gms | German Medical Science

The PerSpat-Project aims to integrate spatial statistics and environmental health and was established a) to evaluate the applicability of different statistical approaches for spatial alignment and b) to investigate associations among Perfluorooctanoic Acid (PFOA) in drinking water and birth outcome.

As part of the PerSpat-Project, we here consider models of the state-wide concentrations of PFOA in drinking water of North Rhine-Westphalia, Germany, in connection with the state-wide comprehensive perinatal data set.

Drinking water samples have been drawn from both the water supply stations and the network of water supply areas. Apart from such comprehensive data, there are regions with no or few measurements, usually non-detects.

Being on postal code level, the birth data are spatially misaligned with respect to the water supply areas, particularly in affected rural regions, so a main task is to assign an exposure level to a given birth based on the mother’s approximate residence.

In order to estimate PFOA concentrations, we formulate spatial models based on the measurements and the spatial connectivity via rivers. Furthermore, the water proportions in the complex relationship of 417 stations supplying 451 areas are estimated based on their supplied and demanded amounts. These weights are used to mediate between station and area level modelling.

Weights based on the census population density are used in the application of various approaches of spatial realignment. Ambiguous postal areas are handled by interpolation of their PFOA concentrations or by further intersection to obtain unique water supply. The latter analysis’ results on the so-called atom level are compared to those on postal area level, obtaining but minor differences.

The applicability of some models is limited to count data; therefore, cases of low birth weight per area are also considered. Using the census data grid, being the finest resolution, leads to considerable random assignment of birth data, whose locations within the postal areas are not actually known.

Generally speaking, interpolation of exposure values as well as uncertainty of outcome locations lead to further variability, which has to be respected when estimating relationships in epidemiological studies based upon these information.

The methods may be adapted to other substances measured in drinking water or in a river network as well as to other spatial datasets based on, e.g., administrative regions, various environmental media such as, e.g., air or noise pollution.

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.