Survey methodology [22]

Since 1997, Sciensano, the Belgian Institute for Health, has conducted six successive Health Interview Surveys (HIS) in Belgium, with additional surveys carried out in 2001, 2004, 2008, 2013 and 2018. The HIS target population compromises all residents of Belgium, utilizing quarterly updates of the National Population Registry (NPR) as the sampling frame. The HIS is a cross-sectional interview survey at household level. Respondents are selected using a multistage sampling design that includes geographical stratification, selection of clusters within each stratum, selection of households within each cluster, and selection of individuals within each household.

The total sample size of the HIS is 10,000 individuals and this sample is representative for Belgium as well as the main regions. Additionally, within both the Flemish and Walloon Regions, a second level of stratification is applied at the provincial level. In each stratum, groups of 50 individuals (12.5 by quarter) are selected from a limited number of municipalities. Within each group, households are selected via a systematic sampling procedure, based on statistical sector, household size and the age of the household’s reference person. A maximum of four members per household are invited for participation. In households with more than four members, the reference person and their partner are always selected, along with two or three other members of the household.

To ensure the target number of interviews is achieved on time, matched substitution is applied. For each selected household, seven consecutive households from the ranked list used in systematic sampling are designated as substitutes. The selected household and its substitutes are collectively referred to as a cluster. The initial selected household and its substitutes are similar in terms of statistical sector, household size and the age group of the reference person. If the first household in the cluster does not participate, the next household in the cluster will be contacted and this process continues until the cluster is exhausted. If the entire cluster is exhausted without securing participation, a substitute cluster is activated. However, the initial and substitute clusters do not share common characteristics regarding the age of the reference person, household size, or statistical sector. The response rate for the survey years ranged from 55 to 61%. The surveys were approved by the ethical committee of Ghent University Hospital (approval number for the last survey cycle 2017/1454). Participants in the HIS provided their consent by taking part in the survey.

Interviewers

In addition to sample selection, the fieldwork for the HIS was managed in collaboration with Statbel, the Belgian statistical office, which specializes in conducting Face-to-Face (F-to-F) interviews. Statbel has a pool of experienced interviewers who also worked on the HIS. Interviewers were required to attend a one-day training session related to the HIS. The practical aspects of the training were provided by staff of Statbel, while the HIS-team from Sciensano covered the content of the questionnaires.

Questionnaire

The HIS consists of a F-to-F questionnaire and a paper based self-administered questionnaire (for participants aged 15+). Until 2008, a Paper and Pencil Interview (PAPI) was applied for the F-to-F interview, after which a Computer Assisted Personal Interview (CAPI) was adopted. The variables used in this study were all included in the F-to-F questionnaire, and the questions pertaining to these variables have remained consistent over time. The data collected are self-reported.

Childhood overweight and obesity

To assess childhood overweight and obesity, participants were asked the following questions about body weight and height: ‘How much do you weigh without clothes and shoes? (kg)’ and ‘How tall are you without shoes? (cm)’. For children under the age of 15, parents provided the responses. Based on this information, the BMI was calculated by dividing weight by the square of height (kg/m²). Extreme values considered as biologically implausible were excluded. Age- and sex-specific BMI cut-off points recommended by the International Obesity Task Force (IOFT) were then applied to classify individuals as (1) overweight (including obesity) and (2) obesity in the age group 2 to 17 years [23]. Hereafter, the term ‘overweight’ will refer to both overweight and obesity.

Parental educational level

Research indicates that parental education shows the strongest association with overweight and obesity compared to parental occupation and income [5, 8, 24, 25]. Additionally, education plays an important role in improving occupational opportunities and income level [24, 26]. Consequently, we have selected the variable ‘highest educational level of the household’ as the indicator of SES, assuming it primarily pertains to the parents. This educational level is assigned to all household members. It is defined according to the International Standard Classification of Education (ISCED), which comprises the following categories: (1) Primary education, (2) Lower secondary education, (3) Upper secondary education, (4) Post-secondary non-tertiary education, (5) Short-cycle tertiary education, (6) Bachelor’s or equivalent level, (7) Master’s or equivalent level and (8) Doctoral or equivalent level [27]. Eurostat proposes three main aggregates to present educational levels: low (categories 1 and 2), intermediate (categories 3 and 4) and high (categories 5–8) [28]. However, to avoid small cell counts, we have dichotomized this variable into low (categories 1–4) and high (categories 5–8) educational levels.

Study population

The sample comprised children aged 2 to 17 years with complete information on BMI and parental educational level. The covariate ‘sex’ was categorized into boys and girls, and the covariate ‘age’ was regrouped in four categories: 2–4 years, 5–9 years, 10–14 years and 15–17 years. These age categories allow for differentiation between toddlers/pre-schoolers, middle childhood, young teens, and teenagers. The total sample size was 10,084 individuals; the distribution by survey year was as follows: 1,584 for 1997, 1,845 for 2001, 1,643 for 2004, 1,477 for 2008 1,677 for 2013 and 1,858 for 2018.

Statistical analysis

The trends in childhood overweight and obesity were assessed by calculating weighted prevalence rates (with their corresponding 95% confidence intervals (95%CI)) stratified by survey year, for the total sample and disaggregated by parental educational level, sex and age of the children. To assess inequalities, the optimal approach is to rely on a set of inequality measures rather than on a single one because different measures show different aspects of the association between health and social status. For this end, we estimated the following two measures:

The prevalence difference, which indicates the absolute difference in the prevalence between low and the high parental education levels.

The odds ratio (OR), which indicates the relative difference between low and high parental education levels.

Both measures, the prevalence difference and the OR, offer insights about the polarization between parental educational levels [32]. They complement each other, as relative inequality measures allow to monitor inequalities by focusing on improving the health of disadvantaged groups, while absolute inequality measures give information on the absolute rate of childhood overweight and obesity within each group, thus allowing to judge the overall burden on public health [29].

Logistic regression models, stratified by survey year, were used to assess the OR and their 95%CI. Overweight or obesity was the dependent variable and parental educational level was the independent variable (with high educational level as reference category). Children’s age group and sex served as covariates. Significant differences were determined based on the p-value (< 0.05). Stratified analyses were deemed justified as the interaction between education level ad survey year showed significance within the general model for overweight, however not for obesity. The SURVEYLOGISTIC procedure was used taking into account the complex survey design (clustering and stratification) and the survey weights for estimating national results. All statistical analyses were performed with SAS 9.4.