This study investigated the clustering of parenting practices across the dietary and activity domain. We also examined whether these clusters are associated with child- and parent-related factors, and with child dietary and activity behaviours. As hypothesised, we found evidence for clustering within the dietary domain (e.g. clustering of SSB- and snack-related parenting practices) and within the activity domain (e.g. clustering of screen time rules and sports rules), which is in line with the few studies that reported on interdependencies between diet-related parenting practices [22, 23] and between activity-related parenting practices . A new finding is that parenting practices cluster across domains: four out of five clusters included both diet- and activity-related parenting practices. In addition, parenting practices cluster on the type of home enviroment: two clusters represented the physical home environment (‘high visibility and accessibility of screens and unhealthy food’ and ‘low availability of unhealthy food’), one represented the policital home environment (the ‘diet- and activity-related rules’ cluster) and the two parental modelling clusters represented the socio-cultural home environment. These new findings are very relevant in terms of broadening the scientific knowledge base on the topic of parenting practices.
In the present study, parental modelling was assessed in two ways: using role modelling scales of the HES  and parent’s own behaviour. The diet- and activity-related positive modelling cluster (cluster 4) included two parental role modelling scales. They referred to parental healthy eating and sedentary behaviour that was directly observed by the child (see example items in Table 1). This might imply the assessment of a more conscious way of parenting (a parenting practice) than when parental modelling is assessed by a parent’s own behaviour.
The diet- and activity-related positive modelling cluster (cluster 4) was more likely to be found in parents of heavier children who are lighter themselves, and express more behavioural control and less psychological control. This suggests that this might be a parental strategy in response to their child’s higher weight, particularly in normal weight parents. Similarly, diet- and activity-related positive modelling may be a stable parental strategy, reflecting normal weight parents’ own way of living , based on health beliefs. Finally, it may not be a parental strategy aimed at healthy dietary and activity behaviour in children, but rather a more unconscious way of parenting based on, for example, habits formed in early life. Similarly, the ‘diet- and activity-related rules’ cluster (cluster 2) might be a parental strategy based on health beliefs, but rule setting in the dietary and activity domain could also be part of a broader parental context of rule setting, based on, for example, parenting beliefs of strictness and involvement. This is supported by the finding that cluster 2 was positively related to behavioural control, which is an indicator of parental involvement.
There is evidence that parental education level indicates a broader parental context in which parenting practices operate [7, 63]. A non-supportive parental context might be reflected in cluster 1, the unhealthy cluster of making screens and unhealthy food visible and accessible at home, which was more likely to be found in low-educated parents, but also in minority groups, parents with a higher BMI and parents who use more psychological control (all found to be associated with a higher child weight and/or unhealthy lifestyle (e.g., [15, 56, 64]). In contrast, healthy clusters are generally more likely to be found in high(er)-educated parents. These findings are consistent with the well-established relationship between socioeconomic position and health, stating that the socioeconomically better-off do better on most measures of health status . Our findings also suggest that low-educated parents are an important target group for intervention development aimed at improving clustered parenting practices. However, because of the explorative nature of our study, the results cannot yet be translated into far reaching implications for public health. Before interventions can be developed, more studies are needed to elucidate how clusters of parenting practices arise (e.g. whether execution of parenting practices is a deliberate or a more unconscious process, whether parents adapt their practices or not and based on which indicators) and how they can be influenced, especially in low-educated parents. Apart from individual factors (e.g. a lack of knowledge and skills about parenting or a lack of health consciousness), exploring the social context of low-educated parents may elucidate why they have less-favourable parenting practices than high-educated parents. Ways in which the social context of low-educated parents can place constraints on their individual choices is by shaping social norms and by providing less opportunities to engage in healthy behaviours. This may influence their own health behaviour , but also their health-related parenting practices. For example, group norms may ensure that low-educated parents pursue other values than health values, and because of neighbourhood safety problems, they may not encourage their children to play outside. To better understand parenting practices in low-educated parents, future studies should explore the influence of the social context.
To indicate the magnitude of their relevance, we examined whether the five clusters were related to child dietary and activity behaviour. We found that the separate clusters were related to both child dietary behaviour and child activity behaviour and, overall, in the hypothesised direction: the ‘high visibility and accessibility of screens and unhealthy food’ cluster was positively related to obesity-inducing behaviour (i.c. child snack intake, SSB intake and screen time) and negatively to obesity-reducing behaviour (i.c. child fruit intake), while the remaining healthy clusters were negatively related to obesity-inducing behaviour and positively to obesity-reducing behaviour. The strongest associations were found in the positive modelling clusters. Diet- and activity-related positive modelling was found to have the strongest associations with child snack intake, SSB intake, active transporting to school, outdoor playing and screen time, while positive modelling on sports and fruit was strongest related to child fruit intake and child sports participation. This underlines the potential of a clustered approach of parental modelling in the dietary and activity domain as a parental strategy to (subtly) improve children’s dietary and activity behaviour. However, in low-educated parents this implies changing their own behaviour, which may be harder to accomplish than, for example, introducing parental rules in the dietary and activity domain. As the diet- and activity-related rules cluster was positively related to cluster 4, setting rules might eventually be an indirect way to change parental role modelling in a positive way.
Our study has the strength of combining diet- and activity-related parenting practices, higher-order parental factors and child dietary and activity behaviours in one study, which is exceptional in this field of research. In addition, our clustering approach, which is new in studies on parenting practices, seems to have potential as a starting point for interventions to assist parents in changing their child’s dietary and activity behaviour. Such interventions could be more efficient because of the synergic effect of a clustered approach. Nevertheless, some limitations should be mentioned. First, diet- and activity-related parenting practices were reported by the primary caregiver (mostly the mother), while research shows that, for example, for child PA paternal and not maternal role modelling is the main determinant . Future studies should (ideally) include both parents to examine whether fathers and mothers have a differential influence on child dietary and activity behaviour. Second, there was low variability in responses for some parenting practices, e.g. fruit availability and accessibility, which might explain why these parenting practices are not part of a cluster. However, this could also be explained by analytical choices, namely choosing a cut-off point for component loadings of 0.4. Although this is in line with recommendations , cut-off points in previous studies ranged from 0.2 to 0.6 . If, for example, a cut-off point of 0.3 had been used in our study, fruit availability, fruit accessibility as well as fruit rules would have been included in the positive modelling on sports and fruit cluster. Third, Cronbach’s alpha values of some of our parenting practices scales were relatively low. Although a Cronbach’s alpha ≥.6 is generally considered acceptable , some authors advocate different cut-off points. Finally, child dietary and activity behaviours were proxy reports of primary caregivers, which may evoke social desirability bias and lead to overestimation of obesity-reducing behaviours and underestemiantion of obesity-inducing behaviours [70–72]. In addition, child activity behaviours were reported in days per week which may not accurately reflect behaviour duration or energy expenditure, especially for outdoor playing and screen time.