9
In search of links between social capital, mental health and sociotherapy: A longitudinal study in Rwanda Femke Verduin a, b, * , Geert E. Smid c, d , Tim R. Wind b, d , Willem F. Scholte a, b, d a Academic Medical Center, Dept. of Psychiatry, University of Amsterdam, Amsterdam, Netherlands b Equator Foundation, Diemen, Netherlands c Foundation Centrum '45, Diemen, Netherlands d Arq Psychotrauma Expert Group, Research Program, Diemen, Netherlands article info Article history: Received 26 October 2013 Received in revised form 25 September 2014 Accepted 29 September 2014 Available online 2 October 2014 Keywords: Rwanda Social capital Post-conict Mental health Sociotherapy Latent growth modeling abstract To date, reviews show inconclusive results on the association between social capital and mental health. Evidence that social capital can intentionally be promoted is also scarce. Promotion of social capital may impact post-conict recovery through both increased social cohesion and better mental health. However, studies on community interventions and social capital have mostly relied on cross-sectional study de- signs. We present a longitudinal study in Rwanda on the effect on social capital and mental health of sociotherapy, a community-based psychosocial group intervention consisting of fteen weekly group sessions. We hypothesized that the intervention would impact social capital and, as a result of that, mental health. We used a quasi-experimental study design with measurement points pre- and post-intervention and at eight months follow-up (2007e2008). Considering sex and living situation, we selected 100 adults for our experimental group. We formed a control group of 100 respondents with similar symptom score distribution, age, and sex from a random community sample in the same region. Mental health was assessed by use of the Self Reporting Questionnaire, and social capital through a locally adapted version of the short Adapted Social Capital Assessment Tool. It measures three elements of social capital: cognitive social capital, support, and civic participation. Latent growth models were used to examine whether effects of sociotherapy on mental health and social capital were related. Civic participation increased with 7% in the intervention group versus 2% in controls; mental health improved with 10% versus 5% (both: p < 0.001). Linear changes over time were not signicantly corre- lated. Support and cognitive social capital did not show consistent changes. These ndings hint at the possibility to foster social capital and simultaneously impact mental health. Further identication of pathways of inuence may contribute to the designing of psychosocial in- terventions that effectively promote recovery in war-affected populations. Trial registration: Nederlands Trial Register 1120. © 2014 Elsevier Ltd. All rights reserved. 1. Introduction Traumatized survivors of war or political violence often have complex mental health problems, with anxiety, depressive and cognitive disturbances (De Jong et al., 2003; Rodin and van Ommeren, 2009). Most patients also suffer from feelings like shame, guilt, distrust and alienation. Such feelings complicate so- cial functioning and interpersonal contacts in communities where social structures and cohesion have already been damaged by hu- man violence (Ager, 2002; Hobfoll et al., 2007). Psychological and behavioral problems hamper daily functioning and the engagement in relations. Learning how to cope with such difculties may not only counter individual suffering, it may also help to prevent additional damage in social relations caused by ongoing behavioral disturbances, and to rebuild meaningful social structures in which people can re-nd and practice (self)respect. Social capital is potentially a key resource supporting post-conict recovery. Pro- motion of social capital may impact post-conict recovery both through increased social cohesion and through better mental health (Scholte and Ager, 2014). This study explores the possible effects of an intervention called sociotherapy on social capital and * Corresponding author. Equator Foundation, Nienoord 5, 1112 XE Diemen, Netherlands. E-mail address: [email protected] (F. Verduin). Contents lists available at ScienceDirect Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed http://dx.doi.org/10.1016/j.socscimed.2014.09.054 0277-9536/© 2014 Elsevier Ltd. All rights reserved. Social Science & Medicine 121 (2014) 1e9

In search of links between social capital, mental health and sociotherapy: A longitudinal study in Rwanda

Embed Size (px)

Citation preview

lable at ScienceDirect

Social Science & Medicine 121 (2014) 1e9

Contents lists avai

Social Science & Medicine

journal homepage: www.elsevier .com/locate/socscimed

In search of links between social capital, mental healthand sociotherapy: A longitudinal study in Rwanda

Femke Verduin a, b, *, Geert E. Smid c, d, Tim R. Wind b, d, Willem F. Scholte a, b, d

a Academic Medical Center, Dept. of Psychiatry, University of Amsterdam, Amsterdam, Netherlandsb Equator Foundation, Diemen, Netherlandsc Foundation Centrum '45, Diemen, Netherlandsd Arq Psychotrauma Expert Group, Research Program, Diemen, Netherlands

a r t i c l e i n f o

Article history:Received 26 October 2013Received in revised form25 September 2014Accepted 29 September 2014Available online 2 October 2014

Keywords:RwandaSocial capitalPost-conflictMental healthSociotherapyLatent growth modeling

* Corresponding author. Equator Foundation, NieNetherlands.

E-mail address: [email protected] (F. Ve

http://dx.doi.org/10.1016/j.socscimed.2014.09.0540277-9536/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

To date, reviews show inconclusive results on the association between social capital and mental health.Evidence that social capital can intentionally be promoted is also scarce. Promotion of social capital mayimpact post-conflict recovery through both increased social cohesion and better mental health. However,studies on community interventions and social capital have mostly relied on cross-sectional study de-signs. We present a longitudinal study in Rwanda on the effect on social capital and mental health ofsociotherapy, a community-based psychosocial group intervention consisting of fifteen weekly groupsessions. We hypothesized that the intervention would impact social capital and, as a result of that,mental health.

We used a quasi-experimental study design with measurement points pre- and post-intervention andat eight months follow-up (2007e2008). Considering sex and living situation, we selected 100 adults forour experimental group. We formed a control group of 100 respondents with similar symptom scoredistribution, age, and sex from a random community sample in the same region. Mental health wasassessed by use of the Self Reporting Questionnaire, and social capital through a locally adapted versionof the short Adapted Social Capital Assessment Tool. It measures three elements of social capital:cognitive social capital, support, and civic participation. Latent growth models were used to examinewhether effects of sociotherapy on mental health and social capital were related.

Civic participation increased with 7% in the intervention group versus 2% in controls; mental healthimproved with 10% versus 5% (both: p < 0.001). Linear changes over time were not significantly corre-lated. Support and cognitive social capital did not show consistent changes.

These findings hint at the possibility to foster social capital and simultaneously impact mental health.Further identification of pathways of influence may contribute to the designing of psychosocial in-terventions that effectively promote recovery in war-affected populations.Trial registration: Nederlands Trial Register 1120.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Traumatized survivors of war or political violence often havecomplex mental health problems, with anxiety, depressive andcognitive disturbances (De Jong et al., 2003; Rodin and vanOmmeren, 2009). Most patients also suffer from feelings likeshame, guilt, distrust and alienation. Such feelings complicate so-cial functioning and interpersonal contacts in communities where

noord 5, 1112 XE Diemen,

rduin).

social structures and cohesion have already been damaged by hu-man violence (Ager, 2002; Hobfoll et al., 2007). Psychological andbehavioral problems hamper daily functioning and the engagementin relations. Learning how to cope with such difficulties may notonly counter individual suffering, it may also help to preventadditional damage in social relations caused by ongoing behavioraldisturbances, and to rebuild meaningful social structures in whichpeople can re-find and practice (self)respect. Social capital ispotentially a key resource supporting post-conflict recovery. Pro-motion of social capital may impact post-conflict recovery boththrough increased social cohesion and through better mentalhealth (Scholte and Ager, 2014). This study explores the possibleeffects of an intervention called sociotherapy on social capital and

F. Verduin et al. / Social Science & Medicine 121 (2014) 1e92

mental health in post-genocide Rwanda. We hypothesized that thisintervention would impact social capital and, as a result of that,mental health.

1.1. Social capital

Social capital is a concept based on the idea that social networksprovide a basis for social cohesion and cooperation. It has beencharacterized as ‘the glue that holds societies together’ (McKenzieet al., 2002). Social capital's most commonly adopted definition inhealth sciences recognizes five characteristics: community net-works, civic engagement, civic identity (belonging, solidarity,equality), reciprocity and norms of cooperation, and trust in thecommunity (Putnam, 1993). Within the literature, studies distin-guish between individual and collective conceptualizations of so-cial capital (Kawachi and Subramanian, 2006), but the definition asa community asset is currently “privileged” over individualdefinitions.

Social capital has been divided into two components, ‘structuralsocial capital’ and ‘cognitive social capital’. Structural social capitalrefers to the existence of relationships, networks, and associationsthat link members together. Cognitive social capital is the ‘drivingforce’; it includes values, norms, civic responsibility, expectedreciprocity, charity, altruism, and trust. Structural and cognitivesocial capital, respectively, can be characterized as what people ‘do’and what people ‘feel’ in terms of social relations (Harpham et al.,2002).

1.2. Promotion of social capital

Until recently, the few existing studies on community in-terventions aiming to improve social capital, tended to rely oncross-sectional study designs. This implies that their ability to drawcausal inferences has been limited (Macinko and Starfield, 2001; DeSilva et al., 2005a). Only very few longitudinal studies evidence thatsocial capital can be fostered by interventions. Coletta and Cullen(2000) discussed changes in social capital resulting from violentconflict in their study of four conflict-affected countries (Cambodia,Rwanda, Guatemala, and Somalia); they provide clear examples ofhow governments and international actors promote decentraliza-tion, civic participation, social inclusion, empowerment, and thestrengthening of grassroots movements. A study by Michael et al.(2008) showed social support and self-rated health improved,while depressive symptoms decreased, after a community-basedparticipatory research intervention, which employed CommunityHealth Workers who used popular education to identify andaddress health disparities in Latino and African American com-munities in a metropolitan area in the United States.

Brune and Bossert (2009) performed a longitudinal study inNicaragua, showing that systematic interventions promotingmanagement and leadership development were effective inimproving some aspects of social capital, in particular the cognitiveattitudes of trust in the communities. Interventions were alsolinked to higher levels of civic participation in governance pro-cesses. As in other empirical studies, they also found that higherlevels of social capital were significantly associated with somepositive health behaviors.

Pronyk et al. (2008) conducted an intervention in rural South-Africa that combined group-based microfinance with participa-tory gender and HIV training in an attempt to catalyze changes insolidarity, reciprocity and social group membership as a means toreducewomen's vulnerability to intimate partner violence and HIV.After two years, adjusted effect estimates indicated higher levels ofstructural and cognitive social capital in the intervention group

than the comparison group, although confidence intervals werewide.

1.3. Social capital and mental health

Social capital may play a role in the incidence and prevalence ofmental illness (McKenzie et al., 2002). The assumed relevance ofsocial capital for mental health has been underscored by nationaland international policies to develop social capital in disaster- orwar-affected communities (Hobfoll et al., 2007; Norris et al., 2008).Over the last decade literature on the salutary association betweensocial capital and mental health is growing (Kawachi and Berkman,2001; Almedom, 2005; De Silva et al., 2005a; De Silva et al., 2007;Engstr€om et al., 2008; Berry and Welsh, 2010; Hamano et al., 2010;Suzuki et al., 2010; Wind et al., 2011). However, systematic reviewsof quantitative studies examining the association between socialcapital andmental illness have shown inconclusive results (De Silvaet al., 2005a,b; Islam et al., 2006). Especially studies that concep-tualized social capital as a community asset (as opposed to an in-dividual asset) found ambiguous associations with individualmental health outcomes (De Silva et al., 2005a,b; Hamano et al.,2010; Eriksson, 2011). At the individual level, several studies haveobserved positive associations with better mental health (Veenstra,2000, 2002; De Silva et al., 2005a,b; Whitley and McKenzie, 2005;Almedom, 2005; Irwin et al., 2008; Patel, 2010).

A study in Rwandan children and families affected by HIV/AIDSshowed that communities which ‘gather people together to discussproblems’, ‘offer advice’ and ‘understand and help solve problems’add to individual coping and strong parentechild relationships inprotecting against mental health problems and promoting resil-ience (Betancourt et al., 2011). A recent study among war-affectedyouth in Sierra Leone revealed the role of the post-conflict socialcontext in shaping mental health in former child soldiers. Findingsunderscored the importance of the social environment and theneed to develop post-conflict interventions that addresscommunity-level processes in addition to the needs of families andindividuals (Betancourt et al., 2014).

Various scholars (Wang et al., 2009; Nakhaie and Arnold, 2010;Wind and Komproe, 2012) assert that the time has come to shedmore light on possible associations between social capital andmental health, and which specific starting points are most impor-tant for interventions.

1.4. The present study

The study presented here was performed within the frameworkof a psychosocial community intervention (community basedsociotherapy) aimed to enhance social bonding (Richters et al.,2005). Sociotherapy has been shown to establish a significantimprovement in mental health in Byumba, Rwanda (Scholte et al.,2011a).

Byumba province is located in the north of Rwanda, borderingUganda. The invasion by the Rwanda Patriotic Front (RPF) fromUganda into Rwanda on 1st October 1990 started a civil war in thenorth of the country. Predominantly of Tutsi origin, many of themembers of the RPF were second generation refugees who had fledto Uganda and settled there from 1959 onwards, escaping ethnicpurges in Rwanda. The RPF went into Rwanda as an army ofliberation but was perceived by the majority of the population(mostly of Hutu origin) as an army of occupation. Low intensityfighting was interrupted by several massacres, including one inByumba. During the 1994 genocide, social capital atrophied as thecountry, communities and families fell prey to hatred and violence(Coletta and Cullen, 2000). The war (1990e1994) and genocide(1994) related problems affected men and women of all ages. The

F. Verduin et al. / Social Science & Medicine 121 (2014) 1e9 3

population experienced atrocities like killings, sexual violence,torture, intimidation, robbery, destruction of property, and socialrejection, leaving the community in a state of trauma. A relativelylarge part of the population came to consist of widows, widowers,orphans, physically handicapped, prisoners and ex-prisoners(Richters et al., 2005, 2008).

Clearly, after years of war and genocide, Rwanda is a contextwhere social fabric has been seriously damaged, which implies lowsocial capital. Before the genocide, potential bridging social capitalexisted in the form of exchange, mutual assistance and reciprocity(Coletta and Cullen, 2000). The ‘success’ of the genocide dependedin part on civilians' sense of civic duty and on the historical strengthof the central government. Vertical social capital, manifested inalmost absolute state power, had historically penetrated Rwandansociety so deeply as to supersede horizontal relations or loyalties.The violent conflict destroyed whatever broad-based forms of so-cial capital had existed (Coletta and Cullen, 2000; Putzel, 1997).

To our knowledge, our study is the first aiming to establishpossible links between social capital and mental health by use of alongitudinal design. We measured social capital at the individuallevel, and looked into what elements of social capital were pro-moted by the community-based sociotherapy program in Rwanda,and if these elements were salutary for mental health. First, weperformed an exploratory factor analysis of the scores on the in-strument used to measure social capital, in order to distinguish therelevant elements of social capital. Then we explored the changeover time in these elements, and examined the relationship be-tween the changes in social capital and mental health. Latentgrowth models (LGM) were used to examine whether the effects ofsociotherapy on three elements of social capital were related to theeffects of sociotherapy on mental health.

2. Methods

Data were collected at the start of the intervention (October2007: T0), directly after (January 2008: T1), and at eight monthsfollow-up (September 2008: T2). Demographic data (sex, age, ed-ucation level and socioeconomic status) were documented.Approval was gained from the Medical Ethics Committee of theAcademic Medical Center in Amsterdam. Trial Registration:Nederlands Trial Register (NTR) 1120.

2.1. Intervention

Sociotherapy seemed to be an approach that could promotesocial capital by creating meaningful social structures in whichpeople can re-find and practice (self)respect. The method thera-peutically uses interaction between individuals and their socialenvironment to help subjects to re-assess and re-define values,norms, relations and possible collaborations. The principal premiseis that reaching a certain level of mutual respect, trust and care ingroup interaction would help to increase the problem solving ca-pacity and subjective mental health in individual group partici-pants (Scholte et al., 2011a).

In sociotherapy with survivors of systematic violence, safety andthe setting of democratic rules are additional primary objectives.Key elements of the working methods in Rwandawere debates andthe exchange of experiences and coping strategies among partici-pants, exercises, games andmutual practical support. The approachdid not primarily aim at sharing or processing traumatic memories.Trauma symptoms were addressed through psycho-education andadvice.

The sociotherapy programme was set up in collaboration withthe Episcopal Church of Rwanda (ECR), funded by the developmentorganization Cordaid and technically supported by the Dutch

agency Equator Foundation. Approval was given by regional andnational authorities in Rwanda. Wide support on community levelwas gained through public acclamation by the ECR (Richters et al.,2008).

The programme was open to any adult (�16 years old) wantingto participate. Groups contained ten to twenty participants andwere mostly mixed: both sexes, various ethnic backgrounds, wideage distribution, who lived in close proximity to one another(Richters et al., 2013). Forty-five groups ran simultaneously, havingweekly meetings over a period of fifteen weeks, lasting threehours each. Group leaders were local people, familiar with theregion's history and current living situation; they had secondaryschool level education and worked as teachers, pastors, localleaders, civil servants or as staff of non-governmental organisa-tions (NGOs). They had received a three months' training fromEquator staff, to lead participants through six phases of socio-therapy; safety, trust, care, respect, rules and memories. Duringtheir training, group leaders had been guided through thesephases as if they were sociotherapy beneficiaries themselves. Atthe end of each session a joint reflection was prompted on theirexperiences and the working methods that had elicited theseexperiences.

Group leaders were allowed to attune the working methods tothe characteristics of their groups (e.g., degree of trust, nature ofproblems) and to their own affinity and experience, puttingdifferent emphases on elements like rules, role plays, and spiritu-ality. The groups were held in a private, undisturbed place, whereall participants could sit down. Democratic rules were applied, andmen and women were treated equally; so were younger and olderpeople.

2.2. Instruments

2.2.1. Mental healthTo measure mental health we used the Self Reporting Ques-

tionnaire (SRQ-20), an instrument developed by the World HealthOrganization (WHO) for screening for common mental disordersin primary health care settings. It consists of twenty yes/noquestions about mood, thinking capacity, feelings of anxiety andphysical well-being. ‘Yes’ answers result in a higher score, mean-ing a poorer mental health condition. We (back-) translated theSRQ-20 to the local language, Kinyarwanda, and validated it for theactual context (Verduin et al., 2010). The capacity of the SRQ-20 toidentify probable psychopathology proved to be sufficient for men(area under the curve: AUC ¼ 0.74) and women (AUC ¼ 0.76).Reliability was considered to be good (Cronbach's alpha ¼ 0.83).We also validated the SRQ-20 for its capacity to assess change insymptom severity over time. The factor structure of the SRQ-20was determined through exploratory factor analysis. Factorialinvariance over time was tested in a multigroup confirmatoryfactor analysis using data of repeated measures (re-assessmentafter a 3-month period) of 230 respondents. The instrumentsfactor structure proved to be time invariant; the number of factors,factor loadings and covariances of factors remained equal overtime (Scholte et al., 2011b).

2.2.2. Social capitalThere is a large variety in the elements of social capital studied

in the various research projects published, depending on whichdistinction or element of social capital is considered most impor-tant. We chose to make use of the short version of the AdaptedSocial Capital Assessment Tool, (Short A-SCAT; Tuan et al., 2005),because of its proven validity in various contexts, its limited lengthand the presumed relevance of its items for the Byumba context.Items of the Short A-SCAT address received support from groups or

F. Verduin et al. / Social Science & Medicine 121 (2014) 1e94

individuals, whether and how people connect with leaders, howthey feel connected to others in their living area and how theyget along.

The Short A-SCAT has been extensively validated in tworesource-poor settings (Vietnam, Peru). Due to time and financialconstraints, we could not validate our Rwandan version in thesame scrupulous way. However, before we used the Short A-SCAT,it was discussed in focus groups, tested and discussed with ourinterviewers. There was overall agreement on the changes to bemade in the phrasing of items in order to capture the intendedmeaning in an easy to understand way for local respondents. Forexample: the word ‘community’ was changed to ‘area, neigh-borhood or hill’; the word ‘majority’ was changed to ‘many peo-ple’, as the term had been contaminated during ethnicpolarization; in the question ‘are you an active member of anygroup?’ the word ‘active’ was discarded because it appeared to betoo ambiguous and would probably bias responders towardsanswering ‘yes’.

After a pilot study, we added some new items in order to raisethe instrument's psychometric qualities, providing more responseoptions for giving and receiving support and for manifestations ofcivic participation. Also, Likert scales were used instead of yes/nooptions, which allowed for a more differentiated response (SeeSupplementary data) (Verduin et al., 2010).

2.3. Participants

We calculated that a minimum of thirty respondents wasneeded to establish a 2.7 effect with a standard of 0.80 power. As atT0 an unexpected high number of sociotherapy group participantsappeared to be willing to be included in the study and as we wereunsure about the future attrition rate, we decided to aim at largernumbers (n ¼ 100) per study group.

2.3.1. Experimental groupThe programme was open to any adult (�16 years) wanting to

participate. It is likely that particularly persons with psychosocialproblems were inclined to apply. Also, community members couldpersonally be invited when considered psychosocial problem casesby sociotherapy group leaders. Out of forty-five sociotherapygroups simultaneously starting sessions, ten groups which in totalcontained an equal number of men and women, were selected inrural as well as urban areas. We then formed our experimentalstudy group by randomly selecting 100 out of the total of 133participants of these ten groups.

2.3.2. Control groupWe applied the following procedure to compose a control

group that was equivalent at baseline with regard to our mainoutcome measure, the SRQ-20 score. During our pilot study, 2.5times more respondents in the experimental group had baselinescores above cut-off than in the control group (Scholte et al.,2011a). For this study we therefore aimed to interview 2.5 times(n ¼ 250) more respondents than in the experimental group, tolater select 100 out of these to compose a control group. Weidentified regions within Byumba district where the programmewas not or had not been running so far, or for practical reasonswould not start over the upcoming eight months. Here, weselected respondents through convenience sampling; interviewersstarted at the top of a hill or in the center of a village and eachwalked down a different footpath towards scattered houses orhuts. An equal number of men and women, at home or in thefields, were randomly chosen and asked to participate. Finally 251respondents were interviewed.

After analysis of the data collected, we selected a group of 100out of these for which the distribution of SRQ-20 scores matchedthat of the intervention group. For this purpose we used eightclusters of scores (0e1, 2e3, 4e5, 6e7, 8e9, 10e12, 13e15, 16e20)and from each cluster randomly selected a number of respondentsequal to the corresponding cluster in the experimental group. Next,we matched clusters on age, and subsequently on sex as much aspossible. The final selection of 100 constituted our definite controlgroup.

2.4. Interviews

Eight local interviewers were recruited; all were sociology stu-dents at the ’Institut Polytechnique de Byumba’ in Byumba. Theirone-week training addressed the principles of a longitudinal studydesign, interviewing techniques and our measuring instrument(Verduin et al., 2010; Scholte et al., 2011a). Informed consent wasobtained from respondents by use of an explanatory text, whichbecause of the high illiteracy rate was read aloud. In case of refusal,demographic data and reasons for refusal would be requested anddocumented, but no-one refused. The interviews were mostly heldoutside, or inside a local church; interviewer and respondent wouldfind a private spot where they could not be overheard by anyone.After the interview a small amount of money, 500 Rwandan francs(0,50 Euros), was given to compensate for the time the respondentusually had to wait.

2.5. Statistical analysis

All psychometric and validation analyses were performed usingSPSS 18. AMOS was used to conduct latent growth modeling.

Before descriptive and inferential analysis (means, standarddeviations (SD), t-tests, chi square tests, correlations), data werescreened for normality in distribution; skewness and kurtosisvalues were between �2 and 2. Missing values were handled usingfull information maximum likelihood (FIML) estimation (N ¼ 200for all analyses). It has been shown that under ignorable missingdata conditions, FIML estimates are unbiased (Enders and Bandalos,2001).

2.5.1. Descriptive and factor analysisWe used descriptive analyses (means, SD) to evaluate possible

changes over time in elements of social capital. Exploratory factoranalysis was performed to distinguish which elements of socialcapital were measured by the Short A-SCAT in our Rwandancontext.

2.5.2. Latent growth modelingWe applied latent growth models (LGMs) to variables that

showed possible changes over time. LGMs can be extended toinclude structural parameters as predictors (Duncan et al., 2006).Thus, we examined the relationship between predictor variables(i.e., sociotherapy), baseline level, and linear change in each vari-able separately. We then examined the association of componentsof social capital and mental health over time using associativeLGMs. Associative LGMs allow researchers to examine the corre-lations among development parameters for pairs of behaviors. Weadded sociotherapy as a predictor variable to the associative LGM toexamine the effects of sociotherapy on mental health and civicparticipation simultaneously. We evaluated model fit using thediscrepancy c2, comparative fit index (CFI), non-normed fit index(NNFI/TLI), root-mean-square error of approximation (RMSEA), andAkaike information criterion (AIC). Models that fit adequately areindicated by CFIs and NNFIs � 0.95[�0.90] and RMSEAs � 0.06[�0.08] (Hu and Bentler, 1999).

F. Verduin et al. / Social Science & Medicine 121 (2014) 1e9 5

3. Results

3.1. Baseline characteristics

The two study groups matched on SRQ-20 score distribution,age and sex at baseline (see Table 1). We compared study partici-pants who completed all three assessments with those whodropped out after the first or second assessment on all variablesassessed at T0.

Drop out did not differ significantly between the experimentaland the control group (p ¼ 0.79). Drop-outs at T1 or T2 from eitherstudy group did not differ significantly in sex, age or level of edu-cation. Neither was there a difference in sex, age, level of education,SRQ-20 scores and scores on Short A-SCAT between actual re-spondents and drop-outs at T1 and T2. A difference between thestudy groups was noted in socioeconomic status (SES) at baseline(see Table 1).

Experimental group Control group

Baseline (T0)

133 participants of 10 sociotherapy groups:

60 men, 73 women

Baseline (T0)

251 interviewed in 5 areas: 127 men, 124 women

Selection of 100, matched by SRQ-20 score, gender and age

47 men, 53 women

T1

81 found

35 men, 46 women

T1

90 found

40 men, 50 women

T1

10 not found

1 deceased 2 ill 7 did not turn up

T1

19 not found

15 7 5 14

Random selection of 100

45 men, 55 women

T2

81 found

34 men, 47 women

T2

73 found

34 men, 39 women

T2

19 not found

5 76

T2

27 not found

12 66

3.2. Reliability and factor analysis of the short A-SCAT

Our adapted version of the Short A-SCAT had reasonable psy-chometric qualities. Using the total sample at baseline (N ¼ 384)and after reversing question 19, Cronbach's alpha was 0.75 for theinstrument as a whole; it did not improve if any item was deleted.

Earlier factor analyses of the Short A-SCAT in Peru and Vietnam (byde Silva and colleagues) yielded three factors: cognitive socialcapital, group membership/(social) support and citizenship(Harpham et al., 2002; Tuan et al., 2005; De Silva et al., 2005b).

An exploratory factor analysis showed our version of the ShortA-SCAT fell apart into similar components as described above. Item1 was omitted because it had a different response format (yes/noquestion) and therefore did not load on one factor specifically.

I. Cognitive social capital (items 11 and 14, 15, 16, 17, 18, 19;Cronbach's alpha ¼ 0.86);

II. Support (items 2, 3, 4, 5a, 6a, 7a, 8a; Cronbach'salpha ¼ 0.68);

III. Civic participation: items 9, 10, 12, 13. Cronbach'salpha ¼ 0.63).

These three factors accounted for 44.8% of the total variance. Theitems can be categorized in three sections; cognitive social capital

and two elements of structural social capital: support manifesta-tions and civic participation. The term support represents distinctmanifestations of received support; help from groups or individualsin comforting/encouraging in good or bad situations, in improvingone's economic situation and in knowing and doing things (SeeSupplementary data). We prefer the term civic participation over

Table 3Model fit indices of linear growth models.

df c2 p RMSEA CFI NNFI AIC

Model 1 Civic Participation LGM 1 0.03 0.859 0.00 1.00 1.10 16.03Model 2 Civic Participation LGM

with predictor2 0.39 0.824 0.00 1.00 1.13 24.39

Model 3 Mental Health LGM 1 0.14 0.711 0.00 1.00 1.03 16.14Model 4 Mental Health LGM

with predictor2 0.61 0.736 0.00 1.00 1.04 24.61

Model 5 Mental Health CivicParticipationAssociative LGM

8 14.79 0.063 0.07 0.97 0.93 52.79

Model 6 Mental Health CivicParticipation AssociativeLGM with predictor

9 14.51 0.105 0.06 0.98 0.94 66.51

Table 1Socio-demographic characteristics of experimental and control group at baseline.

Experimentalgroup n ¼ 100

Control groupn ¼ 100

Tests indep.T-tests/Chi2

GenderMale 45 (45%) 47 (47%) n.s.Female 55 (55%) 53 (53%)Mean age(Minemax) 34.9 (16e76) 38.5 (16e73) n.s.Standard deviation Sd 15.8 Sd 14.1EducationNil 48 (48%) 54 (54%) n.s.Primary 42 (42%) 34 (34%)Secondary 10 (10%) 12 (12%)SESMarginal 6 (6%) 13 (13%) c2(2) ¼ 0.022Poor 83 (83%) 66 (66%)Sufficient 11 (11%) 21 (21%)Mean SRQ-20 score 8.41 8.26 n.s.

Sd 5.05 Sd 4.83

F. Verduin et al. / Social Science & Medicine 121 (2014) 1e96

citizenship (De Silva et al., 2005b), as it entails an active attitudetowards one's community. The items in our scale of civic partici-pation address joining elections and actively discussing and solvingproblems with local leaders or in groups, respectively (SeeSupplementary data).

3.3. Descriptive analyses and correlations of social capital andmental health

Table 2 shows mean scores, standard deviations, and bivariatecorrelations of the three measurements of mental health (SRQ-20scores), cognitive social capital, support and civic participation, forboth the experimental and the control group. Mean scores did notappear to change consistently over time, except for mental healthand civic participation in the experimental group. See boldnumbers in Table 2; in the experimental group, mean SRQ-20scores showed a 10% decrease (from 8.41 at T0 to 6.31 at T2),while the control group showed a 5% decrease (from 8.26 to 7.27)(p < 0.001). For civic participation there was a 7% increase of meanscores (from 6.59 to 7.78) versus a 2% increase (from 7.74 to 8.03) inthe control group (p < 0.001).

3.4. LGM results

Table 3 shows model fit indices of subsequent LGMs. Each in-dividual model showed a good fit. Results of the final associative

Table 2Correlations of mental health and three elements of social capital: cognitive social capitameasurement), T1 (right after intervention, 3 months), T2 (follow-up at 8 months).a,b

1 2 3 4 5 6

1 Mental Health T0 0.67*** 0.64*** �0.16 �0.19 �0.2 Mental Health T1 0.61*** 0.73*** �0.25* �0.25* �0.3 Mental Health T2 0.66*** 0.47*** �0.34** �0.34** �0.4 Cognitive T0 �0.20* �0.01 �0.16 0.41*** 0.365 Cognitive T1 �0.14 �0.06 �0.14 0.36** 0.326 Cognitive T2 �0.15 �0.03 �0.27* 0.33** 0.24*7 Support T0 �0.16 �0.08 �0.11 �0.04 0.01 0.018 Support T1 �0.16 �0.06 �0.20 0.06 �0.14 0.179 Support T2 �0.13 �0.02 �0.21 0.15 0.16 �0.10 Civic Participation T0 �0.07 0.01 �0.16 0.25* 0.22* 0.1911 Civic Participation T1 �0.29** �0.11 �0.24* 0.15 0.40*** 0.1912 Civic Participation T2 �0.19 �0.08 �0.07 0.07 0.09 0.01

M 8.41 7.28 6.31 12.55 12.33 13.8SD (5.05) (4.64) (3.91) (4.35) (4.45) (4.0

*p < 0.05, **p < 0.01, ***p < 0.001.a Sociotherapy group: below diagonal; control group: above diagonal.b Numbers represent correlations unless otherwise indicated.

LGM of mental health and civic participation with sociotherapygroup membership as a predictor variable are reported in Fig. 1 andTable 4. The results showed a significant effect of sociotherapy onboth linear change in mental health and civic participation(Fig. 1: �0.38 for mental health and 0.41 for civic participation). Inaddition, civic participation at baseline was lower in the socio-therapy group than in the control group. Although mental healthand civic participation were correlated at baseline (Fig. 1: �0.26),linear changes in mental health and civic participation over timewere not significantly correlated (�0.21).

4. Discussion

This study aimed to establish the effects of a sociotherapy pro-gram in Rwanda on social capital and mental health, and toinvestigate whether these effects were correlated. We measuredthree elements of social capital: cognitive social capital, support,and civic participation. We hypothesized that the interventionwould positively impact social capital, and that as a result mentalhealth would increase. We found a positive but small effect of theintervention on civic participation andmental health. We could notestablish a correlation between over time changes in civic partici-pation and mental health.

Civic participation refers to an active attitude towards one'scommunity, and to participating in collective actions within thatcommunity. The concept is easily linked to that of collective effi-cacy, which Sampson et al. (1997) defined as ‘social cohesionamong neighbors combined with their willingness to intervene onbehalf of the common good’. They added that whereas social capital

l, support and civic participation in experimental and control group at T0 (baseline

7 8 9 10 11 12 M SD

06 �0.06 0.08** �0.05** �0.18 �0.35** �0.14 8.26 (4.84)07 �0.05 0.13 �0.13 �0.10 �0.30** �0.37** 8.15 (5.69)25* �0.12 0.03 �0.08 �0.13 �0.40** �0.38** 7.27 (4.86)** 0.08 0.02 �0.05 0.15 0.13 0.25* 13.61 (4.03)* 0.29** 0.14 0.12 0.18 0.34** 0.25* 13.09 (4.13)

0.15 0.06 �0.14 0.21 �0.04 0.32** 14.71 (4.25)0.15 0.09 0.30** 0.22* 0.00 3.09 (3.48)

0.23* 0.13 0.19 0.12 0.17 3.47 (4.03)09 0.01 �0.01 �0.08 0.19 0.11 4.26 (4.29)

0.08 0.05 0.16 0.31** 0.37** 7.74 (2.37)0.31** 0.14 0.12 0.49*** 0.30* 7.67 (2.76)0.11 0.08 0.05 0.36** 0.47*** 8.03 (2.39)

0 3.05 3.86 4.05 6.59 7.12 7.788) (3.02) (4.10) (4.02) (2.99) (2.96) (1.82)

.73

1

.57

1

3

.73

1

8

-.45

.49 .37 .41

BaselineLevel

LinearChange

MentalHealth T0

MentalHealth T1

MentalHealth T2

BaselineLevel

LinearChange

CivicParticipation T0

CivicParticipation T1

CivicParticipation T2

1

1

31

8

-.71

-.26*

-.04

-.05

-.21

-.38* -.28**

Sociotherapy

.01 .41*

.14.00 .08 .16

Fig. 1. Associative LGM involving sociotherapy as a predictor of change. For the model: c2(9) ¼ 14.51, p ¼ 0.105, RMSEA ¼ 0.06, CFI ¼ 0.98, NNFI ¼ 0.94, AIC ¼ 66.51. Observed variablesare shown as rectangles, factors as ovals. Covariances are shown as curved lines with correlations. Paths are shown as arrows with standardized regression coefficients or fixedregression weights (without decimal places). Endogenous variables are shown with R2. *p < 0.05, **p < 0.01.

F. Verduin et al. / Social Science & Medicine 121 (2014) 1e9 7

embodies the resource potential of social networks, it should not beoverlooked that the process of activating or converting social ties toachieve desired outcomes is particularly salutary. Collective efficacyrefers to a community's capacity to deal adequately with environ-mental demands and to achieve goals through its social organiza-tion that cannot be achieved by individuals alone. De Silva et al.(2007) hypothesized that through collective actions communitymembers can increase control over their lives and environment,and that this increased control over post-disaster demands may inturn mitigate individual mental health problems. Wind et al. (2011,2012) found that high social capital may foster collective efficacyand decrease the employment of coping strategies and social sup-port by individuals, and that social capital exerts its effect onmental health outcomes via collective efficacy.

In our study a small positive effect of the interventionwas foundon both mental health and civic participation, a concept akin tocollective efficacy. This may indicate that community level psy-chosocial interventions like the sociotherapy program might behelpful in promoting post-conflict recovery in contexts likeRwanda. Such conclusion would be important but untimely, giventhe study's limitations.

Spirituality may have played a role as a confounding variable, asthe sociotherapy program was openly supported by the EpiscopalChurch of Rwanda, and several group facilitators were pastors whoalso used religious exercises duringmeetings. However, there is not

Table 4Effects of 15 weeks of sociotherapy on mental health and civic participation.

Sociotherapy group Control group P

Mean S.E. Mean S.E.

Mental healthBaseline level 8.33 (0.48) 8.28 (0.48) 0.942Linear changea �0.28 (0.05)*** �0.09 (0.05) 0.013

Civic participationBaseline level 6.61 (0.26) 7.70 (0.26) 0.003Linear changea 0.13 (0.04)*** 0.03 (0.04) 0.035

***p < 0.001.a Linear change estimates are per month.

much indication that the effect of the program is attributable tospecific characteristics of individual group leaders. Using a Bayesiananalysis, Scholte et al. (2011a) assessed the extent to which thesociotherapy effectiveness differed between groups. It appearedthat only 14% of the total variability in the score change could beattributed to factors associated with specific sociotherapy groups orany background of group leaders, e.g. spirituality. Notably, groupleaders also had backgrounds other than religious, varying fromteachers, local leaders, or civil servants to NGO staff. The mostrelevant working mechanisms as identified through qualitativeresearch were: *enabling to ‘unburden the heart’; *replacingtraditionally hierarchical relationships successfully by relationshipsbased on equality; and *providing the opportunity to reconnectwith others in the group (Richters et al, 2013).

The validity of outcomes of this study may have been limited bythe use of response categories in questionnaires, as thesemaymasknuances in definitions and understanding of complex concepts.Even though qualitative research and extensive validation of theShort A-SCAT was performed in various countries, and despite ouradaptation of the instrument to the local context, responders maynot have interpreted all items in their intendedmeaning. This studydid not establish change over time for support and cognitive socialcapital; we only found change over time for civic participation. Theintervention may have exerted an effect on civic participation only,without affecting the other elements of social capital wemeasured.In their study on the promotion of social capital in Nicaragua, Bruneand Bossert (2009) already pointed to indications that in non-western cultures cognitive and structural components may bedisconnected. In the highly complex sociocultural setting of post-genocide Rwanda, elements of cognitive social capital like trust,sense of belonging and solidarity are not easily discussed openly.Therefore, items of our questionnaire addressing such sensitiveissues may not have elicited valid responses. Especially men tendnot to trust others easily and to keep problems inside. Qualitativeinformation consistently pointed out that men in Rwanda generallydo not share and seek support for emotional problems (Richterset al., 2005, 2008; Richters, 2010; Verduin et al., 2010).

In turn, lack of trust may impact mutual support. The use ofcredit in exchanges was common in preconflict Rwanda (Coletta

F. Verduin et al. / Social Science & Medicine 121 (2014) 1e98

and Cullen, 2000). This practice has diminished over time, in partdue to decreased levels of trust as a consequence of war, but alsobecause of increasing poverty and the value placed on money andindividualism. In general, people have become more reluctant togive gifts, for they are less confident that these acts will bereciprocated. This should be taken into consideration whenaddressing the issue of support in countries as poor as Rwanda(Verduin et al., 2010). In our study, scores on items of the Short A-SCAT addressing support were not consistent, and no change overtime was established. This may be due to insufficient validity ofthe items concerned. A systematic cognitive validation of theinstrument's version used in this study, performed in accordancewith a method described by De Silva et al. (2005b), revealed thatresponders did not seem to differentiate between the variouskinds of support mentioned in the respective items: comforting,encouraging, or economic support (Verduin et al., 2010). In thisvery poor region, all support seemed to be expressed in materialforms. If no goods were received or given, no comfort orencouragement was experienced or provided. De Silva et al.(2005b) describe similar misunderstandings in respondents'interpretation of items addressing support when using the A-SCAT in Peru and Vietnam. Another limitation of this study is thatanalyses did not include respondents' individual appraisal ofcurrent and past experiences or their coping behavior, both ofwhich may mediate the association between social capital andmental health (Wind et al., 2011). In our study we could not showa correlation between the effect of the intervention on socialcapital and mental health, which was contrary to what might beexpected. Investigation of such possible mediators might have ledto a clearer understanding of the relation between both out-comes. Also, other mediating factors such as social functioning,substance abuse and domestic violence might have beenconsidered.

Our study results may also have beenweakened bymissing data,although it has been shown that under ignorable missing dataconditions FIML estimates are unbiased (Enders and Bandalos,2001). Additionally, larger samples might have increased the val-idity of the multilevel models. And finally, time may have impactedstudy outcomes in various ways: Firstly, time intervals may nothave been long enough to measure real change in this fragile so-ciety. Secondly, a delay may exist between social capital gains andmental health gains, which would affect the analysis of the corre-lation between linear changes.

It is unclear to us why in our study T0 scores on civic partici-pation were higher in the control group than in the interventiongroup. Possibly, individuals with lower scores were not the oneseasily run into during the convenience sampling, while interven-tion group respondents had been actively asked to join the groups.Also, our experimental and control group differed on socioeco-nomic status. We do not think that this seriously impacted theactual equivalence of both groups. Byumba's population isextremely poor in general, and actually there is little real variety inSES. Possibly, the difference is caused by the method of SES deter-mination. Contrary to the control group, participants of theexperimental group described the state of their houses themselves.This may have resulted in a less divergent SES score distribution inthe experimental group.

The intervention method may benefit if groups appoint oneparticipant as being responsible for meetings to continue after theprogram has formally ended. After the intervention in Rwandavarious groups actually continued meeting, sometimes even start-ing joint income generating activities, but not all did so. Notably,organizing communities through the appointment of so-called‘responsables’, has for a long time been common practice inRwanda.

During a seven years period, over 10,000 individuals partici-pated in the sociotherapy program studied here. Over the course oftime the Episcopal Church of Rwanda became the sole co-ordinatorof the intervention, albeit with ongoing external financial support.Lastly, the intervention was also implemented by local NGOs inother regions in Rwanda (http://sociotherapy.org) and in theneighboring countries Burundi and the Democratic Republic ofCongo. Recently a sociotherapy program also started in Liberia.

5. Conclusion

This study addresses the potential of community based socio-therapy to promote recovery in a war-affected population. Evalu-ating the intervention's delivery is complicated by the method'sopenness to contextual determinants, such as the different back-grounds, living situations and wishes of group participants andleaders. Further identification of effective elements requires closemonitoring of practices per group and per session. Also, applicationin strongly differing populations may reveal to what degree themethod's effectiveness is impacted by cultural and contextual fac-tors like religion and history.

The study hints at the possibility to foster one element of socialcapital, i.e. civic participation, and to simultaneously impact mentalhealth. However, further research is needed. The sensitivity of theShort A-SCATas ameasure of social capital in complex socioculturalsettings like Rwanda needs to be confirmed. In particular, the val-idity of items addressing structural social capital and support needsattention (Verduin et al., 2010). Future studies also need to deter-mine which specific elements of sociotherapy are effective, whichparticular elements of social capital can intentionally be promotedin different settings, and what are possible pathways of influenceon mental health. Further elaboration is highly relevant as post-conflict recovery of a population is best served by an increase ofboth societal cohesion and mental health. It may lead the way tocommunity based psychosocial interventions that effectively pro-mote wellbeing in war-effected populations.

Acknowledgments

This study was partly supported by a grant to Femke Verduinfrom the Health Research Development Counsel, Department Pre-vention Program (ZonMW), OOG-Geestkracht (ZonMW: 60-60105-98-117).

We also thank Cordaid for funding the research and PrinsBernhard Cultuurfonds for their grant to F. Verduin. We thankRutayisire Theoneste, Ngendahayo Emmanuel, Sarabwe Emmanueland Hategekimana Balthazar for helping us organise and conductthe research. We also thank our loyal interviewers: Rukundo Ange,Mugisha-Bitanuzire John-Peter, Umuhoza Adelin, Uwizeye Beata,Ingabire Bernadette, Uwizeyimana Albertine, Mukanyiligira MarieGrace, and Muhongwanseko Scholastique.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.socscimed.2014.09.054.

References

Ager, A., 2002. Psychosocial needs in complex emergencies. Lancet 360 (Suppl. l),s43es44.

Almedom, A.M., 2005. Social capital and mental health: an interdisciplinary reviewof primary evidence. Soc. Sci. Med. 61, 943e964.

Berry, H.L., Welsh, J.A., 2010. Social capital and health in Australia: an overview fromthe household, income and labour dynamics in Australia survey. Soc. Sci. Med.70 (4), 588e596.

F. Verduin et al. / Social Science & Medicine 121 (2014) 1e9 9

Betancourt, T.S., Meyers-Ohki, S., Stulac, S.N., Barrera, A.E., Mushashi, C.,Beardslee, W.R., 2011. Nothing can defeat combined hands (Abashize hamwentakibananira): protective processes and resilience in Rwandan children andfamilies affected by HIV/AIDS. Soc. Sci. Med. 73 (5), 693e701.

Betancourt, T.S., McBain, R., Newnham, E.A., Brennan, R.T., 2014. Context matters:community characteristics and mental health among war-affected youth inSierra Leone. J. Child Psychol. Psychiatry 55 (3), 217e226.

Brune, N.E., Bossert, T., 2009. ‘Building social capital in post-conflict communities:evidence from Nicaragua’. Soc. Sci. Med. 68 (5), 885e893.

Coletta, N.J., Cullen, M.L., 2000. Violent Conflict and the Transformation of SocialCapital. Lessons from Cambodia, Rwanda, Guatemala, and Somalia. World Bank,Washington, DC.

De Jong, J.T., Komproe, I.H., van Ommeren, M., 2003. Common mental disorders inpostconfict settings. Lancet 361, 2128e2130.

De Silva, M.J., McKenzie, K., Harpham, T., Huttly, S.R.A., 2005a. Social capital andmental illness: a systematic review. J. Epidemiol. Community Health 59,619e627.

De Silva, M.J., Harpham, T., Tuan, T., Bartolini, R., Penny, M.E., et al., 2005b. Psy-chometric and cognitive validation of a social capital measurement tool in Peruand Vietnam. Soc. Sci. Med. 62 (4), 941e953.

De Silva, M.J., Huttlya, S.R., Harpham, T., Kenward, M.G., 2007. Social capital andmental health: a comparative analysis of four low income countries. Soc. Sci.Med. 64 (1), 5e20.

Duncan, T.E., Duncan, S.C., Strycker, L.A., 2006. An Introduction to Latent VariableGrowth Curve Modeling: Concepts, Issues, and Applications, second ed. In:Quantitative Methodology Series Lawrence Erlbaum Associates, Mahwah (NJ),London.

Enders, C.K., Bandalos, D.L., 2001. The relative performance of full informationmaximum likelihood estimation for missing data in structural equation models.Struct. Equ. Model. 8 (3), 430e457.

Engstr€om, K., Mattsson, F., J€arleborg, A., Hallqvist, J., 2008. Contextual social capitalas a risk factor for poor self-rated health: a multilevel analysis. Soc. Sci. Med. 66(11), 2268e2280.

Eriksson, M., 2011. Social capital and health e implications for health promotion.Glob. Health Action 4, 5611. http://dx.doi.org/10.3402/gha.v410.5611.

Hamano, T., Fujisawa, Y., Ishida, Y., Subramanian, S.V., Kawachi, I., et al., 2010. Socialcapital and mental health in Japan: a multilevel analysis. PLoS One 5 (10),e13214. http://dx.doi.org/10.1371/journal.pone.0013214.

Harpham, T., Grant, E., Thomas, E., 2002. Measuring social capital within healthsurveys: key issues. Health Policy Plan. 17, 106e111.

Hobfoll, S.E., Watson, P., Bell, C.C., Bryant, R.A., Brymer, M.J., Friedman, M.J., et al.,2007. Five essential elements of immediate and mid-term mass trauma inter-vention: empirical evidence. Psychiatry 70, 283e315.

Hu, L.T., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structureanalysis: conventional criteria versus new alternatives. Struct. Equ. Model. 6 (1),1e55.

Irwin, J., LaGory, M., Ritchey, F., Fitzpatric, K., 2008. Social assets and mental distressamong the homeless: exploring the roles of social support and other forms ofsocial capital on depression. Soc. Sci. Med. 76 (12), 1935e1943.

Islam, M.K., Merlo, J., Kawachi, I., Lindstr€om, M., Gerdtham, U.G., 2006. Social capitaland health: does egalitarianism matter? A literature review. Int. J. Equity Health5 (5), 3.

Kawachi, I., Berkman, L.F., 2001. Social ties and mental health. J. Urban Health 78,458e467.

Kawachi, I., Subramanian, S.V., 2006. Measuring and modeling the social andgeographic context of trauma: a multilevel modeling approach. J. Trauma. Stress19 (2), 195e203.

Macinko, J., Starfield, B., 2001. The utility of social capital in research on healthdeterminants. Milbank Q. 79 (3), 387e427. IV. Review.

McKenzie, K., Whitley, R., Weich, S., 2002. Social capital and mental health. Br. J.Psychiatry 181, 280e283.

Michael, Y.L., Farquhar, S.A., Wiggins, N., Green, M.K., 2008. Findings from acommunity-based participatory prevention research intervention designed toincrease social capital in Latino and African American communities. J. Immigr.Minor. Health 10 (3), 281e289.

Nakhaie, R., Arnold, R., 2010. A four year (1996e2000) analysis of social capital andhealth status of Canadians: the difference that love makes. Soc. Sci. Med. 71 (5),1037e1044.

Norris, F.H., Stevens, S.P., Pfefferbaum, B., Wyche, K.F., Pfefferbaum, R.L., 2008.Community resilience as a metaphor, theory, set of capacities, and strategy fordisaster readiness. Am. J. Community Psychol. 41, 127e150.

Patel, V., 2010. Building social capital and improving mental health care to preventsuicide. Int. J. Epidemiol. 39 (6), 1411e1412.

Pronyk, P.M., Harpham, T., Busza, J., Phetla, G., Morison, L.A., Hargreaves, J.R.,Kim, J.C., Watts, C.H., Porter, J.D., 2008. Can social capital be intentionallygenerated? A randomized trial from rural South Africa. Soc. Sci. Med. 67 (10),1559e1570.

Putnam, R., 1993. Making Democracy Work: Civic Traditions in Modern Italy.Princeton University Press, Princeton.

Putzel, J., 1997. Policy arena: accounting for the ‘dark side’ of social capital: readingRobert Putnam on democracy. J. Int. Dev. 9 (7), 939e949.

Richters, A., 2010. Suffering and healing in the aftermath of war and genocide inRwanda: mediations through community-based sociotherapy. In: Kapteijns, L.,Richters, A. (Eds.), Mediations of Violence in Africa: Fashioning New Futuresfrom Contested Pasts. Brill, Leiden.

Richters, A., Dekker, C., De Jonge, K., 2005. Reconciliation in the aftermath of violentconflict in Rwanda. Intervention 3 (3), 203e221.

Richters, A., Dekker, C., Scholte, W.F., 2008. Community based sociotherapy inByumba, Rwanda. Intervention 6 (2), 100e116.

Richters, A., Rutayisire, T., Slegh, H., 2013. Sexual transgression and social discon-nection: healing through community-based sociotherapy in Rwanda. Cult.Health Sex. Int. J. Res. Interv. Care 15 (Suppl. 4), S581eS593.

Rodin, D., van Ommeren, M., 2009. Commentary: explaining enormous variations inrates of disorder in trauma-focused psychiatric epidemiology after majoremergencies. Int. J. Epidemiol. 38 (4), 1045e1048.

Sampson, R.J., Raudenbush, S.W., Earls, F., 1997. Neighborhoods and violent crime: amultilevel study of collective efficacy. Science 277 (5328), 918e924.

Scholte, W.F., Ager, A.K., 2014. Social capital and mental health: connections andcomplexities in contexts of postconflict recovery. Intervention 12 (2), 210e218.

Scholte, W.F., Verduin, F., Kamperman, A.M., Rutayisire, T., Zwinderman, A.H.,Stronks, K., 2011a. The effect on mental health of a large scale psychosocialintervention for survivors of mass violence: a quasi-experimental study inRwanda. PLoS One 6 (8), e21819.

Scholte, W.F., Verduin, F., Van Lammeren, A., Rutayisire, T., Kamperman, A.M., 2011b.Psychometric properties and longitudinal validation of the self-reportingquestionnaire (SRQ-20) in a Rwandan community setting: a validation study.BMC Med. Res. Methodol. 16 (11), 116.

Suzuki, E., Takao, S., Subramanian, S.V., Komatsu, H., Doi, H., Kawachi, I., 2010. Doeslow workplace social capital have detrimental effect on workers' health? Soc.Sci. Med. 70 (9), 1367e1372.

Tuan, T., Harpham, T., Thu Huong, N., de Silva, M., Thi Thuy Huong, V., Thap Long, T.,Thi, N., et al., 2005. Measuring Social Capital and Mental Health in Vietnam: aValidity Study. Young Lives. Working paper no 12.

Veenstra, G., 2000. Social capital, SES and health: an individual-level analysis. Soc.Sci. Med. 50 (5), 619e629.

Veenstra, G., 2002. Social capital and health (plus wealth, income inequality andregional governance). Soc. Sci. Med. 54 (6), 849e868.

Verduin, F., Scholte, W.F., Rutayisire, T., Richters, A., 2010. How qualitative infor-mation helped to shape quantitative research instruments in Rwanda. Inter-vention 8 (3), 233e244.

Wang, H., Schlesinger, M., Wang, H., Hsiao, W.C., 2009. The flip-side of social capital:the distinctive influences of trust and mistrust on health in rural China. Soc. Sci.Med. 68 (1), 133e142.

Whitley, R., McKenzie, K., 2005. Social capital and psychiatry: review of the liter-ature. Harv. Rev. Psychiatry 13 (2), 71e84.

Wind, T.R., Komproe, I.H., 2012. The mechanisms that associate community socialcapital with post-disaster mental health: a multilevel model. Soc. Sci. Med. 75(9), 1715e1720.

Wind, T.R., Fordham, M., Komproe, I.H., 2011. Social capital and post-disaster mentalhealth. Glob. Health Action 4, 6351. http://dx.doi.org/10.3402/gha.v4i0.6351.

Web references

http://www.annemiekrichters.nl/rwanda/new-sociotherapy-progam-2013—2016.http://sociotherapy.org/program_2013_2016.