Introduction/background
The prevalence of childhood overweight and obesity is a public health concern globally. More recent data state the prevalence of overweight or obesity is approximately one quarter of children worldwide [
1,
2]. Ireland is no exception to this issue, a study tracking secular trends in the height and weight of Irish children from 1948–2002, found that the weight of Irish children had increased disproportionally to their height [
3]. Obesity affects the physical and psychosocial wellbeing of children and is a significant risk factor for obesity in adulthood [
4-
8]. It also creates a significant social and financial burden for wider society [
9].
Obesity is a complex issue arising from a myriad of individual and environmental factors. The imbalance between energy intake and energy expenditure is a central causal factor for obesity [
10,
11] and therefore what we eat is important. Currently, there is no consensus on the best approach to measure the association between diet and childhood obesity. To describe this diet-disease association diet quality has gained increasing attention in the literature as a predictor of childhood obesity. However there is no agreement regarding the definition of diet quality [
12]. Much of the adult literature uses dietary quality scores (DQS) or indices as a measure of diet quality [
12]. DQS are also emerging in the paediatric literature [
13] and more recently their association with health related outcomes in children has been reviewed [
14].
Investigating dietary quality using DQS better reflects the real-life scenario that foods are not consumed in isolation [
15,
16]. There are two forms of dietary pattern analysis: an a priori approach DQS, based on current recommendations or a known hypothesis of a healthy diet; and
an a posteriori approach based on data driven statistical analysis [
17,
15]. There is a need to measure the association between whole diet and childhood overweight and obesity. Adherence to dietary recommendations may be useful to provide an insight into dietary patterns in children. The a-priori approach is reproducible across populations and can be used to compare diets across different countries [
18]. Studying children’s dietary patterns could have an important public health impact since results may be easier to translated into public health messages than research on specific foods or nutrients [
19,
20].
Recent studies examining associations between diet quality and health outcomes including obesity in children are predominantly based on detailed dietary assessments [
13] which have high respondent and researcher burden. There is a need to examine the potential value of simple dietary assessment tools from which simple DQS’ can then be developed to measure adherence to dietary recommendations. A systematic literature review stated that brief tools to assess diet quality should include the five major food groups [
21]. These simple tools may be more suitable for use in large-scale epidemiological studies that collect data on all aspects of health and development. Therefore, the aims of this research are firstly to construct a diet quality score from a brief dietary assessment. Secondly, to examine the association between diet quality and childhood overweight or obesity. Thirdly this study also examines the associations between individual DQS components and childhood overweight or obesity.
Discussion
The present research aimed to investigate the association between diet quality and childhood overweight or obesity, constructed from a short dietary assessment tool. The main finding was that a simple DQS constructed from a 20 item, parent reported, FFQ of foods eaten over the past 24-h was associated with childhood obesity. This easy to administer, short FFQ, with a simple application to a DQS, may be a useful tool in other large-scale epidemiological studies to identify children with poor dietary patterns. The findings also suggest that diet quality, as opposed to individual foods or food groups, may act as a useful explanatory variable in the complex myriad of factors associated with the development of obesity. Many foods or nutrients may be involved in the promotion or protection against obesity [
32]. When only one food or nutrient is studied in relation to an outcome such as obesity the findings are often inconsistent across studies [
32]. Individual foods or nutrients may be captured with less precision/reliability especially when investigating obesity as there is a trend to under-report certain foods by overweight participants [
32-
34]. ‘No single nutrient has been unequivocally associated with the development of obesity’ [
32] and hence DQS are being used more frequently as they reflect overall dietary patterns better than single nutrients or food groups [
35] and they take into account that the diet as a whole may be promoting obesity rather than one food/nutrient.
There are several other advantages to assessing the diet-disease relationship using DQS such as; ‘the effect of a single nutrient may be too small to detect, but the cumulative effects of multiple nutrients included in a dietary pattern may be sufficiently large to be detectable’ [
16,
15]. The approach to examining diet-disease relationships using DQS is particularly useful where multiple dietary components are relevant for a particular disease, as it accounts for interactions and correlations between foods and food groups. The approach may be particularly useful for Coronary Heart Disease (CHD), hypertension, diabetes mellitus and obesity where multiple dietary components are established risk factors [
15]. Our DQS demonstrated sufficient power to detect differences in diet quality between normal weight and obese children, differences in diet quality between normal weight and overweight children were less pronounced. The results for overweight are not as pronounced for diet quality but also the other covariates. It is difficult to interpret the comparison of the overweight versus normal weight children as an increased risk of obesity causes a reduction in risk of overweight. It is likely that as a risk factor increases the risk of obesity this has made the children less likely to be overweight as they are more likely to be obese. Whilst our DQS may not provide a reliable estimate of individual diet, it does provide estimates of mean differences in diet quality between groups of normal weight and obese children. Though individuals’ diets vary day to day, we would expect that the mean for the group would be stable. Moreover, variation in time between groups would be uniform. We are of the opinion that our DQS shows construct validity, as stated previously no formal definition of diet quality exists, however DQS are used in the literature as measures of diet quality.
Our findings are similar to those from other studies on diet quality and childhood obesity that used more detailed measures to assess diet such as a 132 item FFQ [
36] 4 day food diary [
37], and a 154 item FFQ [
38]. Feskanich et al. [
36], found that a higher DQS measured using a modified version of the Health Eating Index was associated with a reduced BMI in children [
36]. In the UK, Jennings et al. [
37], found that the Diet Quality Index (DQI) and the Health Diet Indicator (HDI) were both associated with lower waist circumference and lower body fat in a sample of children similar to those included in the current study. The DQI was also associated with lower BMI after controlling for confounding factors such as PA and overall energy density [
37]. The Mediterranean diet KIDMED index was also negatively associated with obesity status though physical activity was suggested as a mediator of the association [
38]. The KIDMED study differs from the current study in a number of important respects. The assessment of PA was more rigorous in the KIDMED study but the investigators relied on parent reported child height and weight. This group investigated BMI as a binary outcome, categorising overweight and obesity together. In the current study where we examined BMI as a continuous predictor and as a 3 level categorical variable, there is more variability in diet quality between obese and normal weight children than overweight and normal weight children. This suggests that it is best (sample size permitting) not to combine the overweight and obese categories, as associations with obesity may be missed. Similar to our study, Lazarou et al. [
38] found that maternal obesity was a strong predictor of childhood obesity. A possible explanation is that parent BMI may be regarded as a proxy measure of diet quality in childhood, reflecting shared eating environments.
A review of short tools to assess children aged 2–5 years’ dietary intake, found that short dietary tools to screen for obesogenic dietary behaviours could be judged useful [
21]. This review also stated that dietary assessment tools from which DQS are developed should include the five major food groups as well as foods deemed unhealthy in order to assess diet quality. The measure we used in the current study, although short, did include the five major food groups, i.e. fruit and vegetables, grains, meat and alternatives as well as measures on sugar sweetened beverages and snacks, and foods high in saturated fat, see Additional file
1.
In order to address our secondary aim we investigated the individual food components of the DQS to identify if any particular foods were driving the DQS. Some of the foods were negatively associated with child obesity as expected. However, foods such as crisps/ savoury snacks and biscuits/ doughnuts/ cakes/ pie/ chocolate were associated with lower levels of obesity, which is counter-intuitive. It may be possible that there is more variability between obese and non-obese children in terms of consumption of the healthy DQS food components. Alternatively, the unexpected findings in relation to some specific food items may reflect reporting bias. Furthermore a lack of variability between normal weight and obese children may be down to the dietary assessment method for example the tool did not define the portion size or whether ‘more than once’ was twice, three or many more times. Finally, it may be that both normal weight and obese children consume a high level of unhealthy snack foods.
The strengths of our research include that the DQS was developed on a large nationally representative sample of children and their families (
n = 8568). All analyses took into account the complex survey design by applying survey weights. We controlled for some important confounders, though residual confounding is possible. The dietary assessment tool was completed with a low level of missing data for individual items, suggesting that this tool is easily administered with low respondent burden. The underlying parent reported FFQ was similar to the child reported FFQ and this may suggest that parents can provide useful data on their child’s diet. Parents as proxy reporters of their child’s diet has been suggested as guideline to adhere to for children less than 10 years old [
39].
Our research also had limitations. Causal inference cannot be implied, given the cross-sectional nature of the study. There is a possibility that a child’s weight status may have introduced response bias from parents. The underlying short FFQ was not tested for validity or reliability. However there are limitations in assessing diets in any population regardless of the assessment technique as detailed by Magarey et al., 2011 [
39]. It should also be noted that there is no gold standard method to measure diet or diet quality. All studies of human food intake, under natural conditions reveal a substantial degree of day to day variability [
40] and misreporting [
34]. Similarly techniques to assess reliability or validity of various short dietary assessment tools also have limitations as described by Bell et al. (2013), in particular the over-reliance on correlations coefficients [
21]. An Australian short dietary assessment tool that tested validity and reliability was conducted [
41] however not all core food groups were assessed. Future research studies that validate DQS’ against more detailed dietary assessments, taking day to day variability, energy misreporting and overall energy intake into account, together with validating it against biological biomarkers, may be the best approximation of diet quality [
42]. To our knowledge no such validated DQS, for use in children, focusing on health outcomes, developed from short dietary assessments, exist.
Simple tools to assess diet quality and identify children who follow unhealthy dietary patterns are important for developing appropriate childhood obesity interventions [
43]. As diet patterns [
44] and obesity [
4] can track into adulthood, promoting healthy eating patterns in childhood may be a useful preventative strategy. Not all studies can afford to assess children’s diet using dietary recalls, food diaries or long FFQ’s. The short FFQ used in this research was easy to understand and cost effective in terms of expertise needed by researchers to both collect and enter the data. This study adds to the emerging area on whole diet analysis and its association with childhood obesity. Research in this area to date is largely based on in-depth scores that, prior to construction, require a more detailed picture of the child’s diet based on methods such as food diaries [
37], 24 h dietary recall [
45] and 132 item FFQ [
36]. These methods although preferred are not always practical to use in large survey designs. Some research using shorter tools are emerging [
21]. However, these instruments, like the DQS used in the current study, need more comprehensive reliability and validity testing.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
CPP is the corresponding author who contributed to the development of the research question, conducted data analysis, and write up of this manuscript. EK is a contributing author who provided help with formulating the research question and co-wrote the manuscript and revised and approved the final draft of the manuscript. APF is a contributing author who provided statistical support and revised and approved final draft of the manuscript. RL is a contributing author who provided feedback on the coding of the underlying score, was involved in the design of the GUI study, revised and approved the final draft of the manuscript. IJP is a contributing author who provided feedback on the abstract, the layout of the overall paper and revised and approved the final draft of the manuscript. JMH is a contributing author, she provided the conception of the research question and additional support with all sections in the manuscript and revised and approved the final draft of the manuscript.