Are family, neighbourhood and school social capital associated with higher self-rated health among Croatian high school students? A population-based study.

OBJECTIVES: We investigated the associations between self-rated health and social capital among Croatian high school students.
DESIGN: A cross-sectional survey among high school students was carried out in the 2013-2014 school year.
SETTING: High schools in Croatia. PARTICIPANTS: Subjects were 3427 high school students (1688 males and 1739 females), aged 17-18 years. MAIN OUTCOME MEASURE: Self-rated health was assessed by the single item: "How do you perceive your health?". Possible responses were arranged along a five-item Likert-type scale: 1 very poor, 2 poor, 3 fair, 4 good, 5 excellent. The outcome was binarised as 'good health' (excellent, good or fair) versus 'poor health' (poor or very poor).
METHODS: We calculated ORs and 95% CIs for good self-rated health associated with family, neighbourhood and school social capital, while adjusting for gender, self-perceived socioeconomic status, psychological distress, physical activity and body mass index. We used generalised estimating equations using an exchangeable correlation matrix with robust SEs.
RESULTS: Good self-rated health was significantly associated with higher family social capital (OR 2.43; 95% CI 1.55 to 3.80), higher neighbourhood trust (OR 2.02; 95% CI 1.48 to 2.76) and higher norms of reciprocity at school (OR 1.79; 95% CI 1.13 to 2.84). When all of the social capital variables were entered simultaneously, good self-rated health remained significantly associated with higher family social capital (OR 1.98; 95% CI 1.19 to 3.30), neighbourhood trust (OR 1.77; 95% CI 1.25 to 2.51) and reciprocity at school (OR 1.71; 95% CI 1.08 to 2.73).
CONCLUSIONS: Higher levels of social capital were independently associated with higher self-rated health among youth. Intervention and policies that leverage community social capital might serve as an avenue for health promotion in youth. Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions.

This study is one of the fewer studies to date that have focused on social capital and health among children and youth.

We used a random sampling approach to select 20 high schools in Zagreb, all of which agreed to take part in the survey. A total of 3427 students (93.8%) responded to the survey which was given during class.

To clarify which source of social capital is likely to affect adolescents’ health, we assessed three different sources of social capital—in the family, in the neighbourhood, and at school—among high school students.

We used the generalised estimating equations using an exchangeable correlation matrix with robust SEs in order to correct SEs for clustering.

Owing to the cross-sectional design, we cannot exclude the possibility of reverse causation. Since we used a subjective measure of health and social capital, there is a possibility of common method bias.

Introduction

Social capital has been defined as the “resources embedded in a social structure which are accessed and/or mobilised in purposive actions.”1 2 Some scholars have conceptualised social capital as the social networks themselves, or as both the network structures and the resources channelled through the networks.3 4 Social capital has garnered increasing attention as a potential influence on the development of youth. In the field of education, research has primarily focused on the role of social capital in children's academic performance;5 6 however, subsequent research has expanded the range of outcomes to include health behaviours and population health outcomes.4 7 Social capital theory posits that interpersonal trust, norms of reciprocity and exchange of social support between members of a networks each constitutes a type of resource, and that access to these resources may facilitate the actions of group members.4 7 8 Investigation of possible influences of social capital on health may be particularly salient in adolescents because previous work has suggested that contextual influences may have effects on the somatic and psychological development of young people throughout the life course.9 10

Associations between social capital and health have been extensively investigated in adult samples.11–13 Social capital has been found in previous studies to be related to physical activity and body mass index (BMI).14–16 At the same time, physical activity and BMI are predictors of self-rated health. Therefore, both are not considered to be confounders of the relation between social capital and self-rated health. Rather, our underlying hypothesised model is: social capital→physical activity/BMI→self-rated health. However, fewer studies to date have focused on social capital and health among children and youth. According to Urie Bronfenbrenner's ecological systems theory of child development, human development is conceptualised as being shaped by the interaction between an individual and his or her environment; furthermore, there are many different levels of environmental influences that can affect a child’s development, starting with people and institutions immediately surrounding the child (ie, parents and families), to school environments, to residential neighbourhoods and eventually the societal culture.17 In-line with this, we hypothesised that family, neighbourhood and school social capital may be associated with an adolescent's good health and that students who report higher levels of social capital in all three domains will have higher self-rated health. However, few studies have simultaneously examined the contribution of different sources of social capital to youth health.

Accordingly, in the present study, we investigated the influences of different sources of social capital—in the family, in the neighbourhood, and at school—on levels of self-rated health among a sample of high school students in Croatia.

Methods

Participants

We administered a survey among high school students in Zagreb, a mid-sized urban city in central Croatia with a population of about 1 000 000 people. A random sampling approach was used to select high schools. All of the 20 schools that we approached agreed to take part in the survey, representing 3650 students enrolled in the 2013–2014 school year. Of these, 3427 students (1688 males and 1739 females, aged 17–18 years) responded to the survey (93.8%) which was given during class. Finally, the data of 3427 students aged 17–18 years were analysed.

Self-rated health

Self-rated health was assessed in these young adolescents using the standard single-item measure: “How do you perceive your health?”. Possible responses were arranged along a five-item Likert-type scale: 1 very poor, 2 poor, 3 fair, 4 good, 5 excellent. We binarised the outcome, that is, fair, good and excellent were collapsed into one category (good health), while poor and very poor were designated as poor health. Perceived health is an easily administered and widely used outcome measure in social epidemiology studies, and it has been shown to be a reliable predictor of mortality and healthcare use in adults.18–20 The measure has also been used in adolescents.21–24

Social capital indicators

On the survey, we inquired about individual perceptions of social capital in the family, neighbourhood and high school settings.3 7 25 Family social capital was assessed by the question: ‘‘Do you feel your family understands and gives attention to you?’’.7 26 Neighbourhood social capital was assessed by using two items; ‘‘Do you feel people trust each other in your neighbourhood (neighbourhood trust)?’’ ‘‘Do you feel that your neighbours step in to criticise someone's deviant behaviour during high school (informal social control)?’’.7 School social capital was assessed by three items; ‘‘Do you feel teachers and students trust each other in your high school (vertical school trust)?’’ ‘‘Do you feel students trust each other in your high school (horizontal school trust)?’’ ‘‘Do you feel students collaborate with each other in your high school (reciprocity at school)?’’ The response options were: ‘strongly agree’; ‘agree’; ‘neither agree or disagree’; ‘disagree’; ‘strongly disagree’. Then, for each response, we created a dichotomous variable (high: ‘strongly agree’, ‘agree’ and ‘neither agree or disagree’; low: ‘disagree’ and ‘strongly disagree’).7 The Cronbach α of the school social capital scale was 0.71, and since other domains have fewer than three questions, we considered it not appropriate to check Cronbach αs for these scales.

Covariates

As a measure of physical activity, we considered students’ total physical activity in the past 7 days. Physical activity was assessed using the validated short version of the International Physical Activity Questionnaire (IPAQ) and was expressed as metabolic equivalent-hours per week).27 As additional potential mediators, we considered BMI based on the calculation from self-reported height and weight (scoring of responses in the range ≥25 kg/m2 vs <25 kg/m2 discriminates between respondents with and without high BMI). Socioeconomic status (SES) was entered in our regression models as a potential confounder, that is, theoretically associated with self-rated health and social capital.28 The classification of SES was based on both parents’ occupation at the time when the research was conducted. Self-perceived SES was categorised into three levels as high SES (ie, managers and professionals), middle SES (white collar) and low SES (blue collar),29 and it was dichotomised as high/middle (responses in the range 2–4) and low (responses in the range 5–6). Psychological distress was also assessed as a potential confounder using the six-item Kessler scale by the questions: ‘‘About how often during the past 30 days did you feel nervous?’’, ‘‘During the past 30 days, about how often did you feel hopeless?’’, ‘‘During the past 30 days, about how often did you feel restless or fidgety?’’, ‘‘How often did you feel so depressed that nothing could cheer you up?’’, ‘‘During the past 30 days, about how often did you feel that everything was an effort?’’ and ‘‘During the past 30 days, about how often did you feel worthless?’’.30 Each question is scored from 0 (none of the time) to 4 (all of the time). Scores of the six questions were then summed (0–24), with a lower score indicating low levels of psychological distress. Previous research has shown that dichotomous scoring of responses in the range 13+ versus 0–12 discriminates between respondents with and without significant psychological distress.30 31 A test of interaction between psychological distress and gender is performed and there is statistical gender difference in psychological distress (p<0.001).

Data analysis

We used the generalised estimating equations using an exchangeable correlation matrix with robust SEs in order to correct SEs for clustering. We calculated ORs and 95% CIs for good self-rated health according to levels of perceived social capital. The association of self-rated health with social capital indicators was examined by conducting a multiple logistic regression analysis. Furthermore, the study included students from 20 schools; thus, one of the assumptions was that the measurements within school might not be independent. Therefore, the cluster effect was considered in the analysis. For this purpose, we adjusted the SEs by computing clustered robust SEs for the coefficients. We also performed a χ2 test to check the statistical significance of the gender difference in proportion of boys versus girls reporting good self-rated health. As potential confounders, we entered gender, self-perceived SES and psychological distress. We also included physical activity and BMI as potential mediators of the association between social capital and self-rated health. We investigated the association between self-rated health and family social capital (model 1), neighbourhood social capital (model 2) and school social capital (model 3). Finally, we entered all of these social capital variables simultaneously (model 4) to assess their independent contributions to self-rated health. The interaction term between social capital and gender was not statistically significant so we dropped the sex-stratified analyses. A p value of <0.05 (two sided) was considered statistically significant. All statistical analyses were performed with Stata, V.12 (StataCorp, Texas, USA).

Results

Boys reported a higher percentage of good self-rated health (85.8%) compared with girls (75.6%). Roughly 20% of the participants reported poor health. It is worth noting that the prevalence of psychological distress in girls was twice as high as that in boys. Males do generally have a higher BMI than females (22.95±2.85 kg/m2 vs 20.81±2.54 kg/m2). Most adolescents evaluated their family SES as high/middle (60.2%) with no statistically significant differences between boys and girls. Boys are significantly more physically active than girls. It is worthwhile to note that a χ2 test shows the statistical significance of the gender difference in proportion of boys versus girls reporting good self-rated health (table 1).

Table 1

Characteristics of the study subjects, Zagreb, Croatia, 2014

	Total(N=3427)	Males(N=1688)	Females(N=1739)	p Value*
	N (%)	N (%)	N (%)
Self-rated health
Poor	664 (19.4)	239 (14.2)	425 (24.4)
Good	2763 (80.6)	1449 (85.8)	1314 (75.6)	<0.001
Family social capital
Low	185 (5.4)	90 (5.4)	95 (5.5)
High	3242 (94.6)	1598 (94.6)	1644 (94.5)	0.808
Neighbourhood trust
Low	1104 (32.2)	466 (27.6)	638 (36.6)
High	2323 (67.8)	1222 (72.4)	1101 (63.4)	<0.001
Informal social control
Low	828 (24.2)	419 (24.8)	409 (23.5)
High	2599 (75.8)	1269 (75.2)	1330 (76.5)	0.342
Vertical school trust
Low	1050 (30.6)	485 (28.7)	565 (32.5)
High	2377 (69.4)	1203 (71.2)	1174 (67.5)	0.014
Horizontal school trust
Low	840 (24.5)	339 (20.1)	501 (28.8)
High	2587 (75.5)	1349 (79.9)	1238 (71.2)	<0.001
Reciprocity at school
Low	459 (13.4)	186 (11.1)	273 (15.7)
High	2968 (86.6)	1502 (88.9)	1466 (84.3)	<0.001
Body mass index
Normal	3001 (87.6)	1367 (80.9)	1634 (93.9)
Overweight/obese	426 (12.4)	321 (19.1)	105 (6.1)	<0.001
Self-perceived socioeconomic status
High/middle	2064 (60.2)	1008 (59.7)	1056 (60.7)
Low	1363 (39.8)	680 (40.3)	683 (39.3)	0.706
Psychological distress
High	848 (24.7)	274 (16.3)	574 (33.0)
Low	2579 (75.3)	1414 (83.7)	1165 (67.0)	<0.001
Physical activity
High/moderate	2943 (85.9)	1499 (88.8)	1444 (83.1)
Low	484 (14.1)	189 (11.2)	295 (16.9)	<0.001

*Univariable, χ2 test.

The association between social capital and self-rated health is shown in table 2. Overall, self-rated health was significantly associated with each domain of social capital. Good self-rated health was significantly associated with higher family social capital (OR 2.43; 95% CI 1.55 to 3.80) and higher neighbourhood trust (OR 2.02; 95% CI 1.48 to 2.76). Regarding school social capital, good self-rated health was significantly associated only with perceptions of reciprocity at school (OR 1.79; 95% CI 1.13 to 2.84). When all social capital variables were entered into the model (model 4), good self-rated health was significantly associated with higher family social capital (OR 1.98; 95% CI 1.19 to 3.30), neighbourhood trust (OR 1.77; 95% CI 1.25 to 2.51) and reciprocity at school (OR 1.71; 95% CI 1.08 to 2.73).

Table 2

ORs for good self-rated health among high school students, Zagreb, Croatia, 2014

	Model 1OR (95% CI)	Model 2OR (95% CI)	Model 3OR (95% CI)	Model 4OR (95% CI)
Family social capital
Low
High	2.43 (1.55 to 3.80)***			1.98 (1.19 to 3.30)**
Neighbourhood trust
Low
High		2.02 (1.48 to 2.76)***		1.77 (1.25 to 2.51)***
Informal social control
Low
High		1.42 (0.99 to 2.04)		1.37 (0.95 to 1.98)
Vertical school trust
Low
High			1.34 (0.99 to 1.81)	1.19 (0.87 to 1.62)
Horizontal school trust
Low
High			1.24 (0.83 to 1.84)	1.19 (0.79 to 1.79)
Reciprocity at school
Low
High			1.79 (1.13 to 2.84)**	1.71 (1.08 to 2.73)*
Gender
Male
Female	0.46 (0.29 to 0.73)***	0.49 (0.31 to 0.78)**	0.48 (0.30 to 0.78)**	0.48 (0.30 to 0.79)**
Body mass index
Normal
Overweight/obese	0.44 (0.25 to 0.77)**	0.43 (0.23 to 0.78)**	0.43 (0.23 to 0.79)**	0.42 (0.23 to 0.77)**
Self-perceived socioeconomic status
High/middle
Low	1.01 (0.64 to 1.60)	1.05 (0.66 to 1.68)	1.03 (0.65 to 1.64)	1.05 (0.66 to 1.65)
Psychological distress
High
Low	0.37 (0.24 to 0.57)***	0.38 (0.24 to 0.59)***	0.38 (0.24 to 0.59)***	0.41 (0.26 to 0.64)***
Physical activity
High/moderate
Low	0.64 (0.42 to 1.00)*	0.65 (0.42 to 1.01)*	0.65 (0.41 to 1.01)*	0.66 (0.42 to 1.04)*

These four models were examined in a sequence of four logistic regression models considering clustering for schools.

Model 1: examine the association between family social capital and youth self-rated health adjusting for gender, body mass index, self-perceived socioeconomic status, psychological distress and physical activity.

Model 2: examine the association between neighbourhood social capital and youth self-rated health adjusting for gender, body mass index, self-perceived socioeconomic status, psychological distress and physical activity.

Model 3: examine the association between school social capital and youth self-rated health adjusting for gender, body mass index, self-perceived socioeconomic status, psychological distress and physical activity.

Model 4: examine the association between all social capital variables and youth self-rated health adjusting for gender, body mass index, self-perceived socioeconomic status, psychological distress and physical activity.

*p<0.05, **p<0.01, ***p<0.001.

Table 3 presents the association between family, neighbourhood and school social capital with good self-rated health among high school students. Good self-rated health was significantly associated with higher family social capital (coefficient 0.88; 95% CI 0.35 to 1.42), higher neighbourhood trust (coefficient 0.70; 95% CI 0.34 to 1.06) and higher norms of reciprocity at school (coefficient 0.58; 95% CI 0.13 to 1.04). When all types of social capital were entered simultaneously, good self-rated health remained significantly associated with higher family social capital (coefficient 0.68; 95% CI 0.13 to 1.23), neighbourhood trust (coefficient 0.57; 95% CI 0.20 to 0.94) and reciprocity at school (coefficient 0.54; 95% CI 0.08 to 0.99).

Table 3

Coefficients for good self-rated health associated with family, neighbourhood and school social capital among high school students, Zagreb, Croatia, 2014

	Model 1Coefficient (95% CI)	Model 2Coefficient (95% CI)	Model 3Coefficient (95% CI)	Model 4Coefficient (95% CI)
Family social capital
Low
High	0.88 (0.35 to 1.42)***			0.68 (0.13 to 1.23)**
Neighbourhood trust
Low
High		0.70 (0.34 to 1.06)***		0.57 (0.20 to 0.94)**
Informal social control
Low
High		0.35 (−0.03 to 0.74)		0.31 (−0.07 to 0.71)
Vertical school trust
Low
High			0.29 (−0.09 to 0.69)	0.17 (−0.22 to 0.57)
Horizontal school trust
Low
High			0.21 (−0.21 to 0.65)	0.17 (−0.25 to 0.61)
Reciprocity at school
Low
High			0.58 (0.13 to 1.04)**	0.54 (0.08 to 0.99)*
Gender
Male
Female	−0.77 (−1.18 to −0.36)***	−0.70 (−1.11 to −0.29)***	−0.71 (−1.12 to −0.30)***	−0.71 (−1.12 to −0.30)***
Body mass index
Normal
Overweight/obese	−0.81 (−1.28 to −0.34)***	−0.83 (−1.30 to −0.36)***	−0.83 (−1.31 to −0.36)***	−0.84 (−1.32 to −0.36)***
Self-perceived socioeconomic status
High/middle
Low	0.01 (−0.35 to 0.38)	0.05 (−0.31 to 0.42)	0.03 (−0.33 to 0.40)	0.04 (−0.32 to 0.42)
Psychological distress
High
Low	−0.97 (−1.33 to −0.60)***	−0.96 (−0.32 to −0.59)***	−0.95 (−1.32 to −0.59)***	−0.87 (−1.24 to −0.50)***
Physical activity
High/moderate
Low	−0.43 (−0.87 to 0.00)*	−0.42 (−0.86 to 0.01)*	−0.42 (−0.87 to 0.01)*	−0.40 (−0.85 to 0.35)

These four models were estimated by generalised estimating equations using an exchangeable correlation matrix with robust SEs.

Model 1: examine the association between family social capital and youth self-rated health adjusting for body mass index, self-perceived socioeconomic status, psychological distress and physical activity.

Model 2: examine the association between neighbourhood social capital and youth self-rated health adjusting for body mass index, self-perceived socioeconomic status, psychological distress and physical activity.

Model 3: examine the association between school social capital and youth self-rated health adjusting for body mass index, self-perceived socioeconomic status, psychological distress and physical activity.

Model 4: examine the association between all social capital variables and youth self-rated health adjusting for body mass index, self-perceived socioeconomic status, psychological distress and physical activity.

*p<0.05, **p<0.01, ***p<0.001.

Discussion

Previous studies in the USA and Europe have suggested that higher levels of informal social control were associated with higher levels of perceived health.10 Drukker et al32 found that higher levels of community informal social control in the Netherlands may directly prevent young people from engaging in deleterious health behaviours as well as indirectly provide them with self-confidence and a sense of protection. Furuta et al7 have shown that the association of social capital with self-rated oral health is not uniform; higher trust is associated with better oral health, whereas higher informal control in the community is associated with worse oral health.

To better understand the findings of this research, it is very important to briefly explain the Croatian social context and the theoretical approach to young people. The mid-1980s in Croatia was a period of socialism before the collapse of the Soviet bloc. The first decade of the 1990s was more turbulent in Croatia compared with other post socialist countries. Croatia experienced armed conflicts that lasted for several years. Finally, at the beginning of the 1990s with the state's declaration of independence, and the abolition of the totalitarian regime, the nation became one of many transition countries. Young people in Croatia are one of the population segments most rapidly affected by these processes and changes. The reasons for this are multiple and related to the timing of political transition, with the transition from childhood to adulthood.33

In this study, we have found a statistically significant association between higher levels of family social capital and higher self-rated health. For young people, family should be important for ‘being there’ in times of need and family members are often regarded as a crucial source of support.34 Morgan and Haglund35 reported that a sense of belonging in family was related to self-rated health and health behaviours in adolescents. In transitional societies, the changes in hierarchical order and value structures accompanying the shift from socialism to free markets meant that families became especially important as a source of social support.36

We also found that those living in high-trust communities reported better health compared with youth living in low-trust communities. According to surveys, Croatian youth frequently spend their time with friends in the neighbourhood engaged in sport or other activities (ie, watching TV and videos, listening to the radio).33

The indicators based in the school social environment suggested that higher reciprocity at school (collaboration between pupils) was associated with higher self-rated health, whereas vertical and horizontal social capital were not significantly associated with self-rated health. Spending time with peers at school may engender a sense of belonging,34 and it may promote better health. The data show that 78% of Croatian youth frequently talk to their school peers about going out and leisure, music, movies and books.33 A previous study in Denmark found that school connectedness and a sense of belonging may have a strong impact on adolescent psychological health.37

Our study has some limitations. First, owing to the cross-sectional design, we cannot exclude the possibility of reverse causation, that is, poor health led to a low level of trust and other indicators of social capital. To mitigate this, we adjusted for psychological distress. In other words, the students with psychological distress would report lower levels of social capital in all three domains simultaneously and psychological distress could potentially affect their self-ratings of health. Therefore, we adjusted for individual differences in psychological distress in order to rule out this possible bias. When we included an interaction term between psychological distress and gender, its coefficient was statistically significant. However, we observed no substantial change in the association between social capital indicators and good health. Moreover, the differential effect estimates of each type of social capital on health cannot be fully explained by reverse causation. Second, we used a subjective measure of health and social capital, and therefore there is a possibility of common method bias which may have resulted in bias away from the null. Again, the differential findings for each type of social capital suggest that this is less likely. Third, since the students responded to the questionnaires during the class, there is a possibility of measurement error of school social capital, in particular vertical social capital. Additionally, the possibility of type 1 error is high because of the clustering. We used the generalised estimating equations using an exchangeable correlation matrix with robust SEs in order to correct SEs for clustering. Fourth, the social capital variables in our study are analysed at the individual level. Therefore, we are referring to the students’ individual perceptions of social capital. Unfortunately, we did not have information about which neighbourhoods the respondents lived in nor about which classrooms the students attended within each school. We cannot fully disentangle the effects of school social capital and neighbourhood social capital in this study, partly because of the lack of information about class and neighbourhood. And fifth, all types of social capital were assessed in the primary sample. Future studies are warranted to assess all three domains (family, neighbourhood and school social capital) by approaching different sample subjects who are not participating in the primary sample.

This study shows that higher levels of family social capital, neighbourhood trust and reciprocity school (ie, collaboration relationships between pupils) were associated with better health among youth. Interestingly, the interaction term between social capital and gender was not statistically significant in this study, although there are some researches showing that, among adolescents, girls tend to report higher levels of social capital, especially school and family belonging than do boys.38 We can speculate that this was found since adolescent girls have a greater number of friends than do boys; they expect and desire more nurturing behaviour from their friends and family members, and experience more empathy, more self-disclosure and less overt hostility in their friendships than do boys.39 Additional studies are needed to identify interventions that can increase social capital to engender healthy habits with the ultimate goal of achieving healthier students. More studies exploring social capital and health in different countries should be conducted since social capital in general and in particular levels of informal social control may depend on different cultural norms and values.40 41