The present study investigated whether subjective neighbourhood social capital and general social trust were associated with self-reported health in a large population of Swedish adolescents. The main finding was that low neighbourhood social capital and low general social trust were associated with higher rates of psychosomatic symptoms, musculoskeletal pain and depression. The association between general social trust and ill health was particularly evident, where the group with low general social trust had a more than three times increased odds of depression, a more than three times increased odds of having many psychosomatic symptoms, and more than double the odds of having many symptoms of musculoskeletal pain compared with the group with high general social trust. The results were controlled for potential confounding factors such as subjective family socioeconomic status, type of residence, parental unemployment and ethnicity, as well as clustering effects of school and housing area.
It is interesting that even in a highly egalitarian country such as Sweden there is an association between neighbourhood social capital, general social trust and ill health in an adolescent population. Our results agree with previous findings of relations between social capital and ill health among adolescents [37, 39, 40, 43]. Social capital has been suggested as one important mediating factor of the relations between income inequality and ill health, as low socioeconomic status and income inequality result in less trusting and reciprocal relationships between citizens, and lower levels of civic and political participation [25, 28–30]. However, the debate regarding the concepts of social capital and social trust and how they are supposed to be measured is not settled. The different schools of sociology, either employing a consensus or a conflict perspective, or a macro or micro perspective, always attract severe criticism from the opposing side . The perspectives employed by sociologists can be organized into groups based on similarities and differences in their theoretical assumptions. For example; Putnam and Bourdieu have different definitions of social capital. Putnam describes social capital by using a functionalist, consensual perspective, whereas Bourdieu's definition of social capital is based more on conflict and exploitation. We aimed to determine whether two forms of bonding and bridging social capital - in this study referred to as "neighbourhood social capital" and "general social trust" - were related to poor health among adolescents. Our measures are in line with Putnam's description of social capital, although this way of using the concept of it can always be criticised.
Several limitations of the study should be noted. Firstly, the analyses are based mainly on self-reports which involves a risk of information bias due to false or inaccurate responses from the participants. Regarding self-reports of health, girls may be more aware of, and/or more inclined to report ill health which might have influenced the higher rates of it in this group. However, since we were interested in associations between individual perspectives of neighbourhood social capital and general social trust, self-reports were deemed the most convenient choice of measurement. Secondly, there is the problem of causality regarding neighbourhood social capital, general social trust and ill health, as the cross-sectional design of the study involved no possibility to distinguish the directions of cause and effect. Individuals suffering from pain and depression may be more inclined to interpret their environment and relations to other individuals in society in a negative way, which might have influenced the associations between neighbourhood social capital, general social trust and ill health. However, when analysing levels of self-reported neighbourhood social capital and general social trust in different housing areas, these plainly differed depending on the income distribution of each housing area as obtained from Statistics Sweden. A higher objective socioeconomic status of a housing area was correlated with a higher level of neighbourhood social capital, a higher level of general social trust, and a higher level of subjective socioeconomic status. Thus, the differences in self-reported neighbourhood social capital and general social trust were associated with the objective measure of socioeconomic status and would not merely be a question of negative interpretation of social relations and society as a cause of illness.
Moreover, even though there were substantial overlaps between the dependent variables (R
≈ 14-34%) there were just minor correlations between the dependent and independent variables (R
≈ 2-6%) when analysed as simple correlations (Table 2). However, in our multivariate models we found considerable odds elevation of psychosomatic symptoms, musculoskeletal pain, and depression (30-330%) in relation to neighbourhood social capital and general social trust. We therefore interpret our findings as being in line with those previously reported among adolescents [20, 37, 39, 40, 43, 44], and suggest that neighbourhood social capital and general social trust may be related to specific health diagnoses among adolescents.
Thirdly, we used different cut-off points for the outcome variables compared with the independent variables. The reason for this is that we did not want the measures of low and high neighbourhood social capital and general social trust to be too strict, and consequently used quartiles (highest and lowest 25%) as cut-offs. However, for the outcome measures of psychosomatic symptoms and musculoskeletal pain, we wanted to identify the participants with the most problems and therefore used the stricter cut-offs of standard deviations (approximately 16% of the individuals with the highest scores). However, the measurement of depression had high sensitivity and low specificity, providing a risk for false-positive classifications. The Depression Self-Rating Scale of the DSM-IV, A-criterion, involves the risk of including participants who would be ruled out according to the more strict B-, C-, D-, and E-criteria. This scale has, nevertheless, been proved to be a useful instrument for defining major depression . The outcome scales were, moreover, used in two different ways: as a summation index of reported symptoms in the linear regression analyses, and as dichotomous variables for the logistic regressions. The dependent variables of the indexes of psychosomatic and musculoskeletal symptoms, as well as depression, were skewed. One way of dealing with this kind of problem would be to transform the data, e. g., by a log or log-log transformation. In the present study, however, neither the log- nor the log-log transformation produced a symmetric distribution of the data, and were therefore not optimal for parametric modelling methods. The procedure with complementary statistical approaches can help to overcome shortcomings with the individual statistical methods and help to eliminate scaling artefacts. Both the hierarchical multi-level linear regressions and the logistic regressions showed similar results, in the same direction, which added further support for the presented findings.
The present study also has several strengths, particularly regarding the large population based sample of adolescents and the high participation rate which might offer an opportunity to generalise the results to other adolescent populations as well. Most previous studies have examined relations between social capital and health in adult populations. The present study focuses on adolescents, to examine whether social capital may be associated with health outcomes among young people as well. Moreover, the few previous studies of social capital that have focused on adolescent populations have often chosen specific areas of low socioeconomic status and high deprivation for their analyses. The present study involves a large adolescent population from a county that is considered to be representative of Sweden as a whole because of its distribution of educational, income, and employment levels as well as urban and rural areas . Thus, our study contributes important information to the research field by its generalizability to other adolescent populations. Further, previous studies have often used general self-rated health as an outcome, which is not a diagnostic instrument but rather a subjective feeling of whether one's health is good or bad [29, 40, 43, 44]. The use of a validated diagnostic instrument for major depression, and specific questions regarding musculoskeletal pain and psychosomatic symptoms, allows for an evaluation of actual differences in the prevalence of specific common health problems as a complement to previous studies of subjective views of general health.