Research Seminar 9.12.2008 The risk classification of pig farms and the benefit of network information based on foot-and-mouth disease simulations
PhD Tapani Lyytikäinen
Risk Assessment Unit, Evira, Helsinki
Evira/Helsinki Viikki, Auditorium C111 Kalevi
Tu 9 December 2008, 3:00-4:00 pm.
The risk classification of pig farms through simulation
The foot-and-mouth disease risk assessment project aims at predicting the extent of individual disease outbreaks if a foot-and-mouth disease outbreak occurred on one farm in Finland. In order to assess it, an epidemiologic simulation model has been under development since 2006. The model simulates the transmission of the disease among farms and makes use of sources such as the Pig Register and, for farm locations, the Farm Register.
As a part of the model’s development, we have tested the part of the model that describes pig production and started disease outbreaks from all pig farms that were in operation in 2006. These simulations do not describe fully the expected value of a foot-and-mouth disease that originates on a pig farm because cattle and sheep production also affect the spread of the disease. However, they do describe its transmission within the area of pig production.
According to the results, disease outbreaks originating on different farms differ in size. The farms can be grouped according to the size of disease outbreaks they are able to cause on the average and by their largest size and according to the general probability of a disease outbreak and its duration.
The farms were divided into four risk classes, the highest of which had slightly over 300 pig farms. When the risk class of a farm was compared with the risk class of the farm where the infection had originated, it was discovered that the farms in the worst group had a greater-than-average tendency to bring about the spread of the disease to other farms in the same group.
Based on the simulation results, risk classification is a tempting method because the anticipated size and duration of an epidemic can be incorporated into the classification. Thus the classification can be assigned a financial value as well. The method should be further developed and studied, however, before it can be exploited in practice in matters such the steering of disease monitoring.
The benefit of network information based on foot-and-mouth disease simulations
The transmission model developed in the project makes direct use of animal transport data contained in the registers and also otherwise attempts to give a network-based description of animal production. This differs from traditional epidemiologic modelling, which assumes that contacts have a random mix in a population studied.
In more developed network models, the network is described both geographically (where from on the farm – where to on the farm) and as a temporal phenomenon (when). Since network models require large amounts of detailed information, it is reasonable to ask whether they provide any actual benefit in risk assessment. This was studied by comparing the simulation results from the foot-and-mouth disease model with predictions without geographic network descriptions (stochastic model) in the assessment of the most essential parameters required for risk assessment.
The results indicate that network description is useful. The stochastic model yielded faulty results in some 40% of the simulation’s starting farms in terms of the average size of the epidemic. The mean prediction at the country level did not deviate much, however, because the results deviated in two directions: the stochastic model either overestimated or underestimated the average size of the epidemic, depending on the farm where the disease outbreak started.
The stochastic model far more clearly overestimated the size of the largest disease outbreak and the variations in the size of the expected epidemic. The results are of significance particularly in the assessment of the size of the worst disease outbreak and of “uncertainty” during the epidemic.
No advance registration to Evira’s research seminars is required.