail previously [9]. This paper examines the correlations between children's daily fruit and vegetable servings with servings of other foods and drinks. In addition, this paper explores differences in daily consumption of food and drinks between the children whose parents were health-care card recipients and those who were not.
Parents were asked about their child's consumption of various food and drinks on a typical day, with descriptor amounts given for a serving size of each, in addition to a colour photograph of an example serving size [10]. Foods were listed separately as categories. Vegetables were described as fresh, frozen, cooked, or raw, and fruit as fresh, dried, or tinned. Snack examples were given as chips, cheezels, and muesli bars. Confectionary was defined as lollies and chocolate. Cake, doughnuts, sweet biscuits, and muffins were listed separately to the other snack examples. Drink examples were listed as water, plain milk, flavoured milk, fruit juice, soft drink, and cordial (un-carbonated sugar-sweetened beverage), again with a descriptor of serving size. A six point serving scale was provided, and included serving values such as 0, 1, 2, 3, 4, and 5 or more, or "I don't know". A copy of the survey is available from the corresponding author.
Ethical approval for the study was granted by the Goulburn Valley Health Human Research Ethics Committee (EC00220) and the Department of Education and Early Childhood Development (2013_002125).

Data Collection
The study population was primary students in Grades 1 or 3 (aged 5 to 11 years) who attended one of the 41 primary schools that chose to participate across the three Local Government Areas of Greater Shepparton, Moira, and Strathbogie Shires in rural or regional North East Victoria. No other inclusion or exclusion criteria were set, but the survey was only provided in English. Surveys were distributed by teachers to the students to take home for their parent/carer to complete within a two-week period and be returned to the school for collection by the research team [9,10].

Data Analysis
Data were analysed using SPSS version 21 (IMB, Chicago, IL, USA). Descriptive statistics were generated for the demographic data and were used to characterise responses to questionnaire items [11], while inferential statistics examined group comparisons. These were undertaken using non-parametric tests such as Spearman's rank-order correlation ( ), Pearson's Chi-Square tests, and Mann-Whitney U tests, as the data, even after being transformed, violated a number of parametric assumptions. Spearman's rank-order correlation was used as the survey elicited information using a six point serving scale that was ordinal rather than continuous level data, where a small correlation was considered at 0.2 [12,13]. Significance was initially two-tailed p ≤ 0.05. However, this p value was adjusted using the Dunn-Šidák correction to account for the possibility of a Type I error rate resulting from multiple comparisons, that is, an increased family-wise error rate. We chose the Dunn-Šidák method over the more common Bonferroni correction because the latter can be overly conservative. That being said, there was a negligible difference in the resulting p values for the two methods in our case. We denote the Dunn-Šidák-adjusted p-value as pDS.

Results
The sample consisted of 550 respondents (28.5% response rate) from the City of Greater Shepparton (n = 258), Moira Shire (n = 233) and Strathbogie Shire (n = 59). Parent or carer respondents ranged in age from 23 to 73 with a mean age of 39.2 years (SD = 6.1). Education levels varied among respondents; 9.0% (n = 48) completed primary school, 24.6% (n = 131) completed secondary school, 61.9% (n = 330) completed tertiary education, and the remaining 4.5% (n = 24) completed other forms of education, such as vocation education. In addition, 32.5% of respondents (n = 174) indicated they were health-care card recipients, and reflective of the percentage of health-care card recipients across the three local government areas [14]. It must be noted that 14.1% (n = 78) of respondents were single parent/carer families, and 31.0% (n = 24) of these single parent/carer households were health-care card recipients and had a higher percentage of male children, as outlined in Table 1.

Overall Correlation between Food and Drink and between Health-Care and Non-Health-Care Card Households
When examining the overall cohort for the association that exists between fruit and vegetable servings and other foods, it was noted that there is a small positive correlation with fruit servings and consumption of water. In addition, there was a small negative correlation between vegetables and consumption of soft drinks (Table 2). When specifically examining the consumption of soft drinks, it was noted that there is a small negative correlation with the consumption of water and milk, while there is a medium positive correlation with sweets and a small positive correlation consumption of snacks, cordial, and fruit juice.
When examining the differences between health-care and non-health-care card recipients in terms of the types of foods that were consumed, both groups had correlations that were significant. For example, among those with a health-care card, the correlation between the number of fruit servings and the number of vegetable servings was smaller ( = 0.204, p < 0.009) than those who did not have a health-care card ( = 0.260, p < 0.001). Conversely, among those with a health-care card, the correlation between eating less healthy food (e.g., confectionary, cake, and snacks) fruits and vegetable servings was larger than among those who did not have a health-care card.
In some cases, the correlation between soft drink and confectionary servings was almost one-and-a-half times larger ( = 0.620, p < 0.001) among health-care card holders than those who did not have a health-care card ( = 0.449, p < 0.001). For example, the correlation between soft drink, snacks, and fruit juice was larger among those households who had a health-care card, ( = 0.276, p < 0.001) and ( = 0.330, p < 0.001) than those who did not ( = 0.190, p < 0.001) and ( = 0.182, p < 0.001), respectively.

Fruit and Vegetables Cost between Health-Care and Non-Health-Care Card Households
There was also a positive correlation between agreement with the statement that fruit and vegetables cost too much and the servings of confectionary and snacks that were consumed by children. However, when examining health-care card and non-health-care card holders, it was indicated that among health-care card holders, the perception that fruit and vegetables cost too much was associated with increased servings of confectionary and snacks, while the same correlation is not observed among non-health-care card holders (Table 3).

Food and Drink Servings between Health-Care and Non-Health-Care Card Households
When examining the differences between health-care and non-health-care card recipients, it was noted that children from households who had a health-care card consumed fewer vegetable servings than those children from households without a health-care card. However, there was no significant difference when examining fruit servings between the household types. The perception of cost of fruit and vegetables was highlighted to be significant among those who were health-care card holders, as outlined in Table 4. Children of health-care card recipients consumed fewer servings of water. Results also showed that these children consume more servings of fruit juice. A Mann-Whitney U test revealed a significant difference between the consumption of fruit juice among households with a health-care card and those that did not have a health-care card (U = 22019, z = −3.30, p < 0.001, = 0.14).
Similarly, greater consumption of sweetened flavoured drinks and soft drink occurred among children from households that were health-care card recipients. A Mann-Whitney U test revealed a significant difference between the consumption of sweetened flavoured drinks (U = 21412, z = −4.65, p < 0.001, r = 0.21, and soft drink, U = 21701, z = −4.40, p < 0.001, = 0.19) among households with a health-care card and those that did not have a health-care card.

Discussion
The main findings were that children who consumed more fruit and vegetables tended to have a higher intake of healthy drinks (plain milk and water) as well as a lower intake of unhealthy snacks and drinks (cake, sweet biscuits, fruit juice, and sugar sweetened drinks). However, respondents who perceived that fruit and vegetables cost too much reported greater consumption of unhealthy snacks and sugar-sweetened beverages. Ascertaining consumption of cordial and soft drinks was an important aspect of this study, as sugar-sweetened beverages have been shown to serve as a good proxy measure of consumption of other unhealthy foods [15]. This finding was more prominent in households that had a health-care card, where children consumed fewer vegetables, less water, and more fruit juice than those who do not have a health-care card.
These findings resonate with current research, where lower-income households do not consume adequate fruit and vegetables and have a greater tendency to purchase and consume energy-dense foods, and children who consumed sweetened beverages also consumed unhealthy snack foods [2,16,17]. It has been suggested these behaviours are centred on the contributing factors such as poor nutritional knowledge, lack of time or confidence in meal planning and preparation, and the perceived high cost of healthier foods [18]. In addition, when examining regional, rural, and remote contexts, other geographical factors may also contribute to this phenomenon. These factors include distance to purchase food, the availability of public transport, the short shelf-life of fresh food, increased transport costs being passed on to consumers, and rural consumers previously experiencing food shortages [5].
However, beyond healthy food access and costs being inhibitory, other studies found that food choice and food consumption behaviours were also influenced by lower-income earners having a relatively low regard for food and the associated health issues, leading to food consumption being less consistent with dietary guideline recommendations [3,19]. Willis and colleagues [20] suggest that individuals and families use "hierarchies" of need and risk, which affect where purchasing nutritious food and health is placed in proportion to other items that are considered more important [21]. This reflects Bourdieu's suggested notion of the "distances from necessity" [22], where individuals with lower incomes eat what is most filling and affordable, while food consumption among those with higher incomes is more nutritious, and they have a greater focus on preparation, presentation service, and entertaining guests. This may explain why popular food and lifestyle programs, such as Master Chef or Gourmet Farmer, are more popular among higher-income viewers, where cooking is considered fun rather than work, and these individuals can vicariously consume and critique food as armchair experts [23][24][25]. Distance from necessity may also indicate why nutritional interventions that promote healthy food consumption are less successful among those that who may be more socioeconomically disadvantaged [17,26].
Engel's [27] law of consumption states that, as income rises, the proportion of money used to purchase food falls; even when food expenditure rises, the percentage of income used to purchase food is much less. However, Bourdieu [22] suggests that, when social conditions, such as high or low income, are experienced in early childhood, there is a propensity for these experiences to shape views of food, influence food consumption behaviours, and even tastes into adulthood. It is through these familial or social-cultural food practices where social order, ideologies, identity, and day-to-day practices are developed and maintained. Even in families, children are not passive recipients but have certain influences on familial food consumption patterns and behaviours [20,28]. Similarly, food consumption behaviours of a family are also influenced and perpetuated by the wider network of social relationships and social divisions or classes that exist in the community.
Despite these factors, individuals and groups may still be influenced by socio-cultural norms, there is the capacity to alter or shape food consumption behaviours, and it is suggested to be reflective of gaining a higher socioeconomic status or moving up the "social ladder" [20]. However, moving from the lowest socioeconomic group to the highest does not always ensure a change in food consumption behaviour. For example, it has been indicated that, even when controlling for income, food preferences and consumption behaviours are intergenerational and persist among children into adulthood [29]. This reiterates that food consumption, meaning and behaviours are social practices that develop, but also sustain self and family identities, social structures, and relationships [30].

Implications for Policy
It has been well established that the causes of unhealthy eating patterns and the resulting population weight gain are complex and inter-related [31]. The Foresight obesity systems map identifies thematic clusters of variables affecting energy balance ranging from individual through to societal and environmental factors, and may assist in the development of a more sophisticated and integrated policy approaches [32]. However, at the community level, community-based system dynamics has been identified as a promising health promotion approach that aims to address the complexity of unhealthy eating patterns. Using a participatory method, known as group model building, this approach seeks to engage local communities in developing real and meaningful solutions to address unhealthy eating behaviours [33,34]. Due to the differences observed between high-and low-income households in relation to eating patterns, a community-based initiative that examines the whole food system within a local context may provide opportunities for developing a better understanding of the reasons behind food and beverage consumption patterns within this target group.
Interventions must also have an intergenerational focus, where there is greater recognition that the changing patterns and behaviours of food consumption occur over a much longer time, or even generations, rather than fleeting and condensed interventions, which have limited sustainability. Research [17,35] has suggested that it begins at the primary school level; however, these changes may not be seen for some time. This requires longitudinal research that evaluates healthy eating systems-based interventions specifically addressing the needs and barriers faced by socioeconomically disadvantaged individuals.

Limitations
The limitations of the study relate to respondents self-selecting to participate in the survey and the season in which the survey was completed. Those who did not respond to the survey were either parents/guardians who did not receive the survey from their child, or who chose not to participate in the study. Those who did not respond were not asked the reason for non-response, as the researchers did not directly recruit participants. It is recognised the non-response creates a level of bias in the results. Presumably, non-response is more likely to come from socioeconomically disadvantaged backgrounds and the survey was completed at a time when fresh fruits and vegetables were more abundant. Consequently, our findings are likely to underestimate the true magnitude of low fruit and vegetable intake and the consumption of unhealthy snacks and sugar-sweetened beverages.

Conclusions
This study provides vital information about childhood fruit and vegetable consumption from one region in Victoria, enabling health promotion interventions specific to the populations included in the study. The results suggest that low-income families in this region should be considered as a sub-group in an effort to reduce sweetened beverage and unhealthy snack consumption. Furthermore, health promotion interventions must incorporate long-term, multi-level solutions, built on policy and community-based system dynamics theory. The survey developed and utilised for this study may be useful for other primary-care service providers wishing to undertake a similar investigation.