Ranking freshwater fish farms for the risk of pathogen introduction and spread.

A semi-quantitative model is presented to rank freshwater rainbow trout farms within a country or region with regards to the risk of becoming infected and spreading a specified pathogen. The model was developed to support a risk-based surveillance scheme for notifiable salmonid pathogens. Routes of pathogen introduction and spread were identified through a process of expert consultation in a series of workshops. The routes were combined into themes (e.g. exposure via water, mechanical transmission). Themes were weighted based on expert opinion. Risk factors for each route were scored and combined into a theme score which was adjusted by the weight. The number of sources and consignments were used to assess introduction via live fish movements onto the farm. Biosecurity measures were scored to assess introduction on fomites. Upstream farms, wild fish and processing plants were included in assessing the likelihood of introduction by water. The scores for each theme were combined to give separate risk scores for introduction and spread. A matrix was used to combine these to give an overall risk score. A case study for viral haemorrhagic septicaemia is presented. Nine farms that represent a range of farming practices of rainbow trout farms in England and Wales are used as worked examples of the model. The model is suited to risk rank freshwater salmonid farms which are declared free of the pathogen(s) under consideration. The score allocated to a farm does not equate to a quantitative probability estimate of the farm to become infected or spread infection. Nevertheless, the method provides a transparent approach to ranking farms with regards to pathogen transmission risks. The output of the model at a regional or national level allows the allocation of surveillance effort to be risk based. It also provides fish farms with information on how they can reduce their risk score by improving biosecurity. The framework of the model can be applied to different production systems which may have other routes of disease spread. Further work is recommended to validate the allocated scores. Expert opinion was obtained through workshops, where the outputs from groups were single point estimates for relative weights of risks. More formal expert opinion elicitation methods could be used to capture variation in the experts' estimates and uncertainty and would provide data on which to simulate the model stochastically. The model can be downloaded (in Microsoft(®)-Excel format) from the Internet at: http://www.cefas.defra.gov.uk/6701.aspx. Crown