Using geographically referenced data on environmental exposures for public health research : a feasibility study based on the German Socio-Economic Panel Study (SOEP)

Background: In panel datasets information on environmental exposures is scarce. Thus, our goal was to probe the use of area-wide geographically referenced data for air pollution from an external data source in the analysis of physical health. Methods: The study population comprised SOEP respondents in 2004 merged with exposures for NO2, PM10 and O3 based on a multi-year reanalysis of the EURopean Air pollution Dispersion-Inverse Model (EURAD-IM). Apart from bivariate analyses with subjective air pollution we estimated cross-sectional multilevel regression models for physical health as assessed by the SF-12. Results: The variation of average exposure to NO2, PM10 and O3 was small with the interquartile range being less than 10µg/m3 for all pollutants. There was no correlation between subjective air pollution and average exposure to PM10 and O3, while there was a very small positive correlation between the first and NO2. Inclusion of objective air pollution in regression models did not improve the model fit. Conclusions: It is feasible to merge environmental exposures to a nationally representative panel study like the SOEP. However, in our study the spatial resolution of the specific air pollutants has been too little, yet.