Neighborhood social capital is associated with participation in health checks of a general population: a multilevel analysis of a population-based lifestyle intervention- the Inter99 study

In this paper we found that higher level of neighborhood social capital was associated with higher probability of participating in the health check phase of a population-based lifestyle intervention. Most of this association was explained by individuals residing in deprived neighborhoods and those with lower socioeconomic position also living in neighborhoods with low social capital.

The results of this paper bring new insights to the evidence on the linkage between social capital and health behavior. To our knowledge no other study has examined the effects of social capital on participation in general health checks. Our results are in line with studies demonstrating a link between social capital and other health maintenance behaviors [6, 21, 23], indicating that high social capital promotes positive health behaviors. One explanation of the results of this paper may be that residents of deprived neighborhoods are more disconnected from sources of information, or lack the social reinforcement and social support to attend a health check. In a previously published paper it was shown that neighborhood deprivation predicts participation above and beyond the effects of individual socioeconomic position [2]. However, in this paper we find no significant effects of neighborhood deprivation in the model that included social capital and individual socioeconomic position. This result may be explained by all of the effect of neighborhood deprivation on health check participation being mediated through social capital.

One argument against our interpretation is that voting participation and participation in a health check are tapping into the same thing – that is, both are measuring a tendency toward higher civic engagement. We argue, however, that political participation and participation in a health check are not interchangeable with each other. Our theoretical model is that the quality of neighborhood social relations (i.e. social capital) is “upstream” of both political participation and health check participation. According to this view, community social capital is the force that mobilizes residents to both get out and vote, as well as to participate in health check. Of course, we hasten to add that factors other than social capital also drive health check participation, such as health status, interest in health, family disposition for ischemic heart disease, previous experiences with the health care system and perceived authority of researchers and health care professionals. We did not consider these factors potential confounders as they are unlikely to influence the level social capital.

If our theoretical model is correct, then strengthening neighborhood social bonds may prove to be a viable approach for boosting the rate of participation in community-based health checks and health promotion interventions. If an intervention was able to increase the neighborhood social capital, measured as either informal socialization or voting turnout, from the lowest to the highest level, we would expect study participation to increase by 9-11 %, after taking into account differences in individual socioeconomic position. We still have limited knowledge on how social capital could be mobilized in local communities. However, a recently published study from Sweden showed that increasing social capital through human interaction is likely to have health promoting effects [3]. The authors suggested a number of interventions which included building neutral meeting places, green areas and walking-friendly neighborhoods.

In this paper, we adopted two measures of neighborhood social capital; informal socializing and voting turnout. A limitation of the measure of contact frequency with friends/family is that non-respondents did not contribute to the mean. Including information from non-participants would have produced more precise estimates, however the method did not seem to bias the results, as the residuals of the outcome did not vary according to the level of informal socializing. Furthermore, if there is a high correlation between participant’s individual social network and that of their neighbors, the measure may in fact not reflect a collective measure of informal socializing but, rather than an average of that of the residents. Thus, even though we demonstrate that high stocks of social capital concentrated within a neighborhood are correlated with participation in a health check, we cannot be sure that this is explained by the social capital of a collective nature. Least, though we assume that high stocks of social capital, reflected in a high contact frequency with family, is likely to raise the stocks of social capital of the neighborhood in general, this is not documented in detail in the literature. Additional statistical analyses (Additional file 1) in which the the measure of informal socializing is based only on contacts with friends and acquaintances show a slight decrease in the correlation between informal socializing and participation in the health check. This result is as expected, as contact with persons within both domains; family as well as friends is likely to contribute to the accumulated social capital. Lack of contact with friends may therefore be buffered from frequent family contacts [19]. Our study lacked information on cognitive aspects of social capital such as levels of interpersonal trust, perceptions of solidarity and cohesion, and willingness to help others. Second, information on individual social capital would have been relevant to include in the analysis. Again, this information was not assessable from non-participants. Including social capital at both the individual and neighborhood levels would have provided a more complete picture of the multi-level influences of social capital on our outcome.

Thanks to the access to national registers with information on all residents of the 73 neighborhoods covering the study catchment area, we were able to calculate aggregated neighborhood measures of social capital in the Inter99 study. The Danish national registers have high quality and validity [26] and such analysis is only possible in Scandinavian countries. The Inter99 study is suitable for conducting multilevel analysis with many covariates due to the large size and relatively large catchment area. Furthermore, using census districts as the neighborhood measure has several strengths. The borders of the districts are typically based on boundaries following the physical division of larger roads, division of urban and rural areas and following borders of housing associations. Because census districts are often equivalent to school districts that represent small communities distinct from one another and the relatively small size of the neighborhoods the dilution of neighborhood effects is minimized. Though the neighborhoods were sharing borders, the population density varied substantially between neighborhoods. Still, we found that including a measure of population density in the statistical model did not change the results significantly (data not presented). While most previous studies looking at effects of social capital adjust only for individual socioeconomic position, we also included neighborhood deprivation in our analysis. This illustrated that the pathway from neighborhood deprivation to participation in a health check is almost entirely mediated through social capital.

A limitation of this study is the missing data on census districts as well as missing data on educational attainment. As census areas are based on 2006 data, changes in road names during the eight year period from study start could be an explanation for the missing data on census district. In all of the 11 municipalities covering the catchment area there were persons with missing census districts, supporting this assumption. When compared to the rest of the study population, a larger proportion of the persons with missing data on education were out of work force and fewer participated in the intervention. However, there were no clear differences in regards to prevalence of ischemic heart disease, income, sex and age distribution. There exists no automatic registration on immigrants’ education level which explains most of the missing data on education. Our analyses are cross-sectional; as social capital was measured at approximately the same point in time as was participation. Voter data from the 2001 parliament election was obtained after the Inter99 health check. However, an analysis comparing the voting turnout at the election for the Danish parliament on March 11^th 1998 with that from 2001 showed a 0.96 correlation at the municipality level. Unfortunately, we did not have asses to voting data at the voting district level from the 1998 election; this would have assured that voting turnout was measured temporality before participation in the Inter99 health check. Nonetheless, we are confident that neighborhood characteristics at the census level are stable over time and we find it unlikely that the Inter99 study should have influenced the neighborhood voting turnout, wherefore we do not believe reverse causality to cause serious bias.