Examining the association between parenting and psychosomatic problems: self-esteem as a mediator across ages in early adolescence

Abstract Supportive parenting is known to protect against psychosomatic manifestations of distress, yet the mechanisms through which this association operates are less clear. The present study evaluates children’s self-esteem as a mediator, partially explaining the association between parent–child relationship quality and psychosomatic problems from late childhood to mid-adolescence. Results from a large, nationally-representative Canadian sample indicated that self-esteem partially mediated these associations, and self-esteem accounted for more shared variance between parent–child relationship quality and psychosomatic problems among younger children. Among older children, shared variance with self-esteem explained a smaller portion of this association, but the remaining direct effect of parent–child relationship on psychosomatic problems was larger. These findings suggest that supportive parents protect against psychosomatic problems, and that they do so by promoting self-esteem for younger children and through other mechanisms as children age. Results are discussed in the context of attachment theory and age-related trends in self-esteem development.


Introduction
The importance of parenting for healthy child development is well documented, and the mechanisms behind this association are numerous and complex. Parents foster healthy development both directly, through emotional support and assistance with problem solving, and indirectly, by enabling the child to develop intrinsic resources such as self-esteem. The effects of parenting behavior are most notable during sensitive periods of development, at which time the child is particularly susceptible to environmental influences (e.g. Feldman, 2015;Whittle et al., 2014). These sensitive periods include infancy, during which attachment representations are built through parent-child interactions (Ainsworth, 1979;Bowlby, 1969;Fraley, 2002), and adolescence, when parents continue to provide a secure base from which the teenager can safely explore independence (Allen et al., 2003). Adolescents who lack this secure base are not only left to navigate an increasingly complex and stressful social world without this 'safety net' , but also tend to believe that other close relationships cannot be relied upon for support (e.g. Pascuzzo, Cyr, & Moss, 2013;Vaughn et al., 2016;Waters, Ruiz, & Roisman, 2017). The relative importance of both self-esteem and support from parents may change as children age into adolescence due to other developmental processes unfolding at this time (e.g. social role change, increased demands, and drive for independence; Baldwin & Hoffmann, 2002;Birkeland et al., 2012;Marsh, Trautwein, Lüdtke, Köller, & Baumert, 2006;Zimmerman, Copeland, Shope, & Dielman, 1997). Indeed, research with older adolescent participants has documented a decreasing effect of parental support on emotional well-being, perhaps reflecting increased reliance on peer support and personal resources such as self-esteem (e.g. Helsen & Vollebergh, 2000;Laible, Carlo, & Raffaelli, 2000;Meeus, Oosterwegel, & Vollebergh, 2002). These age-related trends have not been assessed with a younger sample of adolescents. It is possible that, as with older children, the influence of parents will decrease as children age into the teenage years due to the increasing importance of other influences on well-being (e.g. academic performance, peer relationships, and dating; Helsen & Vollebergh, 2000;Scholte, van Lieshout, & van Aken, 2001). Although a longitudinal approach to this question would be ideal, the lack of existent research on this topic from a developmental perspective means that a cross-sectional approach constitutes a valuable addition to the literature.
The present study aims to evaluate whether self-esteem partially accounts for the association between parent-child relationship quality and psychosomatic problems in the pre-to mid-adolescent period, and whether the relative strength of these influences on psychosomatic problems varies by age. To address this question, a series of moderation and moderated mediation analyses were conducted on cross-sectional data. As depicted in Figure 1, it is hypothesized that: (1) higher parent-child relationship quality will be associated with fewer psychosomatic problems at all ages across pre-, early, and mid-adolescence; (2) this association will be partially accounted for by shared variance with the child's self-esteem (partial mediation), specifically, higher parent-child relationship quality will be associated with higher self-esteem, which in turn will be associated with fewer psychosomatic problems; (3) the association between parent-child relationship quality and psychosomatic problems will be moderated by age, such that the effect size will be smaller among older children; and (4) the association between self-esteem and parent-child relationship quality will be moderated by age, such that the effect size will be smaller among older children.

Participants and procedure
This study used the 2014 Canadian Health Behaviour in School-Aged Children (HBSC) data-set (Freeman, King, & Pickett, 2016). This nationally representative sample consists of 29,784 students in grades 6-10 across 377 schools. Only students who met the following criteria were retained for analysis: 11-16.5 years of age, reported gender, and did not omit more than one item within each scale (Psychosomatic Problems, Parent Relationship Quality, and Self-Esteem). After removing participants who did not meet these criteria, the final sample consisted of N = 25,960 students (51.9% female, M age = 14.01 years). Participants who were excluded due to missing data were younger (t(29,047)= 9.158, p < .001, M age = 13.76) and more likely to be male (χ 2 (1) = 96.71, p < .001, 57.5% male). Questionnaires were completed anonymously during class time. They consisted of 105 items asking about school and community environments, physical and mental health, and interpersonal relationships. Additional information about the HBSC questionnaire design, ethics clearance, sample selection, and demographics is available from the Public Health Agency of Canada (Freeman et al., 2016).

Measures
Scores for each of the measures described below were calculated by computing the mean of all items, and then standardized by converting to z scores. Standardized values were used so that beta weights within the mediation analysis would be interpretable as effect sizes (Hayes, 2013) while preserving the distribution of scores and pattern of significant associations (Hayes & Rockwood, 2016). Further, this approach is recommended for measurements lacking a true zero point, so paths are evaluated using values for individuals who are 'average' on extraneous variables; by using raw scores, this estimate would be erroneously based on a true zero point representing the absence of that construct (Hayes & Rockwood, 2016). In this case, a score of zero would not be possible on the measures of interest, nor would it represent the absence of the construct.

Parent-child relationship quality
The HBSC 2014 questionnaire includes several items assessing family relationships. The parent-child Relationship Quality measure was created by combining items from the Family Support scale (i.e. 'my family really tries to help me' , 'I get the emotional help and support I need from my family' , 'I can talk about my problems with my family' , 'my family is willing to help me make decisions') and the Parental Understanding and Trust section (i.e. 'my parents understand me' , 'my parents expect too much of me' , 'my parents trust me' , 'what my parents think of me is important'). The questionnaire was then reviewed for other items referencing parents, and one additional item was added: 'I get in a lot of arguments with my parents' . Items from the Family Communication section of the HBSC survey were not included (despite referencing parents), as exploratory analyses indicated insufficient shared variance and poor scale reliability when combined with the other items. The resulting parent-child Relationship Quality variable comprises nine items, with strong inter-item reliability (α = .89). Notably, these items do not distinguish between the adolescent's relationship with each parent respectively, but rather respondents independently interpret who is being referenced by the terms 'parents' and 'family' (detailed in Discussion section).

Self-esteem
This measure was based on the self-esteem variable created by Ma (2007) from HBSC 1998 data. Two notable diversions from Ma's scale are that the item 'I like myself' was not available in 2014 so it was excluded, and a Likert scale was used instead of the yes/no response format used in the 1998 survey (Ma, 2007). Students rated the following six statements on a 5-point Likert scale ranging from Strongly Agree to Strongly Disagree: 'I have confidence in myself' , 'I have trouble making decisions' , 'I am often sorry for the things I do' , 'I often wish I were someone else' , 'I would change how I look if I could' , and 'I have a hard time saying no' . This overall measure had acceptable inter-item reliability (α = .71). These items are consistent with the construct of self-esteem, defined in the literature as the degree to which an individual is satisfied with him/herself and believes they are a person of worth (e.g. Greene & Way, 2005).

Psychosomatic symptoms
The HBSC questionnaire included a Psychosomatic Symptoms scale, which asked participants to indicate the frequency with which they experience the following eight symptoms: headache, stomach ache, back ache, feeling low (depressed), irritability or bad temper, feeling nervous, difficulty getting to sleep, and feeling dizzy. HBSC researchers selected these items as common physical manifestations of emotional distress; however, it is not possible to conclusively determine the cause of these symptoms from the questionnaire data. Frequency was rated on a 5-point Likert scale with the following labels: About Every Day, More Than Once a Week, About Every Week, About Every Month, and Rarely or Never. Due to this scale's strong inter-item reliability (α = .84) and previous research documenting frequent somatic presentations of psychological distress in children and adolescents (e.g. Eminson, 2007;Murberg & Bru, 2004;Sweeting et al., 2007), emotional and physical symptoms were treated as a single measure, consistent with the approach taken by the HBSC research team (Freeman et al., 2016).

Descriptive statistics and exploratory analysis
All variables were standardized and subsequent results represent analyses conducted on z-scores. The expected pattern of association was observed across constructs, with higher Parent Relationship Quality associated with higher Self-Esteem and lower rates of Psychosomatic Problems. Parent Relationship Quality and Self-Esteem were lower among older participants, who also reported higher rates of Psychosomatic Problems (see Table 1 for associations with Age). A gender difference emerged across all variables of interest with small to medium effect sizes, wherein males reported fewer Psychosomatic Problems (F(1,25 958) = 1696.9, p < .001, η 2 = .01), higher Parent Relationship Quality F(1,25 958) = 181.63, p < .001, η 2 = .06), and higher Self-Esteem F(1,25 958) = 1595.24, p < .001, η 2 = .06), indicating the importance of controlling for gender in the present analyses (discussed below).

Self-esteem and psychosomatic problems: associations with parent-child relationships across ages (moderation)
The impact of age on the strength of the association between parent-child Relationship Quality and both Self-Esteem and Psychosomatic Problems (respectively) was assessed using the Process macro for SPSS, Model 1 (Hayes, 2013). In each case, R 2 change was small but significant at p < .001. Consistent with hypotheses, the association between parent-child Relationship Quality and Self-Esteem was weaker among older children compared to younger children (See Figure 2 and Table 2). Contrary to hypotheses, older children demonstrated a stronger association between parent-child Relationship Quality and Psychosomatic Problems relative to younger children (See Figure 3 and Table 2).

Self esteem as a mediator across ages (moderated mediation)
Moderated mediation was evaluated using the Process macro for SPSS, model 8 (Hayes, 2013). Gender was expected to exert a main effect on each variable, but not to moderate associations between variables, and was therefore entered as a covariate. To confirm that this approach was appropriate, the model was re-run with males and females separately, and the same pattern of results emerged across genders (See Supplementary Material). All pathways depicted in the model are significant at p < .001 (Figure 4), and the confidence interval for the index of moderated mediation (CI: .01-.02) did not overlap zero, indicating statistical significance. The analysis indicated that Self-Esteem partially mediates the association between parent-child Relationship Quality and Psychosomatic Problems, moderated by Age (Figure 4 and Table 2). The Psychosomatic Problems variable includes both emotional and physical items, so the two alternative models were re-run with these items separately: one model including only emotional items (i.e. feeling low (depressed), irritability or bad temper, and feeling nervous) and

Discussion
The current results support the hypothesis that a high quality parent-child relationship protects against psychosomatic problems, in part through promotion of high self-esteem. This pattern of associations was significant at all ages from pre-to mid-adolescence; however, effect sizes in the model varied as a function of age. Self-esteem accounted for a smaller portion of this association among older children, and older children showed a stronger association between parent-child relationship quality and psychosomatic problems relative to younger children (contrary to the hypothesized direction). Together, these results suggest that although children's self-esteem becomes increasingly independent from their family context with age, they also rely more heavily on their parents to support well-being as they age. These findings are consistent with principles of attachment theory, which posit that the child develops a sense of self-worth out of the relationship with the primary caregiver (Ainsworth, 1979;Bowlby, 1980;Fraley, 2002), which then exerts an ongoing influence on mental health throughout development (Sroufe, 2005). Although attachment theory and empirical evidence support this conceptualization (Ainsworth et al., 1974;Bowlby, 1982;Metalsky et al., 1993;Sowislo & Orth, 2012), the present data  is cross-sectional and therefore cannot establish causal directionality. It is possible that chronic and severe psychosomatic problems and corresponding negative mood states may impair self-esteem via affect-induced cognitive biases and social impairment (Christodoulou & Burke, 2016;Joiner, 2000;Shahar & Davidson, 2003) and also strain the parent-child relationship. Results indicated that parental support protects against psychosomatic problems even after accounting for self-esteem. This pattern demonstrates the increasingly important role for parents during the early-and mid-adolescent period.
Above and beyond fostering self-esteem, supportive parents may provide strategies for emotion regulation, a secure base from which to explore independence, and the opportunity for 'fun' experiences that induce positive mood states (e.g. Allen et al., 2003;Buckholdt, Parra, & Jobe-Shields, 2014;Wilson et al., 2014). This association is probably reciprocal; children with lower rates of psychosomatic distress are more likely to elicit positive interactions with their parents (Hummel, Kiel, & Zvirblyte, 2016;Kochanska, Kim, & Boldt, 2015). Children who experience frequent negative mood states (which may manifest  physiologically or emotionally) are more challenging to parent, and therefore elicit more negative parent-child interactions (Brooker et al., 2015;Edwards et al., 2017;Klein et al., 2016). The moderating effect of age underscores the importance of investigating these constructs developmentally. A reduction in the strength of the indirect effect through self-esteem was expected due to additional, extra-familial influences on self-esteem (e.g. academics, peers, body-image) which become more salient during adolescence (Baldwin & Hoffmann, 2002;Birkeland et al., 2012;Hay & Ashman, 2003). Note that this pattern does not reflect a decrease in the association between self-esteem and psychosomatic problems, but that the portion of shared variance between parent-child relationship quality and psychosomatic problems that can be attributed to self-esteem declines with age. Documented associations between self-esteem and psychosomatic symptoms (e.g. Garaigordobil et al., 2005;Sweeting et al., 2007) suggest the presence of additional shared variance separate from the parent-child relationship. Thus, relationships with parents become increasingly important as children develop into adolescence, despite the fact that self-esteem becomes less tied to familial relationships at this time.
The age-related increase in the influence of parents on physical and emotional well-being may be attributable to the age range in the present sample, which spanned late childhood to mid-adolescence. Previous studies documenting an age-related decline in the impact of parent-child relationships on well-being have done so with older samples (e.g. Helsen & Vollebergh, 2000;Laible et al., 2000;Meeus et al., 2002), suggesting that reliance on support from parents may peak in mid-adolescence and then decline later in development. This pattern may be partly attributable to internal and external stressors faced during the first half of adolescence. Prior studies with this age range have found decreases in self-esteem during this period (Ogihara, Uchida, & Kusumi, 2016;Robins & Trzesniewski, 2005), a trend which was replicated in this sample. It is possible that, in the face of declining self-esteem, children with a positive relationship with their parents use this source of support to partially buffer against corresponding emotional distress. When relationships with parents are poor, adolescents face this age-related reduction in self-esteem without effective support from parents, and experience adverse effects in the form of psychosomatic problems.
This conceptualization accounts for the moderation of age on both pathways: the decreasing indirect effect of self-esteem (which declines with age regardless of familial relationship quality, resulting in decreased shared variance) and increasing direct effect of parent-child relationship quality on psychosomatic problems (possibly as a compensatory mechanism for the effects of declining self-esteem). The increasing direct effect of parent-child relationship quality with age is consistent with the sensitive period conceptualization of adolescence (e.g. Blakemore & Mills, 2014;Fuhrmann et al., 2015). As children transition into adolescence, environmental effects (in this case, relationship with parents) become more influential for well-being. Longitudinal research suggests that the effects of environmental influences during adolescence are likely to persist into adulthood (e.g. Blakemore & Mills, 2014;Raphael, 2013;Sawyer et al., 2012;Weinfield, Sroufe, & Egeland, 2000) via mechanisms such as HPA axis development, cognitive styles, and regulatory skills (Fuhrmann et al., 2015;Romeo, 2010;Whittle et al., 2014;Zahn-Waxler, Klimes-Dougan, & Slattery, 2000).
This conceptualization is also consistent with the limited empirical evidence for this pattern of associations among older children. The present study replicates and extends findings by Wilkinson (2004), who documented an association between parental support and depression, partially mediated by self-esteem. The current results replicated these findings using a younger sample, with a stronger residual direct effect of parental support. A developmental trend was identified wherein this direct effect was stronger among older participants, whereas the mediating effect of self-esteem becomes slightly weaker with age as self-evaluation is less tied to the relationship with one's parents. Further, this model was extended to psychosomatic problems, a common manifestation of psychological distress among youth.
These findings must be interpreted in the context of study limitations, such as reliance on self-report and cross-sectional data. The HBSC questionnaire is not able to distinguish medically unexplained symptoms (i.e. psychosomatic problems) from those symptoms attributable to physiological conditions (e.g. lactose intolerance, scoliosis). Further, it is not possible to distinguish the unique ongoing contribution of early-life experience with the attachment figure from the ongoing effect of parent-child interactions over time, a distinction emphasized in the attachment literature (e.g. Fraley, 2002;Weinfield et al., 2000). Future research may extend these findings using longitudinal data. Comparison across family structure and relative impact of mothers and fathers as a function of child's gender and age is another promising direction as existent research has explored the role of parent and child gender in these relationships (e.g. Garside & Klimes-Dougan, 2002;Merrill, Gallo, & Fivush, 2015;Updegraff, Delgado, & Wheeler, 2009). Although participant gender was entered as a covariate, unfortunately it was not possible to differentiate this information in this study, as students responded to questions about their parents. It is also unknown whether self-esteem as a mechanism across ages operates the same way in other cultures. For example, in collectivist cultures, a strong ongoing impact of family relationships on self-esteem may occur, manifesting as greater shared variance and a larger indirect effect on psychosomatic problems. Finally, the cognitive mechanisms underlying the effects of self-esteem and family support (e.g. negative self-schemas with concurrent positive relationship-schemas) may further illustrate these influences.

Conclusions
Results from the 2014 HBSC survey suggest that supportive parents protect against psychosomatic problems in part by facilitating the development of self-esteem. These relationships are particularly important for children in early and mid-adolescence, at which time self-esteem declines and they rely more strongly on their parents for support. These findings highlight the importance of prioritizing parent-child relationships in this developmental period and the value of considering somatic expressions of psychological distress.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by Public Health Agency of Canada.

Notes on contributors
Amy C. O'Neill is a PhD candidate in Clinical Psychology at Queen's University and research interests are parent-child interactions throughout development, early intervention parenting skills programs.