Journal of Happiness Studies

, Volume 17, Issue 5, pp 1847–1872

A Population-Based Study of Children’s Well-Being and Health: The Relative Importance of Social Relationships, Health-Related Activities, and Income

  • Anne M. Gadermann
  • Martin Guhn
  • Kimberly A. Schonert-Reichl
  • Shelley Hymel
  • Kimberly Thomson
  • Clyde Hertzman
Research Paper

DOI: 10.1007/s10902-015-9673-1

Cite this article as:
Gadermann, A.M., Guhn, M., Schonert-Reichl, K.A. et al. J Happiness Stud (2016) 17: 1847. doi:10.1007/s10902-015-9673-1

Abstract

This study investigated how various risk and protective factors interface with child health and well-being at the population level. Specifically, we examined the association of income, social-contextual variables, and indicators of health-related habits and activities to children’s life satisfaction and perceived overall health. Child data were collected via a self-report survey, the Middle Years Development Instrument, which was administered in three demographically diverse Canadian school districts to 5026 grade 4 students (83 % of the students had complete data and were included in the analyses). Multiple regression and mediation analyses were conducted to examine the joint associations of social relationships with adults and peers, nutrition and sleep habits, and after school sports activities with children’s satisfaction with life and perceived health. Results indicate that peer belonging and relationships with adults at home and school were the strongest predictors of life satisfaction. Furthermore, the (small) association between income and life satisfaction was mediated by social relationship variables. Child reports of perceived health were predicted by peer belonging, adult relationships (home, school, neighborhood), after-school team sports, and nutrition habits. The (small) association between income and health was mediated by social relationships and team sports participation. Findings are discussed in light of previous research on social determinants and socio-economic gradients of children’s health and life satisfaction.

Keywords

Life satisfaction Health Children Social relationships SES 

1 Introduction

There is an extensive body of literature on subjective well-being and its correlates and predictors in adults (Diener et al. 1999; Luhman et al. 2012), but far less research on subjective well-being in childhood (Gilligan and Huebner 2007; Holder 2012). Over the last decade, however, as strengths-based approaches, such as the positive youth development and developmental assets approach, have garnered increased research attention, a burgeoning theoretical and empirical literature has begun to emerge, delineating the developmental processes and mechanisms that promote thriving (Damon 2004; Lerner and Benson 2003). The focus of the positive youth development approach is on the developmental potential of children and youth to learn and thrive in the diverse contexts in which they live (Damon 2004; Oberle et al. 2011; Theokas et al. 2005). This emphasis on positive emotions, health, and well-being, and fostering resources, skills, and strengths reflects an important paradigm shift in psychology—one in which the focus is on the study of the positive aspects of human nature (Seligman and Csikszentmihalyi 2000). This complements decades of research that primarily focused on negative emotions and mental health disorders, based on a problem- and deficit-oriented approach to child and youth development.

At the same time, the child indicators movement has underscored the importance of measuring and monitoring children’s and adolescents’ well-being from their own perspective (Ben-Arieh 2005). Such an approach aligns with the UN Convention on the Rights of the Child, Article 12 (www.unicef.org/crc/), and emphasizes the notion that children and adolescents should be included as active research participants in studies examining their well-being (Ben-Arieh 2008). Moreover, this approach draws from a framework of children’s empowerment, which is in line with research that has shown that children would like to take an active role in research and be asked and heard on matters that impact them and their well-being (Ben-Arieh 2005). This is supported by studies demonstrating that 10-year-old children can report on their health and satisfaction with life in psychometrically reliable and valid ways (Gadermann et al. 2011, 2010; Riley 2004; Schonert-Reichl et al. 2013). Despite this focus on obtaining information from the children themselves, research has often relied on adult (e.g., parent or caregiver) reports of children’s well-being (Ben-Arieh 2012). One potential reason for this may be due to the relative paucity of child-oriented indicators of well-being (Ben-Arieh et al. 2001). In recent years, however, this situation has changed with the development of psychometrically sound measures for assessing children and adolescents’ subjective well-being (Gadermann et al., 2010, 2011; Gilligan and Huebner 2007; Huebner and Diener 2008; Seligson et al. 2005; see Gilman and Huebner 2000). In accord with the focus on self-reports of children’s well-being, the present study explored the association between children’s self-reported well-being and health, social connectedness, health- and well-being-related habits and activities, and tax filer information on income.

Looking at child well-being and health from a rights of the child perspective is, of course, not just a matter of measurement (i.e., self-report), but also a matter of equity (cf. Marmot et al. 2008). This issue has received increased attention over the past decades. Specifically, debates in public health have centered on social determinants of health and the notion that “the true upstream drivers of health inequities reside in the social, economic, and political environments” (p. 1, Blas and Kurup 2010). Accordingly, researchers and policy makers have pursued the question of how to effectively reduce inequities via social programs and policies. As will be discussed below, empirical research has consistently found socioeconomic gradients in health outcomes. However, it is not yet well understood via which ‘pathways’—mechanisms and processes—socioeconomic factors are associated with health outcomes (Russ et al. 2014; Braveman et al. 2011; Maggi et al. 2010). A better understanding of such pathways may help identify possibilities for deliberate interventions that may “improve health and reduce health disparities” (Braveman et al. 2011; Östlin et al. 2011).

1.1 Children’s Subjective Well-Being and Health

Health and subjective well-being in childhood and adolescence are associated with important later outcomes. The roots of many chronic mental and physical health issues are already well underway by the time children reach high school. Indeed, the median age of onset for anxiety and impulse control disorders is age 11 and half of all lifetime mental health disorders start by age 14 (Kessler et al. 2005). Moreover, subjective well-being among adolescents has been found to predict mental health problems the following year (Huebner et al. 2000) and to serve as a protective influence for later substance use in a 3-year longitudinal study (Wills et al. 1999). In terms of physical health, a recent Statistics Canada report indicated that a third of Canadian children ages 5–17 are considered overweight or obese (Roberts et al. 2012). Childhood physical fitness persists into adulthood, putting individuals at risk for further health problems such as cardiovascular and metabolic disease (Roberts et al. 2012; Shields et al. 2010).

1.2 Correlates of Children’s Subjective Well-Being

Similar to findings from research with adults (Diener and Diener 1996), the majority of children and adolescents report that they are satisfied with their lives across different cultures (Gilman and Huebner 2003); yet, a notable proportion of children perceive their lives negatively. For example, in a study with US high school students, 11 % chose the response options ‘terrible’, ‘unhappy,’ or ‘mostly dissatisfied’ to rate their satisfaction with life (Huebner et al. 2000). Researchers have examined the relation of life satisfaction to a range of variables in children and youth, and have found results similar to those found with adult populations, namely that the associations of demographic variables, such as gender, age, and ethnic background, to life satisfaction are weak (Gilman and Huebner 2003). Findings with regard to the association between parental socio-economic status (SES) and child life satisfaction have been inconsistent; with some studies reporting a significant association, favoring students from high SES families or higher relative SES position (e.g., Ash and Huebner 2001; Burton and Phipps 2008), whereas others found no differences (e.g., Huebner 1991), which may be due to varying measures of SES and sample composition. Personal characteristics, such as global self-esteem and self-efficacy, are strongly associated with student life satisfaction (Gilman and Huebner 2003). Positive daily experiences (e.g., interacting with friends) significantly predict student life satisfaction, even after controlling for self-concept (McCullough et al. 2000). Previous research also underscores the importance of close social relationships, particularly relationships with guardians/family members and peers, to children’s well-being (Chaplin 2009; Guhn et al. 2013, 2012; Holder and Coleman 2009; Uusitalo-Malmivaara and Lehto 2013). Furthermore, social capital, defined as the ‘connections among individuals—social networks and the norms of reciprocity and trustworthiness that arise from them’ (Putnam 2000, p. 19), has shown to be strongly associated with subjective well-being in adults (Helliwell and Putnam 2004). Similarly, research with Canadian youth aged 12–17 years old indicates that a sense of ‘belonging to the local community’ is strongly associated with life satisfaction (Burton and Phipps 2010).

In addition, involvement in hobbies and sports, success in school, and material possessions have been found to positively correlate with children’s well-being (Chaplin 2009; Holder et al. 2009; McHale et al. 2001; Uusitalo-Malmivaara 2012). In a nation-wide sample of over 80,000 Finnish adolescents (aged 14–16 years), Konu et al. (2002) found that well-being (1) was most strongly associated with social relationships with parents and peers, (2) was also strongly associated with health-related behaviors (exercise, healthy eating, sleep), and (3) had statistically significant but relatively smaller associations with socio-economic background variables (parental employment and education).

1.3 Correlates of Children’s Health

There is a large literature on global self-rated health in adult populations. The construct of self-rated health has been widely used in survey research, including large-scale international comparative studies (Bambra and Eikemo 2009). A consistent finding is that self-rated global health predicts subsequent chronic disease status (Shadbolt 1997) and mortality, after controlling for specific health status indicators (DeSalvo et al. 2006; Idler and Benyamini 1997). Also, self-rated health correlates with ‘objective’ measures of health status, such as disease status and blood markers (Wu et al. 2013), and a range of social, socio- economic, psychological, and health behavior related factors (Amstadter et al. 2010; Cano et al. 2003; Kawachi et al. 1999; Kunst et al. 2005; Operario et al. 2004).

For children and adolescents, self-rated health has only been examined more recently. For instance, in research with Canadian students in grades 4 and 7 (Human Early Learning Partnership 2014a, b), most children rated their health as ‘excellent’ or ‘good’ (94 and 91 % for grades 4 and 7, respectively), and a relatively small proportion rated their health as ‘poor’ or ‘fair’ (7 and 9 % for grades 4 and 7, respectively). A survey conducted with 22,827 children and adolescents aged 8–18 years in 13 European countries has shown positive associations between self-reported physical well-being and psychological well-being, social support, positive self-perception, autonomy, financial resources and socio-economic status (Ravens-Sieberer et al. 2008). Furthermore, children with physical or mental health problems reported significantly lower physical well-being in this study. Also, parental support and family connectedness have been shown to be protective factors against health-risk behaviors (Carter et al. 2007; Simantov et al. 2000). In turn, health-related behaviors such as poor nutrition habits, lack of physical activity, and inadequate sleep in children and adolescents are associated with negative health outcomes, such as chronic diseases (Smaldone et al. 2007; World Health Organization 2003). Two recent reviews (Umberson et al. 2010; Umberson and Montez 2010) highlight the importance of social relationships in influencing physical and mental health, health behavior, and mortality risk, and summarize evidence of how social experiences in early life shape long lasting social and health trajectories.

Research has also examined the link between parental SES and child health outcomes (Blair and Raver 2012; Bradley and Corwyn 2002; McLoyd 1998). Specifically, a gradient has been documented indicating that lower SES is associated with an increased health risk. This has been shown for various indices of SES (e.g., parental education, income, or occupation) and specific child and adolescent health outcomes (e.g., injuries, asthma, and risk factors related to cardiovascular disease) (Bor et al. 1993; Boulton et al. 1995; Lowry et al. 1996; Weitzman et al. 1990). A review of SES-related differences in children’s health (Chen et al. 2002) identified several potential mechanisms that might mediate this association, including social relationships with family and peers, with family relationships assumed to have a stronger influence during childhood, and peer relationships to have a stronger influence in adolescence. Specifically, low socioeconomic status is associated with some family risk factors (e.g., conflict and cold, unsupportive relationships), which in turn are associated with detrimental health outcomes in children (Dodge et al. 1994; Repetti et al. 2002). Other potential mechanisms that have been proposed to mediate the association of income and health are nutrition and health-relevant lifestyles (Bradley and Corwyn 2002).

1.4 Purpose and Hypotheses of the Study

To date, there is a paucity of research jointly examining the relative importance of socio-economic, demographic, and social factors, as well as health-related habits and activities to child well-being and health. Such analyses may help to determine the relative importance of different social and contextual factors in regard to children’s health and well-being. Furthermore, exploring the nature of such relationships may help to identify potential pathways and mechanism via which social and contextual factors are associated with health and well-being outcomes. Such understanding, in turn, will help focusing attention for program initiatives and policy decisions on those social and contextual factors that are both (1) modifiable and (2) that appear to have the greatest leverage in terms of enhancing children’s health and well-being. To address this gap, we examined the joint associations and relative importance of these variables in a large, middle childhood sample. Middle childhood (6–12 years; Collins 1984) is an important, but often overlooked (Moore and Theokas 2008) developmental period marked by biological, cognitive, and social changes that set the stage for adolescence and adulthood (Eccles 1999). In addition, given interest in mediators that can account for the long-recognized association between income and health (Chen et al. 2002), we examined whether social relationships and/or health-related activities and habits mediate the association between income and health or well-being. As Bradley and Corwyn (2002; p. 386) suggest, “one of the main limitations of research on SES is the failure to simultaneously consider correlated mediating processes…”. Thus, we used an approach that allowed the simultaneous testing of multiple mediators.

Specifically, we hypothesized that our study would replicate previously found associations as described above (cf. Bradley and Corwyn 2002; Burton and Phipps 2008; Holder and Coleman 2009; Konu et al. 2002; Ravens-Sieberer et al. 2008) when examining the (zero-order) correlations between our predictor variable (income), the mediator variables (connectedness with peers, connectedness with adults, health-related activities and habits), and the outcome variables perceived (health, satisfaction with life). In particular, we hypothesized that the correlations between the social connectedness variables and the outcome variables would be the largest of all the zero-order correlations.

In regard to our main research questions, we hypothesized that the association between income and satisfaction with life would be mediated by children’s connectedness to adults and peers. Secondly, we hypothesized that the association between income and perceived health would be mediated by children’s connectedness to adults and peers as well as health-related activities and habits. These hypotheses are based on developmental theories of children’s well-being, which propose that social relationships and associated proximal processes—the frequently re-occurring interactions between a developing child and his/her social and physical environments—are the most critical factors for children’s development (Bronfenbrenner 1979, 2005; Masten and Coatsworth 1998). In contrast, context factors, such as socio-economic status, are assumed to influence children’s development indirectly, affecting some of the proximal processes that children experience in their different ecologies. For example, repeated parent–child interactions may have direct effects on children’s development and well-being, whereas income may have an indirect rather than an immediate effect (e.g., limited availability of financial resources leads to parental stress which in turn affects the nature and frequency of parent–child interactions). In our study, measures of children’s connectedness to peers and adults served as indicators of their social relationships. With regard to proximal processes, we used measures of children’s health-related activities and habits, namely nutrition, sleep, and physical activity. A proxy of family income served as measure for children’s socioeconomic background, and thus represents a measure of context. According to our theoretical model, we hypothesized that the association between income (context factor) and children’s health and satisfaction with life (developmental outcomes) would be mediated by indicators of social relationships (adult connectedness and peer connectedness) and proximal processes (health-related activities and habits). Hence, our study was designed to test one of the main propositions of Bronfenbrenner’s bioecological model of human development, for two developmental outcome measures—self-rated health and satisfaction with life.

2 Method

2.1 Participants

The sample included 5026 4th graders (49 % girls; Mage = 9.7, SD = .3) from 121 elementary schools in three public school districts in British Columbia, Canada. The first of these school districts was large and ethnically and socioeconomically diverse (n = 3026), the second one was an ethnically diverse (sub)urban district (n = 1921), and the third was a small rural district (n = 79). In the first school district, 93 % of the students in participating elementary schools (72 schools out of the district’s 91) took part in the study, representing 80 % of the district’s total grade 4 population. In the second and third school districts, all elementary schools (45 and 4, respectively) were involved in the study, with participation rates of 88 and 99 %, respectively. English was reported as the first language learned at home by 72 % of the students; other languages first learned were Kantonese (14 %), Mandarin (8 %), Korean (4 %), Tagalog (4 %), Punjabi (3 %), Spanish (3 %), Vietnamese (3 %), French (2 %), Japanese (2 %), Hindi (1 %), and other (11 %). These languages are representative of the communities in which students were residing. Participation was voluntary for each school and teacher. Passive parental/guardian consent procedures were used; parents/caregivers were informed via letters about the research project 4 weeks prior to its start, and could withdraw their children by sending a note to the school. Student assent was also obtained (see further description in the procedure section).

2.2 Instruments

All child data were collected using the Middle Years Development Instrument (MDI; Schonert-Reichl et al. 2013), a 71-item, self-report survey for children aged 9–14. The MDI includes items and subscales that have been previously used in developmental research with children and/or adolescents (e.g., the adult connectedness scale adapted from the California Healthy Kids Survey; see details below). Previous research has documented the psychometric properties of the subscales of the MDI, providing evidence for the subscales’ reliabilities, factor structure, and convergent and discriminant validity (Guhn et al. 2012; Schonert-Reichl et al. 2013). The following subscales and items from the MDI were used in this study (for each of the subscales, Cronbach’s alphas from the present data are provided in Table 1).

2.2.1 Satisfaction with Life

Life satisfaction has been defined as “…a report of how a respondent evaluates or appraises his or her life taken as a whole” (Diener 2006, p. 401). The Satisfaction with Life Scale adapted for Children (SWLS-C; Gadermann et al. 2011, 2010) is an adaptation of the Satisfaction with Life Scale (Diener et al. 1985), containing five items (sample item: ‘In most ways my life is close to the way I would want it to be.’). Children are asked to respond on a 5-point Likert type response scale ranging from 1 = ‘disagree a lot’ to 5 = ‘agree a lot’. Total satisfaction scores, averaged across the five items, were computed, with higher scores reflecting greater life satisfaction.

2.2.2 Overall Health

Perceived overall health refers to a global subjective rating of a person’s health and is commonly defined as “a state of […] physical, mental and social well-being and not merely the absence of disease or infirmity” (World Health Organization 2006). Children rated their overall health on a single item, ‘In general, how would you describe your health?’ with a 4-point Likert type response scale (1 = ‘poor’ to 4 = ‘excellent’ (McCreary Centre Society 2009).

2.2.3 Adult Connectedness

Three 3-item subscales, adapted from the California Healthy Kids SurveyMiddle School Questionnaire (Constantine and Benard 2001), were used to assess the extent to which children feel having caring relationships with adults at home, in the neighborhood, and at school. Students were asked to rate statements concerning the degree to which there was a supportive/caring adult in each context on a 4-point Likert scale ranging from 1 = ‘not at all true’ to 4 = ‘very much true’ (sample items: ‘In my home, there is a parent or another adult who believes that I will be a success,’ ‘In my neighborhood, there is an adult who really cares about me,’ ‘At my school, there is a teacher or another adult who listens to me when I have something to say’). For each subscale, scores were averaged across relevant items, with higher scores reflecting greater adult connectedness.

2.2.4 Connectedness with Peers

Connectedness with peers was measured with a 3-item scale on peer belonging, which assesses the extent to which children feel they belong to a peer group (sample item: ‘I feel part of a group of friends that do things together’). This scale was adapted from the Relational Provisional Loneliness Questionnaire (RPLQ) (Hayden-Thomson 1989). The 5-point Likert type response scale ranges from 1 = ‘disagree a lot’ to 5 = ‘agree a lot’. Higher total scores (averaged across items) reflected greater peer connectedness.

2.2.5 Nutrition

Two nutrition-related habits were assessed with two single items, ‘How often do you eat breakfast?’ and ‘How often do you eat food like pop, candy, potato chips, or something else?’, both with 5-point response scales [‘never’ (1), ‘1 or 2 times a week’ (2), …, ‘every day’ (5)] (Schonert-Reichl 2007).

2.2.6 Sleep

One item on sleep habits was included, ‘What time do you usually go to bed during the weekdays?’ with a 5-point response scale ranging from ‘before 9:00 pm’ (1) to ‘after 12:00 am/midnight’ (5) (Schonert-Reichl et al. 2013).

2.2.7 Physical Activities

Children reported how many days they participated in after school (3:00 to 6:00 pm) team sports or individual sports activities during the previous school week on a 4-point rating scale [‘never’ (0), 1–2 times a week (1), 3–4 times a week (2), 5 times a week (every day) (3)] (Goerge and Chaskin 2004), with higher scores reflecting greater sports participation.

2.2.8 English as a Second Language (ESL)

Children’s home language background, serving as a proxy for cultural/ethno-cultural background, was assessed with one item ‘What is the first language you learned at home?’ The response options include 12 frequently used languages and an option ‘other’. Children are asked to check more than one if applicable. Children who only learned a language other than English as their first language were classified as ESL. ESL was included as a control variable in the regression and mediation analyses as previous research has found statistically significant differences (of small effect size) for ESL versus non-ESL children, with non-ESL tending to report higher satisfaction with life (Gadermann et al. 2010).

2.2.9 Socioeconomic Status

Socioeconomic status (SES) data were obtained from 2006 tax filer data (Statistics Canada) and linked to the MDI data. In the analyses, the variable ‘median equivalized disposable income’ at census enumeration area was assigned to each child according to the child’s residential 6-digit level postal code, and served as a proxy for a child’s family’s SES background. This variable represents the equivalized disposable income per person within an enumeration area (Ebert 1999), taking into account family size (because the living costs of running a family household do not linearly increase with increasing family size) and the age of the family members (because children and teenagers have, on average, lower living costs than adults). The variable is calculated from (individual) census family income information, but due to privacy laws, Statistics Canada only provides the data at an aggregated enumeration area code level. In densely populated urban and suburban areas—such as the two large districts in this study—the size of an enumeration area typically spans 1–3 street blocks, consists of about 200–300 census families,1 and coincides with the 6-digit postal code area. In rural areas—such as the rural school district in this study—the census enumeration area spans a wider geographical area, and may include fewer census families. Validation studies in Canada (Mustard et al. 1999) and the US (Krieger 1992) indicated that census income data aggregated at the enumeration area level (in the Canadian context) or at the census tract and/or block level (in the US context) serve as a reasonable proxy for individual family-level income in large-scale analyses. Accordingly, previous Canadian research studies have employed the same methodology (Guhn et al. 2010; Oliver et al. 2007).

2.3 Procedure

The MDI survey was administered by classroom teachers in January 2010 in one school district and in February 2011 in the other two. Teachers received a manual describing procedures for administering the MDI. They read aloud a verbal student assent script, informing students that participation was voluntary and confidential, and that there would be no consequences if a student chose not to participate. To guard against biases due to variability in children’s reading proficiencies, the teacher read each item aloud, and students marked their responses accordingly. The completion of the MDI took, on average, two 40–min class periods. The study and all associated procedures were approved by the University of British Columbia’s Human Subjects Institutional Review Board and by each of the School Board administrations.

2.4 Data Analysis

2.4.1 Missing Data

The degree of missing data ranged between .6 and 5.4 % for the predictor variables. The outcome variables had 2.6 (health) and 1.8 % (life satisfaction) missing values. Of the total sample (N = 5026), 83 % had complete data for all variables in the regression analyses (n = 4157 for life satisfaction and n = 4168 for health as outcome variable). Cases with complete data were included in the regression analyses.2 Using independent samples t tests, we compared the means of the outcome and predictor variables for cases that had complete data on all predictor variables and cases that had one or more missing values. There was a significant difference for the variable ‘English as first language learned at home’, in that the proportion of children reporting English as their first language was significantly higher among the included cases than among the excluded cases (73 vs. 66 % for the satisfaction with life analysis: t(4918) = 3.7; p < .001; and 73 vs. 65 % for the health analysis: t(4872) = 4.0, p < .001).

2.4.2 Analyses

The data were analyzed in SPSS, version 18. Given the nested structure of the data (children in schools), we calculated the intraclass correlation for each base line model. The intraclass correlations for the intercept-only model with satisfaction with life as well as health was .03. Given that intraclass correlations greater than .01 can inflate alpha levels (Cohen et al. 2003), we analyzed the data via multiple linear regression as well as via multi-level regression models using the MIXED procedure.

A sequence of nested models was fitted, with one set of models for the outcome health, the other for satisfaction with life (all predictors were the same and entered in the same order for the two outcome variables). In the multilevel analyses, each model was run with the variable ‘school’ as the (level 2) grouping variable to examine whether a significant amount of variance in satisfaction with life or health scores was associated with children’s school context. Model 1 included gender, first language learned at home, and median equivalized disposable income as predictors. Model 2 included gender, first language learned at home, connectedness with adults at home, in the neighborhood and at school, and peer belonging as predictors. Model 3 combined predictors from model 1 and 2 (i.e., median equivalized disposable income was added to Model 2). This allows one to see to what degree the coefficients change for the income variable and the relationship variables in the presence of each other. In Model 4, we included gender and first language learned at home, together with several health-related habits and activities variables (i.e., the two variables related to eating habits, one variable on sleep habits, and two variables related to after school sports activities). Model 5 was identical to Model 4, except that we added the income variable as predictor. As with Models 2 and 3, this allows one to examine to what degree the coefficients for the income variable and the variables on health-related habits and activities change in the presence of each other. Model 6 included all predictor variables. To examine the relative importance of predictor variables, we calculated the relative Pratt index (Thomas et al. 1998). Finally, in order to test the mediation hypotheses formally we used the SPSS-compatible version of the INDIRECT macro, developed by Preacher and Hayes (2008), which allows the examination of path coefficients of multiple mediators and provides bootstrap confidence intervals (CI; 95 % with 5000 bootstrap resamples) for the total and specific indirect effects.

3 Results

3.1 Descriptive Statistics and Bivariate Findings

Descriptive statistics and Cronbach’s alphas (where applicable) for the study variables are provided in Table 1. As in previous studies, we found that the distributions of the scales on life satisfaction, connectedness to adults and peers, and health were negatively skewed (ranging from −.4 to −1.6) (Gadermann et al. 2010; Schonert-Reichl et al. 2013), indicating that children tended to rate these dimensions favorably. The majority of children reported having breakfast ‘every day’ (80 %) and close to half the children reported consuming ‘junk food’ like candy or chips, ‘1 or 2 times per week’ (49 %) and going to bed on a usual weekday between ‘9:00 and 10:00 pm’ (46 %). With regard to after-school sports, less than half of the students reported having participated on 1 or more days in after school individual (42 %) or team sports (39 %). All scales showed satisfactory internal consistencies, ranging between .71 and .86. The bivariate associations among the study variables are shown in Table 2. Health and life satisfaction were significantly positively correlated with each other and, with the exception of gender (in regard to health), all predictor variables were significantly associated with the outcome variables and in a theoretically expected direction (i.e., all coefficients were positive, except for frequency of ‘junk food’ consumption and bed time). The largest zero-order correlations were found between the social connectedness variables and the life satisfaction and health outcome measures.
Table 1

Descriptive statistics and Cronbach’s alphas of study variables

Measure

Means (SD)

Skewness (kurtosis)

Cronbach’s alpha

Satisfaction with life (5 items)a

4.0 (.9)

−1.2 (1.1)

.83

Health (1 item)

3.4 (.7)

−.9 (.6)

Median equivalized disposable income

25,715 (9719)

 

Adult connectedness, home (3 items)a

3.5 (.6)

−1.6 (2.3)

.73

Adult connectedness, neighborhood (3 items)a

2.8 (1.0)

−.4 (−1.0)

.86

Adult connectedness, school (3 items)a

3.1 (.7)

−.6 (−.2)

.71

Peer belonging (3 items)a

4.0 (.9)

−1.0 (.7)

.80

Measure

Frequency (%)

Breakfast frequency (1 item)

 

 Never

95 (2)

 1 or 2 times a week

254 (5)

 3 or 4 times a week

209 (4)

 5 or 6 times a week

423 (9)

 Every day

3915 (80)

Junk food frequency (1 item)

 

 Never

358 (7)

 1 or 2 times a week

2400 (49)

 3 or 4 times a week

1242 (26)

 5 or 6 times a week

498 (10)

 Every day

360 (8)

What time do you usually go to bed during the weekdays?

 

 Before 9:00 pm

1276 (26)

 Between 9:00 and 10:00 pm

2266 (46)

 Between 10:00 and 11:00 pm

804 (17)

 Between 11:00 pm and midnight

312 (6)

 After 12:00 am/midnight

234 (5)

Individual sports after school, past week (1 item)

 Never

2752 (58)

 1–2 times a week

1385 (30)

 3–4 times a week

439 (9)

 5 times a week

149 (3)

Team sports after school, past week (1 item)

 Never

2894 (61)

 1–2 times a week

1198 (25)

 3–4 times a week

471 (10)

 5 times a week

175 (4)

aScale scores are calculated as mean scores (not sum scores) across all items

Table 2

Intercorrelations among study variables

 

1

2

3

4

5

6

7

8

9

10

11

12

13

Outcome variables

1. Health

             

2. Life satisfact.

.36**

            

Predictor variables

Demographics

  3. Gendera

.02

.05**

           

  4. First language

.10**

.10**

.00

          

  5. Med. Incomeb

.10**

.09**

.00

.21**

         

Relationshipsc

 Connectedness w/adults at/in…

             

  6. Home

.29**

.42**

.05**

.10**

.15**

        

  7. Neighborhd

.23**

.24**

.07**

.13**

.12**

.31**

       

  8. School

.26**

.33**

.08**

.08**

.11**

.43**

.44**

      

  9. Peer belong.

.31**

.41**

−.02

.10**

.13**

.37**

.31**

.39**

     

Health related activities

 10. Breakfast

.13**

.13**

−.01

.04**

.13**

.16**

.08**

.10**

.11**

    

 11. Junk food

−.10**

−.06**

−.03

.02

−.03*

−.05**

−.04**

−.02

−.01

−.03

   

 12. Bed time

−.12**

−.13**

−.09**

−.16**

−.15**

−.17**

−.10**

−.11**

−.09**

−.18**

.13**

  

 13. Indiv. sports

.09**

.05**

.07**

−.01

.09**

.06**

.13**

.09**

.05**

.03

.00

.02

 

 14. Team sports

.14**

.07**

−.19**

.17**

.22**

.09**

.17**

.09**

.14**

.05**

−.01

−.06**

.25**

aBoys coded as 0 and girls coded as 1

bESL coded as 0 and non-ESL coded as 1

cMedian equivalized disposable income

N = 4133

p < .05; ** p < .01

3.2 Multiple Regression and Mediation Analyses3

As noted above, both multiple linear and multilevel regression analyses were conducted. However, given that the regression coefficients for all predictor variables were equivalent regardless of whether the clustering of the data was taken into account, and all level-2 variance was accounted for by differences in the individual level predictors, we only report the multiple linear regression results (Tables 3, 4).
Table 3

Multiple regression results (standardized beta coefficients; p values in parentheses; Pratt scores) for the Satisfaction with Life Scale Adapted for Children

Predictor variables

Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Pratt scores (%)

Gendera

.05 (.00)

.02 (.01)

.03 (.01)

.05 (.00)

.05 (.00)

.03 (.03)

1

First languageb

.09 (.00)

.04 (.01)

.04 (.01)

.08 (.00)

.07 (.00)

.04 (.01)

1

Median equivalized disposable income

.07 (.00)

 

.00 (.73)

 

.04 (.02)

−.01 (.39)

0

Adult connectedness (home)

 

.27 (.00)

.27 (.00)

  

.26 (.00)

41

Adult connectedness (neighborhood)

 

.03 (.06)

.03 (.06)

  

.03 (.09)

2

Adult connectedness (school)

 

.10 (.00)

.10 (.00)

  

.10 (.00)

12

Peer belonging

 

.26 (.00)

.26 (.00)

  

.26 (.00)

39

Breakfast frequency

   

.10 (.00)

.10 (.00)

.04 (.00)

2

Junk food frequency

   

−.05 (.00)

−.05 (.00)

−.04 (.00)

1

Bedtime

   

−.08 (.00)

−.08 (.00)

−.03 (.07)

1

Individual sports

   

.03 (.10)

.02 (.12)

.00 (.78)

0

Team sports

   

.05 (.00)

.04 (.01)

−.01 (.64)

0

R2 (%)

2

26

26

4

4

27

 

Notes Predictors of meaningful relative importance according to the Pratt index are bolded

N = 4157 for all models

aBoys coded as 0, girls as 1

bESL coded as 0 and non-ESL coded as 1

3.2.1 Satisfaction with Life

As shown in Table 3, the income variable was significantly related to satisfaction with life scores in Model 1 and in the presence of the health-related habits and activities (Model 5), but was statistically non-significant in the presence of the connectedness predictor variables in Models 3 and 6. The results for Models 2–5 show that connectedness with adults at home and at school, peer belonging, breakfast frequency, and after-school team sports were statistically significant, positive predictors of satisfaction with life scores, and that junk food frequency and late bedtime were statistically significant, negative predictors of satisfaction with life scores. It needs to be noted that the models including the connectedness variables (Models 2 and 3) explained substantially more variance (26 %) than the models with the health-related habits and activities (Models 4 and 5, 4 %).

In the final model, Model 6, the predictor variables jointly explained 27 % of the variance in life satisfaction scores. For this model, Pratt indices are provided, indicating the relative importance of each predictor in the model. Relative Pratt indices of dj > 1/(2 × p) (p = number of predictors in the regression equation) are considered to reflect meaningful, interpretable relative importance. In the present study, although most of the predictors were statistically significant, only three of the connectedness variables were of meaningful importance in explaining variance in the life satisfaction outcome variable: Connectedness with adults at home (41 % of the explained variance in the model), peer belonging (39 %), and connectedness with adults at school (12 %).

Finally, in the mediation model, we specified (a) the income variable as the independent predictor variable, (b) the predictors from the multiple regression with Pratt indices of meaningful relative importance (i.e., >4 %) as mediators, (c) gender and first language as covariates, and (d) the satisfaction with life variable as outcome variable. The path coefficients, indirect effects and corresponding confidence intervals, and the total effect are shown in Fig. 1. The total effect of income on satisfaction with life was statistically significant (.06; p < .001), but largely mediated by the significant mediator variables. This is indicated by the non-significant direct effect of income on satisfaction with life (.02; p = .09).
Fig. 1

Mediation of the effect of income on satisfaction with life through social relationships

3.2.2 Health

Similar to the results obtained for life satisfaction, the income variable was a statistically significant predictor in Model 1, which included the demographic variables gender and ESL as control variables (see Table 4). However, income was a non-significant predictor in the presence of the connectedness variables or health-related habits and activities variables in Models 3, 5, and 6. In Models 2–5, all predictor variables were statistically significant. As in the models with life satisfaction as outcome variable, the models including the connectedness variables explained more variance in perceived health than the health-related habits and activities, though the difference was less pronounced (15 vs. 6 %). The predictor variables jointly explained 17 % of the variance in perceived health in Model 6. The Pratt indices indicated that the following predictors were of meaningful relative importance: Peer belonging (36 % of the explained variance in the model), connectedness with adults at home (23 %) at school (11 %) and in the neighborhood (9 %), after-school team sports (5 %), junk food frequency (5 %, with a negative association to overall health), and breakfast frequency (4 %).

In the mediation regression model, we again included all variables with Pratt Index values above 4 % as mediators (see Fig. 2). Similar to the mediation analysis for satisfaction with life, the total effect of income on health was significant (.06; p < .001), but it was largely mediated by the mediator variables. The direct effect of income on health was very small (.01) and not significant (p = .58) (see Fig. 2).
Table 4

Multiple regression results (standardized beta coefficients; p values in parentheses; Pratt scores) for the perceived overall health variable

Predictor variables

Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Pratt scores (%)

Gendera

.02 (.13)

.01 (.68)

.01 (.67)

.03 (.04)

.03 (.04)

.01 (.36)

1

First languageb

.08 (.00)

.05 (.00)

.04 (.00)

.07 (.00)

.06 (.00)

.04 (.01)

2

Median equivalized disposable income

.08 (.00)

 

.02 (.10)

 

.03 (.05)

.00 (.93)

0

Adult connectedness (home)

 

.15 (.00)

.15 (.00)

  

.13 (.00)

23

Adult connectedness (neighborhood)

 

.08 (.00)

.08 (.00)

  

.07 (.00)

9

Adult connectedness (school)

 

.07 (.00)

.07 (.00)

  

.07 (.00)

11

Peer belonging

 

.19 (.00)

.19 (.00)

  

.19 (.00)

36

Breakfast frequency

   

.10 (.00)

.10 (.00)

.06 (.00)

4

Junk food frequency

   

−.09 (.00)

−.09 (.00)

−.08 (.00)

5

Bedtime

   

−.07 (.00)

−.06 (.00)

−.03 (.05)

2

Individual sports

   

.05 (.00)

.05 (.00)

.03 (.02)

2

Team sports

   

.11 (.00)

.10 (.00)

.07 (.00)

5

R2 (%)

2

15

15

6

6

17

 

Notes Predictors of meaningful relative importance according to the Pratt index are bolded

N = 4168 for all models

aBoys coded as 0, girls as 1

bESL coded as 0 and non-ESL coded as 1

Fig. 2

Mediation of the effect of income on health through social relationships and health-related behaviors

4 Discussion

Drawing conceptually from an ecological model of human development, and corresponding research design, the present study explored correlates of children’s own perspective on their well-being and health. Addressing a critical gap in research to date, (a) we examined the relative importance of multiple social and contextual predictor variables with regard to life satisfaction and perceived health, and (b) tested whether the association between SES and life satisfaction and health is mediated by social and contextual variables. Our findings illustrate the central importance of social relationships in children’s perceptions of their life satisfaction and perceived health in middle childhood.

With regard to satisfaction with life, three predictors were of significant relative importance: Connectedness with adults at home, peer belonging, and connectedness with adults at school. Previous research also indicates the critical role of social relationships for life satisfaction, especially with family members, in middle childhood (Gadermann et al. 2011). In our study, relationships with adults at home and peer belonging were of similar relative importance in predicting satisfaction with life, accounting for 41 and 39 %, respectively, of the explained variance in life satisfaction in the model including all predictor variables (R2 = 27 %). In previous research, relationships with parents/family and peers have also been shown to be significant predictors of life satisfaction for children and adolescents, although several studies with students across grades 3–10 suggest a stronger association of parent than peer relationships with life satisfaction (Goswami 2012; Ma and Huebner 2008; Man 1991; Terry and Huebner 1995). Furthermore, though of less relative importance than the relationships to adults at home and peers in our joint model, relationships with teachers and connectedness to school have been shown to be relevant for the wellbeing and the social and emotional adjustment among 11-14 year-olds (Konu et al. 2002; Murray and Greenberg 2000). The significance of social relationships is also reflected in the finding that social connectedness in adolescence was associated with well-being in adulthood over a decade later (Olsson et al. 2013). The SES variable and health-related activities and behaviors included in our study showed statistically significant bivariate associations of small effect size with life satisfaction, though none of these were of meaningful relative importance in the final multivariable model.

With regard to perceived health, we found that peer belonging was the strongest overall predictor of perceived health, accounting for 36 % of the explained variance in perceived health. This finding corroborates previous studies, which showed that peer influences are strongly associated with adolescent health behavior (Umberson et al. 2010). Much of the literature examining the association between peer relations and child and adolescent health has focused on (the initiation of) health risk-behaviors such as smoking, alcohol consumption and substance use, particularly during the transition to high school and during the high school years (Allen et al. 2012; Fujimoto et al. 2012; Go et al. 2010; Trucco et al. 2011). These studies have identified approval seeking, behavior modeling, peer group affiliation, and active selection of similar friends as pathways through which these influences operate (Fujimoto et al. 2012; Go et al. 2010; Trucco et al. 2011). However, there is also evidence that peer relationships are positively associated with physical activity, healthy eating, and avoidance of health risk behaviors (Fletcher et al. 2011; Salvy et al. 2012; Smith 1999; Teunissen et al. 2012). Peer relationships may also be associated with health by reducing barriers to physical activity involvement. For example, Stanley et al. (2012) found that 10–13-year-olds reported that support and acceptance from their peers facilitated their enthusiasm for play and participation in sports at lunchtime. Peer acceptance may also be associated with both mental and physical health to the extent that it promotes children’s self-esteem, sense of control, and belonging and companionship (Thoits 2011).

In addition to positive connections with peers, connectedness to adults remains critical during this developmental period for setting children on healthy trajectories into late adolescence and adulthood (Umberson et al. 2010; Viner et al. 2012). In middle childhood, children begin to spend more time away from home, both in school and in the community, where they make connections with non-related adults including teachers, sports coaches, and instructors of after-school activities (Schonert-Reichl 2007; Umberson et al. 2010). It has been suggested that children’s connectedness to adults may promote their physical and mental health in many ways, including buffering stress, increasing self-esteem, providing a sense of mattering, and increasing a sense of efficacy and perceived control (Thoits 2011; Umberson et al. 2010). Past research has also shown that adolescents in grades 7–12 who feel more connected to their schools are less likely to engage in health risk behaviors (Resnick et al. 1997, 1993; Shochet et al. 2006). Our results demonstrate that positive relationships with adults in the home, school, and neighbourhood are each uniquely associated with children’s perceived health. This finding is consistent with research by Youngblade et al. (2007) who found that youth aged 11–17 who received positive reinforcement within multiple contexts (family, school, neighborhood) experienced fewer externalizing and internalizing symptoms and showed higher health-promoting behaviors and self-esteem.

Apart from social relationships, several of the health-related behaviors emerged as significant predictors of children’s self-reported health, namely the frequency of eating junk food like pop, candy, or potato chips (with an inverse association to health), having breakfast, and participating in team sports. This is also in line with previous research with children in middle childhood and adolescents (Fletcher et al. 2003; Linver et al. 2009; McHale et al. 2001; O’Dea 2003; Rampersaud et al. 2005; Zarrett et al. 2009). The finding that participating in team sports was of meaningful relative importance in predicting perceived health, but participation in individual sports was not, again underscores the importance of the social component in predicting children's self-rated health. It is therefore of interest for future research to explore how these associations compare to research with objective indicators of health, as well as different facets of self-rated health (e.g., mental health vs. physical health).

The bivariate associations between income and life satisfaction as well as health were statistically significant, although the effect size was relatively small. The mediation analyses showed that the association between income and satisfaction with life was mediated by the social relationship variables. Similarly, the association between income and self-rated health was mediated by social relationship and health-related behavior variables. Specifically, in both models, the indirect effects of income via connectedness with adults at home and via peer belonging were significant. Furthermore, for self-rated health, the indirect effect of income via participating in team sports was significant. These findings are in line with theories that propose that social processes mediate the association between indicators of socio-economic status and different indicators of children’s positive development, health, and well-being (Bradley and Corwyn 2002; Chen et al. 2002). These findings also support the notion that the importance of considering SES for child well-being may primarily lie in examining how it is related to children’s social relationships and associated proximal processes (e.g., the re-occurring interactions between a child and his/her social and physical environment).

Previous studies indicate that family experiences, such as negative parent–child interaction patterns, are involved in mediating the association of SES and the social and emotional adaptation of youth. Parents with low SES are faced with many stressors (e.g., time poverty; cf. Harvey and Mukhopadhyay 2007) and uncertainties, including family life stressors, such as divorce or legal problems (Dodge et al. 1994), which in turn can impact children’s health and well-being due to differing parental attitudes, expectations and interaction styles. For example, children from families with low SES are more likely to be exposed to harsh discipline, conflict communications and aggressive adult models, and less parental warmth and social support (Dodge et al. 1994; McLoyd 1998), and these in turn are related to detrimental outcomes for children and adolescents (Mechanic and Hansell 1989; Montgomery et al. 1997; Wickrama et al. 1997). This is further supported by previous research demonstrating that the associations of SES and various aspects of children’s psychosocial adjustment are (partially) mediated through parenting practices (Bolger et al. 1995; Dodge et al. 1994; Felner et al. 1995; McCoy et al. 1999).

Furthermore, the peer relationship variable significantly mediated the association between SES and life satisfaction and health. Youth in disadvantaged neighborhoods may have less opportunity to engage in organized activities and instead participate in unsupervised peer groups, which may be associated with negative peer group affiliation (Sampson and Groves 1989). Similarly, research indicates that peer support tends to have detrimental effects in high-risk neighborhoods, but positive ones in low-risk neighborhoods (Leventhal et al. 2000). As described above, peer group affiliation, in turn, is associated with the initiation of various health risk-behaviors such as smoking, alcohol consumption and substance use (Allen et al. 2012; Fujimoto et al. 2012; Go et al. 2010; Trucco et al. 2011). In contrast, positive peer influences may improve children’s health and well-being by reducing barriers to physical activity involvement. For example, in a study conducted by Stanley et al. (2012), 10–13-year-olds reported that support and acceptance from their peers facilitated their enthusiasm for play and participation in sports at lunchtime

In their model of mediating mechanisms for the association between SES and health (Chen et al. 2002) proposed that early childhood would be a period of strong influence of family relationships (and less so for peers), whereas the influence of peer relationships would grow stronger during adolescence. Our findings indicate that, in middle childhood (a transitional period to adolescence), relationships to peers was a stronger mediator than connectedness with adults at home, in regard to the association between SES and health. In contrast, relationships to adults at home and to peers were equally strongly associated with children’s life satisfaction.

Limitations of our study include the cross-sectional design of our study and the block-level census data income variable used as a proxy for child-level SES. However, previous studies in Canada and the US indicate that similar analyses with block-level income variables provided valid results (Krieger 1992; Mustard et al. 1999). Furthermore, we only used one indicator of SES in our analyses, whereas (a combination of) others, such as educational background, employment, and social status of the children’s families, could also be considered. The association between income and the outcome variables could have been attenuated in our study because of our use of the block-level income rather than parental income or other SES variables, such as overall family educational attainment/highest degree earned in the family or wealth, which may be more relevant during middle childhood. For example, one previous study with families with a child in middle childhood examined the association between a range of SES measures (assessments of poverty, income, wealth, maternal and overall family educational attainment, subjective social status, and cumulative social risk), and child developmental outcomes (including physical and mental health, social and school functioning) using parents’ reports for all measures (Nuru-Jeter et al. 2010). Of the SES measures used in this study, wealth and overall family educational attainment showed the strongest associations across most of the child health/functioning domains. The study did not, however, examine any potential mediators/pathways. It will be important for future studies to examine to what extent each of the associations of various SES indicators to children’s self report of health and life satisfaction may be mediated via children’s social relationships as well as the health- and well-being-related proximal processes in which the children engage on a regular basis.

Bradley and Corwyn (2002; p. 379) state “the literature mostly provides bits and pieces of the larger person-process-context-time tableaux described by Bronfenbrenner”. Our findings add to the extant literature by providing information about the child-process-context aspects. Future research should further examine how these processes may differ by age as well as how they change over time in longitudinal research. Also, these mediating pathways may vary by ethnic background and gender (McLoyd 1998). Our findings underline the essential importance of social relationships for children’s health and life satisfaction. The findings suggest that settings like schools or after-school programs that wish to foster children’s and youth’s well-being and thriving may need to be places of relatedness and caring—a notion that underlies humanist theories of education, schooling, and learning (Noddings 1988; Ryan and Deci 2000; Solomon et al. 1996). A growing body of research indicates that creating schools as places of caring and relatedness is possible. Several evidence-based programs exist to support children’s social relationships and promote prosocial behavior (see http://findyouthinfo.gov for an overview of various programs as well as the Collaborative for Academic, Social, and Emotional Learning’s Program Guide of evidence-based social and emotional learning programs: http://www.casel.org/guide). For example, the Caring School Community program, an elementary school program, aims to foster children’s prosocial development and their connectedness to school, including positive student–teacher relationships and peer relationships by creating a caring classroom and school community in which children’s fundamental needs regarding autonomy, relatedness and competence (Deci and Ryan 2000) are met (Battistich et al. 1997; Battistich and Hom 1997). Similarly, Raising Healthy Children is a school-based preventive intervention for elementary through high school students with the aim of supporting positive youth development with a focus on social relationships, incorporating school and family environment (Catalano et al. 2003; Hawkins et al. 2005). Furthermore, a meta-analysis of social-emotional learning (SEL) programs (Durlak et al. 2011) found that programs with specific characteristics (i.e., programs are sequenced, include active learning activities, and explicitly focus on specific outcome skills) significantly enhance children’s social and emotional competences as well as children’s academic achievement. Finally, a review of evaluation studies of the Comer Child Development Project and the School Development Program—two whole-school reform programs that focus on relationship building—found that these programs’ effectiveness and sustainability was largely dependent on multiple interconnected context and process factors, such as prolonged leadership support, teacher training, school-family-community collaboration, and relationship building opportunities (Guhn 2009). In sum, our findings corroborate that social connectedness is essential for children’s health and well-being—and the research evidence on SEL programs indicates that fostering social connectedness through targeted initiatives is not only possible, but also beneficial for children’s academic achievement. Remaining challenges toward implementing programs and initiatives that comprehensively address children’s social relationships exist. Teacher programs commonly do not comprehensively cover SEL theories and practices. Also, school curricula usually remain focused on academic content. Furthermore, compared to a generation ago, caregivers and parents spend an increasing number of hours working to make ends meet (Kershaw 2005; Zuberi 2006)—making it harder for school, families, and communities to create the time and space to build social connections and coalitions that could enhance the sustainability and effectiveness of SEL initiatives. These challenges highlight that school-based and after-school-based initiatives are embedded within a social, cultural, and political context—and that initiatives that promote children’s social relationships, well-being, and health are only going to be effective and sustainable to the extent that our society creates the social, cultural, and political environments and resources that support the implementation and sustainability of such initiatives.

Footnotes
1

A ‘census family’ frequently coincides with a household; but in some cases, a household contains multiple census families.

 
2

The results were replicated using pairwise deletion.

 
3

The multiple regresssion analyses were also run separately for the data from the urban school district, as well as the (sub)urban and rural school districts combined. As the results were equivalent, we report the combined analyses.

 

Acknowledgments

Dr. Anne Gadermann acknowledges support from the Michael Smith Foundation for Health Research and Canadian Institutes of Health Research. Dr. Martin Guhn acknowledges support from the Lawson Foundation.

Compliance with Ethical Standards

Conflict of interest

The authors declare that they have no competing interests.

Copyright information

© Springer Science+Business Media Dordrecht 2015

Authors and Affiliations

  • Anne M. Gadermann
    • 1
  • Martin Guhn
    • 1
  • Kimberly A. Schonert-Reichl
    • 1
  • Shelley Hymel
    • 1
  • Kimberly Thomson
    • 1
  • Clyde Hertzman
    • 1
  1. 1.Human Early Learning Partnership, School of Population and Public HealthUniversity of British ColumbiaVancouverCanada

Personalised recommendations