Traffic-related air pollution and socioeconomic status: a spatial autocorrelation study to assess environmental equity on a small-area scale

Abstract

Background: : Most ecologic studies of environmental equity show that groups with lower socioeconomic status (SES) are more likely to be exposed to higher air pollution levels than groups of higher SES. However, these studies rarely consider spatial autocorrelation in the data. We investigated the associations between traffic-related air pollution and SES on a small-area level in Strasbourg (France) and assessed the impact of spatial autocorrelation on the results.

Methods: : We used a deprivation index, constructed from census data, to estimate SES at the block level. Average ambient nitrogen dioxide (NO2) levels during year 2000, modeled at the block level by a dispersion model, served as a marker of traffic exhaust. We estimated the association between exposure to NO2 and the deprivation index by using an ordinary least squares model and a simultaneous autoregressive model that controls for the spatial autocorrelation of data.

Results: : The association between the deprivation index and NO2 levels was positive and nonlinear with both regression models; the midlevel deprivation areas were the most exposed. Control of spatial autocorrelation strongly reduced the strength of the association but clearly improved the model's goodness-of-fit; the most pronounced reduction was observed for the midlevel deprivation areas (regression coefficients decreased by 67%).

Conclusions: : This study confirms the need to take spatial autocorrelation into account in ecologic studies and shows that failure to do so may lead to biased and unreliable estimates and thus to erroneous conclusions. This may be especially important in studying the role of air pollution on social inequalities in health.