Use of Bayesian Belief Network techniques to explore the interaction of biosecurity practices on the probability of porcine disease occurrence in Canada

Identification and quantification of pathogen threats need to be a priority for the Canadian swine industry so that resources can be focused where they will be most effective. Here we create a tool based on a Bayesian Belief Network (BBN) to model the interaction between biosecurity practices and the probability of occurrence of four different diseases on Canadian swine farms. The benefits of using this novel approach, in comparison to other methods, is that it enables us to explore both the complex interaction and the relative importance of biosecurity practices on the probability of disease occurrence.

In order to build the BBN we used two datasets. The first dataset detailed biosecurity practices employed on 218 commercial swine farms across Canada in 2010. The second dataset detailed animal health status and disease occurrence on 90 of those farms between 2010 and 2012. We used expert judgement to identify 15 biosecurity practices that were considered the most important in mitigating disease occurrence on farms. These included: proximity to other livestock holdings, the health status of purchased stock, manure disposal methods, as well as the procedures for admitting vehicles and staff. Four diseases were included in the BBN: Porcine reproductive and respiratory syndrome (PRRS), (a prevalent endemic aerosol pathogen), Swine influenza (SI) (a viral respiratory aerosol pathogen), Mycoplasma pneumonia (MP) (an endemic respiratory disease spread by close contact and aerosol) and Swine dysentery (SD) (an enteric disease which is re-emerging in North America).

This model indicated that the probability of disease occurrence was influenced by a number of manageable biosecurity practices. Increased probability of PRRS and of MP were associated with spilt feed (feed that did not fall directly in a feeding trough), not being disposed of immediately and with manure being brought onto the farm premises and spread on land adjacent to the pigs. Increased probabilities of SI and SD were associated with the farm allowing access to visiting vehicles without cleaning or disinfection. SD was also more likely to occur when the health status of purchased stock was not known. Finally, we discuss how such a model can be used by the Canadian swine industry to quantify disease risks and to determine practices that may reduce the probability of disease occurrence.

Development of a network based model to simulate the between-farm transmission of the porcine reproductive and respiratory syndrome virus

Contact structure within a population can significantly affect the outcomes of infectious disease spread models. The objective of this study was to develop a network based simulation model for the between-farm spread of porcine reproductive and respiratory syndrome virus to assess the impact of contact structure on between-farm transmission of PRRS virus. For these farm level models, a hypothetical population of 500 swine farms following a multistage production system was used. The contact rates between farms were based on a study analyzing movement of pigs in Canada, while disease spread parameters were extracted from published literature. Eighteen distinct scenarios were designed and simulated by varying the mode of transmission (direct versus direct and indirect contact), type of index herd (farrowing, nursery and finishing), and the presumed network structures among swine farms (random, scale-free and small-world). PRRS virus was seeded in a randomly selected farm and 500 iterations of each scenario were simulated for 52 weeks. The median epidemic size by the end of the simulated period and percentage die-out for each scenario, were the key outcomes captured. Scenarios with scale-free network models resulted in the largest epidemic sizes, while scenarios with random and small-world network models resulted in smaller and similar epidemic sizes. Similarly, stochastic die-out percentage was least for scenarios with scale-free networks followed by random and small-world networks. Findings of the study indicated that incorporating network structures among the swine farms had a considerable impact on the spread of PRRS virus, highlighting the importance of understanding and incorporating realistic contact structures when developing infectious disease spread models for similar populations.

Simulation of between-farm transmission of porcine reproductive and respiratory syndrome virus in Ontario, Canada using the North American Animal Disease Spread Model

Porcine reproductive and respiratory syndrome (PRRS), a viral disease of swine, has major economic impacts on the swine industry. The North American Animal Disease Spread Model (NAADSM) is a spatial, stochastic, farm level state-transition modeling framework originally developed to simulate highly contagious and foreign livestock diseases. The objectives of this study were to develop a model to simulate between-farm spread of a homologous strain of PRRS virus in Ontario swine farms via direct (animal movement) and indirect (sharing of trucks between farms) contacts using the NAADSM and to compare the patterns and extent of outbreak under different simulated conditions. A total of 2552 swine farms in Ontario province were allocated to each census division of Ontario and geo-locations of the farms were randomly generated within the agriculture land of each Census Division. Contact rates among different production types were obtained using pig movement information from four regions in Canada. A total of 24 scenarios were developed involving various direct (movement of infected animals) and indirect (pig transportation trucks) contact parameters in combination with alternating the production type of the farm in which the infection was seeded. Outbreaks were simulated for one year with 1000 replications. The median number of farms infected, proportion of farms with multiple outbreaks and time to reach the peak epidemic were used to compare the size, progression and extent of outbreaks. Scenarios involving spread only by direct contact between farms resulted in outbreaks where the median percentage of infected farms ranged from 31.5 to 37% of all farms. In scenarios with both direct and indirect contact, the median percentage of infected farms increased to a range from 41.6 to 48.6%. Furthermore, scenarios with both direct and indirect contact resulted in a 44% increase in median epidemic size when compared to the direct contact scenarios. Incorporation of both animal movements and the sharing of trucks within the model indicated that the effect of direct and indirect contact may be nonlinear on outbreak progression. The increase of 44% in epidemic size when indirect contact, via sharing of trucks, was incorporated into the model highlights the importance of proper biosecurity measures in preventing transmission of the PRRS virus. Simulation of between-farm spread of the PRRS virus in swine farms has highlighted the relative importance of direct and indirect contact and provides important insights regarding the possible patterns and extent of spread of the PRRS virus in a completely susceptible population with herd demographics similar to those found in Ontario, Canada.

Analysis of Swine Movement in Four Canadian Regions: Network Structure and Implications for Disease Spread

Direct and indirect contacts among animal holdings are important in the spread of infectious diseases. The objectives of this study were to describe networks of pig movements and the sharing of trucks used for those movements between swine farms in four Canadian regions using network analysis tools and to obtain contact parameters for infectious disease spread simulation models. Four months of swine movement data from a pilot pig traceability programme were used. Two types of networks were created using three time scales (weekly, monthly and the full study period): one-mode networks of farm-to-farm direct contact representing animal shipments and two-mode networks representing the sharing of trucks between farms. Contact patterns among farms were described by estimating a range of relevant network measures. The overall network neglecting the four regions consisted of 145 farms, which were connected by 261 distinct links. A total of 184 trucks were used to transport 2043 shipments of pigs during the study period. The median in- and out-degree for the overall one-mode network was 1 and ranged from 0 to 26 and 0 to 10, respectively. The overall one-mode network had heterogeneous degree distribution, a high clustering coefficient and shorter average path length than would be expected for randomly generated networks of similar size. On average one truck was shared by four farms in the overall network, or by three farms when considered the monthly and weekly networks. Degree distribution of the two-mode overall network demonstrated characteristics of power-law distribution. For more than 50% of shipments on any given day, the same truck was used for at least one other shipment. Findings from this study are in agreement with previous work, which suggested that swine movement networks exhibit small-world and scale-free topologies. Furthermore, trucks used for the shipment of pigs can play an important role in connecting otherwise unconnected farms and may increase the spread of disease.

Description of empirical movement data from Canadian swine herds with an application to a disease spread simulation model

We report the methods and findings of a survey of Canadian swine producers summarizing farm-types at-risk of foreign animal disease (FAD) and the routine movement of animals, semen and workers among swine farms, as observed during a 42-day period. Of the 311 producers who returned completed questionnaires, 17% represented swine-herds with no swine or semen movement on or off the farm during the 42 days, 57% were sow herds or farrow-to-finish herds with limited movement onto the farm but movement off the farm, and 26% were swine-herds with movements on and off the farm. A substantial number of premises (>50% in some provinces) with swine also kept other animal species on the same premises. We applied the empirical movement data from the survey in a stochastic simulation model to estimate the number of herds infected and the basic regional distribution of infection that could be expected to occur if the FAD was not detected and routine movements were permitted to occur up to 42 days after infection with a FAD of a single randomly selected herd. Forty-five percent of the simulations did not involve spread beyond the index farm, whereas 34.8% involved spread among five or more farms after 42 days of routine movement.

The role of immunostimulation in the development of postweaning multisystemic wasting syndrome in pigs under field conditions

We evaluated the effects of immunostimulation in the development of postweaning multisystemic wasting syndrome (PMWS) in 930 pigs 53 to 54 d old in a grower/finisher barn with a history of PMWS. The pigs were allocated to 5 treatment groups: 4 groups received a single intramuscular injection of RespiSure-ONE (a commercial Mycoplasma hyopneumoniae vaccine; n = 197), Emulsigen (an oil-based adjuvant; n = 172), Alhydrogel (an aluminum-hydroxide-based adjuvant; n = 172), or physiologic saline (n = 218); 1 group received no treatment (n = 171). Pigs affected by PMWS were found in all the groups. Antigen to Porcine circovirus type 2 (PCV-2) was detected by immunohistochemical testing within lesions of mesenteric lymph node, spleen, Peyer's patch, and lung of affected pigs. There was no significant difference in the incidence of PMWS among the groups. The findings indicate that immunostimulation did not influence the expression of PMWS in this study. Thus, routine vaccination against swine diseases may not significantly contribute to the occurrence of PMWS under field conditions.

Growth performance, carcass quality, meat quality and fatty acid composition of pigs fed diets

Two hundred gilts and 200 barrows, housed within sex in pens of 25, were randomly allotted to two replications of four dietary treatments to determine the effects of incorporating 30, 20, 10 or 0% extruded soybeans (ESB), displacing a commercial protein supplement, in barley-based grower and finisher diets for pigs. Growth, feed intake and carcass quality of the pigs, and meat quality and fatty acid composition of the pork from a random subset of the pigs on test were determined. No sex × diet interactions were observed. ESB inclusion rate had no effect on growth rate; however, per-pen feed consumption decreased numerically with increasing ESB resulting in an improvement in feed efficiency. The 30% ESB inclusion rate increased carcass fat content (P < 0.05) compared with the control, whereas lean content was unaffected. Meat colour and marbling score were similar across all treatments whereas fat and lean firmness was reduced by the 30% ESB inclusion rate (P < 0.05) compared with all other treatments. Increasing ESB in the diet altered the fatty acid content of the pork by decreasing the amount of short-chain saturated and monounsaturated fatty acids and increasing the amount of long-chain polyunsaturated fatty acids (PUFA). The results of this study indicate that ESB can be used as the sole source of supplemental protein in barley-based diets for pigs with no detrimental effects on performance and minimal negative effects on carcass and meat quality. Alteration of fatty acid content of pork from feeding ESB has both positive and negative implications for consumer acceptance by increasing PUFA content while concomitantly increasing the risk of premature oxidation. Key words: Extruded soybeans, pigs, pork, growth, fatty acids, meat quality

An overview of techniques for dealing with large numbers of independent variables in epidemiologic studies

Many studies of health and production problems in livestock involve the simultaneous evaluation of large numbers of risk factors. These analyses may be complicated by a number of problems including: multicollinearity (which arises because many of the risk factors may be related (correlated) to each other), confounding, interaction, problems related to sample size (and hence the power of the study), and the fact that many associations are evaluated from a single dataset. This paper focuses primarily on the problem of multicollinearity and discusses a number of techniques for dealing with this problem. However, some of the techniques discussed may also help to deal with the other problems identified above. The first general approach to dealing with multicollinearity involves reducing the number of independent variables prior to investigating associations with the disease. Techniques to accomplish this include: (1) excluding variables after screening for associations among independent variables; (2) creating indices or scores which combine data from multiple factors into a single variable; (3) creating a smaller set of independent variables through the use of multivariable techniques such as principal components analysis or factor analysis. The second general approach is to use appropriate steps and statistical techniques to investigate associations between the independent variables and the dependent variable. A preliminary screening of these associations may be performed using simple statistical tests. Subsequently, multivariable techniques such as linear or logistic regression or correspondence analysis can be used to identify important associations. The strengths and limitations of these techniques are discussed and the techniques are demonstrated using a dataset from a recent study of risk factors for pneumonia in swine. Emphasis is placed on comparing correspondence analysis with other techniques as it has been used less in the epidemiology literature.

Interstitial pneumonia in piglets on Prince Edward Island

Images

An investigation of enzootic Glasser's disease in a specific-pathogen-free grower-finisher facility using restriction endonuclease analysis

Enzootic Glassers's disease was investigated to study the epidemiology of the disease strains on a farm where it presented a problem. Restriction endonuclease fingerprinting (REF) analysis technique was used, as all strains of Haemophilus parasuis are biochemically similar and many strains are biochemically untypable. After young weaned pigs were moved from farm A to farm B, Glasser's disease routinely occurred despite the use of antibiotics and a commercial bacterin. Isolates were taken from the nasal passages and from carcasses of clinically affected cases and subjected to REF analysis. Haemophilus parasuis was not isolated from any of the pigs on farm A, but it was isolated from 7/10 and 5/10 nasal swabs taken from farm B. Two H. parasuis strains isolated from clinical cases of Glasser's disease from farm B had an identical REF pattern, but were different from the nasal swabs and the H. parasuis strain contained in the bacterin. The subsequent use of a custom autogenous bacterin made from a clinical isolate of H. parasuis reduced the mortality rate on farm B. This investigation indicates that nasal isolates of H. parasuis are different than those causing clinical disease, and not all bacterin strains are cross protective for other strains.

Evaluation of rapid gross visual appraisal of swine lungs at slaughter as a diagnostic screen for enzootic pneumonia

A rapid gross visual appraisal of enzootic pneumonia lesions was made of 87 lungs at a local abattoir and the lungs were then set aside and examined in more detail. A sample of pulmonary tissue was taken from each lung and submitted for bacterial and histological examination. The principal investigator who had performed the gross and detailed lung scores was then used to assess the agreement of two inspectors who were scoring lung lesions in the abattoir. The Pearson's correlation coefficient between grossly scored lungs and scores derived from a detailed examination was 0.94. Using histological examination as the gold standard, rapid gross examination had a sensitivity of 76% and a specificity of 71%. Using bacterial recovery as a gold standard yielded a sensitivity of 77% and a specificity of 51%. The sensitivity and specificity of inspectors 1 and 2 compared to the principal investigator were: sensitivity = 97.5% and specificity = 97.4% for inspector 1, and 97% and 98% for inspector 2. The kappa values for both of the inspectors compared to the principal investigator were 0.95 suggesting that designated lay inspectors consistently agreed with the principal investigator.

Benefits of using a photo-electric alert system for swine farrowing operations

A system has been designed to alert swine producers when the farrowing process has begun. The results indicate a potential for the alert system to reduce stillbirths. No differences in body weight of piglets, nor differences in bacterial growth in the sows as a result of using the alert system tampon, were observed. Key words: Farrowing, alert system, piglet, mortality, stillborn