Spatially Modelling and Ranking the Risks of Helicobacter Pylori Infection

Abstract

P-777 Introduction: The Helicobacter pylori infection (H. pylori), a disease of the gastrointestinal tract, influences health and well being of people. Even though this disease is widespread, infection routes of H. pylori are not completely detected. Hygienic conditions play a central role for the infection pathway, but there are also environmental and socio-economic risk factors. There are spatial differences of the disease occurrence within a geographical area, possibly caused by spatial heterogeneity in risk factors and the spatial differences in population distribution. Spatial risk assessment, considering the heterogeneity in risk factors as well as the spatial correlation between the data, allows studying the relationship of those spatial differences. Methods: We used epidemiological data for H. pylori from a cross sectional study in Leipzig and 2 rural districts, Germany, including socio-economic and also hygienic risk factors. We applied conditional autoregressive models (CAR), which take into account spatial heterogeneity of risk factors and the population distribution and, additionally, consider spatial correlation between neighbored or nearby data. This method splits the total spatial disease risk into (1) exposure related risk (exposure risk) and (2) risk related to spatial locations of the sub areas (area-specific location risk). Results: Exposure to poor hygienic conditions, to environmental tobacco smoke (ETS) and to further factors is associated with increased risk of H. pylori infection in children. We ranked all risk factors by the intensity of their health effect, accounting for correlation between the risk factors. Before adjusting the spatial disease risks for important risk factors, we observed several hot spots for the H. pylori infection over the area of interest. After taking the heterogeneities in exposure distribution into account, only some of the previously identified hot spots of the area-specific location risk remained statistically significant. Conclusions: Spatial differences in risk may be caused by a heterogeneous distribution of risk factors. The presented CAR model splits the total risk of infection into risks attributable to several risk factors and offers extensive possibilities to estimate unbiased spatial risks by taking into account the specific characteristics of spatial data. Aknowledgements: This research was financially supported by the European Union (European Commission, FP6 Contract No. 003956). Key Words: risk ranking, CAR, conditional autoregressive model, spatial correlation, spatial risk modeling