Despite widespread use of social capital in research and public health interventions to decrease HIV transmission, we believe this to be the first report to describe levels of total social capital in a large, international sample of PLWH. This study helps fill a number of gaps in the literature including providing a description of levels of social capital and its correlates in an international sample of PLWH. With further development our findings can be used to help develop evidence-based, public health interventions, seeking to modify social capital among PLWH.
Our participants reported a higher than average total social capital score compared with previous research using this scale. While Bullen and Onyx (2000) do not give guidance on how to interpret the mean score, previous investigations of individual-level social capital using the Social Capital Scale have classified a mean social capital score greater than 2.5 as high social capital and anything less than 2.5 as low social capital . In our study, we had an aggregate total social capital score of 2.68, with all individual sites reporting at least a mean total social capital score of 2.55. There are several plausible explanations for our higher than average social capital scores. All participants were recruited through HIV clinics or through AIDS service organizations in urban settings. By virtue of this recruitment method, participants were already engaged in their health care and possibly with related social services. This level of access, could lead them to perceive more trust in organizations, and to perceive that they have access to more social resources than other PLWH who are not as engaged in their health care. These factors could constitute a higher level of social capital. Additionally, due to anticipated low employment in our sample (22% of participants were employed) we did not include five work-related items on the original Social Capital Scale, which may have skewed our results upward. This is similar to the approach that Onyx and Bullen adopted, when addressing significant levels of unemployment in their original sample . We believe this is an appropriate approach to the measurement of social capital given the sociodemographic composition of PLWH around the globe. However, taken together, our findings support recent evidence suggesting that PLWH may not be as marginalized as previously argued [3, 14, 15] and that the relationship between HIV and underlying structural factors in society is complex . But as we discussed above, most of this literature is based on persons at risk for HIV infection not those currently living with HIV/AIDS, thus more evidence on social capital in PLWH is needed before drawing final conclusions in this regard.
The measurement of social capital is challenging [42, 55]. Consistent with our aims, we assessed individual-level social capital using Onyx and Bullen's Social Capital Scale, but psychometric properties for this scale in PLWH were lacking. We observed evidence of reliability and validity of a modified Social Capital Scale in an international sample of PLWH. Our analysis of the reliability of the Social Capital Scale indicated that the scale measures a single latent construct of individual-level social capital among all sites, suggesting this scale is a reliable assessment of social capital in PLWH. However, we found differences between our sample and Onyx and Bullen's original data, when we examined the scale's validity. Our data support a five factor solution explaining 65% of the variance in total social capital, in contrast to Onyx and Bullen's original eight factor solution explaining 49% of variance . These differences lie in two factors upon which the items did not load: the value of life factor and the social agency or proactivity in a social context factor. These results are surprising because the factor of social agency was one of the more explanatory factors that Onyx and Bullen found in their original scale development work. In our study, the value of life items loaded on the friends and family connections factor, suggesting that participant's perceived self-value, may be related the friends and family connections. Our observation is interesting because it harkens back to Durkheim's work on social isolation, anomie, and suicide, in that those who are more socially connected may perceive their life to be of more value and may take action to improve their health [56, 57]. For PLWH, this may also translate into engaging in other risk reducing behaviors including antiretroviral therapy adherence. Another difference we observed was that items that were originally loaded on what Onyx and Bullen described as social agency, or proactivity in the social context factor, loaded on two different factors including, friends and family connections and tolerance of diversity. This may have been explained by our study samples being drawn from sites where they may perceive themselves and their peers as members of a proactive social context. They may perceive these connections as bonds between friends and families. These findings also suggest that individual-level social capital may be heavily based on the personal connections with friends and family, and the resources they provide . This theory is also supported by the strong correlation between perceived social support and social capital in our sample and suggests interventions to build these connections, i.e. family or peer group-based interventions, may be helpful in facilitating behaviors to enhance the health of PLWH [58–60].
In summary, among our sample of PLWH, we observed that more of the items on the Social Capital Scale appeared to represent a subscale of friends and family connections, followed by participation in the local community, which suggests a modification to the factor structure for anyone wishing to explore the individual factors or subscales in this population in the future. However, in our analyses, we only examined relationships between social capital and health-related outcomes using the total Social Capital Scale. This alternative strategy is advantageous because it can identify factors that contribute to health outcomes among PLWH. For example, those who perceive themselves to be healthy and in possession of social capital may be empowered to collaborate with public health researchers, clinicians, and policy-makers to participate in HIV prevention, HIV treatment, and health promotion strategies [61, 62]. Challenges to this strategy may be that a focus on physical health and biomedical interventions only first limits our understanding of other complex factors that are critical for individuals to access and use the social capital available to them and their community.
The moderate relationships we observed between total social capital score and self-reported physical and psychological health condition underscore the importance of perceived social resources and trust in organizations when assessing personal health. Previous investigators examining the relationship between perceived health and social capital observed similar findings in large, national samples . It is possible that this observed relationship may, again, be attributed to our sampling methods, but this does not diminish the implications of these findings. Since participants were already engaged in health care, they may have had more trust in social organizations and access to necessary social resources, than persons who are less engaged with the health care system. This, in turn, gives participants an avenue from which they more easily receive information about their health and to trust that the information will be helpful and not harmful. This may increase the individual's likelihood of enacting this received health information and will allow them to more efficiently address any deviations from perceived "good health". These findings suggest a potential role for social capital in public health interventions targeting health and wellness in PLWH around the globe. This could include refining existing interventions to help PLWH and their communities build trust in medical and social service organizations before recommending challenging health-related behavioral changes including medication adherence, dietary and physical activity changes, and decreasing substance use [61, 62, 64, 65]. Our observations clearly suggest individual health is one essential element in the complex web of social and structural factors that constitute social capital and the overall health and wellbeing of PLWH around the globe.
While this study has several advantages, including filling a significant gap in the literature, there are limitations that must be considered by the reader. The first limitation is that we used a convenience sampling method, not random sampling. Therefore, our data are only representative of the samples surveyed and the information cannot be extrapolated to the entire population in any country where the samples were obtained. With the exception of the United States, every country only had one site where data were collected. Therefore, it would be inappropriate to base conclusions about an entire country on a single site, especially when considering the geographic size and demographic diversity of the countries studied (Canada, China, Namibia, and Thailand). However, even though country-level extrapolation is not appropriate, our study is among the first to describe individual-level social capital in some of these sites, which allows for tempered cross-national comparison. An additional limitation may be the modification of the Social Capital Scale by the removal of the 5 work-related items. While this strategy was similar to the one adopted by Onyx and Bullen, it is possible that this strategy could have led to an upward bias in our overall summary statistics. Finally, most of our data collection sites were in the United States and in our summary statistics, the U.S. estimates exerted more weight, leading to an overall U.S. bias in these statistics. These analyses also assume that there is a level of homogeneity among the participants simply because they are all PLWH, which may be an unjustified assumption. We tried to address these concerns by providing the data at both individual site level and at the country level. Additionally, to better determine the influence that country may have had on our results, we explored this issue with multiple regression analyses (Additional file 4: Table S4). These results indicate that, despite the overrepresentation of participants living in the United States, country of origin does not influence our model. Thus, the risk of U.S. bias on results appears to be minimal.