Neighbourhood-level effects on psychoses: re-examining the role of context

Abstract

The incidence of schizophrenia varies by individual-level characteristics and neighbourhood-level attributes. Few specific socio-environmental risk factors (SERFs) have been identified at the neighbourhood level. Cross-level interactions are poorly understood. We investigated these issues using data from the Aetiology and Ethnicity in Schizophrenia and Other Psychoses (AESOP) study. All incidence cases of ICD-10 schizophrenia (F20) and other non-affective psychoses (F21-29), aged 16-64 years, across 33 wards in Southeast London were identified over a 2-year period (1997-1999). Census data provided the denominator for each ward. Multilevel Poisson regression simultaneously modelled individual- and neighbourhood-level SERFs, including socio-economic deprivation, voter turnout (proxy for social capital), ethnic fragmentation (segregation) and ethnic density. A total of 218 subjects were identified during 565 576 person-years at risk. Twenty-three per cent of variance in incidence of schizophrenia across wards could be attributed to neighbourhood-level risk factors [95% confidence interval (CI) 9.9-42.2]. Thus, 1% increases in voter turnout [incidence rate ratio (IRR) 0.95, 95% CI 0.92-0.99] and ethnic segregation (IRR 0.95, 95% CI 0.92-0.99) were both independently associated with a reduced incidence of 5%, independent of age, sex, ethnicity, deprivation and population density. This was similar for other non-affective psychoses. There was some evidence that ethnic minority individuals were at greater risk of schizophrenia in areas with smaller proportions of minority groups (p=0.07). SERFs at individual and neighbourhood levels were implicated in the aetiology of psychosis, but we were unable to determine whether these associations were causal. Individual risk may be mediated by social capital, which could operate as a protective factor, perhaps moderating social stress in the onset of psychoses.