A protocol to identify and minimise selection and information bias in abattoir surveys estimating prevalence, using Fasciola hepatica as an example

Abattoir surveys and findings from post-mortem meat inspection are commonly used to estimate infection or disease prevalence in farm animal populations. However, the function of an abattoir is to slaughter animals for human consumption, and the collection of information on animal health for research purposes is a secondary objective. This can result in methodological shortcomings leading to biased prevalence estimates. Selection bias can occur when the study population as obtained from the abattoir is not an accurate representation of the target population. Virtually all of the tests used in abattoir surveys to detect infections or diseases that impact animal health are imperfect, leading to errors in identifying the outcome of interest and consequently, information bias. Examination of abattoir surveys estimating prevalence in the literature reveals shortcomings in the methods used in these studies. While the STROBE-Vet statement provides clear guidance on the reporting of observational research, we have not found any guidelines in the literature advising researchers on how to conduct abattoir surveys. This paper presents a protocol in two flowcharts to help researchers (regardless of their background in epidemiology) to first identify, and, where possible, minimise biases in abattoir surveys estimating prevalence. Flowchart 1 examines the identification of the target population and the appropriate study population while Flowchart 2 guides the researcher in identifying, and, where possible, correcting potential sources of outcome misclassification. Examples of simple sensitivity analyses are also presented which approximate the likely uncertainty in prevalence estimates due to systematic errors. Finally, the researcher is directed to outline any limitations of the study in the discussion section of the paper. This protocol makes it easier to conduct an abattoir survey using sound methods, identifying and, where possible, minimizing biases.