Statistics and sample design in epidemiological studies of Echinococcus multilocularis in fox populations

In this paper possible sampling strategies for estimating the prevalence of Echinococcus multilocularis infections in foxes are discussed. To draw valid conclusions from the analysis of fractions of a total fox population, each member of the total population must have the same chance of being selected for the investigation (random sampling), the sample must be representative with respect to all epidemiologically relevant conditions in the population (e.g. age, endemic status, seasonal effects, population density), and it must be large enough to obtain results with the required precision. For detection/exclusion of infections at a pre-specified prevalence threshold and confidence level (e.g. 99%), the required sample is rather small, but the information obtained from the data is limited. For prevalence estimates, the required sample sizes depend on the expected prevalence, the desired precision of the estimate and the chosen confidence level (e.g. 90, 95, or 99%). The samples need to be taken in spatial units where the variation of the conditions potentially influencing the infection can be neglected. A first impression of the spatial distribution of E. multilocularis infections in foxes can also be obtained by mapping the investigated sample (infected and uninfected animals) using the municipalities where they were shot or found as a spatial grid. To analyse the local influence of environmental factors, data on the geographical positions where the animals were sampled need to be collected and analysed in the context of a Geographic Information System (GIS).