Abstrakt
In epidemiology, it is very important to estimate the baseline incidence of infectious diseases, but the available data are often subject to outliers due to epidemic outbreaks. Consequently, the estimate of the baseline incidence is biased and so is the predicted epidemic threshold which is a crucial reference indicator used to suspect and detect an epidemic outbreak.
Another problem is that the "usual" incidence varies in a season dependent manner, i.e. it may not be constant throughout the year, is often periodic, and may also show a trend between years. To take account of these factors, more complicated models adjusted for outliers are used.
If not adjusted for outliers, the baseline incidence estimate is biased. As a result, the epidemic threshold can be overestimated and thus can make the detection of an epidemic outbreak more difficult.
Classical Serfling's model is based on the sine function with a phase shift and amplitude. Multiple approaches are applied to model the long-term and seasonal trends.
Nevertheless, none of them controls for the effect of epidemic outbreaks. The present article deals with the adjustment of the data biased by epidemic outbreaks.
Some models adjusted for outliers, i.e. for the effect of epidemic outbreaks, are presented. A possible option is to remove the epidemic weeks from the analysis, but consequently, in some calendar weeks, data will only be available for a small number of years.
Furthermore, the detection of an epidemic outbreak by experts (epidemiologists and microbiologists) will be compared with that in various models.