Addi van Bergen

172 Study Setting Population Sample size Indicators SE/SI * Health measure ** Statistical analysis Richter & Hoffmann [48] Switzerland General population 18-64 for men / 63 for women 14,969 h) S1: living alone; S2 living without a partner; S3 no person to talk to; S4 attendance of social events; S5 feeling lonely; S6 social support (3) E: low income MH1: severe mental illness or disability h) Logistic regression MH2: common mental illness h) * S= social dimension; E=economic dimension; P=political dimension; C=cultural dimension; between brackets the number of items (if more than 1). A group of letters e.g. EP or SEP indicates an aggregate measure based on the listed dimensions. ** MH=mental health. *** Code for results: + hypothesis confirmed i.e. high SE/low SI associated with adverse health outcome; ns no significant association; - hypothesis rejected i.e. low SE/high SI associated with adverse health outcome; +? high SE/low SI combined with adverse health, but no statistical testing; na=not applicable. C/C = Case/Control. $ P-value <.05 unless stated otherwise. OR’s and HR’s are given with the 95% confidence interval between brackets. SD=standard deviation. Adj=adjusted for potential confounders. $$ Code for results: + - 0 see ***; +/0 hypothesis confirmed for 30-70% of SE/SI indicators and the remaining 70-30% not significant; x no statistical testing or no associations reported. # Specific methodological limitations: Limitations: 1, no theoretical motivation of the concept SE/SI; 2, data set not originally designed to measure SE/SI; 3, not all dimensions of SE/SI measured; 4, no composite measure SE/ SI; 5, no existing SE/SI measure; 6, testing of association SE/SI -H was not a stated objective; 7, no adjustment for demographic and other potential confounding factors; ”, limitation partly present. For more details please see Supplementary file 4. ## General study quality was appraised with the CASP=Critical Appraisal Skills Programme tool for cohort studies (Supplementary file 2) or cross-sectional studies (Supplementary file 3). a) A social exclusion index was constructed with three underlying domains; S1=Civic participation (4 items e.g. participation in cultural, sports and leisure activities), S2=Social relations and resources (5 items e.g. living alone, no close friendship) and P= Service provision and access (5 items e.g. poor quality of local medical facilities). b) Psychological distress was measured with the 12 item General Health Questionnaire (GHQ). c) SAH=self-assessed health. LLTI=limiting long-term illness/disability. Transitions in SAH and LLTI were entered in the regression model with MH1 as independent variable and SP as dependent variable. d) Jehoel Gijsbers & Vrooman [35].