Applications of machine learning in animal and veterinary public health surveillance.

Machine learning (ML) is an approach to artificial intelligence characterised by the use of algorithms that improve their own performance at a given task (e.g. classification or prediction) based on data and without being explicitly and fully instructed on how to achieve this. Surveillance systems for animal and zoonotic diseases depend upon effective completion of a broad range of tasks, some of them amenable to ML algorithms. As in other fields, the use of ML in animal and veterinary public health surveillance has greatly expanded in recent years. Machine learning algorithms are being used to accomplish tasks that have become attainable only with the advent of large data sets, new methods for their analysis and increased computing capacity. Examples include the identification of an underlying structure in large volumes of data from an ongoing stream of abattoir condemnation records, the use of deep learning to identify lesions in digital images obtained during slaughtering, and the mining of free text in electronic health records from veterinary practices for the purpose of sentinel surveillance. However, ML is also being applied to tasks that previously relied on traditional statistical data analysis. Statistical models have been used extensively to infer relationships between predictors and disease to inform risk-based surveillance, and increasingly, ML algorithms are being used for prediction and forecasting of animal diseases in support of more targeted and efficient surveillance. While ML and inferential statistics can accomplish similar tasks, they have different strengths, making one or the other more or less appropriate in a given context.

L'apprentissage automatique (AA) est une approche de l'intelligence artificielle caractérisée par l'utilisation d'algorithmes qui améliorent leurs propres performances sur une tâche donnée (par exemple, la classification ou la prédiction) sur la base de données et sans avoir reçu d'instructions spécifiques ou complètes concernant la marche à suivre. Les systèmes de surveillance des maladies animales et des zoonoses sont tributaires de la mise en oeuvre efficace d'un large éventail de tâches, parmi lesquelles certaines sont susceptibles de fonctionner avec des algorithmes d'AA. Comme dans d'autres domaines, l'utilisation de l'AA s'est beaucoup développée ces dernières années dans le secteur de la surveillance de la santé animale et de la santé publique vétérinaire. Les algorithmes d'AA sont utilisés pour accomplir des tâches qui ne sont devenues possibles que grâce à l'arrivée de grandes séries de données, de nouvelles méthodes d'analyse et de capacités informatiques accrues. Parmi les exemples, on peut citer la capacité à déceler une structure sous-jacente dans de grands volumes de données provenant d'un flux continu de registres de saisies d'abattoirs, l'utilisation de l'apprentissage profond pour identifier les lésions révélées par les images numériques obtenues pendant l'abattage et l'extraction de texte libre à partir des registres sanitaires électroniques des cabinets vétérinaires à des fins de surveillance sentinelle. L'AA est cependant également appliqué dans des tâches qui s'appuyaient précédemment sur une analyse classique de données statistiques. Les modèles statistiques ont été largement utilisés pour déduire des relations entre prédicteurs et maladie afin d'étayer la surveillance fondée sur le risque ; les algorithmes d'AA sont de plus en plus utilisés pour prédire et pronostiquer des maladies animales en vue d'une surveillance plus ciblée et efficace. S'il est vrai que l'AA et la statistique inférentielle peuvent accomplir des tâches similaires, chaque approche présente ses propres atouts et pourra se révéler plus ou moins pertinente selon le contexte spécifique.

El aprendizaje automático es una vertiente de la inteligencia artificial que se caracteriza por el uso de algoritmos capaces de mejorarse a sí mismos en la ejecución de una determinada tarea (p.ej., procesos de clasificación o predicción) con empleo de datos y sin necesidad de recibir instrucciones explícitas y completas sobre la manera de lograrlo. Los sistemas de vigilancia de enfermedades animales y zoonóticas dependen de la ejecución eficaz de numerosas y muy diversas tareas, algunas de las cuales se prestan al uso de algoritmos de aprendizaje automático. Al igual que en otros campos, la aplicación del aprendizaje automático en sanidad animal y salud pública veterinaria se ha extendido sobremanera en los últimos años. Ahora se utilizan algoritmos de aprendizaje automático para realizar tareas que solo han empezado a ser factibles con el advenimiento de ingentes conjuntos de datos, nuevos métodos para analizarlos y una mayor capacidad de tratamiento informático. Entre otros ejemplos, cabe citar la determinación de la estructura subyacente de grandes volúmenes de datos procedentes de un flujo continuo de registros de los descartes de matadero; la utilización del aprendizaje profundo para detectar lesiones en imágenes digitales obtenidas durante las operaciones de sacrificio, o el análisis del texto libre de registros sanitarios electrónicos de procedimientos veterinarios con fines de vigilancia centinela. Con todo, el aprendizaje automático se está aplicando también a tareas que anteriormente reposaban en el análisis estadístico clásico de los datos. Los modelos estadísticos han sido extensamente utilizados para inferir relaciones entre una enfermedad y uno u otro predictor y alimentar a partir de ahí la vigilancia basada en el riesgo. Por otro lado, cada vez más se vienen empleando algoritmos de aprendizaje automático para predecir y anticipar enfermedades animales y conferir así más eficacia y especificidad a las actividades de vigilancia. Aunque el aprendizaje automático y la estadística inferencial realizan tareas parecidas, sus puntos fuertes son distintos, con lo cual, en función del contexto de que se trate, será preferible recurrir a uno u otro método.

Surveillance and risk assessment for early detection of emerging infectious diseases in livestock.

Those who work in the area of surveillance and prevention of emerging infectious diseases (EIDs) face a challenge in accurately predicting where infection will occur and who (or what) it will affect. Establishing surveillance and control programmes for EIDs requires substantial and long-term commitment of resources that are limited in nature. This contrasts with the unquantifiable number of possible zoonotic and non-zoonotic infectious diseases that may emerge, even when the focus is restricted to diseases involving livestock. Such diseases may emerge from many combinations of, and changes in, host species, production systems, environments/habitats and pathogen types. Given these multiple elements, risk prioritisation frameworks should be used more widely to support decision-making and resource allocation for surveillance. In this paper, the authors use recent examples of EID events in livestock to review surveillance approaches for the early detection of EIDs, and highlight the need for surveillance programmes to be informed and prioritised by regularly updated risk assessment frameworks. They conclude by discussing some unmet needs in risk assessment practices for EIDs, and the need for improved coordination in global infectious disease surveillance.

Les personnes travaillant dans le domaine de la surveillance et de la prévention des maladies infectieuses émergentes (MIE) sont confrontées à la difficulté de prédire avec exactitude le lieu d'émergence d'une maladie, ainsi que l'espèce, le système ou le site affectés. La mise en place de programmes de surveillance et de lutte contre les MIE exige une mobilisation conséquente et durable de ressources nécessairement limitées. Par contraste, le nombre des maladies infectieuses zoonotiques et non zoonotiques pouvant se déclarer est impossible à quantifier, même si l'on s'en tient aux seules maladies affectant les animaux d'élevage. Ces maladies surviennent à la faveur des nombreuses et diverses configurations, associations ou modifications qui peuvent se produire parmi les espèces hôtes, les systèmes de production, les environnements ou habitats et les types d'agents pathogènes. Compte tenu de la multiplicité de ces éléments, il devrait être fait plus largement appel à des cadres de priorisation du risque afin de soutenir les processus de prise de décision et d'allocation des ressources en matière de surveillance. Les auteurs s'appuient sur des exemples récents d'événements liés à des MIE pour faire le point sur les méthodes de surveillance appliquées pour la détection précoce de ces maladies et soulignent l'importance de documenter et de prioriser les programmes de surveillance en procédant à des mises à jour régulières des cadres utilisés pour l'évaluation du risque. Ils concluent en évoquant certains aspects importants que les pratiques actuelles d'évaluation du risque ne permettent pas de couvrir lorsqu'il s'agit de MIE, ainsi que l'importance d'améliorer la coordination de la surveillance des maladies infectieuses au niveau mondial.

Cuantos trabajan en el ámbito de la vigilancia y la prevención de enfermedades infecciosas emergentes (EIE) tienen dificultades para predecir con precisión dónde va a surgir y a quién (o qué) afectará una infección. La instauración de programas de vigilancia y control de EIE exige una inversión sustancial y duradera de recursos que por definición son escasos, sobre todo teniendo en cuenta el número incalculable de enfermedades infecciosas zoonóticas y no zoonóticas que pueden aparecer, aun considerando solo aquellas que afectan al ganado. Este tipo de enfermedades pueden surgir como resultado de muchas combinaciones distintas de especie hospedadora, sistema productivo, medio/hábitat y tipo de patógeno o por efecto de cambios que se den en cualquiera de estos elementos. En vista de la multiplicidad de factores que concurren, convendría emplear de modo más generalizado un sistema de jerarquización de los riesgos en el cual fundamentar las decisiones de vigilancia y la distribución de los recursos destinados a ella. Los autores, valiéndose de ejemplos recientes de episodios infecciosos emergentes que afectaron al ganado, pasan revista a distintos métodos de vigilancia para la detección temprana de EIE y recalcan que los programas de vigilancia deben reposar en procedimientos de determinación del riesgo periódicamente actualizados y en las prioridades fijadas a partir de estos procedimientos. Por último, los autores se detienen en algunas necesidades desatendidas en la praxis de la determinación del riesgo de EIE y en la necesidad de una mejor coordinación de la vigilancia mundial de las enfermedades infecciosas.

Assessing the quality of data for drivers of disease emergence.

Drivers are factors that have the potential to directly or indirectly influence the likelihood of infectious diseases emerging or re-emerging. It is likely that an emerging infectious disease (EID) rarely occurs as the result of only one driver; rather, a network of sub-drivers (factors that can influence a driver) are likely to provide conditions that allow a pathogen to (re-)emerge and become established. Data on sub-drivers have therefore been used by modellers to identify hotspots where EIDs may next occur, or to estimate which sub-drivers have the greatest influence on the likelihood of their occurrence. To minimise error and bias when modelling how sub-drivers interact, and thus aid in predicting the likelihood of infectious disease emergence, researchers need good-quality data to describe these sub-drivers. This study assesses the quality of the available data on sub-drivers of West Nile virus against various criteria as a case study. The data were found to be of varying quality with regard to fulfilling the criteria. The characteristic with the lowest score was completeness, i.e. where sufficient data are available to fulfil all the requirements for the model. This is an important characteristic as an incomplete data set could lead to erroneous conclusions being drawn from modelling studies. Thus, the availability of good-quality data is essential to reduce uncertainty when estimating the likelihood of where EID outbreaks may occur and identifying the points on the risk pathway where preventive measures may be taken.

Les facteurs d'émergence sont des éléments ayant le potentiel direct ou indirect d'influencer la probabilité d'émergence ou de réémergence d'une maladie infectieuse. Il est probablement rare qu'une maladie infectieuse émergente apparaisse en raison d'un seul facteur ; c'est plutôt un faisceau de sous-facteurs (éléments pouvant avoir une influence sur un même facteur) qui contribue à ce que les conditions soient réunies pour qu'un agent pathogène puisse (ré)émerger et s'établir. Les concepteurs de modèles ont donc utilisé les données relatives aux sous-facteurs pour identifier les zones sensibles où les prochaines maladies infectieuses émergentes pourraient survenir, ou pour faire une estimation des sous-facteurs ayant la plus grande influence sur la probabilité de leur occurrence. Les chercheurs ont besoin de données de qualité pour décrire ces sous-facteurs, afin de minimiser le risque d'erreur et de biais lors de la modélisation de l'interaction entre les différents sous-facteurs, et de contribuer ainsi à mieux prédire la probabilité d'apparition d'une maladie infectieuse émergente. Les auteurs présentent une étude de cas qui a consisté à évaluer la qualité des données disponibles relatives aux sous-facteurs d'émergence du virus de la fièvre de West Nile au regard de différents critères. Il est apparu que la qualité des données était variable au regard des critères examinés. Le paramètre dont le score était le plus bas est celui de la complétude - le fait que suffisamment de données soient disponibles pour répondre à toutes les exigences du modèle. Il s'agit pourtant d'un paramètre important car des données incomplètes peuvent inciter à tirer des conclusions erronées des études de modélisation. La disponibilité de données de bonne qualité est essentielle pour réduire l'incertitude lors de l'estimation de la probabilité d'apparition de maladies infectieuses émergentes dans des zones déterminées, ainsi que pour identifier les points critiques de concrétisation du risque où des mesures préventives pourraient être mises en place.

Los inductores o factores de inducción [drivers] son aquellos que, directa o indirectamente, pueden influir en la probabilidad de que surjan o resurjan enfermedades infecciosas. Todo indica que rara vez una enfermedad infecciosa emergente aparece por efecto de un solo factor de inducción, sino que es probable que haya más bien una combinación de "subfactores de influencia" [sub-drivers] (factores que pueden influir en un inductor) que cree condiciones propicias para que un patógeno (re)surja y logre asentarse. Los creadores de modelos, por consiguiente, se han servido de datos sobre estos subfactores de influencia para localizar aquellas zonas donde con mayor probabilidad puedan aparecer próximamente enfermedades infecciosas emergentes o para determinar cuáles son los subfactores que más influyen en la probabilidad de que ello ocurra. Para reducir al mínimo los errores y sesgos al modelizar la interacción entre los subfactores y ayudar así a calcular la probabilidad de que surja una enfermedad infecciosa emergente, los investigadores necesitan datos de buena calidad para caracterizar estos subfactores. En el análisis expuesto por los autores se utilizó el virus del Nilo Occidental como ejemplo de estudio para evaluar, con arreglo a diversos criterios, la calidad de los datos existentes sobre los subfactores que inciden en la aparición de este virus. Lo que se constató, en relación con el grado de cumplimiento de los criterios, es que esos datos eran de calidad variable. La característica o parámetro que deparó la puntuación más baja fue la completud, es decir, la existencia de datos suficientes para aportar al modelo toda la información requerida para que este funcione bien. Se trata de una característica importante, pues un conjunto incompleto de datos podría llevar a extraer conclusiones erróneas de los estudios de modelización. Por ello, para reducir la incertidumbre a la hora de calcular la probabilidad de que en cierto lugar surjan brotes de enfermedades infecciosas emergentes y de determinar, dentro de la cadena de materialización del riesgo, aquellos eslabones en los que cabe adoptar medidas preventivas, es indispensable disponer de datos de buena calidad.

Assessing the risks of SARS-CoV-2 in wildlife.

The novel coronavirus SARS-CoV-2 likely emerged from a wildlife source with transmission to humans followed by rapid geographic spread throughout the globe and severe impacts on both human health and the global economy. Since the onset of the pandemic, there have been many instances of human-to-animal transmission involving companion, farmed and zoo animals, and limited evidence for spread into free-living wildlife. The establishment of reservoirs of infection in wild animals would create significant challenges to infection control in humans and could pose a threat to the welfare and conservation status of wildlife. We discuss the potential for exposure, onward transmission and persistence of SARS-CoV-2 in an initial selection of wild mammals (bats, canids, felids, mustelids, great apes, rodents and cervids). Dynamic risk assessment and targeted surveillance are important tools for the early detection of infection in wildlife, and here we describe a framework for collating and synthesising emerging information to inform targeted surveillance for SARS-CoV-2 in wildlife. Surveillance efforts should be integrated with information from public and veterinary health initiatives to provide insights into the potential role of wild mammals in the epidemiology of SARS-CoV-2.

A cost-benefit assessment of Salmonella-control strategies in pigs reared in the United Kingdom.

Pork and pork products are a major source of human salmonellosis in the United Kingdom (UK). Despite a number of surveillance programmes, the prevalence of Salmonella in the UK slaughter pig population remains over 20%. Here, we present the results of a Cost-Benefit Analysis comparing five on-farm control strategies (where the cost is the cost of implementation and the benefits are the financial savings for both the human health and pig industries). The interventions considered were: wet feed, organic acids in feed, vaccination, enhanced cleaning and disinfection and movement of outdoor breeding units. The data originate from published papers and recent UK studies. The effectiveness was assessed by adapting a previous risk assessment, originally developed for the European Food Safety Authority. Using this method, none of the intervention strategies produced a net cost-benefit. Our results suggest that the cost of implementation outweighed the savings for all interventions, even if the effectiveness could be improved. Therefore, to achieve a net cost-benefit it is essential to reduce the cost of interventions. Analyses concluded that large cost reductions (up to 96%) would be required. Use of organic acids required the smallest reduction in cost (22.7%) to achieve a net cost benefit. Uncertainty analysis suggested that a small net gain might be possible, for some of the intervention measures. But this would imply that the model greatly underestimated some key parameters, which was considered unlikely. Areas of key uncertainty were identified as the under-reporting factor (i.e. the proportion of community cases of Salmonella) and the source attribution factor (i.e. the proportion of human Salmonella cases attributable to pork products).

Quantitative risk assessment of the introduction of rabies into Japan through the importation of dogs and cats worldwide.

Japan has been free from rabies since 1958. A strict import regimen has been adopted since 2004 consisting of identification of an animal with microchip, two-time rabies vaccination, neutralizing antibody titration test and a waiting period of 180 days. The present study aims to quantitatively assess the risk of rabies introduction into Japan through the international importation of dogs and cats and hence provide evidence-based recommendations to strengthen the current rabies prevention system. A stochastic scenario tree model was developed and simulations were run using @RISK. The probability of infection in a single dog or cat imported into Japan is estimated to be 2·16 × 10-9 [90% prediction interval (PI) 6·65 × 10-11-6·48 × 10-9]. The number of years until the introduction of a rabies case is estimated to be 49 444 (90% PI 19 170-94 641) years. The current import regimen is effective in maintaining the very low risk of rabies introduction into Japan and responding to future changes including increases in import level and rabies prevalence in the world. However, non-compliance or smuggling activities could substantially increase the risk of rabies introduction. Therefore, policy amendment which could promote compliance is highly recommended. Scenario analysis demonstrated that the waiting period could be reduced to 90 days and the requirement for vaccination could be reduced to a single vaccination, but serological testing should not be stopped.

The challenge of using experimental infectivity data in risk assessment for Ebola virus: why ecology may be important.

Analysis of published data shows that experimental passaging of Zaire ebolavirus (EBOV) in guinea pigs changes the risk of infection per plaque-forming unit (PFU), increasing infectivity to some species while decreasing infectivity to others. Thus, a PFU of monkey-adapted EBOV is 10(7) -fold more lethal to mice than a PFU adapted to guinea pigs. The first conclusion is that the infectivity of EBOV to humans may depend on the identity of the donor species itself and, on the basis of limited epidemiological data, the question is raised as to whether bat-adapted EBOV is less infectious to humans than nonhuman primate (NHP)-adapted EBOV. Wildlife species such as bats, duikers and NHPs are naturally infected by EBOV through different species giving rise to EBOV with different wildlife species-passage histories (heritages). Based on the ecology of these wildlife species, three broad 'types' of EBOV-infected bushmeat are postulated reflecting differences in the number of passages within a given species, and hence the degree of adaptation of the EBOV present. The second conclusion is that the prior species-transmission chain may affect the infectivity to humans per PFU for EBOV from individuals of the same species. This is supported by the finding that the related Marburg marburgvirus requires ten passages in mice to fully adapt. It is even possible that the evolutionary trajectory of EBOV could vary in individuals of the same species giving rise to variants which are more or less virulent to humans and that the probability of a given trajectory is related to the heritage. Overall the ecology of the donor species (e.g. dog or bushmeat species) at the level of the individual animal itself may determine the risk of infection per PFU to humans reflecting the heritage of the virus and may contribute to the sporadic nature of EBOV outbreaks.

A Transport and Lairage Model for Salmonella Transmission Between Pigs Applicable to EU Member States.

A model for the transmission of Salmonella between finisher pigs during transport to the abattoir and subsequent lairage has been developed, including novel factors such as environmental contamination and the effect of stress, and is designed to be adaptable for any EU Member State (MS). The model forms part of a generic farm-to-consumption model for Salmonella in pigs, designed to model potentially important risk factors and assess the effectiveness of interventions. In this article, we discuss the parameterization of the model for two case study MSs. For both MSs, the model predicted an increase in the average MS-level prevalence of Salmonella-positive pigs during both transport and lairage, accounting for a large amount of the variation between reported on-farm prevalence and reported lymph-node prevalence at the slaughterhouse. Sensitivity analysis suggested that stress is the most important factor during transport, while a number of factors, including environmental contamination and the dose-response parameters, are important during lairage. There was wide variation in the model-predicted change in prevalence in individual batches; while the majority of batches (80-90%) had no increase, in some batches the increase in prevalence was over 70% and in some cases infection was introduced into previously uninfected batches of pigs. Thus, the model suggests that while the transport and lairage stages of the farm-to-consumption exposure pathway are unlikely to be responsible for a large increase in average prevalence at the MS level, they can have a large effect on prevalence at an individual-batch level.

Q fever through consumption of unpasteurised milk and milk products - a risk profile and exposure assessment.

Q fever is a zoonotic disease caused by the bacterium Coxiella burnetii which is endemic in cattle, sheep and goats in much of the world, including the United Kingdom (UK). There is some epidemiological evidence that a small proportion of cases in the developed world may arise from consumption of unpasteurised milk with less evidence for milk products such as cheese. Long maturation at low pH may give some inactivation in hard cheese, and viable C. burnetii are rarely detected in unpasteurised cheese compared to unpasteurised milk. Simulations presented here predict that the probability of exposure per person to one or more C. burnetii through the daily cumulative consumption of raw milk in the UK is 0·4203. For those positive exposures, the average level of exposure predicted is high at 1266 guinea pig intraperitoneal infectious dose 50% units (GP_IP_ID50 ) per person per day. However, in the absence of human dose-response data, the case is made that the GP_IP_ID50 unit represents a very low risk through the oral route. The available evidence suggests that the risks from C. burnetii through consumption of unpasteurised milk and milk products (including cheese) are not negligible but they are lower in comparison to transmission via inhalation of aerosols from parturient products and livestock contact.

Pathways for entry of livestock arboviruses into Great Britain: assessing the strength of evidence.

The emergence of bluetongue virus and Schmallenberg virus in Great Britain (GB) during the last decade has highlighted the need for understanding the relative importance of the various pathways of the entry of livestock arboviruses so as to help focus surveillance and mitigation. This study summarizes what is known for the main routes of entry and assesses the strength of the current evidence for and against. Entry through infected arthropod vectors is considered at the level of each life cycle stage for tick-, biting midge- and mosquito-borne viruses, and while there is evidence that this could happen through most tick and mosquito stages, strong evidence that only exists for entry through adult midges. There is also strong evidence that entry through immature midge stages could not happen. The weight of supporting evidence is strongest for importation of viraemic livestock including horses. While there is some indication of a common pathway for midge-borne viruses from sub-Saharan Africa to GB via Continental Europe, other factors such as maternal transmission in dogs and sheep need to be considered in the light of recent findings.

Entry of H5N1 highly pathogenic avian influenza virus into Europe through migratory wild birds: a qualitative release assessment at the species level.

AIMS: To estimate qualitatively the probabilities of release (or entry) of Eurasian lineage H5N1 highly pathogenic avian influenza (HPAI) virus into Great Britain (GB), the Netherlands and Italy through selected higher risk species of migratory water bird. METHODS AND RESULTS: The probabilities of one or more release events of H5N1 HPAI per year (Pre(lease)) were estimated qualitatively for 15 avian species, including swans, geese, ducks and gulls, by assessing the prevalence of H5N1 HPAI in different regions of the world (weighted to 2009) and estimates of the total numbers of birds migrating from each of those regions. The release assessment accommodated the migration times for each species in relation to the probabilities of their surviving infection and shedding virus on arrival. Although the predicted probabilities of release of H5N1 per individual bird per year were low, very low or negligible, Pre(lease) was high for a few species reflecting the high numbers of birds migrating from some regions. Values of Pre(lease) were generally higher for the Netherlands than for GB, while ducks and gulls from Africa presented higher probabilities to Italy compared to the Netherlands and GB. CONCLUSIONS: Bird species with high values of Pre(lease) in GB, the Netherlands and Italy generally originate from within Europe based on data for global prevalence of H5N1 between 2003 and 2009 weighted to 2009. Potential long-distance transfer of H5N1 HPAI from North Asia and Eurasia to GB, the Netherlands and Italy is limited to a few species and does not occur from South-East Asia, an area where H5N1 is endemic. SIGNIFICANCE AND IMPACT OF THE STUDY: The approach accommodates biogeographical conditions and variability in the estimated worldwide prevalence of the virus. The outputs of this release assessment can be used to inform surveillance activities through focusing on certain species and migratory pathways.

Estimating the time at which commercial broiler flocks in Great Britain become infected with Campylobacter: a Bayesian approach.

Campylobacter is a common cause of intestinal disease in humans and is often linked to the consumption of contaminated poultry meat. Despite considerable research on the topic there is a large amount of uncertainty associated with Campylobacter epidemiology. A Bayesian model framework was applied to multiple longitudinal datasets on Campylobacter infection in UK broiler flocks to estimate the time at which each flock was first infected with Campylobacter. The model results suggest that the day of first infection ranges from 10 to 45 days; however, over half had a time of infection between 30 and 35 days. When considering only those flocks which were thinned, 48% had an estimated day of infection within 2 days of the day of thinning, thus suggesting an association between thinning and Campylobacter infection. These results demonstrate how knowledge of the time of infection can be correlated to known events to identify potential risk factors for infection.

Epidemic threats to the European Union: expert views on six virus groups.

In recent years, several animal disease epidemics have occurred within the European Union (EU). At the 4th Annual Meeting of the EPIZONE network (7-10 June 2010, St. Malo, France), an interactive session was run to elicit the opinions of delegates on a pre-defined list of epidemic threats to the EU. Responses from over 190 delegates, to questions relating to impact and likelihood, were used to rank six virus groups with respect to their perceived threat now (2010) and in 2020. The combined opinions of all delegates suggested that, from the pre-selected list of virus groups, foot-and-mouth disease and influenza are currently of most concern. Delegates thought that influenza would be less of a threat and zoonotic arboviruses would be more of a threat in 2020. Although the virus group rankings should not be taken as definitive, the results could be used in conjunction with experimental and field data, by scientists, policy-makers and stakeholders when assessing and managing risks associated with these virus groups.

A quantitative release assessment for the noncommercial movement of companion animals: risk of rabies reintroduction to the United kingdom.

In 2004, the European Union (EU) implemented a pet movement policy (referred to here as the EUPMP) under EU regulation 998/2003. The United Kingdom (UK) was granted a temporary derogation from the policy until December 2011 and instead has in place its own Pet Movement Policy (Pet Travel Scheme (PETS)). A quantitative risk assessment (QRA) was developed to estimate the risk of rabies introduction to the UK under both schemes to quantify any change in the risk of rabies introduction should the UK harmonize with the EU policy. Assuming 100 % compliance with the regulations, moving to the EUPMP was predicted to increase the annual risk of rabies introduction to the UK by approximately 60-fold, from 7.79 × 10(-5) (5.90 × 10(-5), 1.06 × 10(-4)) under the current scheme to 4.79 × 10(-3) (4.05 × 10(-3), 5.65 × 10(-3)) under the EUPMP. This corresponds to a decrease from 13,272 (9,408, 16,940) to 211 (177, 247) years between rabies introductions. The risks associated with both the schemes were predicted to increase when less than 100 % compliance was assumed, with the current scheme of PETS and quarantine being shown to be particularly sensitive to noncompliance. The results of this risk assessment, along with other evidence, formed a scientific evidence base to inform policy decision with respect to companion animal movement.

Fecal carriage and shedding density of CTX-M extended-spectrum {beta}-lactamase-producing escherichia coli in cattle, chickens, and pigs: implications for environmental contamination and food production.

The number and proportion of CTX-M positive Escherichia coli organisms were determined in feces from cattle, chickens, and pigs in the United Kingdom to provide a better understanding of the risk of the dissemination of extended-spectrum ß-lactamase (ESBL) bacteria to humans from food animal sources. Samples of bovine (n = 35) and swine (n = 20) feces were collected from farms, and chicken cecal contents (n = 32) were collected from abattoirs. There was wide variation in the number of CTX-M-positive E. coli organisms detected; the median (range) CFU/g were 100 (100 × 10(6) to 1 × 10(6)), 5,350 (100 × 10(6) to 3.1 × 10(6)), and 2,800 (100 × 10(5) to 4.7 × 10(5)) for cattle, chickens, and pigs, respectively. The percentages of E. coli isolates that were CTX-M positive also varied widely; median (range) values were 0.013% (0.001 to 1%) for cattle, 0.0197% (0.00001 to 28.18%) for chickens, and 0.121% (0.0002 to 5.88%) for pigs. The proportion of animals designated high-density shedders (≥1 × 10(4) CFU/g) of CTX-M E. coli was 3/35, 15/32, and 8/20 for cattle, chickens, and pigs, respectively. We postulate that high levels of CTX-M E. coli in feces facilitate the dissemination of bla(CTX-M) genes during the rearing of animals for food, and that the absolute numbers of CTX-M bacteria should be given greater consideration in epidemiological studies when assessing the risks of food-borne transmission.

Zoonoses action plan Salmonella monitoring programme: an investigation of the sampling protocol.

The Zoonoses Action Plan (ZAP) Salmonella Programme was established by the British Pig Executive to monitor Salmonella prevalence in quality-assured British pigs at slaughter by testing a sample of pigs with a meat juice enzyme-linked immunosorbent assay for antibodies against group B and C(1) Salmonella. Farms were assigned a ZAP level (1 to 3) depending on the monitored prevalence, and ZAP 2 or 3 farms were required to act to reduce the prevalence. The ultimate goal was to reduce the risk of human salmonellosis attributable to British pork. A mathematical model has been developed to describe the ZAP sampling protocol. Results show that the probability of assigning a farm the correct ZAP level was high, except for farms that had a seroprevalence close to the cutoff points between different ZAP levels. Sensitivity analyses identified that the probability of assigning a farm to the correct ZAP level was dependent on the sensitivity and specificity of the test, the number of batches taken to slaughter each quarter, and the number of samples taken per batch. The variability of the predicted seroprevalence was reduced as the number of batches or samples increased and, away from the cutoff points, the probability of being assigned the correct ZAP level increased as the number of batches or samples increased. In summary, the model described here provided invaluable insight into the ZAP sampling protocol. Further work is required to understand the impact of the program for Salmonella infection in British pig farms and therefore on human health.

Assessing the impact of climate change on vector-borne viruses in the EU through the elicitation of expert opinion.

Expert opinion was elicited to undertake a qualitative risk assessment to estimate the current and future risks to the European Union (EU) from five vector-borne viruses listed by the World Organization for Animal Health. It was predicted that climate change will increase the risk of incursions of African horse sickness virus (AHSV), Crimean-Congo haemorrhagic fever virus (CCHFV) and Rift Valley fever virus (RVFV) into the EU from other parts of the world, with African swine fever virus (ASFV) and West Nile virus (WNV) being less affected. Currently the predicted risks of incursion were lowest for RVFV and highest for ASFV. Risks of incursion were considered for six routes of entry (namely vectors, livestock, meat products, wildlife, pets and people). Climate change was predicted to increase the risk of incursion from entry of vectors for all five viruses to some degree, the strongest effects being predicted for AHSV, CCHFV and WNV. This work will facilitate identification of appropriate risk management options in relation to adaptations to climate change.

Factors influencing the presence and concentration of E. coli O157 and E. coli in farm waste on six cattle farms in North-West England.

AIMS: To investigate the factors influencing the presence and burden of Escherichia coli O157 in farm wastes. METHODS AND RESULTS: Wastes from six cattle farms were screened for the presence and concentration of E. coli O157 and E. coli on three occasions over a year and waste management data were collected. Sixty-three of 878 (7.1%) samples were positive for verocytotoxigenic Escherichia coli O157 and 664/875 (75.9%) for E. coli with detectable levels greater in fresh waste than in stored waste, pasture or dirty water. CONCLUSIONS: The turning/stirring of stored waste and the use of more than one store (allowing longer storage times) reduced the proportion of E. coli O157 positive samples. The presence of E. coli O157 significantly reduced from a high prevalence found in fresh faeces and stored waste to lower proportions in dirty water and pasture samples. Escherichia coli O157 was only detected on pasture when waste was spread from contaminated stores the day before sampling. A high prevalence of positive E. coli O157 samples were detected when cattle were re-housed. SIGNIFICANCE AND IMPACT OF THE STUDY: These findings help to support the importance of treating and storing farm waste, as well as providing evidence for the level of dilution of E. coli O157 from fresh waste to recently spread pastures.

Dynamics of Salmonella transmission on a British pig grower-finisher farm: a stochastic model.

Previous modelling studies have estimated that between 1% and 10% of human salmonella infections are attributable to pig meat consumption. In response to this food safety threat the British pig industry have initiated a salmonella monitoring programme. It is anticipated that this programme will contribute to achieving a UK Food Standards Agency target for reducing salmonella levels in pigs at slaughter by 50% within 5 years. In order to better inform the monitoring programme, we have developed a stochastic transmission model for salmonella in a specialist grower-finisher pig herd, where data from a Danish longitudinal study have been used to estimate some of the key model parameters. The model estimates that about 17% of slaughter-age pigs will be infected with salmonella, and that of these infected pigs about 4% will be excreting the organism. In addition, the model shows that the most effective control strategies will be those that reduce between-pen transmission.