Longitudinal association between neighbourhood-level social capital and incidence of major psychiatric disorders: a cohort study of 1.4m people in Stockholm County, Sweden

Background: Greater neighbourhood-level social capital is cross-sectionally associated with lower incidence of non-affective psychotic disorders [NAPD], but longitudinal evidence is missing, and specificity to this outcome or across population groups is unclear. We investigated these issues in a large, population-based cohort in Stockholm County, Sweden, linked to prospectively-collected ratings of neighbourhood-level social capital.Methods: All Swedish-born people living in Stockholm County between 2002-2016, aged 14-64 years, were followed from cohort entry until first diagnosis of ICD-10 severe mental illness [SMI] (NAPD, affective psychotic disorder [APD] or non-psychotic bipolar disorder [NP-BPD]), out-migration from Stockholm County, 65th birthday, death, or 31 December 2016, whichever was sooner. We derived three neighbourhood-level social capital exposures (political trust, welfare trust, personal trust) for 813 neighbourhoods, averaged from individual scores from 23,510 independent, randomly-selected respondents to the 2002 Stockholm County Public Health [SPHC] survey; most SPHC respondents were Swedish-born to two Swedish-born parents (72.2%). We used multilevel proportional hazards models of each outcome-exposure association, adjusted for extensive confounder sets informed by directed acyclic graphs. We tested a priori interactions between social capital exposures and parental region-of-origin. Results: Our cohort included 1,467,128 participants, of whom 17,760 (1·2%) received their first SMI diagnosis during follow-up. Greater neighbourhood-level personal trust was associated with reduced NAPD incidence, but only amongst those of Swedish (adjusted hazard ratio [aHR] associated with a one standard deviation increase in trust: 0·88; 95%CI: 0·82-0·95) or European (aHR: 0·85; 95%CI: 0·74-0·97) parental origin (interaction p-value: <0·0001). Conversely, greater neighbourhood-level personal trust increased NAPD rates for those of North African and Middle-Eastern (aHR: 1·61; 95%CI: 1·17-2·02) or Sub-Saharan African (aHR: 1.84; 1.09-3.09) parental origins. We observed similar effect modification for NP-BPD (p<0·0001), but not APD (p<0·80), nor any effect of political or welfare trust on outcomes.Interpretation: Social capital had longitudinal effects on several mental health outcomes, but could be either protective or harmful depending parental region-of-origin, a potential indicator of outgroup position.