First detection of Edwardsiella ictaluri (Proteobacteria: Enterobacteriaceae) in wild Australian catfish.

The bacterium Edwardsiella ictaluri is considered to be one of the most significant pathogens of farmed catfish in the United States of America and has also caused mortalities in farmed and wild fishes in many other parts of the world. E. ictaluri is not believed to be present in wild fish populations in Australia, although it has previously been detected in imported ornamental fishes held in quarantine facilities. In an attempt to confirm freedom from the bacterium in Australian native fishes, we undertook a risk-based survey of wild catfishes from 15 sites across northern Australia. E. ictaluri was detected by selective culturing, followed by DNA testing, in Wet Tropics tandan (Tandanus tropicanus) from the Tully River, at a prevalence of 0.40 (95% CI 0.21-0.61). The bacterium was not found in fishes sampled from any of the other 14 sites. This is the first report of E. ictaluri in wild fishes in Australia.

Assessing the delay to detection and the size of the outbreak at the time of detection of incursions of foot and mouth disease in Australia.

The time delay to detection of an outbreak of an emergency animal disease directly affects the size of the outbreak at detection and the likelihood that the disease can be eradicated. This time delay is a direct function of the efficacy of the surveillance system in the country involved. Australia has recently completed a comprehensive review of its general surveillance system examining regional variation in both the behaviour of modelled outbreaks of foot and mouth disease and the likelihood that each outbreak will be detected and reported to government veterinary services. The size of the outbreak and the time delay from introduction to the point where 95% confidence of detection was reached showed significant (p < 0.05) regional variation with the more remote northern areas experiencing smaller outbreaks that are less likely to spread and less likely to be reported to government services than outbreaks in the more developed southern areas of Australia. Outbreaks in the more densely populated areas may take up to 43 days until a 95% confidence of detection is achieved and at that time, the outbreak may involve up to 53 farms.

Assessing the efficacy of general surveillance for detection of incursions of livestock diseases in Australia.

Australia, as a relatively isolated country with a high level of agricultural production, depends on, and has the opportunity to maintain, freedom from a range of important diseases of livestock. Occasional incursions of such diseases are generally detected by 'passive', general surveillance (GS). In current surveillance planning, a risk-based approach has been taken to optimising allocation of resources to surveillance needs, and having mapped the relative risk of introduction and establishment of diseases of concern, a means of mapping the efficacy of GS for their detection was required, as was a means of assessing the likely efficacy of options for improving GS efficacy if needed. This paper presents the structure and application of a tool for estimating the efficacy of Australia's GS, using the example of foot and mouth disease (FMD). The GS assessment tool (GSAT) is a stochastic spreadsheet model of the detection, diagnosis and reporting of disease on a single infected farm. It utilises the output of an intraherd disease spread model to determine the duration and prevalence of infection on different types of farm. It was applied separately to each of twelve regions of Australia, demarcated by dominant livestock production practices. Each region supplied estimates of probabilities relevant to the detection of FMD, for each of fourteen farm types and all species susceptible to the disease. Outputs of the GSAT were the average probability that FMD on the farm would be detected (single farm sensitivity), the average time elapsed from incursion of the disease to the chief veterinary officer (CVO) being notified (time to detection), and the number of average properties that would need to be infected before the CVO could be 95% confident of detecting at least one. The median single farm sensitivity for FMD varied among regions from 0.23 to 0.52, the median time to detection from 20 to 33 days, and the number of properties infected for 95% confidence of detecting at least one from 4 to 12. The GSAT has proved a valuable tool in planning surveillance for detection of exotic livestock disease in Australia, and it provides a practical example of the use of probabilistic modelling to answer important questions in the face of imperfect information.

Use of a multi-criteria analysis framework to inform the design of risk based general surveillance systems for animal disease in Australia.

Australia is a major exporter of livestock and livestock products; a trade assisted by a favourable animal health status. However, increasing international travel and trade, land use changes and climatic change increase the risks of exotic and emerging diseases. At the same time, public sector resources for managing these risks are static or declining. Animal health authorities in Australia identified the need to develop a consistent national approach to surveillance that allocates resources according to risk. A study was undertaken to assess the relative likelihood of occurrence of eight significant diseases of concern to animal health authorities with the aim of producing risk maps to better manage animal disease surveillance. The likelihood of disease occurrence was considered in terms of the likelihood that a disease is introduced and the likelihood that the disease establishes and spreads. Pathways for introduction and exposure and for establishment and spread were identified and data layers representing the factors contributing to each pathway produced as raster maps. A multi-criteria analysis process was used to combine data layers into pathways and pathways into likelihood maps using weightings that reflect the relative importance of each layer and pathway. The likelihood maps for introduction and exposure and for establishment and spread were combined to generate national likelihood maps for each disease. To inform Australia's general surveillance system that exists to detect any disease of importance, the spatial profiles of the eight diseases were subsequently combined using weightings to reflect their relative consequences. The result was a map of relative likelihood of occurrence of any significant disease. Current surveillance activity was assessed by combining data layers for government disease investigations, proximity to vets and wildlife disease investigations. Comparison of the overall risk and current surveillance maps showed that the distribution of current effort was well matched to the distribution of risk.

A novel approach to assess the probability of disease eradication from a wild-animal reservoir host.

Surveying and declaring disease freedom in wildlife is difficult because information on population size and spatial distribution is often inadequate. We describe and demonstrate a novel spatial model of wildlife disease-surveillance data for predicting the probability of freedom of bovine tuberculosis (caused by Mycobacterium bovis) in New Zealand, in which the introduced brushtail possum (Trichosurus vulpecula) is the primary wildlife reservoir. Using parameters governing home-range size, probability of capture, probability of infection and spatial relative risks of infection we employed survey data on reservoir hosts and spillover sentinels to make inference on the probability of eradication. Our analysis revealed high sensitivity of model predictions to parameter values, which demonstrated important differences in the information contained in survey data of host-reservoir and spillover-sentinel species. The modelling can increase cost efficiency by reducing the likelihood of prematurely declaring success due to insufficient control, and avoiding unnecessary costs due to excessive control and monitoring.

Evaluation of post-farm-gate passive surveillance in swine for the detection of foot and mouth disease in Australia.

Pigs are considered high risk for the introduction and spread of foot and mouth disease (FMD) in Australia. One of the most likely pathways of introduction of FMD into Australia would be through the illegal importation of FMD-contaminated meat, which is then fed to feral or domestic pigs. Locations where animals from different origins are commingled, such as livestock markets and abattoirs, pose a risk for disease spread. Early detection of exotic diseases at these locations is crucial in limiting the spread of an outbreak. The aims of this study were to evaluate the likelihood of exotic disease detection with current passive disease surveillance activities for pigs at saleyards and abattoirs in eastern Australia, and make recommendations for improving surveillance. Sensitivity (Se) of the current post-farm-gate passive surveillance for detection of exotic diseases was estimated using the scenario tree modelling methodology (Martin et al., 2007a). Four surveillance system components were identified: (i) domestic saleyard, (ii) export saleyard, (iii) domestic abattoir, and (iv) export abattoir. Pig farms were classified according to herd size (Small vs. Large) and subsequently into two risk categories depending on the probability of swill feeding (Swill feeding vs. Not swill feeding). A scenario tree representing the pathways by which infected animals could be detected was developed and the Se of detection in each surveillance system component was estimated. Industry statistics, information on previous exotic disease outbreaks, and interviews with pig producers were used to estimate herd category proportions and the relative risk of swill feeding. Quantitative estimates for probabilities of detection were sourced from State legislation and policies, stakeholder consultation and observational studies at saleyards and abattoirs. Results of a FMD case study showed a Se of detection at a representative location for each surveillance system component during a 2-week period of 0.19 at domestic saleyards, 0.40 at export saleyards, 0.32 at domestic abattoirs and, 0.53 at export abattoirs. This output assumed the country was infected with herd and unit design prevalences of 1% and 30%, respectively. Improving disease awareness of saleyard and abattoir stockmen, increasing the presence of inspectors at these venues and identifying those herds posing a higher risk for FMD introduction, could improve the capacity of the country for early detection of emerging animal diseases.

Assessment of different surveillance systems for avian influenza in commercial poultry in Catalonia (North-Eastern Spain).

Compulsory surveillance programmes for avian influenza (AI) have been implemented in domestic poultry and wild birds in all the European Member States since 2005. The implementation of these programmes is complex and requires a close evaluation. A good indicator to assess their efficacy is the sensitivity (Se) of the surveillance system. In this study, the sensitivities for different sampling designs proposed by the Spanish authorities for the commercial poultry population of Catalonia were assessed, using the scenario tree model methodology. These samplings were stratified throughout the territory of Spain and took into account the species, the types of production and their specific risks. The probabilities of detecting infection at different prevalences at both individual and holding level were estimated. Furthermore, those subpopulations that contributed more to the Se of the system were identified. The model estimated that all the designs met the requirements of the European Commission. The probability of detecting AI circulating in Catalonian poultry did not change significantly when the within-holding design prevalence varied from 30% to 10%. In contrast, when the among-holding design prevalence decreased from 5% to 1%, the probability of detecting AI was drastically reduced. The sampling of duck and goose holdings, and to a lesser extent the sampling of turkey and game bird holdings, increased the Se substantially. The Se of passive surveillance in chickens for highly pathogenic avian influenza (HPAI) and low pathogenicity avian influenza (LPAI) were also assessed. The probability of the infected birds manifesting apparent clinical signs and the awareness of veterinarians and farmers had great influence on the probability of detecting AI. In order to increase the probability of an early detection of HPAI in chicken, the probability of performing AI specific tests when AI is suspected would need to be increased.

Alternative methods for computing the sensitivity of complex surveillance systems.

Stochastic scenario trees are a new and popular method by which surveillance systems can be analyzed to demonstrate freedom from pests and disease. For multiple component systems-such as a combination of a serological survey and systematically collected observations-it can be difficult to represent the complete system in a tree because many branches are required to represent complex conditional relationships. Here we show that many of the branches of some scenario trees have identical outcomes and are therefore redundant. We demonstrate how to prune branches and derive compact representations of scenario trees using matrix algebra and Bayesian belief networks. The Bayesian network representation is particularly useful for calculation and exposition. It therefore provides a firm basis for arguing disease freedom in international forums.

Current value of historical and ongoing surveillance for disease freedom: surveillance for bovine Johne's disease in Western Australia.

'Confidence' in freedom from disease is generally derived from multiple sources of varied surveillance information, and typically this surveillance evidence has been accumulated over time. In the state of Western Australia (WA) the main surveillance evidence supporting Free Zone status in the national bovine Johne's disease (BJD) program comprises periodic surveys and the ongoing clinical diagnostic system. This paper illustrates a simple approach to current valuation of historical surveillance information, based on the calculated sensitivity of the surveillance processes, the time elapsed since the data were accumulated, and the probability of new introduction of disease into the population during that elapsed time. Surveillance system components (SSCs) contributing to the overall sensitivity of the surveillance system were the clinical diagnostic system and periodic targeted surveys. Sensitivity of each component was estimated using a stochastic scenario tree model of the surveillance process as implemented. Probability of introduction of BJD into WA during each time period was estimated retrospectively from a stochastic import risk analysis model applied to actual cattle importation data. The probability that the WA cattle population was free from infection (at design prevalences of 0.2% of herds and 2% of animals within an infected herd) was estimated following each of 11 years, giving a median probability that the State was free of BJD (at these design prevalences) at the end of 2005 of 0.89. The meaning of this result is discussed.

Demonstrating freedom from disease using multiple complex data sources 2: case study--classical swine fever in Denmark.

A method for quantitative evaluation of surveillance for disease freedom has been presented in the accompanying paper (Martin et al., 2007). This paper presents an application of the methods, using as an example surveillance for classical swine fever (CSF) in Denmark in 2005. A scenario tree model is presented for the abattoir-based serology component of the Danish CSF surveillance system, in which blood samples are collected in an ad hoc abattoir sampling process, from adult pigs originating in breeding herds in Denmark. The model incorporates effects of targeting (differential risk of seropositivity) associated with age and location (county), and disease clustering within herds. A surveillance time period of one month was used in the analysis. Records for the year 2005 were analysed, representing 25,332 samples from 3528 herds; all were negative for CSF-specific antibodies. Design prevalences of 0.1-1% of herds and 5% of animals within an infected herd were used. The estimated mean surveillance system component (SSC) sensitivities (probability that the SSC would give a positive outcome given the animals processed and that the country is infected at the design prevalences) per month were 0.18, 0.63 and 0.86, for among-herd design prevalences of 0.001, 0.005 and 0.01. The probabilities that the population was free from CSF at each of these design prevalences, after a year of accumulated negative surveillance data, were 0.91, 1.00 and 1.00. Targeting adults and herds from South Jutland was estimated to give approximately 1.9, 1.6 and 1.4 times the surveillance sensitivity of a proportionally representative sampling program for these three among-herd design prevalences.