At the census tract level, we examined the relationship between outdoor concentrations of NO2 and socioeconomic status in an area of northern Spain. Outdoor concentrations of NO2 are higher for higher level of education and with higher socioeconomic index based on occupation in census tracts with over 50% urban area. By contrast, in census tracts in more rural areas, we found higher NO2 concentrations with a lower socioeconomic index and no relationship with the mean educational level.
The strength of the association with outdoor NO2 concentration was different between the socioeconomic indicator based on occupational status and that based on education. The need for a careful definition of socioeconomic variables has been identified in previous studies as an important issue [3, 6]. Different socioeconomic indicators were also found to be associated with mortality and cancer incidence in a US study . In general, socioeconomic position is determined through such variables as occupation, education, income and wealth . In the current study, we did not have information on income distribution. Our study illustrates the importance of gathering as much information as possible from a specific population if we wish to assess a potential confounding by area-level socioeconomic position in environmental epidemiology studies. In general, socioeconomic position is associated with individual health both at the individual and area level . It is also important to highlight the potential impact of the spatial autocorrelation on the association estimates. Introducing the spatially lagged variable into the model allowed controlling for the presence of spatial autocorrelation.
Furthermore, in the same region, we discovered different sizes and directions of the associations, which underline the complexity of assessing the spatial correlation between exposure levels and socioeconomic patterns. This finding is consistent with that of other recent studies in Spain, in which it was established that environmental inequalities in spatially determined exposures may not always be great and may not always be negative in direction . These findings may indicate that this is a national issue, rather than one typical only for a studied region. Further research is needed to clarify the uncertain relationship between socioeconomic indexes, especially in non-urban areas, where little is known about this issue.
The positive correlations found in mostly urban areas are in line with the findings in recent publications [26, 42], in which populations with higher socioeconomic positions tended to be more exposed. This observation is in contrast with those of many other studies, which reported environmental disadvantages for groups with low socioeconomic status [1, 3, 20]. The inconsistent results across studies may be due to methodological differences or reflect different processes that underlie the relationship between pollution sources and socioeconomic factors .
Our analysis was performed at the census-tract scale, which is generally preferable to using zip codes . In the urban area, census-tract scale is a fairly fine spatial scale and reflects neighborhood exposure. In sparsely populated rural areas, census-tract scale is on a large scale. Nevertheless, our study did not reflect small-scale variations related to the amount of traffic on the nearest road, which has been carried out in several investigations .
We used NO2 to represent the complex mixture of outdoor air pollution mixture; we employed NO2 as a surrogate for traffic-related exposure to ambient air pollutants, especially particulates, as has been done in previous studies [20, 26]. NO2 was calculated from a LUR model , which was developed to assess precisely the risks of exposure, as have been suggested in numerous studies [43, 44]. With this assessment, the mean levels for all the census tracts were below the annual limit of 40 μg/m3 recommended for NO2 by the World Health Organization air-quality guides  and established by European Directive 2008/50/CE .
Other studies have also used dispersion models . The model used in the latter study included predictor variables, which have been used in other LUR models. It is very unlikely that these variables artificially induced a correlation, particularly in the urban areas. That model also included percentage of agricultural land cover, which can be inversely related to the variable used to split the analyses; however, land cover was categorized into continuous urban, discontinuous urban, agricultural, and industrial, and so in that case the variable percentage of industrial land could be used as a weighting variable. Moreover, we do not think that this type of relationship could have had an influence on the associations found in the present study.
One limitation of our study is that we evaluated outdoor exposures, not personal exposure. Hence differences in time activity patterns between different socioeconomic groups could not be accounted for. A French study suggested that while subjects in the least deprived neighborhoods in the suburbs experienced lower outdoor NO2 concentrations, their commuting exposures could be higher .
A further limitation is the combination of socioeconomic data for 2001 and pollution data for 2005. However, it is unlikely that both socioeconomic and pollution spatial patterns changed appreciably over the space of four years.
Associations between socioeconomic position and environmental exposure may be due to a variety of processes, such as housing prices and political decisions . In the twentieth century, enormous growth in the population of the study region occurred owing to the construction of several large factories in the Avilés urban nucleus and its surroundings. In 1953, construction work began on the ENSIDESA factory—a large steel mill that is currently owned by Arcelor Mittal Heavy Steel Industry. More recently, other major companies in the area have included Saint Gobain Glass Ltd.; this company together with ENDASA (currently owned by Alcoa Inespal Aluminium Industry Ltd.), Asturian Zinc Industry Ltd., DuPont Industry, and Fertiberia Ltd. Have transformed Avilés into one of Spain’s main industrial centers (Additional file 3). This could explain the urban structure of the population studied, the great variability found in the rural areas, and the low correlation between pollution and educational level in this area.
Even though air pollution has become a major concern for its impact on health, and it may vary under different socioeconomic and demographic conditions, few studies in Spain have examined the distribution of air pollution levels by census tract, and related it to a socioeconomic index. With the present study, we were able to obtain maps of the pollution in Asturias and determine how the population is distributed with regard to demographic characteristics and different levels of NO2 exposure. From an epidemiological point of view, this study is important because socioeconomic characteristics may have an impact on the association between exposure levels and health outcomes.