Using multi-level data to estimate the effect of social capital on hazardous alcohol consumption in the former Soviet Union
Abstract
Background: Hazardous alcohol consumption is a leading cause of mortality in the former Soviet Union (fSU), but little is known about the social factors associated with this behaviour. We set out to estimate the association between individual- and community-level social capital and hazardous alcohol consumption in the fSU. Methods: Data were obtained from Health in Times of Transition 2010, a household survey of nine fSU countries (n = 18 000 within 2027 communities). Individual-level indicators of social isolation, civic participation, help in a crisis and interpersonal trust were aggregated to the community level. Adjusting for demographic factors, the association of individual- and community-level indicators with problem drinking (CAGE) and episodic heavy drinking was estimated using a population average model for the analysis of multi-level data. Results: Among men, individual social isolation [odds ratio (OR) = 1.20], community social isolation (OR = 1.18) and community civic participation (OR = 4.08) were associated with increased odds of CAGE. Community civic participation (OR = 2.91) increased the odds of episodic heavy drinking, while community interpersonal trust (OR = 0.89) decreased these odds. Among women, individual social isolation (OR = 1.30) and community civic participation (OR = 2.94) increased odds of CAGE. Conclusion: Our results provide evidence of the role of some elements of social capital in problem drinking in the fSU, and highlight the importance of community effects. The nature of civic organizations in the fSU, and the communities in which civic participation is high, should be further investigated to inform alcohol policy in the region.
Introduction
The former Soviet Union (fSU) region experienced a sharp decline in life expectancy in the 1990s, from which it has yet to fully recover.1 Although there is now compelling evidence that alcohol has played a major proximal role in this mortality crisis,2 driven by rapid social change,3 the factors determining individual vulnerability, or conversely, resilience, are still being worked out in detail. A recent systematic review of research from the fSU on social factors and alcohol consumption found little on the role of commonly studied factors such as education and income, with what exists providing mixed results, and no published research examining the role of the social environment on consumption.4
One social factor that has recently gained attention in public health research from other regions is ‘social capital’, defined as ‘those features of social organization—such as density of civic associations, levels of interpersonal trust and norms of reciprocity—that act as resources for individuals, and facilitate collective action’.5 Specific mechanisms via which social capital may affect health, such as by reducing the negative impacts of stress,6 or facilitating the dissemination of health-related information,7 have been hypothesized.8 With regard to health behaviours specifically, the hypothesis that communities with higher levels of social capital are better able to exercise social control over health behaviours7 has found some empirical support in evidence linking elements of social capital (namely, civic engagement, trust and social support) to individual health behaviours,8 including alcohol consumption.9,10 Further research showed that the association between social capital and mortality was attenuated when differences in health behaviours were accounted for, suggesting that health behaviours may mediate the effect of social capital on overall health.11
While consensus regarding the importance of social capital in health behaviour research has grown, there is persisting disagreement in the literature as to whether social capital should be treated as an individual attribute or a collective one (e.g. at the level of the community or state).12 In their summary of the various conceptualizations of social capital in public health research, Kawachi et al.12 argue that the most theoretically appropriate level for analysis of its association with health is both the individual and collective level, within a multi-level framework. They provide evidence for the legitimacy of aggregating individual survey responses to obtain collective measures of social capital,12 an approach now commonly used.10 Several studies have found a positive association between aggregate social capital measures and individual health outcomes;13 however, as pointed out by d’Hombres and colleagues,14 these studies did not simultaneously include individual-level measures of social capital, thereby failing to eliminate the possibility that the positive effect of community-level social capital was due to its positive correlation with individual-level social capital.14 Some subsequent studies that measured both individual- and community-level social capital simultaneously found no residual association between community-level social capital and health once individual-level social capital was adjusted for, leading d’Hombres and colleagues to conclude that ‘community social capital does not have an independent effect on self-reported health’ once individual-level social capital is accounted for and therefore ‘affects health only indirectly’;14 however, studies from elsewhere have reported independent associations between community-level social capital and self-reported health,15 as well as alcohol consumption.10
That social capital, either at the community or individual level, might have an effect on alcohol consumption in the fSU is plausible, given what we know of the region. The Soviet regime suppressed civil society, leading its citizens to rely on informal networks, such as friends and family, for financial or other means of support, leaving socially isolated individuals vulnerable.5 This lack of social capital has been linked to worse health outcomes generally among individuals in Russia,5 elsewhere in the fSU14 and in the wider post-communist world;16 however, any association between social capital and alcohol consumption in the region has not yet been explored. Recognizing the need for research on social determinants of alcohol consumption in the fSU, and building on existing evidence of the specific role of social capital in health in the fSU, and in alcohol consumption elsewhere, we set out to analyse the association between individual- and community-level social capital and hazardous alcohol consumption in nine fSU countries.
Methods
Data
Data were obtained from the Health in Times of Transition 2010 study (HITT). The HITT conducted nationally representative surveys in nine fSU countries—Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Russia and Ukraine—between March and June 2010 (data collection in Kyrgyzstan was postponed by one year due to political violence). Multistage random sampling with stratification by region and rural/urban settlement type was used; within each primary sampling unit (PSU; local administrative unit), households were selected by standardized random route procedures. Using a standardized survey instrument, trained fieldworkers interviewed survey participants in their homes. The response rates for the HITT varied from 47.3% in Kazakhstan to 83% in Moldova. There were 1800 respondents in each country, except in Russia (3000) and Ukraine (2200) to reflect the larger and more regionally diverse populations in these two countries, and in Georgia (2200) where a booster survey of 400 additional interviews was undertaken in November 2010 to ensure a more representative sample. The final sample size was N = 18 000.
Measuring social capital
Social capital is still a relatively new concept in public health research and, as yet, there is no standard approach to its measurement. One model regards social capital as consisting of two components: a structural component, which includes the ‘extent and intensity of associational links and activity’, and a cognitive component, which includes ‘perceptions of support, reciprocity, sharing and trust’.17 Using this framework, we operationalized social capital using the following indicators: social isolation (structural), active civic participation (structural), having someone to turn to for help in a crisis (cognitive) and interpersonal trust (cognitive). Using self-reported survey responses, ‘social isolation’ (‘How often do you feel lonely?’) and ‘interpersonal trust’ (‘To what degree do you feel that people can be trusted?’) were measured as continuous variables, while ‘help in a crisis’ (‘Is there anyone who you can really count on to help you out in a crisis?’) and ‘active civic participation’ (‘Are you an active member of at least one of these organizations?’) were measured as binary variables. More detailed information on the survey questions used and response options can be found in Appendix table 1.
Characteristica | Frequency (%)b |
---|---|
Gender | |
Male | 7828 (43.5) |
Age category | |
18-29 | 5042 (28.0) |
30-39 | 3411 (19.0) |
40-49 | 3380 (18.8) |
50-59 | 2755 (15.3) |
60+ | 3410 (19.0) |
Marital status | |
Married | 11129 (62.1) |
Single | 3691 (20.6) |
Divorced | 1152 (6.4) |
Widowed | 1962 (10.9) |
Religion | |
Muslim (vs. Non-Muslim) | 4436 (24.7) |
Education | |
Incomplete secondary or lower | 2345 (13.1) |
Incomplete higher or lower | 11543 (64.3) |
Complete higher | 4066 (22.65) |
Occupation | |
Employed | 15766 (88.2) |
Unemployed (not seeking work) | 608 (3.4) |
Unemployed (seeking work) | 1499 (8.4) |
Household economic status | |
Very bad/bad | 3616 (20.3) |
Average | 10195 (57.3) |
Very good/good | 3984 (22.4) |
Place of residence | |
Urban (vs. Rural) | 10864 (60.4) |
Smoking status | |
Smoker (vs. Non-smoker) | 4642 (25.8) |
Social isolation | |
Never | 8454 (47.6) |
Rarely | 3723 (21.0) |
Sometimes | 3892 (21.9) |
Often | 1702 (9.6) |
Active civic participation | |
Yes (vs. No) | 1,149 (6.4) |
Help in a crisis | |
Yes (vs. No) | 16233 (91.5) |
Interpersonal trust | |
1 (Low) | 954 (5.4) |
2 | 781 (4.4) |
3 | 1638 (9.2) |
4 | 2079 (11.7) |
5 | 3674 (20.7) |
6 | 2912 (16.4) |
7 | 2650 (14.9) |
8 | 1950 (11.0) |
9 | 680 (3.8) |
10 (High) | 487 (2.7) |
We used the PSUs in the HITT survey to represent communities: 2027 PSUs were included with approximately 8–10 individuals per PSU. To estimate simultaneously the association between community-level and individual-level social capital and our outcomes of interest, we followed standard multilevel practice18 and introduced both the individual-level score as well as the average of all scores in the community into the linear predictor. Unlike recent studies of social capital and health that have used ‘self-excluded’ measures of community-level social capital (i.e. the individual’s score is not included in the average community-level score),15 we used a ‘self-included’ measure (i.e. the individual’s score is included in the average community-level score). This approach decomposes the collective effect of social capital into its within- and between-group components, and allows us to estimate the expected changes in hazardous drinking of individual i in community j associated with a unit change in individual-level social capital, expressed as a deviation from the community mean, and with a unit change in community-level social capital, respectively.
Measuring hazardous alcohol consumption
Two measures of hazardous drinking were used. The first used a validated standard measure of problem drinking—the CAGE 4-item questionnaire for assessing alcohol dependence.19 Cronbach’s alpha for the CAGE questionnaire in the HITT data was 0.75.
The second measure- episodic heavy drinking (EHD)- is more specific to the post-Soviet context where this pattern of drinking is widespread and a major driver of mortality, being linked to increased risk of sudden cardiac death20 as well as injury and violence.21 It is particularly common among working-age men in the fSU.22 As noted by Pomerleau et al.,22 researchers in countries of the fSU have used different definitions of EHD; for consistency we use Pomerleau et al.’s definition (i.e. >2 L of beer, 750 g of wine or 200 g of strong spirits on one occasion).
Statistical analysis
Our dataset consists of 18 000 individuals nested in 2027 communities, thus calling for a modelling approach that accounts for the non-independence of individuals within the same community. Though much of the recent research on community-level social capital and health has favoured ‘random effects’ multilevel modelling,12,13,15 which uses maximum likelihood estimation, we have opted for a ‘population average model’, which uses a generalized estimating equation approach. Applied to our research question, a random-effects approach would provide us with an estimate of the average odds ratio (OR) of CAGE problem drinking or EHD associated with a unit change in social capital, within a given community (i.e. comparing two individuals from the same community, or from two communities with all relevant characteristics equal), whereas the population-average approach provides an estimate of the ORs of these outcomes associated with a unit change in social capital across all communities (i.e. comparing two individuals taken from the whole population, irrespective of community). The latter provides estimates that represent average effects over the whole population, and so reflects population-level changes in social capital. Such models require fewer assumptions than the corresponding random-effects models in terms of the distribution of unobservable community random effects. Hubbard et al.23 have provided a more detailed discussion of the use of population average models vs. random effects models to study community-level effects on health.
We began with the following model, which has been used in previous research on community-level social capital15: where Aij is the dependent variable (CAGE or EHD) for individual i in community j, Sij is the social capital indicators measured for individual i in community j, Sj is the average of social capital indicators in community j and Xij is the set of socio-demographic potential confounders for individual i and community j.
We then re-parameterized this model in the following way: where (β2 − β1) represents the effect of community-level social capital over and above any individual level effect (i.e. if β2 = β1, there is no effect of social capital at the level of the community).
Both models give the same individual-level coefficient, but while the community-level coefficient in the first model represents the combined effect of individual- and community-level social capital on drinking, in our model it represents the contribution of community-level social capital variables over and above individual-level variables. Men and women were analysed separately given the large differences in consumption patterns between them,22 and the following variables were controlled for in the analysis: age, marital status, religion, education, occupation, household economic status, place of residence (urban vs. rural), country of residence and smoking status.
Results
Table 1 shows the characteristics of the study sample and the distribution of social capital indicators. Roughly 44% of the sample was male; most were married (62%), employed (88%) and living in urban areas (60%). Almost 10% of the sample reported being lonely ‘often’; roughly 6% reported being active in a civic organization; 92% had someone to go to in a crisis, and about 5% reported low trust in others. Social capital indicators were not highly correlated with each other.
The prevalence of CAGE problem drinking and EHD in our sample, by age category and gender, are shown in table 2. As expected, men were much more likely to report CAGE-defined problem drinking and EHD than women, and our estimates are similar to the earlier study by Pomerleau, et al.22 that examined the prevalence of EHD in this population.
Age category | CAGE problem drinking | Episodic heavy drinking | Both | |||
---|---|---|---|---|---|---|
Men N (%) | Women N (%) | Men N (%) | Women N (%) | Men N (%) | Women N (%) | |
18–29 | 421 (18.7) | 160 (7.2) | 488 (19.8) | 102 (4.0) | 182 (8.1) | 29 (1.3) |
30–39 | 383 (27.7) | 120 (7.0) | 403 (27.6) | 107 (5.5) | 191 (14.0) | 39 (2.3) |
40–49 | 408 (30.7) | 104 (6.30) | 387 (27.0) | 78 (4.0) | 179 (13.5) | 21 (1.3) |
50–59 | 353 (33.1) | 77 (5.5) | 299 (26.1) | 42 (2.6) | 160 (15.0) | 18 (1.3) |
60+ | 278 (22.8) | 50 (2.7) | 183 (13.9) | 20 (1.0) | 85 (7.0) | 4 (0.2) |
Total | 1843 (25.4) | 511 (5.8) | 1760 (22.5) | 349 (3.4) | 797 (11.0) | 111 (1.3) |
Source: Health in Times of Transition Study 2010, http://www.hitt-cis.net/
The results from our population average model in table 3 show the additional effect of community-level social capital variables on CAGE problem drinking and EHD among men, over and above the individual effect. Adjusting for possible socio-demographic confounders, we found that in addition to the increased odds of individual CAGE problem drinking associated with individual-level social isolation, community-level social isolation also increased the odds of this behaviour, as did community-level civic participation. The odds of EHD also increased with community-level civic participation but were not significantly associated with individual-level social isolation, while the odds of engaging in EHD among men decreased with higher levels of community interpersonal trust.
Social capital indicators | CAGE | EHD | ||
---|---|---|---|---|
OR | 95% C.I. | OR | 95% C.I. | |
Community-level variables | ||||
Social isolation | 1.18 | 1.00–1.38 | 1.02 | 0.87–1.19 |
Active civic participation | 4.08 | 2.23–7.47 | 2.91 | 1.51–5.59 |
Help in a crisis | 1.36 | 0.72–2.54 | 1.17 | 0.66–2.10 |
Interpersonal trust | 0.97 | 0.92–1.03 | 0.89 | 0.83–0.95 |
Individual-level variables | ||||
Social isolation | 1.20 | 1.11–1.29 | 1.06 | 0.97–1.15 |
Active civic participation | 0.94 | 0.72–1.22 | 0.91 | 0.69–1.19 |
Help in a crisis | 0.99 | 0.78–1.26 | 1.05 | 0.81–1.36 |
Interpersonal trust | 0.98 | 0.94–1.02 | 1.02 | 0.98–1.05 |
↵a: Adjusted for age, education, occupation, marital status, religion, household economic status, country of residence, place of residence (urban v. rural) and smoking status. Source: Health in Times of Transition Study 2010, http://www.hitt-cis.net/
The results of the same analysis for women are found in table 4. A similar pattern was observed among women as among men for CAGE problem drinking. Higher odds were observed for individual-level social isolation and community-level civic participation, although, unlike men, civic participation was also associated with increased risk of CAGE problem drinking at the individual level. Also unlike men, social isolation at the community level was associated with increased risk of EHD among women.
Social capital indicators | CAGE | EHD | ||
---|---|---|---|---|
OR | 95% C.I. | OR | 95% C.I. | |
Community-level variables | ||||
Social isolation | 1.09 | 0.86–1.37 | 1.34 | 1.01–1.79 |
Active civic participation | 2.94 | 1.20–7.21 | 1.01 | 0.35–2.89 |
Help in a crisis | 0.34 | 0.11–1.03 | 1.24 | 0.30–5.17 |
Interpersonal trust | 1.06 | 0.96–1.17 | 0.98 | 0.87–1.10 |
Individual-level variables | ||||
Social isolation | 1.30 | 1.16–1.46 | 0.99 | 0.86–1.14 |
Active civic participation | 1.47 | 1.02–2.11 | 0.72 | 0.44–1.20 |
Help in a crisis | 0.96 | 0.68–1.37 | 1.28 | 0.70–2.35 |
Interpersonal trust | 0.96 | 0.90–102 | 0.95 | 0.86–1.02 |
↵a: Adjusted for age, education, occupation, marital status, religion, household economic status, country of residence, place of residence (urban v. rural) and smoking status. Source: Health in Times of Transition Study 2010, http://www.hitt-cis.net/
Discussion
To the best of our knowledge, this is the first analysis of both individual- and community-level social capital and their relation with hazardous alcohol consumption in countries of the fSU. We used two measures of hazardous consumption, both relevant to health but addressing different constructs. Responses to the CAGE instrument capture the role of alcohol in aspects of the individual’s daily life, in particular the extent to which they are dependent on it. EHD captures a particular behaviour that may be seen in those who are not necessarily dependent but which, nonetheless, has profound health consequences. The associations with the two measures differ.
Individual-level social isolation is associated with CAGE problem drinking among both men and women. One possible explanation is that socially isolated individuals are less well equipped to cope, particularly given the shock of the social and economic transition that occurred in the fSU, leading them to turn to problem drinking as a coping mechanism. This hypothesis is supported by previous research linking social isolation to poor self-reported health14 and to psychological stress,24 which may in turn lead to hazardous alcohol consumption,25 and is consistent with the excess mortality observed among single men in post-communist societies compared with married men,16 and among the socially marginalized.26 It is important to note the possibility of reverse causality, as individuals who engage in problem drinking may in fact be more likely to experience family conflicts,27 withdraw from society28 and become psychologically distressed.29 Qualitative research in Russia, using narratives provided by widows of men who died of alcohol-related causes, indicates a bi-directional relationship, with hazardous drinking and psychological distress mutually reinforcing each other, although either can start the process off.30
We found that higher interpersonal trust was associated with lower odds of EHD among men at the community level. This is consistent with previous reports of a strong association between community-level trust and self-rated health,15 (although there is only limited support thus far for the hypothesis that the relationship between social capital and health is mediated by health behaviours such as alcohol consumption).8 The relationship between community-level trust and EHD in the fSU might be explained, in part, by fear of crime. There was a sharp rise in crime in many fSU countries in the immediate post-Soviet period,31 and crime has been associated with worse health outcomes,3 including increased psychological distress.32 This is important because, as mentioned above, psychological distress may in turn have increased the risk of hazardous alcohol consumption.25 It is possible that communities with higher levels of mistrust are those in which crime, and resulting psychological distress, is more prevalent. Research from the United States provides evidence of increased community-level social mistrust in communities with higher rates of firearm homicides.33 A simple regression of community-level interpersonal trust on community-level fear of crime in our data showed that the former was strongly negatively associated with the latter (β = −6.822); however, a more in-depth analysis of these factors is required before drawing conclusions on their relationship.
Perhaps our most surprising finding is that of a positive association between community-level civic participation and CAGE problem drinking among men and women, and EHD among men. This finding challenges the theory that membership in groups may encourage the dissemination of health information and curtail deviant and hazardous health behaviours,7 including alcohol consumption among college students in the United States.10 However, our study differs from the college study in terms of study population and context. Another key difference is the nature of the organizations to which study participants belong, which likely differs significantly between the fSU and American college campuses.
Further analysis of our data (not shown) indicated that the most commonly reported organization in which individuals were members was a ‘trade union’, and community-level trade union membership was significantly associated with an increased risk of both CAGE problem drinking and EHD. What is it about trade union membership that results in increased hazardous drinking in the fSU? One possible explanation is that trade unions are an example of ‘single issue organizations’ that entail a narrow ‘radius of trust’ and are not likely to improve generalized trust in others.9 This form of civic participation has been coined the ‘miniaturization of community’34 and has been linked to alcohol consumption.9 Another, perhaps more likely explanation, given the weakened role of trade unions in people’s lives since the fall of the Soviet Union, is that communities with high levels of union membership simply represent communities where many inhabitants are in industrial employment where there is mandatory union membership. This latter hypothesis is consistent with research from Ukraine, which has shown that alcohol consumption is higher in the industrial South and East regions of the country compared with the agrarian West.35 This might also explain why we found an association between membership and hazardous alcohol consumption only at the community level for men and not at the individual level (although there was an association at the individual level for women).
One other possible explanation for the association between community-level civic participation and hazardous alcohol consumption is that communities where there is a high level of membership in organizations may offer frequent opportunities to gather at social events where drinking is common and expected. This explanation was offered by an earlier study in Taiwan that found a similar association between community social participation and frequent drinking.36 The potential for social capital to create demands for conformity among community members has been described by Portes37 and Ferlander,38 and is plausible in the fSU context; however, further qualitative research is required to better understand the nature of civic organization membership (especially trade union membership) in the fSU and the role that it plays in alcohol consumption.
There are some limitations to our study. First, the cross-sectional nature of the data prohibits us from making conclusions about causality, as discussed above with regards to social isolation and alcohol use. Second, there is a tendency for respondents, especially in the fSU, to under-report their own alcohol consumption.39 As such, our estimate of the prevalence of EHD may be an underestimate; however, there is no reason to believe that this potential underestimate would create spurious associations between our indicators of interest and EHD. The measurement of CAGE problem drinking may be less vulnerable to bias, as it does not focus on perceived alcohol consumption.19 Also, this study will have probably missed the most severe drinkers (e.g. intoxicated, homeless), who may also be the most socially isolated, thereby producing somewhat conservative estimates of the relationship between social capital and hazardous alcohol consumption across these countries. Third, as there is almost no existing research on the relationship between social capital and alcohol consumption among adults, we were forced to compare and contrast our findings with those from studies of social capital and general health, despite inconsistent evidence thus far that alcohol plays a mediating role between them.8,11 Fourth, we were not able to distinguish between different forms of social capital that may be important in the association between civic engagement and alcohol consumption, namely, ‘bonding’, ‘bridging’ and ‘linking’, which have shown to be important to health outcomes in other contexts.40 Lastly, because of the resources required for conducting a multi-country study, the number of individuals in each community in the HITT survey was small (an average of nine per community). Although this does not invalidate our findings, it does highlight the need for further research within individual countries using larger community samples.
Conclusions
Our results provide evidence of the independent association between individual-level social isolation, as well as community-level civic participation and interpersonal trust, and hazardous alcohol consumption in the fSU. The finding that community-level civic participation is associated with increased odds of hazardous alcohol consumption seems to contradict evidence from other regions that links civic participation to improvements in health and should be investigated further.
Funding
This work (part of the Health in Times of Transition Project of the European Centre on Health of Societies in Transition) was supported by the European Union’s 7th Framework Program, project HEALTH-F2-2009-223344. The European Commission cannot accept any responsibility for any information provided or views expressed.
Conflicts of interest: None declared.
Key points
To our knowledge, this is the first study that investigates the effect of individual- and community-level social capital on hazardous alcohol consumption in the former Soviet Union, using multi-level data from nine former Soviet countries.
We show that some elements of social capital—social isolation, social mistrust and active civic participation— are positively associated with hazardous consumption in the region, and we highlight the importance of community effects.
The strong positive relationship between community-level civic participation and hazardous alcohol consumption is particularly surprising, given evidence from other regions of the world pointing to the protective effect of civic participation on drinking.
The nature of civic organizations in the former Soviet Union and the role they play in alcohol consumption should be researched further in order to inform public health interventions at the community level.
- © The Author 2014. Published by Oxford University Press on behalf of the European Public Health Association. All rights reserved.