Our cross-sectional study used data from the population-based ‘Water campaign’ intervention [21]. This controlled trial aims to assess the effects of a combined school- and community-based intervention on children’s SSB consumption. The Medical and Ethical Review Committee of the Erasmus Medical Centre issued a ‘declaration of no objection’ (i.e. formal waver) for this study (reference number MEC-2011-183). Four primary schools located in multi-ethnic, socially deprived neighbourhoods in Rotterdam, the Netherlands, were included in the 'Water campaign' intervention study. This resulted in a total of 1288 children aged 6–13 years (grades 3–8) who were invited to participate. Passive parental consent was obtained. Parents (and children) received an information brochure to notify and inform them about the intervention and study participation. The study was also announced by the school, via the school letter and through the teachers and by flyers which were visible throughout the school. Parents (and children) were free to refuse participation without giving any explanation. They could do so by informing one of the teachers at their school or one of the researchers when present at school. At all times, the researchers could be contacted by a special phone-number or e-mail, for instance to decline participation [21].
The questionnaire items were based on previously widely used questionnaires, mainly used in earlier Dutch studies [22, 23]. We used two questionnaires to assess habitual consumption: one version was completed by children in grades 6–8 and another was completed by the parents of children in grades 3–8. Children filled in their questionnaire at school in the presence of a researcher and their teacher. The parent questionnaire was to be completed at home by the main caregiver of the child, within a period of maximum two months.
To objectively record what children brought to school to consume during break-time we used observation forms. These forms have been frequently applied for these kinds of information gathering by the Youth Health Care in recent years. Observations (unobtrusive) at school for children in grades 3–8 were conducted by trained public health professionals.
For the present study, we used baseline data from children in grades 6–8 only (aged 9–13 years). Pairing of data from the child questionnaires with data from the parent questionnaires or with data from the observations generated a study population of 539 children.
Population characteristics
Socio-demographic characteristics were obtained from parent and/or child reports: the parent and child questionnaires included items on child’s gender, age, grade and ethnic background. Ethnic background was determined by country of birth of the parents according to definitions given by Statistics Netherlands [24]. The child’s ethnic background was defined as Dutch only if both parents had been born in the Netherlands; if one of the parents had been born in another country, ethnic background was defined according to that country; and if both parents had been born in different foreign countries, ethnic background was defined as the mother’s country of birth. Ethnic background was categorised as either Dutch, Surinamese/Antillean, Moroccan/Turkish or other/missing.
Gender, age and educational level of the caregiver were recorded. The caregiver’s highest educational level was categorised as either ‘high’; ‘mid-high’; ‘mid-low’; or ‘low’, based on standard Dutch cut-off points [25].
Trained personnel measured the child’s height and weight at baseline. Weight status was determined by calculating Body Mass Index in kg/m2 with height measured to the nearest 0.1 cm and weight measured to the nearest 0.2 kg, in light clothing or gym clothes, according to a national standardized protocol for Youth Health Care [26]. Children were categorised as being either ‘non-overweight’ or ‘overweight or obese’, based on cut-off points published by the International Obesity Task Force [27].
Data pairs from observations and child reports
Trained public health professionals observed and registered which food items children brought along for the 10:00 am break at school. During the same morning, children completed a questionnaire in which they were asked to recall the food(s) they brought to school to consume during break-time. In Additional file 1: S1 and S2, the observation form and the child’s questionnaire item are shown. Data from the observations and child reports were grouped into three categories (‘sandwiches’, ‘fruit’ and ‘snacks’) by two researchers independently of each other. Any inconsistencies were discussed and where necessary a third researcher was included in order to reach agreement. If food items did not fit into one of the categories, these items were coded as ‘missing’ (<5 %). A ‘nothing’ category was added for those children who had brought nothing to eat during break-time. The four categories were dichotomised into (0) ‘not brought along’, and (1) ‘yes, brought along’. Paired data (observed and child-reported) were included in the analyses (n = 407 pairs).
Data pairs from parent and child reports
Children and their parents completed parallel questionnaires regarding the child’s fruit, water and SSB intake. We assessed ‘daily consumption’ and ‘average amounts consumed’. Data collection took place over a period of two months, in April and May. Paired data (parent-reported and child-reported) were included in the analyses (n = 275 pairs).
Daily consumption of fruit was measured using the question 'Does your child/do you consume fruit on a daily basis?', with answer categories ‘no, not every day’ or ‘yes, every day’. Average amounts of fruit consumed was measured using the question ‘On a day your child eats/you eat fruit, how many pieces of fruit does your child/do you consume on average?’. Answer possibilities ranged from ‘half a piece of fruit’ to ‘two or more pieces of fruit’. Examples were given to assist respondents in determining the number of fruit pieces (e.g. tangerine as a half piece; an apple as one piece).
Daily consumption of water was assessed using the question ‘Does your child/do you drink water on a daily basis?’, with answer categories ‘no, not every day’ or ‘yes, every day’. Average amounts of water consumed was measured using the question ‘On a day your child drinks/you drink water, how many glasses of water does your child/do you consume on average?’. Answer possibilities ranged from ‘none’ to ‘five or more glasses of water’. The total water intake per day, converted to litres (for comparison with SSB), was calculated by multiplying the number of glasses by an estimated average volume of 200 ml.
Daily consumption of SSB was measured using the question ‘Does your child/do you drink SSB on a daily basis?’, with answer categories ‘no, not every day’ or ‘yes, every day’. Average amounts of SSB consumed was measured using the question ‘On a day your child drinks/you drink SSB, how many glasses (250 ml), cans (330 ml) or bottles (500 ml) does your child/do you consume on average a day?’. Answer categories ranged from ‘none’ to ‘five or more’. The total SSB intake per day, converted to litres, was calculated by adding up the volumes of the total number of glasses, cans and bottles that were consumed. Examples of SSB were provided, based on our definition of SSB: Beverages containing added sugar, sweetened dairy products (e.g. chocolate milk), fruit juice (e.g. apple juice), soft drinks (e.g. cola) and energy drinks (e.g. sport energy drinks). In Additional file 1: S3, an overview of the questionnaire items used to assess SSB intake are given.
Analysis
For the dichotomous variables in the observed-child data pairs and parent-child data pairs, we calculated overall level of agreement (% same quartile). Kappa was used as an effect-size measure for the level of agreement, ranging from ‘0’ (agreement as expected by chance) to ‘1’ (perfect agreement) [28]. Agreement strength was based on the following criteria: 0.00–0.20 = ‘poor’; 0.21–0.40 = ‘fair’; 0.41–0.60 = ‘moderate’; 0.61–0.80 = ‘good’; 0.81–1.00 = ‘very good’ [29].
To explore the relationship between the different measurements methods with regard to consumption, Intra-class Correlation Coefficients (ICC) were calculated for each of the analysed behaviours. This generated a measure of absolute agreement for the dichotomous variables. For the continuous variables in the parent-child data pairs, the calculated ICC was a measure of consistency. The ICC is a value that ranges between '0' and '1', with a higher ICC corresponding to a better correlation. The following ICC cut-points were used: 0.00–0.20 = ‘poor’; 0.21–0.40 = ‘fair’; 0.41–0.60 = ‘moderate’; 0.61–0.80 = ‘good’; 0.81–1.00 = ‘very good’ [30, 31]. The mean (SD) of the difference and the limits of agreement were also calculated using Paired T-tests and used for input for the Bland Altman plots [32].
The McNemar test was used to compare the child’s reports with that of the observed reports or the reports by parents (level of significance set at 5 %) [32].