Background Research has consistently shown an inverse relationship between socioeconomic status (SES) and both fatal and non-fatal injuries (e.g., Laflamme et al., 2009). Many studies using multilevel analysis assume independence between geographical areas, thus ignoring any spatial dependencies of injury rates. Likewise, studies using spatial modelling of rates will not capture the degree to which the observed rates are explained by individual (composition) and neighbourhood (context). This study aimed to examine SES by using both a multilevel and a spatial modelling approach. We examined the rates of all-cause non-fatal unintentional injuries among the adult population in 94 neighbourhoods in Oslo, Norway, adjusting for individual demographic and socioeconomic indicators.

Methods Multilevel Poisson regression models were used to analyse the relationship between neighbourhood SES and individual SES for 150 000 non-fatal injuries from hospital data from the Norwegian Patient Registry for residents in Oslo in the period 2009–2011. Additional registry information on each individual was linked using a unique personal identification number. Area-level information was linked from Statistics Norway. In addition we used geographically weighted regression (GWR) to capture the spatial heterogeneity in associations between injury and the explanatory variables.

Results Our analyses of hospital data showed strong evidence of socioeconomic differences at both individual and neighbourhood levels. However, the magnitude and direction of these differences was not uniform, but varied as a function of gender, age, activity and location at time of injury, diagnosis and severity of injury.

Conclusions These findings highlight that both contextual and compositional effects of socioeconomic status should be considered in allocating resources for injury prevention. Given the population-based nature of this study, these findings are likely to generalise to other settings.