Background
Identifying the primary determinants of children’s physical activity (PA), is an important public health issue [
1]. Undoubtedly, parents play a key role in the development of their children’s health behaviors. For instance, the Integrated Model of Physical Activity Parenting (IMPAP) describes how parents’ attributes and parenting practices influence their children’s PA attributes and outcomes [
2]. The literature to date has focused largely on two parent attributes: parental support for PA and parental PA (often described as parental modeling of PA). A recent meta-analysis [
3] found that parental support was moderately related to children’s PA (
r = .38). Parental PA, on the other hand, was a weak predictor of child PA (
r = .16). Several previous reviews have also reported mixed findings across studies for parental PA [
4,
5]. This is surprising considering parents are routinely encouraged to be active role models for their children by health professionals [
6].
Several potential reasons for the discrepancy across studies have been proposed including the over reliance on subjective measures of PA (i.e., parental-proxy or self-report questionnaires) which may bias results [
6,
7], differences according to the intensity of PA (e.g., moderate-to-vigorous PA [MVPA], total PA) [
8], and parents having less influence as children become older and gain independence [
3,
9]. Parental influence could also vary by within and between family factors such as parent and child gender, and socioeconomic status (SES).
In support of the importance of using objective measures, the meta-analysis by Yao and Rhodes found a trend towards studies using objective assessments of youths’ PA (
r = .24) having a stronger parent–child PA relationship compared to those using subjective measures (
r = .13) [
3]. Heterogneity was also present, particularly in the questionnaire-based studies, suggesting that questionnaires tend to introduce measurement error that attenuates effects. Comparing the parent–child PA relationship using objective and subjective measures in the same sample would provide evidence to test this assumption.
Several studies have examined the relationship between parents’ and their children’s objectively measured steps/day [
10‐
14], accelerometer-measured counts [
9,
15‐
17], light PA [
18], and/or MVPA [
9,
17,
19]. Most studies have reported a significant relationship with at least one parent, however many have also observed differences by parent or child gender. For example, Jacobi et al. [
10] observed a mother effect whereby mother’s steps (but not the father’s steps) were correlated to their offspring’s steps in a sample of French nuclear families. In contrast, significant relationships between steps/day in father-sons, father-daughters, and mother-sons were found in a representative sample of Canadian children and their parents, yet the mother-daughters PA relationship did not reach statistical significance (
p = .08) [
8]. Thus it is unclear from the existing literature whether the parent–child PA relationships differs by child or parent gender. Studies that formally test whether gender specific relationships are significantly different from one another could provide some clarity.
Total PA performed throughout the day, as well as higher intensity activity, are important for the health of young people [
20], and thus it is important to understand the correlates and determinants of both of these outcomes. Pedometers are a reliable and valid measure total ambulatory activity performed throughout the day [
21‐
23]. They are also affordable and accessible to practitioners and families, making the findings of pedometer studies easy to apply to real-life settings [
8]. As mentioned, several studies have assessed the parent–child PA relationship using pedometers. For example, 10 additional minutes of parental MVPA resulted in one additional minute of MVPA in 1267 children aged 5–6 years in the UK [
24]. Similarly, a 2500 step/day increase in parent’s steps resulted in 752–1143 step increase in a sample of Czech children aged 4–7 years [
11].
As illustrated in the IMPAP, the influence of parental variables on children’s PA can vary depending on different child and parent attributes. Examining moderators (e.g., education, income) is important because it can provide information about whom and under what circumstances specific interventions may be effective [
2,
25]. Thus, testing a number of potential moderators of the parent–child PA relationship could provide important insights into whether family-based interventions will be effective for everyone or more effective for specific subgroups. Disentangling potential moderators of this relationship could also shed some light on the mixed findings in the literature. SES may be an important moderator to consider as lower SES families often have less access to quality facilities and less time and resources to be active [
26‐
28]. For example, with 286 nuclear families including children aged 8–18 years, Jacobi et al. [
10] found significant mother-offspring step count correlations with employed mothers (ICC = .24) and non-significant correlations with unemployed mothers (ICC = .11), however the difference was not formally tested. Considering overweight/obese children and adults are often found to be less active than those who are non-overweight [
20,
29,
30], the body size of the children and parents could impact the parent–child PA relationship. It is also possible that children may relate to and imitate their parent(s) if they are of a similar body size, indicating a “weight status homogeneity effect”.
The purpose of our study was to examine the relationship between parents’ and children’s PA in a sample of 7- to 8-year-olds and one of their parents. Research question 1 examined the relationship between pedometer-measured steps/day in parents and their children. Research question 2 examined whether gender (parent, child), gender homogeneity, weight status (parent, child), weight status homogeneity, and SES variables (parent education, household income, area-level SES) moderated this relationship. Research question 3 examined the parent–child PA relationship as measured by questionnaires. Consistent with Yao and Rhodes [
3], we hypothesized a small but positive relationship would exist between PA levels in parents and their children. We also hypothesized a stronger parent-child PA relationship with pedometer-determined compared to questionnaire-determined PA. Due to the inconsistencies in the existing literature, the moderation analyses were exploratory.
Discussion
The purpose of this study was to examine the relationship between pedometer-measured steps/day of parents and their children, and whether this relationship varied by gender (parent, child), gender homogeneity, weight status (parent, child), weight status homogeneity, parent education, household income, and area-level SES. We also assessed the parent–child PA relationship as measured by questionnaires. When PA was measured via pedometers, we observed a significant relationship between parents’ and children’s PA. Further, this relationship was stronger for higher income families and parents with a graduate degree, however the effects did not reach statistical significance. None of the other variables moderated this relationship. Using questionnaires, a relatively smaller parent–child PA relationship was found.
We found a 260 step/day increase in the children’s steps/day for every 1000 step/day increase in the parents’ steps/day, which was a small to medium sized effect (
r = .25,
rpartial = .24). Several studies in recent years have assessed the parent–child PA relationship using pedometers in both children and parents and all have observed significant findings with at least one parent [
8,
10‐
13]. For example, significant father-child and mother-child step relationships were observed in a slightly younger sample than ours (aged 4–7 years) in the Czech Republic, however the effect sizes were larger than in our sample [
11]. A 2500 step/day increase in mothers’ weekday/weekend day steps was related to an increase of 1143/928 weekday/weekend day steps in the children, and a 2500 step/day increase in fathers’ weekday/weekend day steps resulted in an increase of 903/753 weekday/weekend day step increase in the children. Combined these findings suggest that children and their parents accumulate similar amounts of ambulatory activity throughout the day. Thus, interventions designed to get children moving more throughout the day could be enhanced by including a parent component. Indeed a meta-analysis of family-based interventions found significant but small effects across 19 studies [
54].
Our study suggests that the parent–child PA relationship as measured by pedometers does not exhibit a same gender (i.e., gender homogeneity), child gender, or parent gender effect. The literature on gender specific parent–child PA relations using pedometers is quite mixed. For example, significant mother-offspring but not father-offspring step correlations were found in a sample of 8–18 year-old youth in France [
10]. Contrarily, fathers’ steps (but not mothers’ steps) was related to the steps of their children in a sample of Spanish children aged 8–9 years and their parents [
14]. In contrast, father-son, father-daughter, and mother-son step relationships were observed in a sample of 5–19 year-old Canadian youth and their parents [
8]. Yet the relationship between mothers and their daughter’s steps did not reach significance (
p = .08). Another study found father-daughter, father-son, mother-daughter, and mother-son step relationships on weekends, however the father’s steps was not related to his daughter’s steps on weekdays [
13]. A limitation of these studies is that they did not formally test whether the slopes in each groups were significantly different from one another, but rather subjectively compared the size of effects and the significance of the coefficient. Because we formally tested these interactions, our findings provide more conclusive evidence. A study by Jago and colleagues [
24] that assessed MVPA using accelerometers, illustrates the importance of formally testing interactions. From an inspection of the beta coefficients, the magnitude of effects of the father-son and father-daughter PA relationships appeared to be similar. For mothers, the effects appeared to be stronger for daughters than sons. A formal test however showed that the mother-child PA relationship was not significantly different for boys and girls.
In addition to testing differences by gender, and guided by the IMPAP, we also explored SES and weight status variables as potential moderators of the parent–child PA relationship. Correlations were higher for parents who had completed graduate school and for those who made > $80,000/year, but the results did not reach statistical significance. Weight status of the child and parent, and gender homogeneity were not effect modifiers. A few studies have examined SES moderators of the parent–child PA relationship. Jacobi and colleagues [
10] observed higher correlations between mother-offspring pedometer-determined PA for employed mothers compared to unemployed mothers however the differences were not empirically tested. In a sample of 5–6 year-olds, mothers’ and children’s accelerometer-measure MVPA was related regardless of the mothers’ education. However, fathers’ and children’s MVPA was only significant related for fathers with high education (i.e., had attended university) [
19]. To the contrary, Fridlund Dunton and colleagues [
55] found that children and parents whose household income was < $30,000/year performed more MVPA together (indicating PA co-participation) than those with a household income >$100,000/year. Similar to our study, Jago and colleagues [
9] did not find that parent BMI modified the parent–child PA relationship with 431 parent–child dyads. Taken together, our exploration of several potential moderators provides a unique contribution to the literature and suggests that interventions for families with children aged 7–8 years do not need to be individually tailored by SES, weight status, or gender. Future research should explore if other parental attributes or parenting styles are moderating factors.
A stronger child–parent PA relationship was found when PA was measured objectively using pedometers, compared to when measured subjectively using questionnaires. Similar to our study, a meta-analysis recently reported slightly higher effects when PA was measured using objective
vs subjective measures, yet the difference was not significant [
3]. One explanation is that the higher degree of measurement error that comes with using self- or proxy-report surveys attenuated the effects [
6]. Thus, the parent–child PA relationship as measured by pedometers is more precise estimate of this relationship. A second potential explanation is there is stronger familial aggregation for total PA (captured by pedometers) compared to sport and volitional activities which are often captured by questionnaires [
8]. In support of this, a significant mother-daughter PA relationship as found for accelerometer-measured counts/min in a sample of 5–12 year old daughters and their mothers, but not for MVPA. Another study however, found non-significant parent-child PA relationships for both counts/min and MVPA [
9]. A final consideration is that we measured parent leisure time MVPA during the last seven days and child usual PA. The slight differences in the measures may partially explain the smaller effects found in the questionnaire analysis. Regardless, our results reinforce the importance of using objective measures of PA in both parents and children when possible.
It is important to consider the potential reasons for why active parents have active children. Though this relationship is often described as parental “modeling” or observational learning [
4], it is likely due to many factors including genetics, co-activity, and parenting practices and beliefs. It is also possible that children influence their parents [
2,
56]. Further, parents could influence their children through several mechanisms including children’s enjoyment, motivation, perceived competence, and/or self-efficacy for PA [
2]. A greater understanding of these complex relationships is important for advancing theory in this area. Further, new advances in accelerometry such as GPS tracking and proximity tagging will be useful objective tools for teasing apart when parents and children are being active together, and hence when observational learning is likely occurring.
Strengths of the study include the objective measure of PA used with both parents and children, the large sample size, and the examination of several moderators. Multiple imputation of missing data allowed us to retain a large sample, reduce biases, and to test several potential moderators. But, some limitations should also be acknowledged. First, because the participants were volunteers, and only 39 % of the participants from the baseline phase of the SHAPEs study completed follow-up phase, a self-selection bias could exist. Indeed, the parents in our sample were more educated and higher income earners than the general population in the region. Second, the study is cross-sectional and non-experimental; therefore, we cannot assume that the parents’
caused their children to take more steps themselves. To date, no studies have examined the parent–child PA relationship over time using an objective measure of PA in both parents and children. As such, longitudinal and experimental studies are needed to establish temporal precedence and determine if there is evidence of a causal relationship. Third, despite pedometers providing an objective measure of PA, they have limitations. Because the pedometers were unsealed and parents’ recorded theirs and their children’s steps each evening, they could have unknowingly increased their activity or made a recording error. Pedometers are also not able to provide an indication of missing wear time, and thus we had to rely on the logbooks to determine if the monitor was not worn. Further, reactivity could have occurred whereby the parents and children increased their activity level in response to wearing an activity monitor. Several studies have examined the patterns of PA across the week to determine if there is an initial increase in PA and then a leveling off. No evidence of reactivity was found with sealed and unsealed pedometers in both children and adults [
36,
57‐
59]. Two studies have also shown no differences in step counts with sealed and unsealed pedometers in children and adults [
60,
61]. In adults, however, evidence of reactivity has been demonstrated when a covert condition (i.e., participants were unaware that their PA is being measured) was compared to a condition where participants were aware their PA was being monitored [
62‐
64]. Finally, we acknowledge the current recommendation for pedometers is seven days of monitoring [
65]. When this study was designed, there was no consensus on the number of days of monitoring required to measure habitual PA, and several studies have shown that four days of monitoring provide reliable estimates of habitual PA [
21].
Acknowledgements
We would to thank and acknowledge the contributions of Mildred Masimira, Ciara Kallal, Katelynn Theal, Megan Lanager, Lynne Wong, Annie Selzler, Eun-Young Lee, Jayantha Dassanayake, Riva Benditt, Valerie Carson, and Joreen Tamayo, Melissa Wong, Stephanie Voaklander to this project.