Using Combined Diagnostic Test Results to Hindcast Trends of Infection from Cross-Sectional Data.

Infectious disease surveillance is key to limiting the consequences from infectious pathogens and maintaining animal and public health. Following the detection of a disease outbreak, a response in proportion to the severity of the outbreak is required. It is thus critical to obtain accurate information concerning the origin of the outbreak and its forward trajectory. However, there is often a lack of situational awareness that may lead to over- or under-reaction. There is a widening range of tests available for detecting pathogens, with typically different temporal characteristics, e.g. in terms of when peak test response occurs relative to time of exposure. We have developed a statistical framework that combines response level data from multiple diagnostic tests and is able to 'hindcast' (infer the historical trend of) an infectious disease epidemic. Assuming diagnostic test data from a cross-sectional sample of individuals infected with a pathogen during an outbreak, we use a Bayesian Markov Chain Monte Carlo (MCMC) approach to estimate time of exposure, and the overall epidemic trend in the population prior to the time of sampling. We evaluate the performance of this statistical framework on simulated data from epidemic trend curves and show that we can recover the parameter values of those trends. We also apply the framework to epidemic trend curves taken from two historical outbreaks: a bluetongue outbreak in cattle, and a whooping cough outbreak in humans. Together, these results show that hindcasting can estimate the time since infection for individuals and provide accurate estimates of epidemic trends, and can be used to distinguish whether an outbreak is increasing or past its peak. We conclude that if temporal characteristics of diagnostics are known, it is possible to recover epidemic trends of both human and animal pathogens from cross-sectional data collected at a single point in time.

Two major ruminant acute phase proteins, haptoglobin and serum amyloid A, as serum biomarkers during active sheep scab infestation.

Two ruminant acute phase proteins (APPs), haptoglobin (Hp) and serum amyloid A (SAA), were evaluated as serum biomarkers (BMs) for sheep scab-a highly contagious ectoparasitic disease caused by the mite Psoroptes ovis, which is a major welfare and production threat worldwide. The levels of both APPs increased in serum following experimental infestation of sheep with P. ovis, becoming statistically significantly elevated from pre-infestation levels at 4 weeks post-infestation. Following successful treatment of infested sheep with an endectocide, Hp and SAA serum levels declined rapidly, with half lives of less than 3 days. In contrast, serum IgG levels which specifically bound the P. ovis-derived diagnostic antigen Pso o 2 had a half-life of 56 days. Taking into account pre-infestation serum levels, rapidity of response to infestation and test sensitivity at the estimated optimum cut-off values, SAA was the more discriminatory marker. These studies illustrated the potential of SAA and Hp to indicate current sheep scab infestation status and to augment the existing Pso o 2 serological assay to give disease-specific indications of both infestation and successful treatment.

Social isolation of unfamiliar cattle by groups of familiar cattle.

Domestic herbivores show a strong motivation to form associations with conspecifics and the social dynamics of any group is dependant on the individuals within the group. Thus, common farm management practices such mixing may cause social disruption. Social integration of new group members has previously been defined as a lack of aggressive interactions within the group. However, a lack of aggression among group members may not represent full integration into the social group. Here we observe the impact of disrupting groups of cattle via the introduction of an unfamiliar individual, on the social network patterns of six groups of cattle. Cattle contacts between all individuals in a group were recorded before and after the introduction of the unfamiliar individual. Pre-introduction, resident cattle showed preferential associations with specific individuals in the group. Post-introduction, resident cattle reduced the strength of their contacts (e.g., frequency) with each other relative to the pre-introduction phase. Unfamiliar individuals were socially isolated from the group throughout the trial. The observed social contact patterns suggest that new group members are socially isolated from established groups longer than previously thought, and common farm mixing practices may have negative welfare consequences on introduced individuals.

Using sequence data to identify alternative routes and risk of infection: a case-study of campylobacter in Scotland.

BACKGROUND: Genetic typing data are a potentially powerful resource for determining how infection is acquired. In this paper MLST typing was used to distinguish the routes and risks of infection of humans with Campylobacter jejuni from poultry and ruminant sources METHODS: C. jejuni samples from animal and environmental sources and from reported human cases confirmed between June 2005 and September 2006 were typed using MLST. The STRUCTURE software was used to assign the specific sequence types of the sporadic human cases to a particular source. We then used mixed case-case logistic regression analysis to compare the risk factors for being infected with C. jejuni from different sources. RESULTS: A total of 1,599 (46.3%) cases were assigned to poultry, 1,070 (31.0%) to ruminant and 67 (1.9%) to wild bird sources; the remaining 715 (20.7%) did not have a source that could be assigned with a probability of greater than 0.95. Compared to ruminant sources, cases attributed to poultry sources were typically among adults (odds ratio (OR) = 1.497, 95% confidence intervals (CIs) = 1.211, 1.852), not among males (OR = 0.834, 95% CIs = 0.712, 0.977), in areas with population density of greater than 500 people/km2 (OR = 1.213, 95% CIs = 1.030, 1.431), reported in the winter (OR = 1.272, 95% CIs = 1.067, 1.517) and had undertaken recent overseas travel (OR = 1.618, 95% CIs = 1.056, 2.481). The poultry assigned strains had a similar epidemiology to the unassigned strains, with the exception of a significantly higher likelihood of reporting overseas travel in unassigned strains. CONCLUSIONS: Rather than estimate relative risks for acquiring infection, our analyses show that individuals acquire C. jejuni infection from different sources have different associated risk factors. By enhancing our ability to identify at-risk groups and the times at which these groups are likely to be at risk, this work allows public health messages to be targeted more effectively. The rapidly increasing capacity to conduct genetic typing of pathogens makes such traced epidemiological analysis more accessible and has the potential to substantially enhance epidemiological risk factor studies.

Patterns of mortality in domesticated ruminants in Ethiopia.

Background: Premature death of livestock is a problem in all ruminant production systems. While the number of premature ruminant deaths in a country is a reasonable indicator for the nation's health, few data sources exist in a country like Ethiopia that can be used to generate valid estimates. The present study aimed to establish if three different data sets, each with imperfect information on ruminant mortality, including abortions, could be combined into improved estimates of nationwide mortality in Ethiopia. Methods: We combined information from a recent survey of ruminant mortality with information from the Living Standards Measurement Study and the Disease Outbreak and Vaccination Reporting dataset. Generalized linear mixed and hurdle models were used for data analysis, with results summarized using predicted outcomes. Results: Analyses indicated that most herds experienced zero mortality and reproductive losses, with rare occasions of larger losses. Diseases causing deaths varied greatly both geographically and over time. There was little agreement between the different datasets. While the models aid the understanding of patterns of mortality and reproductive losses, the degree of variation observed limited the predictive scope. Conclusions: The models revealed some insight into why mortality rates are variable over time and are therefore less useful in measuring production or health status, and it is suggested that alternative measures of productivity, such as number of offspring raised to 1 year old per dam, would be more stable over time and likely more indicative.

Geographic determinants of reported human Campylobacter infections in Scotland.

BACKGROUND: Campylobacteriosis is the leading cause of bacterial gastroenteritis in most developed countries. People are exposed to infection from contaminated food and environmental sources. However, the translation of these exposures into infection in the human population remains incompletely understood. This relationship is further complicated by differences in the presentation of cases, their investigation, identification, and reporting; thus, the actual differences in risk must be considered alongside the artefactual differences. METHODS: Data on 33,967 confirmed Campylobacter infections in mainland Scotland between 2000 and 2006 (inclusive) that were spatially referenced to the postcode sector level were analysed. Risk factors including the Carstairs index of social deprivation, the easting and northing of the centroid of the postcode sector, measures of livestock density by species and population density were tested in univariate screening using a non-spatial generalised linear model. The NHS Health Board of the case was included as a random effect in this final model. Subsequently, a spatial generalised linear mixed model (GLMM) was constructed and age-stratified sensitivity analysis was conducted on this model. RESULTS: The spatial GLMM included the protective effects of the Carstairs index (relative risk (RR) = 0.965, 95% Confidence intervals (CIs) = 0.959, 0.971) and population density (RR = 0.945, 95% CIs = 0.916, 0.974. Following stratification by age group, population density had a significant protective effect (RR = 0.745, 95% CIs = 0.700, 0.792) for those under 15 but not for those aged 15 and older (RR = 0.982, 95% CIs = 0.951, 1.014). Once these predictors have been taken into account three NHS Health Boards remain at significantly greater risk (Grampian, Highland and Tayside) and two at significantly lower risk (Argyll and Ayrshire and Arran). CONCLUSIONS: The less deprived and children living in rural areas are at the greatest risk of being reported as a case of Campylobacter infection. However, this analysis cannot differentiate between actual risk and heterogeneities in individual reporting behaviour; nevertheless this paper has demonstrated that it is possible to explain the pattern of reported Campylobacter infections using both social and environmental predictors.

Considering appropriate replication in the design of animal social network studies.

Social network analysis has increasingly been considered a useful tool to interpret the complexity of animal social relationships. However, group composition can affect the contact structure of the network resulting in variation between networks. Replication in contact network studies is rarely done but enables determination of possible variation in response across networks. Here we explore the importance of between-group variability in social behaviour and the impact of replication on hypothesis testing. We use an exemplar study of social contact data collected from six replicated networks of cattle before and after the application of a social disturbance treatment. In this replicated study, subtle but consistent changes in animal contact patterns were detected after the application of a social disturbance treatment. We then quantify both within- and between-group variation in this study and explore the importance of varying the number of replicates and the number of individuals within each network, on the precision of the differences in treatment effects for the contact behaviour of the resident cattle. The analysis demonstrates that reducing the number of networks observed in the study would reduce the probability of detecting treatment differences for social behaviours even if the total number of animals was kept the same.

A novel, high-welfare methodology for evaluating poultry red mite interventions in vivo.

Optimisation and use of a device for the on-hen in vivo feeding of all hematophagous stages of Dermanyssus gallinae is described. The sealed mesh device contains the mites and is applied to the skin of the hen's thigh where mites can feed on the bird through a mesh which has apertures large enough to allow the mites' mouth-parts to access to the bird but small enough to contain the mites. By optimising the depth and width of the mesh aperture size we have produced a device which will lead to both reduction and refinement in the use of animals in research, allowing the pre-screening of new vaccines and systemic acaricides/insecticides which have been developed for the control of these blood-feeding parasites before progressing to large field trials. For optimal use, the device should be constructed from 105 µm aperture width, 63 µm depth, polyester mesh and the mites (irrespective of life stage) should be conditioned with no access to food for 3 weeks at 4 °C for optimal feeding and post-feeding survival.

Evaluating Diagnostic Tests With Near-Perfect Specificity: Use of a Hui-Walter Approach When Designing a Trial of a DIVA Test for Bovine Tuberculosis.

Active surveillance of rare infectious diseases requires diagnostic tests to have high specificity, otherwise the false positive results can outnumber the true cases detected, leading to low positive predictive values. Where a positive result can have economic consequences, such as the cull of a bovine Tuberculosis (bTB) positive herd, establishing a high specificity becomes particularly important. When evaluating new diagnostic tests against a "gold standard" reference test with assumed perfect sensitivity and specificity, calculation of sample sizes are commonly done using a normal approximation to the binomial distribution, although this approach can be misleading. As the expected specificity of the evaluated diagnostic test nears 100%, the errors arising from this approximation are appreciable. Alternatively, it is straightforward to calculate the sample size by using more appropriate confidence intervals, while precisely quantifying the effect of sampling variability using the binomial distribution. However, regardless of the approach, if specificity is high the sample size required becomes large, and the gold standard may be prohibitively costly. An alternative to a gold standard test is to use at least two imperfect, conditionally independent tests, and to analyse the results using a variant of the approach initially proposed by Hui and Walter. We show how this method performs for tests with near-perfect specificity; in particular we show that the sample size required to deliver useful bounds on the precision becomes very large for both approaches. We illustrate these concepts using simulation studies carried out to support the design of a trial of a bTB vaccine and a diagnostic that is able to "Differentiate Infected and Vaccinated Animals" (DIVA). Both test characteristics and the efficacy of the bTB vaccine will influence the sample size required for the study. We propose an improved methodology using a two stage approach to evaluating diagnostic tests in low disease prevalence populations. By combining an initial gold standard pilot study with a larger study analyzed using a Hui-Walter approach, the sample size required for each study can be reduced and the precision of the specificity estimate improved, since information from both studies is combined.

The dynamic influence of genetic variation on the susceptibility of sheep to gastrointestinal nematode infection.

The interaction between sheep and the nematode Teladorsagia circumcincta is one of the best understood of all host-parasite interactions. Following infection, there is considerable variation among lambs in the number of nematode eggs produced, the number of early fourth-stage larvae and the number of adult worms in the mucosa. These traits have a high variance to mean ratio (i.e. they are overdispersed or aggregated among hosts), they are skewed and approximately negative binomially distributed. The sources of overdispersion are differences among lambs in the ingestion of infective larvae and the immune response. Both forces can produce aggregation but their relative importance is unknown. The key components of variation can be identified by variance analysis. The sum of the average effects of polymorphic genes is known as additive genetic variation and this increases essentially from zero at one month of age to quite high values at six months of age. The major mechanism underlying genetic variation appears to be the differences among individuals in immune responses. Two of the major sources of variation in immune responses are differences in antigen recognition and differences in the type of cytokines produced. Genes that influence both these sources of variation are associated with differences in resistance to nematode infection. Therefore, much of the heterogeneity among animals in parasite transmission appears to be due to genetic variation in immune responsiveness.

Assessing the convergence of Markov Chain Monte Carlo methods: an example from evaluation of diagnostic tests in absence of a gold standard.

The accessibility of Markov Chain Monte Carlo (MCMC) methods for statistical inference have improved with the advent of general purpose software. This enables researchers with limited statistical skills to perform Bayesian analysis. Using MCMC sampling to do statistical inference requires convergence of the MCMC chain to its stationary distribution. There is no certain way to prove convergence; it is only possible to ascertain when convergence definitely has not been achieved. These methods are rather subjective and not implemented as automatic safeguards in general MCMC software. This paper considers a pragmatic approach towards assessing the convergence of MCMC methods illustrated by a Bayesian analysis of the Hui-Walter model for evaluating diagnostic tests in the absence of a gold standard. The Hui-Walter model has two optimal solutions, a property which causes problems with convergence when the solutions are sufficiently close in the parameter space. Using simulated data we demonstrate tools to assess the convergence and mixing of MCMC chains using examples with and without convergence. Suggestions to remedy the situation when the MCMC sampler fails to converge are given. The epidemiological implications of the two solutions of the Hui-Walter model are discussed.

Combining Slaughterhouse Surveillance Data with Cattle Tracing Scheme and Environmental Data to Quantify Environmental Risk Factors for Liver Fluke in Cattle.

Liver fluke infection causes serious disease (fasciolosis) in cattle and sheep in many regions of the world, resulting in production losses and additional economic consequences due to condemnation of the liver at slaughter. Liver fluke depends on mud snails as an intermediate host and infect livestock when ingested through grazing. Therefore, environmental factors play important roles in infection risk and climate change is likely to modify this. Here, we demonstrate how slaughterhouse data can be integrated with other data, including animal movement and climate variables to identify environmental risk factors for liver fluke in cattle in Scotland. We fitted a generalized linear mixed model to the data, with exposure-weighted random and fixed effects, an approach which takes into account the amount of time cattle spent at different locations, exposed to different levels of risk. This enabled us to identify an increased risk of liver fluke with increased animal age, rainfall, and temperature and for farms located further to the West, in excess of the risk associated with a warmer, wetter climate. This model explained 45% of the variability in liver fluke between farms, suggesting that the unexplained 55% was due to factors not included in the model, such as differences in on-farm management and presence of wet habitats. This approach demonstrates the value of statistically integrating routinely recorded slaughterhouse data with other pre-existing data, creating a powerful approach to quantify disease risks in production animals. Furthermore, this approach can be used to better quantify the impact of projected climate change on liver fluke risk for future studies.

Spatial distribution of Escherichia coli O157-positive farms in Scotland.

Using a sample of 949 Scottish farms with finishing cattle, the spatial distribution of Escherichia coli O157-positive farms was investigated using disease mapping models. The overall prevalence of E. coli O157-positive farms was estimated as 22%. The regions used in this study were the 16 postcode areas of Scotland. For each region, the posterior relative risk (RR) was estimated as a model-based alternative to the saturated standardized morbidity ratio (SMR), i.e., the ratio between observed and expected cases in a region. Three Bayesian hierarchical models with generalized linear modeling of the area-specific risks were used to estimate the posterior relative risk of E. coli O157-positive farms in the postcode areas: a random-effects model incorporating only spatially uncorrelated heterogeneity; a model incorporating both spatially correlated and uncorrelated heterogeneity; and a pseudo-mixture model with unstructured correlation and a weighted mix of two variance components representing the spatial correlation and a jump structure. None of the models identified any areas with a significant increase or decrease in risk. The deviance information criteria slightly favored the simplest model (RR range: 0.92--1.09). However, this model appeared to smooth out more of the variation in the RR compared to the pseudo-mixture model, which gave a more informative pattern of the posterior relative risks (range: 0.81--1.22).

Implications of within-farm transmission for network dynamics: consequences for the spread of avian influenza.

The importance of considering coupled interactions across multiple population scales has not previously been studied for highly pathogenic avian influenza (HPAI) in the British commercial poultry industry. By simulating the within-flock transmission of HPAI using a deterministic S-E-I-R model, and by incorporating an additional environmental class representing infectious faeces, we tracked the build-up of infectious faeces within a poultry house over time. A measure of the transmission risk (TR) was computed for each farm by linking the amount of infectious faeces present each day of an outbreak with data describing the daily on-farm visit schedules for a major British catching company. Larger flocks tended to have greater levels of these catching-team visits. However, where density-dependent contact was assumed, faster outbreak detection (according to an assumed mortality threshold) led to a decreased opportunity for catching-team visits to coincide with an outbreak. For this reason, maximum TR-levels were found for mid-range flock sizes (~25,000-35,000 birds). When assessing all factors simultaneously using multivariable linear regression on the simulated outputs, those related to the pattern of catching-team visits had the largest effect on TR, with the most important movement-related factor depending on the mode of transmission. Using social network analysis on a further database to inform a measure of between-farm connectivity, we identified a large fraction of farms (28%) that had both a high TR and a high potential impact at the between farm level. Our results have counter-intuitive implications for between-farm spread that could not be predicted based on flock size alone, and together with further knowledge of the relative importance of transmission risk and impact, could have implications for improved targeting of control measures.

Reliability of a quantitative rating scale for assessment of horses with distal tarsal osteoarthritis.

Various radiographic rating scales have been described for use in horses with distal tarsal osteoarthritis but little information is available on their reliability. The aim of this study was to develop a radiographic rating scale based on the results of an expert consultation process (the Delphi process), and to test the reliability of the radiographic rating scale. Seven radiographic features were identified as important indicators of distal tarsal osteoarthritis and these were then incorporated in the radiographic rating scale, which used a 100-mm-long visual analog scale. On two occasions nine equine veterinarians applied the radiographic rating scale, and a verbal descriptive rating scale, to three sets of tarsal radiographs, each comprising four standard radiographic views. Reliability was assessed using Bland-Altman plots and by calculating the 95% agreement limits. Analysis of variance (ANOVA) was used to identify significant interactions between the ratings of different assessors made from different views and at each assessment. Rating of distal tarsal osteoarthritis was different for the nine assessors. The most precise second ratings were between 16 mm higher and 18 mm lower than the first. Significant variables were "joint," "assessor," and "assessment" (univariable ANOVA); and "joint and assessor" and "assessor and assessment" (multivariable ANOVA). The radiographic rating scale developed for interpretation of distal tarsal osteoarthritis was less reliable than a verbal descriptive rating scale.

Dairy cow cleanliness and milk quality on organic and conventional farms in the UK.

A subjective cow cleanliness scoring system was validated and used to assess the cleanliness score of dairy cows at different times in the year. A longitudinal study followed a number of farms from summer to winter, and a larger, cross-sectional study assessed a greater number of farms during the housed winter period. The scoring system was demonstrated to be both a repeatable and practical technique to use on-farm and showed that cows become dirtier in the transition from summer grazing to winter housing. Although farming system (organic or conventional) had no effect on cow cleanliness when cows were at grass, when housed in the winter, organic cows were significantly more likely to be cleaner. There was a link between cow cleanliness scores and milk quality, with herds having lower bulk tank somatic cell counts (BTSCC) tending to have a lower (cleaner) median cow cleanliness score; with this relationship strongest for the organic herds. There was no significant link between cleanliness score and Bactoscan (BS) count or clinical mastitis incidence. No major mastitis pathogens were cultured from bulk tank milk samples from the quartile of herds with the cleanest cows in contrast to the quartile of herds with the dirtiest cows, where significant mastitis pathogens were cultured. Based on this study, all farms, especially organic systems, should attempt to keep cows clean as part of subclinical mastitis control.

Investigation of the vitamins A and E and beta-carotene content in milk from UK organic and conventional dairy farms.

During a 12-month longitudinal study, bulk-tank milk was collected from organic (n=17) and conventional (n=19) dairy farms in the UK. Milk samples were analysed for vitamin A (retinol), vitamin E (alpha-tocopherol) and beta-carotene content. The farming system type, herd production level and nutritional factors affecting the milk fat vitamin content were investigated by use of mixed model analyses. Conventionally produced milk fat had a higher mean content of vitamin A than organically produced milk fat, although there were no significant differences in the vitamin E or beta-carotene contents between the two types of milk fat. Apart from farming system, other key factors that affected milk fat vitamin content were season, herd yield and concentrate feeding level. Milk vitamin content increased in the summer months and in association with increased concentrate feeding, whilst higher-yielding herds had a lower milk vitamin E and beta-carotene content. Thus, conventional dairy farms in the UK produced milk with a higher vitamin A content, possibly owing to increased vitamin A supplementation in concentrate feeds. However, knowledge of the effects of season, access to fresh grazing or specific silage types and herd production level may also be used by all producers and processors to enhance the vitamin content in milk.

The Gamma-Poisson model as a statistical method to determine if micro-organisms are randomly distributed in a food matrix.

The Gamma-Poisson model, i.e., a Poisson distribution where the parameter lambda is Gamma distributed, has been suggested as a statistical method for determining whether or not micro-organisms are randomly distributed in a food matrix. In this study, we analyse the Gamma-Poisson model to explore some of the properties of the Gamma-Poisson model left unexplored by the previous study. The conclusion of our analysis is that the Gamma-Poisson model distinguishes poorly between variation at the Poisson level and the Gamma level. Estimated parameter values from simulated data-sets showed large variation around the true values, even for moderate sample sizes (n=100). Furthermore, at these sample sizes the likelihood ratio is not a good test statistic for discriminating between the Gamma-Poisson distribution and the Poisson distribution. Hence, to determine if data are randomly distributed, i.e., Poisson distributed, the Gamma-Poisson distribution is not a good choice. However, the ratio between variation at the Poisson level and the Gamma level does provide a measure of the amount of overdispersion.