Table (B)

Characteristics of the seven studies measuring social capital at the ecological level

Study refSettingStudy designPopulationSample sizeValiditySocial capital measureMental illness measureStatistical analysisResults
NS, age not specified. Key to type of study: *study published in peer reviewed journal or accepted for publication, †unpublished manuscript, ‡report/book chapter, §PhD thesis. Study design: XS, cross sectional; L, longitudinal; E, ecological. Social capital measure: C, cognitive social capital; S, structural social capital. Mental illness measure: CMD, common mental disorders. Key to methodological limitations: Measurement of social capital—1, secondary analysis of survey questions not originally designed to measure social capital; 2, not all aspects of social capital measured, or combined different aspects of social capital into one score; 3, no information in validity of social capital measure; 4, clusters may not represent respondents’ views of their community. Methodological limitations of study that may bias results—5, testing the relation between social capital and mental illness not a stated objective of the study; 6, sampled from one community type so little variation in social capital scores between individuals; 7, response rate less than 60%. Features of analysis that may bias results—8, no control for confounding by socioeconomic status; 9, hierarchical data structure (individual and community level variables, or clustered data), but only single level modelling used; 10, neighbourhood disorder, violence or psychological resources adjusted for. These variables may be on the causal pathway between social capital and mental illness, thus making the relation non-significant. Key to results: ++, positive association between social capital and mental illness significant at p<0.05; +, positive association between social capital and mental illness not significant at p = 0.05; − −, inverse association between social capital and mental illness significant at p<0.05; −, inverse association between social capital and mental illness not significant at p = 0.05; ., non-significant association, direction of association not known.
Stafford 200356§England, Scotland National sampleXS16+ years9223 adults from 239 electoral wards/postcode sectors2, 4C  =  trust, attachment to neighbourhood, reliance on neighbours for practical helpCMD measured using GHQ12MultilevelC  =  Trust  =  + Neighbourhood attachment  =  +
S  =  participation in organised activitiesHelp from neighbours  =  −
Aggregated to electoral ward levelS  =  −
Boydell 200257*England UrbanEAll agesPopulation from two electoral wards2, 3, 4, 7, 8C  =  social cohesionPsychotic: standardised incidence rates of schizophrenia made by clinical teamsBi-variateC  =  − −
Aggregated to electoral ward level
Drukker 200358*Netherlands UrbanXS10–12 years576 from 36 neighbourhoods2, 3, 7C  =  social cohesion and trust.Child Health Questionnaire mental illness and behaviour subscalesMultilevelC  =  Mental illness  =  + Behaviour  =  +
Aggregated to neighbourhood level
Harper 2003†USAEAll agesTotal population of 311 metropolitan areas1, 2, 3, 4S  =  number of membership organisations per capitalDeath rate from suicideMultilevelS  =  Males  =  + + Females  =  +
Aggregated to metropolitan area level
Desai 200359*USALPatients discharged from a veterans’ psychiatric inpatient program (age NS)1219331, 3, 4, 8Level of community organisational life, engagement in public affairs, community volunteerism, informal sociability, social trust.Death from suicide within one year of dischargeMultilevel− −
Aggregated to state level
Rosenheck 200160*USA UrbanLHomeless people with severe mental illness (age NS)3293 respondents’ from 18 communities1, 2, 3Thin trust, volunteering, community project involvement, club meetings attended, and proportion of adults who voted Aggregated to county levelCMD and psychotic problems measured using C-DIS-R and PERI.MultilevelCMD  =  +
Alcohol and drug problems assessed Addiction Severity IndexAlcohol  =  + + Drug  =  +
Cutrona 200061*USAXSAfrican American women, primary caregiver for a 10–12 year old child (age NS)700 women from 31 clusters created from259 block census areas2, 3, 4, 10C  =  social cohesionCMD measured using Mini-MASQMultilevelC  =  −
Aggregated to cluster level