Precise estimates of disease transmission rates are critical for epidemiological simulation models. Most often these rates must be estimated from longitudinal field data, which are costly and time-consuming to conduct. Consequently, measures to reduce cost like increased sampling intervals or subsampling of the population are implemented. To assess the impact of such measures we implement two different SIS models to simulate disease transmission: A simple closed population model and a realistic dairy herd including population dynamics. We analyze the accuracy of different methods for estimating the transmission rate. We use data from the two simulation models and vary the sampling intervals and the size of the population sampled. We devise two new methods to determine transmission rate, and compare these to the frequently used Poisson regression method in both epidemic and endemic situations. For most tested scenarios these new methods perform similar or better than Poisson regression, especially in the case of long sampling intervals. We conclude that transmission rate estimates are easily biased, which is important to take into account when using these rates in simulation models.Simulation models are widely used to model spread and control of many different infectious diseases in both human and veterinary medicine 1 , for instance malaria 2 , SARS 3 , STDs 4 , influenza 5 , MRSA 6 , Ebola 7 , rabies 8 , scrapie 9 and mastitis 10. In these models, transmission rates are used to describe the flow of individuals in a population going from a susceptible state to an infected state, and it is important to obtain a realistic estimate of the transmission rates in order to create a useful and realistic simulation models for decision support [11][12][13] . Accurate estimation of this rate is important because it can have a major influence on model predictions and conclusions (e.g. refs 11 and 14). The rate may have different names depending on the specific mathematical representation of transfer between susceptible and disease states, e.g. infection rate, transmission coefficient and transmission rate. This key parameter is difficult to estimate for most host-pathogen models because natural processes are stochastic, and transmission events are influenced by other parameters than what can be included in a transmission model 1,15 . Thus, large datasets are often needed to reach a good estimate.Experiments that can be used to estimate the transmission rate are often both difficult and time-consuming to conduct 12 . A means to obtain longitudinal data for estimation of the transmission rate can be to sample a subpopulation instead of the whole population, as for instance the study of Backer et al. 16 where field data were used to estimate transmission rate of hepatitis E virus in pigs, subsampling down to 5% of the population. This can be the case for instance if animals must be caught prior to sampling, if the population is large, or if the test for disease is expensive 17 . In such situations subsampling can be a convenie...

No abstract

No abstract

No abstract

No abstract