Disease surveillance is the systematic collection, analysis, interpretation and distribution of health data for preventing health related problems. The primary purpose of disease surveillance is early detection of disease outbreaks for prevention of further morbidity and mortality. In Denmark disease surveillance is carried out by Statens Serum Institut, which denes disease outbreaks as an unusual high number of incidences of a disease. The objective of this thesis is to compare dierent statistical models for prospective detection of possible outbreaks. Adjustments for seasonal variations, secular trends and past outbreaks should be incorporated into the model. Three dierent models are used: Farringtons algorithm, a dynamic linear model and a multi-process dynamic linear model. Comparison of the models is presented applying data from Statens Serum Institut consisting of all samples tested positive for Mycoplasma pneumoniae infections from July 1994 to July 2005. The analysis indicate that the dynamic linear model and the multi-process dynamic linear model are superior to Farringtons algorithm. The threshold value in Farringtons algorithm is highly affected by the baseline values used in the calculations, where the dynamic linear model and the multi-process dynamic linear model are better at adapting to the seasonal variations and past outbreaks. The multi-process dynamic linear model has the advantage that it can identify outliers.