Spatial analysis of binary health indicators with local smoothing techniques: The Viadana study

Abstract

IntroductionWhen pollution data from a monitoring network is not available, mapping the spatial distribution of disease can be useful to identify populations at risk and to suggest a potential role for suspected emission sources. We aimed at obtaining a continuous spatial representation of the prevalence of symptoms that are potentially associated with the exposure to the pollutants emitted from the wood factories in the children who live in the district of Viadana (Northern Italy). MethodsIn 2006, all the parents of the children aged 3–14years residing in the Viadana district (n=3854), filled in a questionnaire on respiratory symptoms, irritation symptoms of the eyes and skin, use of health services. The children's residential addresses were also collected and geocoded. Generalized additive models and local weighted regression (LOWESS) were used to estimate the distribution of the symptoms, to test for spatial trends of the symptoms' prevalence and to control for potential confounders. Permutation tests were used to identify the areas of significantly increased risk (“hot spots”). ResultsThe prevalence of respiratory symptoms, eye symptoms and the use of health services showed a statistically significant spatial variation (p<0.05), but skin symptoms did not. Symptoms' prevalence was lower in the northern part of the district, where no wood factories were present, and it was higher in the southern part, where the two big chipboard industries were located. Hot spots were identified fairly near to one of the two chipboard industries in the district. ConclusionsThe north-to-south trend in the prevalence of respiratory and eye symptoms, but not of skin symptoms, as well as the location of hot spots, are consistent with the potential exposure to air pollutants both emitted by the wood factories and related to traffic. In these “high risk areas” monitoring of pollution and preventive actions are clearly needed.