Suitability of sentinel abattoirs for syndromic surveillance using provincially inspected bovine abattoir condemnation data.

BACKGROUND: Sentinel surveillance has previously been used to monitor and identify disease outbreaks in both human and animal contexts. Three approaches for the selection of sentinel sites are proposed and evaluated regarding their ability to capture overall respiratory disease trends using provincial abattoir condemnation data from all abattoirs open throughout the study for use in a sentinel syndromic surveillance system. RESULTS: All three sentinel selection criteria approaches resulted in the identification of sentinel abattoirs that captured overall temporal trends in condemnation rates similar to those reported by the full set of abattoirs. However, all selection approaches tended to overestimate the condemnation rates of the full dataset by 1.4 to as high as 3.8 times for cows, heifers and steers. Given the results, the selection approach using abattoirs open all weeks had the closest approximation of temporal trends when compared to the full set of abattoirs. CONCLUSIONS: Sentinel abattoirs show promise for integration into a food animal syndromic surveillance system using Ontario provincial abattoir condemnation data. While all selection approaches tended to overestimate the condemnation rates of the full dataset to some degree, the abattoirs open all weeks selection approach appeared to best capture the overall seasonal and temporal trends of the full dataset and would be the most suitable approach for sentinel abattoir selection.

Suitability of bovine portion condemnations at provincially-inspected abattoirs in Ontario Canada for food animal syndromic surveillance.

BACKGROUND: Abattoir condemnations may play an important role in a food animal syndromic surveillance system. Portion condemnation data may be particularly useful, as these data can provide more specific information on health outcomes than whole carcass condemnation data. Various seasonal, secular, disease, and non-disease factors have been previously identified to be associated with whole carcass condemnation rates in Ontario provincial abattoirs; and if ignored, may bias the results of quantitative disease surveillance methods. The objective of this study was to identify various seasonal, secular, and abattoir characteristic factors that may be associated with bovine portion condemnation rates and compare how these variables may differ from previously identified factors associated with bovine whole carcass condemnation rates. RESULTS: Data were collected from the Ontario Ministry of Agriculture, Food and Rural Affairs (OMAFRA) and the Ontario Cattlemen's Association regarding "parasitic liver" and pneumonic lung condemnation rates for different cattle classes, abattoir compliance ratings, and the monthly sales-yard price for commodity classes from 2001-2007. To control for clustering by abattoirs, multi-level Poisson modeling was used to investigate the association between the following variables and "parasitic liver" as well as pneumonic lung condemnation rates: year, season, annual abattoir audit rating, geographic region, annual abattoir operating time, annual total number of animals processed, animal class, and commodity sales price. CONCLUSIONS: In this study, "parasitic liver" condemnation rates were associated with year, season, animal class, audit rating, and region. Pneumonic lung condemnation rates were associated with year, season, animal class, region, audit rating, number of cattle processed per year, and number of weeks abattoirs processed cattle. Unlike previous models based on whole carcass condemnations, commodity price was not associated with partial condemnations in this study. The results identified material-specific predictor variables for condemnation rates. This is important for syndromic surveillance based on abattoir data and should be modeled and controlled for during quantitative surveillance analysis on a portion specific basis.

Factors associated with whole carcass condemnation rates in provincially-inspected abattoirs in Ontario 2001-2007: implications for food animal syndromic surveillance.

BACKGROUND: Ontario provincial abattoirs have the potential to be important sources of syndromic surveillance data for emerging diseases of concern to animal health, public health and food safety. The objectives of this study were to: (1) describe provincially inspected abattoirs processing cattle in Ontario in terms of the number of abattoirs, the number of weeks abattoirs process cattle, geographical distribution, types of whole carcass condemnations reported, and the distance animals are shipped for slaughter; and (2) identify various seasonal, secular, disease and non-disease factors that might bias the results of quantitative methods, such as cluster detection methods, used for food animal syndromic surveillance. RESULTS: Data were collected from the Ontario Ministry of Agriculture, Food and Rural Affairs and the Ontario Cattlemen's Association regarding whole carcass condemnation rates for cattle animal classes, abattoir compliance ratings, and the monthly sales-yard price for various cattle classes from 2001-2007. To analyze the association between condemnation rates and potential explanatory variables including abattoir characteristics, season, year and commodity price, as well as animal class, negative binomial regression models were fit using generalized estimating equations (GEE) to account for autocorrelation among observations from the same abattoir. Results of the fitted model found animal class, year, season, price, and audit rating are associated with condemnation rates in Ontario abattoirs. In addition, a subset of data was used to estimate the average distance cattle are shipped to Ontario provincial abattoirs. The median distance from the farm to the abattoir was approximately 82 km, and 75% of cattle were shipped less than 100 km. CONCLUSIONS: The results suggest that secular and seasonal trends, as well as some non-disease factors will need to be corrected for when applying quantitative methods for syndromic surveillance involving these data. This study also demonstrated that animals shipped to Ontario provincial abattoirs come from relatively local farms, which is important when considering the use of spatial surveillance methods for these data.

Investigation of exposure to swine influenza viruses in Ontario (Canada) finisher herds in 2004 and 2005.

The epidemiology of influenza in the North American swine population has changed since the emergence of a triple-reassortant H3N2 influenza virus. Although seen previously in North America, the Ontario swine population had likely been free of viruses of the reassortant H3N2 lineage until 2005. The objective of this study was to investigate the frequency and distribution of exposure to H1N1 and H3N2 subtypes in the Ontario finisher pig population prior to and after the H3N2 outbreak that occurred in 2005. This included investigating prevalence and spatial distribution of positive herds, assessing proportion of random variation at different hierarchical levels, and evaluating selected demographic factors and management procedures as potential risk factors. In total, 919 and 978 sera collected in cross-sectional studies from 46 and 49 finisher herds in 2004 and 2005 were tested by a H1N1 subtype-specific and a H3N2 subtype-specific commercial ELISA. For the H1N1 subtype, the point prevalence of positive herds (>3 reactors) was 19.5% and 30.6% in 2004 and 2005, respectively. For the H3N2 subtype the point prevalence of positive herds (>3 reactors) was 6.5% and 40.8% in 2004 and 2005, respectively. Sera from 2004 that were positive on H3N2 ELISA did not cross-react with any of the H3N2 variants used as antigen on a sequential HI test. Only herds positive for H3N2 subtype in 2005 clustered in space (P<0.01). The H1N1 status in 2005 was associated with the H1N1 status in 2004, and with reported distance to the nearest herd. The H3N2 status in 2005 was associated with reported distance to the nearest herd and a type of replacement gilt source. For H3N2, distance seemed to be important even after controlling for type of gilt source. Most variability in seropositivity was between herds with little variability between pens. This study confirms that in 2005, the epidemic H3N2 subtype co-circulated with endemic H1N1 subtype in the Ontario finisher herds. We concluded that in Ontario, the endemic H1N1 subtype was likely maintained through circulation within herds and sites with common flow. Whereas the transmission of epidemic H3N2 subtype was attributed to local spread, which could include different modes of direct, indirect, and airborne transmission. We emphasize the importance of establishing routine monitoring systems that would allow using molecular tools, and maintaining serum banks as a useful resource for retrospective comparisons.

The North American Animal Disease Spread Model: a simulation model to assist decision making in evaluating animal disease incursions.

The North American Animal Disease Spread Model is a stochastic, spatial, state-transition simulation model for the spread of highly contagious diseases of animals. It was developed with broad international support to assist policy development and decision making involving disease incursions. User-established parameters define model behavior in terms of disease progression; disease spread by animal-to-animal contact, contact with contaminated personnel or equipment, and airborne dissemination; and the implementation of control measures such as destruction and vaccination. Resources available to implement disease control strategies, as well as the direct costs associated with these strategies, are taken into consideration. The model records a wide variety of measures of the extent of simulated outbreaks and other characteristics. The graphical interface and output visualization features also make it a useful tool for training and preparedness exercises. This model is now being used to evaluate outbreak scenarios and potential control strategies for several economically important exotic animal diseases in the United States, Canada, and elsewhere. NAADSM is freely available via the Internet at http://www.naadsm.org.

Estimating the under-reporting rate for infectious gastrointestinal illness in Ontario.

BACKGROUND: In Ontario, infectious gastrointestinal illness (IGI) reporting can be represented by a linear model of several sequential steps required for a case to be captured in the provincial reportable disease surveillance system. Since reportable enteric data are known to represent only a small fraction of the total IGI in the community, the objective of this study was to estimate the under-reporting rate for IGI in Ontario. METHODS: A distribution of plausible values for the under-reporting rate was estimated by specifying input distributions for the proportions reported at each step in the reporting chain, and multiplying these distributions together using simulation methods. Input distributions (type of distribution and parameters) for the proportion of cases reported at each step of the reporting chain were determined using data from the Public Health Agency of Canada's National Studies on Acute Gastrointestinal Illness (NSAGI) initiative. RESULTS: For each case of enteric illness reported to the province of Ontario, the estimated number of cases of IGI in the community ranged from 105 to 1,389, with a median of 285, and a mean and standard deviation of 313 and 128, respectively. CONCLUSIONS: Each case of enteric illness reported to the province of Ontario represents an estimated several hundred cases of IGI in the community. Thus, reportable disease data should be used with caution when estimating the burden of such illness. Program planners and public health personnel may want to consider this fact when developing population-based interventions.

Syndromic surveillance using veterinary laboratory data: data pre-processing and algorithm performance evaluation.

Diagnostic test orders to an animal laboratory were explored as a data source for monitoring trends in the incidence of clinical syndromes in cattle. Four years of real data and over 200 simulated outbreak signals were used to compare pre-processing methods that could remove temporal effects in the data, as well as temporal aberration detection algorithms that provided high sensitivity and specificity. Weekly differencing demonstrated solid performance in removing day-of-week effects, even in series with low daily counts. For aberration detection, the results indicated that no single algorithm showed performance superior to all others across the range of outbreak scenarios simulated. Exponentially weighted moving average charts and Holt-Winters exponential smoothing demonstrated complementary performance, with the latter offering an automated method to adjust to changes in the time series that will likely occur in the future. Shewhart charts provided lower sensitivity but earlier detection in some scenarios. Cumulative sum charts did not appear to add value to the system; however, the poor performance of this algorithm was attributed to characteristics of the data monitored. These findings indicate that automated monitoring aimed at early detection of temporal aberrations will likely be most effective when a range of algorithms are implemented in parallel.

Syndromic surveillance using veterinary laboratory data: algorithm combination and customization of alerts.

BACKGROUND: Syndromic surveillance research has focused on two main themes: the search for data sources that can provide early disease detection; and the development of efficient algorithms that can detect potential outbreak signals. METHODS: This work combines three algorithms that have demonstrated solid performance in detecting simulated outbreak signals of varying shapes in time series of laboratory submissions counts. These are: the Shewhart control charts designed to detect sudden spikes in counts; the EWMA control charts developed to detect slow increasing outbreaks; and the Holt-Winters exponential smoothing, which can explicitly account for temporal effects in the data stream monitored. A scoring system to detect and report alarms using these algorithms in a complementary way is proposed. RESULTS: The use of multiple algorithms in parallel resulted in increased system sensitivity. Specificity was decreased in simulated data, but the number of false alarms per year when the approach was applied to real data was considered manageable (between 1 and 3 per year for each of ten syndromic groups monitored). The automated implementation of this approach, including a method for on-line filtering of potential outbreak signals is described. CONCLUSION: The developed system provides high sensitivity for detection of potential outbreak signals while also providing robustness and flexibility in establishing what signals constitute an alarm. This flexibility allows an analyst to customize the system for different syndromes.

Exploratory analysis of methods for automated classification of laboratory test orders into syndromic groups in veterinary medicine.

BACKGROUND: Recent focus on earlier detection of pathogen introduction in human and animal populations has led to the development of surveillance systems based on automated monitoring of health data. Real- or near real-time monitoring of pre-diagnostic data requires automated classification of records into syndromes--syndromic surveillance--using algorithms that incorporate medical knowledge in a reliable and efficient way, while remaining comprehensible to end users. METHODS: This paper describes the application of two of machine learning (Naïve Bayes and Decision Trees) and rule-based methods to extract syndromic information from laboratory test requests submitted to a veterinary diagnostic laboratory. RESULTS: High performance (F1-macro = 0.9995) was achieved through the use of a rule-based syndrome classifier, based on rule induction followed by manual modification during the construction phase, which also resulted in clear interpretability of the resulting classification process. An unmodified rule induction algorithm achieved an F(1-micro) score of 0.979 though this fell to 0.677 when performance for individual classes was averaged in an unweighted manner (F(1-macro)), due to the fact that the algorithm failed to learn 3 of the 16 classes from the training set. Decision Trees showed equal interpretability to the rule-based approaches, but achieved an F(1-micro) score of 0.923 (falling to 0.311 when classes are given equal weight). A Naïve Bayes classifier learned all classes and achieved high performance (F(1-micro)= 0.994 and F(1-macro) = .955), however the classification process is not transparent to the domain experts. CONCLUSION: The use of a manually customised rule set allowed for the development of a system for classification of laboratory tests into syndromic groups with very high performance, and high interpretability by the domain experts. Further research is required to develop internal validation rules in order to establish automated methods to update model rules without user input.

Retrospective time series analysis of veterinary laboratory data: preparing a historical baseline for cluster detection in syndromic surveillance.

The practice of disease surveillance has shifted in the last two decades towards the introduction of systems capable of early detection of disease. Modern biosurveillance systems explore different sources of pre-diagnostic data, such as patient's chief complaint upon emergency visit or laboratory test orders. These sources of data can provide more rapid detection than traditional surveillance based on case confirmation, but are less specific, and therefore their use poses challenges related to the presence of background noise and unlabelled temporal aberrations in historical data. The overall goal of this study was to carry out retrospective analysis using three years of laboratory test submissions to the Animal Health Laboratory in the province of Ontario, Canada, in order to prepare the data for use in syndromic surveillance. Daily cases were grouped into syndromes and counts for each syndrome were monitored on a daily basis when medians were higher than one case per day, and weekly otherwise. Poisson regression accounting for day-of-week and month was able to capture the day-of-week effect with minimal influence from temporal aberrations. Applying Poisson regression in an iterative manner, that removed data points above the predicted 95th percentile of daily counts, allowed for the removal of these aberrations in the absence of labelled outbreaks, while maintaining the day-of-week effect that was present in the original data. This resulted in the construction of time series that represent the baseline patterns over the past three years, free of temporal aberrations. The final method was thus able to remove temporal aberrations while keeping the original explainable effects in the data, did not need a training period free of aberrations, had minimal adjustment to the aberrations present in the raw data, and did not require labelled outbreaks. Moreover, it was readily applicable to the weekly data by substituting Poisson regression with moving 95th percentiles.

Evaluation of a veterinary-based syndromic surveillance system implemented for swine.

Practicing veterinarians play an important role in detecting the initial outbreak of disease in animal populations. A pilot study was conducted to determine the feasibility of a veterinary-based surveillance system for the Ontario swine industry. A total of 7 practitioners from 5 clinics agreed to submit information from July 1, 2007 to June 30, 2008. The surveillance program was evaluated in terms of timeliness, compliance, geographic coverage, and data quality. Our study showed that the veterinary-based surveillance system was acceptable to practitioners and produced useful data. The program obtained information from 25% of pig farms in Ontario during this time period. However, better communication with practitioners, more user-friendly recording systems that can be adapted to each clinic's management system, active involvement of the clinics' technical personnel, and the use of financial incentives may help to improve compliance and timeliness.

Simple models to assist in communicating key principles of animal disease control.

Aggressive actions for disease eradication, including animal disposal, can have serious impacts on the livestock industry, environment and public confidence. Coordinated efforts are required for effective, efficient and acceptable disease control and eradication. The authors summarise concepts from previous publications into a group of simple examples, schematic diagrams and basic equations. These simplified models may be used to communicate principles of disease control to livestock owners and workers, and to regulatory officials and policy-makers. Such stakeholders may not have time to study more complex models. It is hoped that a broader appreciation of key principles will compel stakeholders to act routinely in a manner that improves the prevention and control of infectious animal diseases.

Application of an Automated Hydrophobic Grid Membrane Filter Interpreter System at a Poultry Abattoir.

The performance of a system to measure broiler carcass hygiene was investigated in the abattoir environment. The system involved: whole carcass rinses aided by a mechanical carcass shaker; filtration of rinse solutions though hydrophobic grid membrane filters (HGMF) (ISO-GRIDR, QA Laboratories Ltd.); and use of an automated HGMF interpreter. The interpreter recorded culture results in units of most probable number (MPN) of aerobic bacteria, in electronic data files (MI-100 HGMF Interpreter System, Richard Brancker Research Ltd.). Set-up and operation of the system by government inspection staff at an abattoir ran relatively smoothly with minimal interference to normal plant operation. The system demonstrated good repeatability in measuring log10 most probable number per gram of carcass (LgMPN/g), between repeat readings of the same filters (r=0.993 p<0.001), and good repeatability between repeat filters within the same carcass rinses (r=0.970 p<0.001). Overall, the LgMPN/g ranged from 0.258 to 3.955 with a mean of 2.276 and a variance of 0.324. These corresponded to MPN/g counts in the range of 2 to 9000 and a geometric mean of 188.8 MPN/g. A regression model was developed to investigate poultry supplier and abattoir effects on the variability of counts. A significant supplier effect was observed. The addition of two more carcass showers located just after the venting machine along the evisceration line was not associated with a change in carcass hygiene.

Variability and Determinants of Carcass Bacterial Load at a Poultry Abattoir.

Roaster chicken carcasses (2,928) were collected from the evisceration line of a poultry abattoir over a 5-month period and identified as to the lot (truck load) and supplier. Bacterial load was determined by mechanically rinsing each eviscerated carcass in sterile water and then using an automated hydrophobic grid membrane interpreter system to obtain the log10 most probable number of aerobic bacteria per gram of carcass. Analysis of variance demonstrated that the between-carcass, between-lots-within-supplier, and between-supplier components of variability in bacterial load represented 73.2, 14.2, and 12.6% of the total variability, respectively. There was a significant (p < 0.001) supplier and lots-within-supplier effect on bacterial load of carcasses. A regression model demonstrated that bacterial load of lots significantly (p ≤ 0.05) decreased with increasing hours of operation of the evisceration line. Factors in the model which were significantly (p ≤ 0.05) associated with increased bacterial load included longer crating and holding times, higher visible contamination scores, slaughter during winter months, higher outdoor temperatures, and slaughter of lots composed of only pullets. The model explained about 23% of the variability in bacterial load.

Variability in Broiler Carcass Bacterial Load at Three Abattoirs, as Measured by a Hydrophobic Grid Membrane Filter Interpreter.

In this observational study, the variability of broiler carcass bacterial load was investigated at three federally inspected abattoirs, using an automated hydrophobic grid membrane filter interpreter system. The measurement protocol involved: whole carcass rinses aided by a mechanical carcass shaker; filtration of rinse solutions through hydrophobic grid membrane filters (HGMF) (ISO-GRID®, QA Laboratories, Ltd., Toronto, Ont.); and use of an automated HGMF interpreter (MI-100 HGMF Interpreter System, Richard Brancker Research, Ltd., Ottawa, Ont.). Carcass and lot mean bacterial loads were measured, respectively, in units of log10 most probable number (MPN) of mesophylic aerobic colony forming units per gram of carcass (LgMPN/g), and slaughter lot mean LgMPN/g (LMLgMPN/g). Whole carcass rinses were conducted on a total of 1,917 carcasses, among 96 slaughter lots from three abattoirs. Overall, the LgMPN/g ranged from 1.054 to 4.180 with a mean of 2.585 and a variance of 0.263. These corresponded to MPN/g counts from 11 to 15,135 and a geometric mean of 385 MPN/g. Statistically significant differences were observed between abattoirs and between lots within abattoirs. The intra-abattoir correlation coefficient of LgMPN/g was r = 0.180 (p < 0.001). The within abattoir intralot correlation coefficient was r = 0.259 (p < 0.001). In this data set, approximately 56, 26, and 18% of the variability in LgMPN/g were attributed to factors operating at the individual bird, lot, and abattoir levels of organization, respectively. Factors significantly associated with LMLgMPN/g included: abattoir (p < 0.001), transportation time from farm to abattoir (p < 0.001), and waiting time from arrival at the abattoir yard to actual slaughter (p = 0.002). Analysis of a series of five repeat rinses, conducted on one bird from each of the 96 study lots, demonstrated that bacterial counts in the second to fifth sequential rinses were positively associated with the bacterial count of the first rinse. Also, after adjusting for the initial count, a pattern of decreasing counts was observed in subsequent rinses.

Survey of Ontario Bulk Tank Raw Milk for Food-Borne Pathogens.

Raw (unpasteurized) milk can be a source of food-borne pathogens. Raw milk consumption results in sporadic disease outbreaks. Pasteurization is designed to destroy all bacterial pathogens common to raw milk, excluding spore-forming bacteria and possibly Mycobacterium paratuberculosis , but some people continue to drink raw milk, believing it to be safe. Current methods for assessing the bacteriological quality of raw milk, such as aerobic plate counts, are not usually designed to detect specific pathogens. The objective of this study was to estimate the proportion of pick-ups (loads of raw milk from a single farm bulk tank) from Ontario farm bulk tanks that contained Listeria monocytogenes . Salmonella spp., Campylobacter spp., and/or verotoxigenic Escherichia coli (VTEC). Samples from 1,720 pick-ups of raw milk were tested for the presence of these pathogens, and 47 L. monocytogenes , three Salmonella spp., eight Campylobacter spp., and 15 VTEC isolates were detected, representing 2.73, 0.17, 0.47, and 0.87% of milk samples, respectively. Estimates of the proportion of theoretical tanker truck loads of pooled raw milk contaminated with pathogens ranged from a low of 0.51 % of tankers containing raw milk from 3 bulk tanks being contaminated with Salmonella spp. to a high of 34.41 % of tankers containing raw milk from 10 bulk tanks being contaminated with at least one of the pathogens. Associations between the presence of pathogens and raw milk sample characteristics were investigated. The mean somatic cell count was higher among VTEC- or L. monocytogenes -positive samples, and the mean aerobic plate count was found to be higher among L. monocytogenes -positive samples. These results confirm the presence of bacterial food pathogens in raw milk and emphasize the importance of continued diligence in the application of hygiene programs within dairies and the separation of raw milk from pasteurized milk and milk products.