Agreement between parent and child report on parental practices regarding dietary, physical activity and sedentary behaviours: the ENERGY cross-sectional survey

The EuropeaN Energy balance Research to prevent excessive weight Gain among Youth (ENERGY) project aimed to develop and evaluate a theory-informed and evidence based multi-component school-based and family involved intervention program (http://​www.​projectenergy.​eu). [31]. In 2010 a school-based, parent-involved cross sectional study was carried out as part of the ENERGY project in eight European countries: Belgium, Greece, Hungary, Netherlands, Norway, Slovenia, Spain and Switzerland [32]. This cross-sectional study aimed to provide information regarding overweight, obesity and waist circumference in representative samples in these countries, as well as information regarding EBRBs and their personal, social and physical environmental correlates. A minimum sample of 1000 children aged 10 to 12 years per country and one parent for each child was required to detect a between country difference of 5% in obesity prevalence as statistically significant with a power of 90%. The sampling of schools was random, multi-staged and stratified by degree of urbanization [32]. In total, 202 schools were recruited and 7915 children and 6463 parents/caregivers filled in a questionnaire. Response rates of the children were in general high (≥64%) except for Hungary (33%), Norway (45%) and Spain (43%) [2]. The response rate of parents was also high (>60%) except for the Netherlands (41%). The survey in Switzerland was conducted later than in other participating countries and its results have been published elsewhere [33]. For the current analysis we excluded data from 38 caregivers other than mother or father who filled in the parental questionnaire, resulting in 6425 parent–child dyads.

The project adhered to the Helsinki Declaration and the conventions of the Council of Europe on human rights and biomedicine. Ethical clearance was obtained from the relevant ethical committees and ministries in all participating countries (a detailed list including names and affiliations is provided in Additional file 1: Table S5). In addition, research permission was obtained from local school authorities if necessary. Passive informed consent from the parents was allowed in the Netherlands [32]. In all remaining countries parents were asked for written consent for their child’s and their own participation.

The selection of EBRBs and correlates assessed in the questionnaires was based on literature reviews and secondary data analyses conducted within ENERGY [34, 35]. Both child and parent questionnaires were developed using items from validated European and country-specific questionnaires and if needed new items were added. Once created, the questionnaires were translated back and forward into the language of each participating country and were pretested in small samples in all participating countries (available in all languages at http://​www.​projectenergy.​eu).

Descriptive statistics were used to characterize parent–child dyads available for analysis. Factors from the child questionnaire that were associated with non-response of parents were identified using multilevel logistic regression including gender of the child, age and BMI, the living situation and whether the family spoke the national language at home, overall (level 1 country, level 2 schools) and for each country (level 1 schools) separately. To assess overall agreement between child and parent reports on parental practices we created an agreement score of all 14 practices ranging from 0 (no agreement at all) to 100 (agreement in all practices). An average agreement score was calculated for parent–child dyads that answered at least 10 out of 14 practices (6358 out of 6425; 99%). We calculated percentage agreement between child and parent report for each parental practice in total and for each of the participating countries separately. The established criteria for percentage agreement were “good to excellent” (≥75%), “moderate” (60%-74%) and “poor” (<60%) [38, 39]. Country specific percentage agreement was explicitly mentioned in the text if agreement between countries differed by more than 10%, a cut off arbitrarily chosen, but roughly distinguishing relevant from small irrelevant differences, to facilitate reading of the large Table 1. Furthermore we looked at the direction of disagreement (i.e. did children on average score higher or lower than parents) by calculating the mean difference in scores of parent and child reports. To report the size of disagreement we then calculated the root of the squared difference for each practice. In addition, we calculated weighted Cohen’s kappa coefficients to assess the agreement between child and parent report. Weighted Cohen’s kappa takes into account the degree of disagreement, i.e. that the disagreement is greater if categories are further apart (e.g. ‘completely agree’ vs. ‘completely disagree’ rather than ‘completely agree vs. ‘agree somewhat’). We assigned a linear set of weights. In case of items with 5 answer categories, this means that weights were 1, 0.75, 0.50, 0.25 and 0 when there is a difference of 0 (=total agreement) or 1, 2, 3, and 4 categories respectively. Cohen’s kappa coefficients were defined as following: <0.0 poor, 0.00-0.20 slight, 0.21- 0.40 fair, 0.41-0.60 moderate, 0.61-0.80 substantial and 0.81-1.00 almost perfect agreement [43]. We repeated the agreement analysis recoding the 5 answer categories into a 3-point scale, i.e. always or often versus sometimes versus not often or never.

Factors associated with the agreement score were assessed using univariate multilevel linear regression for the combined dataset (level 1: country, level 2: school) as well as for each country separately (level 1: school). The assumption of normally distributed residuals was checked and met. Next, a multilevel linear regression was conducted including all factors that were found to be significantly associated with agreement in univariate analysis. Stata version 11 (StataCorp, College Station, TX, USA) was used for all analyses. P-value < 0.05 was considered as statistically significant.

Table 2 presents the characteristic of the children and their parents. Of the children 23% were overweight or obese and 8% underweight. Country specific results are presented in supplemental tables. Table 2 also shows the results of the non-response analyse, showing that parents of boys, non-native speakers of the national language and parents who were single mothers or fathers and parents of children with higher BMI were more likely to be non-responders. Similar associations were observed across countries with a few exceptions that are shown in Additional file 3: Table S2.

Considerable variation in parental practices was observed across countries (Additional file 4: Table S3). For example 47% of children in Greece reported to never or less than once a week have breakfast with their parents while this was true for 13% of children in the Netherlands. In Swiss children 11% reported that their parents would allow them to often or always watch TV if they asked for it as compared to 46% in Hungarian children. Dutch and Swiss children (40% and 41%, respectively) reported the lowest parental encouragement to be physically active or to do sports; Slovenian children the highest encouragement (83%). Children in Belgium and the Netherlands more often reported soft drinks to be available at home than children from other countries.

The mean total agreement score, i.e. including all parental practices, between child and parent reports was 43% (Standard deviation = 14; Figure 1). Mean total agreement score was similar in all countries ranging from 41% in the Netherlands to 45% in Norway and Spain.

Agreement between child and parent reports varied by parental practice.

Using the recoded 3-point scales resulted in higher agreement between parent and child responses. For all parental practices the mean total agreement score reached 60% (SD = 15) and was similar in all countries ranging from 58% in the Netherlands and Slovenia to 63% in Belgium (Figure 1). Furthermore, agreement between parent and child report was higher for joint sport activities (59%), and to some extent for having breakfast (50%) and watching TV (47%) together when using the recoded 3-point scale compared to when using the original 5-point scale (Table 1).

In multilevel univariate analyses agreement was lower for boys, for younger children and for younger parents and parents with less years of education (Additional file 5: Table S4). In addition, agreement between child and parent report was lower for overweight children and for children whose parents perceived them to be either over- or underweight.

Our results indicate that agreement was significantly lower in boys (regression coefficient “b” = -1.84; 95% Confidence interval “CI” = -2.59 to -1.10; p-value < 0.0001) (Table 3). Further, agreement improved with increasing age of the child (b = 1.06; 95% CI -0.04 to 2.16 and 2.41; 95% CI 0.12 to 4.71 for age 12 and 13 years, respectively as compared to age 10 years; p-value = 0.013) and of the parent (b = 0.08 per year; 95% CI 0.00 to 0.15; p-value = 0.041). More than 14 years of parental education was associated with better agreement between child and parent reports as compared to fewer years of education (b = 2.08; 95% CI 1.28 to 2.88; p-value < 0.0001). Lower agreement was observed for single parents as compared to two parent households (b = -1.31; 95% CI -2.62 to -0.00; p-value = 0.050). Parents with a higher self-reported BMI (b = -0.11 per kg/m²; 95% CI -0.20 to -0.02; p-value = 0.013) and parents who perceived their child to be underweight showed lower agreement scores (b = -1.11; 95% CI -2.28 to 0.07; p-value = 0.039). In general, the directions of these associations were observed in all countries, although not always statistically significant. In Belgium, parents who perceived their child to be overweight agreed more with their child on parental practices (b = 3.81; 95% CI 0.39 to 7.22; p-value = 0.029). The direction of this association varied across countries. Furthermore, speaking the national language at home increased agreement between children and parents in Belgium and Slovenia (b = 4.35; 95% CI -0.05 to 8.75; p-value = 0.052 and b = 4.28; 95% CI 0.38 to 8.18; p-value = 0.032, respectively).

In general we found that parents and children responded differently to questions assessing parenting practices. This disagreement may be explained by several factors.

Firstly, parents and children may have different perspectives on their relationship and behaviours [5, 10, 21, 27]. Despite differences in methodologies and questionnaires, considerable disagreement between parents and children has been reported before. A study by van Assema et al. [27] reported low agreement for perceived availability and accessibility of snack, fruit and breakfast products. Barr-Anderson et al. found low levels of agreement between child and parent reports regarding family support for physical activity and parental limitations on child’s TV use (ranging from 25%-42%) and good agreement (≈70%) if child and parent responses were allowed to differ by one point in the 4-point scale [21]. These results suggest that future studies should take both perspectives into account.

A second explanation for the disagreement may be that parents are more likely to give socially desirable answers [18]. However, we did not find that parents consistently reported more favourable scores. On the contrary Barr-Anderson et al. reported that in their study children perceived their parents to be more supportive than parents rated themselves suggesting that parents may have viewed their small actions of support (encouraging words, providing rides) as part of their parental duties [21]. Similarly, joint child–parent activities may be perceived differently by children and parents. If social desirability was at stake we expect parents to report joint sports activities more frequent and watching TV together less frequent than their children. However, such a pattern was only observed in Switzerland. On the other hand, parents may also perceive watching TV together, i.e. social co-viewing, as a favourable behaviour as in this way they can better monitor what they children watch counteract undesirable effects of television programs and commercials [44]. Likewise, we do not know whether children in this study more often reported being allowed to consume soft drinks or fruit juice and to watch TV than their parents due to different perceptions or rather due to socially desirable answers of their parents. Parents may be aware of potential unfavourable effects of excessive soft drink consumption and TV viewing and realize that they should limit these behaviours and therefore provide a desirable answer. Children, on the other hand, may have given more honest answers, resulting in disagreement with their parent’s answer especially in children who drink more soft drinks and watch more TV.

Thirdly, a methodological explanation for low agreement, especially for some items, can be the low test-rest reliability and validity. Practices with low agreement i.e. engaging in physical activity or doing sports together, watching TV together, allowing the child to take soft drinks, fruit juices or watch TV also had lower test-retest reliability and validity (Additional file 2: Table S1) [38, 39], especially in the child questionnaire. A possible explanation is that these practices vary more on a daily basis and are thus less stable and clear. In addition, questions on joint parent–child activities as well as on availability of a TV in the bedroom were framed slightly different in the parental and child questionnaire, which may also partly explain the lower agreement found for these parental practices as compared to others. Another explanation may be the limitations in memory capacities and cognitive competencies in children [18‐20].

Finally, the sensitivity analyses with the recoded 3-point score for the parenting practices showed that this recoding strongly influenced the agreement between child and parent reports and reached moderate levels (60%). We assessed this because many studies (including studies based on the ENERGY data) recode original 5-point scales into 3-point scales for methodological reasons (e.g. skewness, empty cells). These 5-catogory variables cannot be used as continuous variables in analyses and are usually recoded into a lower number of categories to make analyses and interpretation easier. As the majority of parents and children tended to choose the same answer category, remaining cells in a 5 × 5 table of parent and child scores on parental practices were left with very low numbers, resulting in low kappa values. Kappa is affected by the prevalence of the finding under consideration and very low values of kappa may not necessarily reflect low rates of overall agreement [45]. Therefore we decided to present both, kappa values and percentage of agreement. This problem can be avoided by using two separate indexes of proportionate agreement: i.e. the observed proportion of positive agreement and the observed proportion of negative agreement. These two indexes are analogous to sensitivity and specificity for concordance in a diagnostic marker test [46].

Our findings show that agreement increases with children’s age. Cognitive competencies of children develop with age, i.e. the information process mechanisms that enable them to attend to, select, represent and act on information as well as working memory functions that support the mental representation of information [19]. The lower agreement found for boys and their parents in our study might be explained by an earlier start of psychosocial maturation in girls compared to boys [47]. A low capability of self-reflection and self-concept in adolescents have been associated with poor communication with their parents [48]. Likewise, in two previous studies, girls agreed more with their parents on the availability of specific fruits at home [49] and on the frequency of family meals than boys [24]. It may also be that girls are more inclined to give socially desirable answers than boys and therefore their answers may be more in agreement with their parents [50]. Agreement was higher in older parents. Older parents may have more routine as well as more clearly defined parental practices. In addition, the child who has older parents is more likely to have an older sibling who might remind him of the parental practices in their family. The observed higher agreement between parent and child reports with higher level of parental education corresponds to findings of previous studies on agreement between parent and child reports regarding parental practices [26, 51]. A higher BMI of parents was associated with lower agreement in our study. Parents with a higher BMI may use different parental practices and may be more prone to giving social desirable answers. In an observational study parents with a higher BMI applied more controlling feeding practices and maternal BMI was related to greater reported than observed restriction of their children’s intake [52]. A significant proportion of parents tend to underestimate their child’s weight [53] especially if children and parents themselves are overweight [54], which might be partly explained by the parental desire to not be judged as a bad parent. However, more research is needed to understand the mechanisms of how parental perception of weight status in their child influences agreement between child and parent report on parental practices. Our data suggests the existence of underlying cultural factors, which we have not assessed.

Agreement between child and parent reports was similar across countries despite a large variation in parental practices. This indicates that such agreement is rather stable and robust, and influenced by other factors than cultural differences or country differences in parenting practices between the included countries. Similarly, in a study comparing parent-reported and adolescent self-reported problems in 25 societies found that parent-adolescent item agreement varied widely across dyads but mean dyadic agreement was modest in every society [55].

As part of the ENERGY project [31] eight European countries participated in the cross-sectional survey [32] resulting in a large dataset reflecting different cultures. This is important since parental practices and the perception of these practices are likely to vary between cultures [30]. A set of variables potentially influencing agreement between child and parent reports, such as objectively assessed BMI of children and the parental perception of their child’s weight status, were available for analysis. Test-retest reliability was good for both parent and child questionnaires and construct validity was good and moderate to good for most but parental practices items in the parent and child questionnaire, respectively [38, 39]. In general, response rates for children were high (above 80% in all countries). Response rates in parents were generally high as well, except in Belgium (62%) and the Netherlands (44%), which may have biased the results, especially because non-response was related to the child’s BMI [32].

Questions on joint parent–child activities (i.e. having breakfast, participating in physical activity/doing sports and watching television) were framed slightly different for children and parents in that children were asked to report on the frequency of these activities they did together with their parents while parents were asked to report the frequency that either both parents or one parent did together with their child. This may partly explain the lower agreement for these practices. Not all constructs could be compared between children and their parents. For instance, no direct corresponding question of whether the parents have rules for each of the five EBRBs (soft drink and fruit juice consumption, having breakfast, taking part in physical activities/doing sports and watching TV) was asked to the parents. Further, no corresponding question was asked to the child on whether the parents negotiate with them about the five EBRBs. Further, we could not compare whether parents gave soft drinks or fruit juices to the child as a reward or to comfort the child and whether parents would prohibit the child to take part in his or her physical activity/sports sessions as a punishment since it was only included in the parent questionnaire.

In this ENERGY sample very few parents with a low education, i.e. less than 7 years, were included. Therefore, our findings on agreement cannot be generalized to parents with less than 7 years of education and their children. Finally, country specific analyses often lacked power and results therefore did not reach statistical significance.

Ideally, parental practices should be directly observed in parent -child interactions in their home environment, which of course is unfeasible in large samples due to many reasons including ethical concerns as well as financial and time constraints. A review of studies examining psychometric properties of available parental practice questionnaires would be useful in guiding the design of future questionnaires. Better understanding of the role of parental practices in weight and weight-related behaviours of their children is needed to improve obesity prevention programs. Qualitative research may contribute to understanding the mechanisms’ that link parental practices to EBRB and weight status of children [14] including parental perceptions of EBRBs of their children or the weight status of their child.