AUTHOR=de Blas Ignacio , Muniesa Ana , Vallejo Adriana , Ruiz-Zarzuela Imanol TITLE=Assessment of Sample Size Calculations Used in Aquaculture by Simulation Techniques JOURNAL=Frontiers in Veterinary Science VOLUME=7 YEAR=2020 URL=https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2020.00253 DOI=10.3389/fvets.2020.00253 ISSN=2297-1769 ABSTRACT=

An adequate sampling methodology is the key to knowing the health status of aquatic populations. Usually, the aims of epidemiological surveys in aquaculture are to detect an infection and estimate the disease prevalence, and different formulas are used to calculate the sample size. The main objective of this study was to assess if the sample sizes calculated using classical epidemiological formulas are valid considering the sampling methodology, the population size, and the spatial distribution of diseased animals in the population (non-clustered or clustered). However, the use of sample sizes of 30, 60, and 150 fish is widely accepted in aquaculture, due to the requirements of the World Organization for Animal Health (OIE) for epidemiological surveillance. We have developed a specific software using ASP (Active Server Pages) language and MySQL database in order to generate aquatic populations from 100 to 10 000 brown trouts infected by Aeromonas salmonicida with different levels of prevalence: 2, 5, 10, and 50%. Then we implemented several Monte Carlo simulations to estimate empirically the sample sizes corresponding to the different scenarios. Furthermore, we compared these results with the values calculated by classical formulas. We determined that simple random sampling was more accurate in detecting an infection, because it is independent of the distribution of infected animals in the population. However, if diseased animals are non-clustered it is more efficient to use systematic methods, even in the case of small populations. Finally, the formula to calculate sample size to estimate disease prevalence is not valid when the expected prevalence is far from 50%, and it is necessary to increase the sample size to reach the desired precision.