Development of Reporting Guidelines for Animal Health Surveillance-AHSURED.

With the current trend in animal health surveillance toward risk-based designs and a gradual transition to output-based standards, greater flexibility in surveillance design is both required and allowed. However, the increase in flexibility requires more transparency regarding surveillance, its activities, design and implementation. Such transparency allows stakeholders, trade partners, decision-makers and risk assessors to accurately interpret the validity of the surveillance outcomes. This paper presents the first version of the Animal Health Surveillance Reporting Guidelines (AHSURED) and the process by which they have been developed. The goal of AHSURED was to produce a set of reporting guidelines that supports communication of surveillance activities in the form of narrative descriptions. Reporting guidelines come from the field of evidence-based medicine and their aim is to improve consistency and quality of information reported in scientific journals. They usually consist of a checklist of items to be reported, a description/definition of each item, and an explanation and elaboration document. Examples of well-reported items are frequently provided. Additionally, it is common to make available a website where the guidelines are documented and maintained. This first version of the AHSURED guidelines consists of a checklist of 40 items organized in 11 sections (i.e., surveillance system building blocks), which is available as a wiki at https://github.com/SVA-SE/AHSURED/wiki. The choice of a wiki format will allow for further inputs from surveillance experts who were not involved in the earlier stages of development. This will promote an up-to-date refined guideline document.

A modified TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) applied to choosing appropriate selection methods in ongoing surveillance for Avian Influenza in Canada.

To achieve an appropriate and efficient sample in a surveillance program, the goals of the program should drive a careful consideration of the selection method or combination of selection methods to be applied. Therefore, the ongoing analysis and assessment of a surveillance system may include an assessment of the ability of the applied selection methods to generate an appropriate sample. There may be opinions from many technical experts (TEs) and many criteria to consider in a surveillance system so there is a need for methods to combine knowledge, priorities and preferences from a group of TEs. This paper proposes a modified weighted and unweighted TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) analysis to choose selection methods in surveillance. An example from the Canadian Notifiable Avian Influenza surveillance (CanNAISS) is used to illustrate the method as this surveillance offers unique data with multiple selection methods and subpopulations. The primary objective was to assess the performance of the different selection methods applied in CanNAISS, from 2008 to 2013, in three subpopulations (A-C). A modified TOPSIS (weighted and unweighted) analyses is proposed to aggregate preferences from three TEs and to identify the selection method that was closest to the ideal solution agreed upon by the TEs. Criteria weights were provided individually by three TEs. For the group decision making, internal and external aggregation approaches were used with arithmetic and geometric means. The results of the weighted modified TOPSIS analysis showed that the selection methods that used farm registries yielded high estimates of the relative closeness to ideal-solution. The ranking of selection methods based on the modified TOPSIS weighted analysis, conducted at the individual and group decision making levels were similar. Regardless of the aggregation approach used (internal or external) in group decision making, the use of the arithmetic and geometric means yielded similar estimates of relative closeness to ideal-solution. The unweighted modified TOPSIS analysis yielded similar estimates of the relative closeness to the ideal-solution and therefore making the interpretation of the results difficult. The weighted modified TOPSIS analysis contributed to an informed decision on the best selection method to apply in CanNAISS. The weighted modified TOPSIS analysis is a straightforward and suitable technique to address decision making problems where the profile of the ideal and non-ideal solutions is known a priori by the decision makers.

Assessing the sensitivity of bovine tuberculosis surveillance in Canada's cattle population, 2009-2013.

The objectives of this study were (1) to estimate the annual sensitivity of Canada's bTB surveillance system and its three system components (slaughter surveillance, export testing and disease investigation) using a scenario tree modelling approach, and (2) to identify key model parameters that influence the estimates of the surveillance system sensitivity (SSSe). To achieve these objectives, we designed stochastic scenario tree models for three surveillance system components included in the analysis. Demographic data, slaughter data, export testing data, and disease investigation data from 2009 to 2013 were extracted for input into the scenario trees. Sensitivity analysis was conducted to identify key influential parameters on SSSe estimates. The median annual SSSe estimates generated from the study were very high, ranging from 0.95 (95% probability interval [PI]: 0.88-0.98) to 0.97 (95% PI: 0.93-0.99). Median annual sensitivity estimates for the slaughter surveillance component ranged from 0.95 (95% PI: 0.88-0.98) to 0.97 (95% PI: 0.93-0.99). This shows that slaughter surveillance to be the major contributor to overall surveillance system sensitivity with a high probability to detect M. bovis infection if present at a prevalence of 0.00028% or greater during the study period. The export testing and disease investigation components had extremely low component sensitivity estimates-the maximum median sensitivity estimates were 0.02 (95% PI: 0.014-0.023) and 0.0061 (95% PI: 0.0056-0.0066) respectively. The three most influential input parameters on the model's output (SSSe) were the probability of a granuloma being detected at slaughter inspection, the probability of a granuloma being present in older animals (≥12 months of age), and the probability of a granuloma sample being submitted to the laboratory. Additional studies are required to reduce the levels of uncertainty and variability associated with these three parameters influencing the surveillance system sensitivity.

Comparing capture-recapture methods for estimation of the size of small and medium-sized populations using empirical data on commercial turkey farms in Canada.

The study objectives were (1) to conduct a systematic review of the performance of capture-recapture methods; (2) to use empirical data to estimate population size in a small-sized population (turkey breeder farms) and a medium-sized population (meat turkey farms) by applying two-source capture-recapture methods (the Lincoln-Petersen, the Chapman, and Chao's lower-bound estimators) and multi-source capture-recapture methods (the log-linear modeling and sample coverage approaches); and (3) to compare the performance of these methods in predicting the true population sizes (2007 data). Our set-up was unique in that we knew the population sizes for turkey breeder farms (99) and meat turkey farms (592) in Canada in 2007, which we applied as our true population sizes, and had surveillance data from the Canadian Notifiable Avian Influenza Surveillance System (2008-2012). We defined each calendar year of sampling as a data source. We confirmed that the two-source capture-recapture methods were sensitive to the violation of the local independence assumption. The log-linear modeling and sample coverage approaches yielded estimates that were closer to the true population sizes than were the estimates provided by the two-source methods for both populations. The performance of both multi-source capture-recapture methods depended on the number of data sources analyzed and the size of the population. Simulation studies are recommended to better understand the limits of each multi-source capture-recapture method.

Estimation of the population size of Canadian commercial poultry farms by log-linear capture-recapture analysis.

The objective of this study was to estimate the population size of Canadian poultry farms in 3 subpopulations (British Columbia, Ontario, and Other) by poultry category. We used data for 2008 to 2011 from the Canadian Notifiable Avian Influenza (NAI) Surveillance System (CanNAISS). Log-linear capture-recapture models were applied to estimate the number of commercial chicken and turkey farms. The estimated size of farm populations was validated by comparing sizes to data provided by the Canadian poultry industry in 2007, which were assumed to be complete and exhaustive. Our results showed that the log-linear modelling approach was an appropriate tool to estimate the population size of Canadian commercial chicken and turkey farms. The 2007 farm population size for each poultry category was included in the 95% confidence intervals of the farm population size estimates. Log-linear capture-recapture modelling might be useful for estimating the number of farms using surveillance data when no comprehensive registry exists.

L'objectif de cette étude était d'estimer le nombre de ferme de volaille au Canada dans trois sous-populations (Colombie-Britannique, Ontario et Autre) par catégorie de volaille. Nous avons utilisé des données du Système canadien de surveillance de l'influenza aviaire à déclaration obligatoire (SCSIADO) de 2008 à 2011. Nous avons utilisé des modèles log-linéaires pour estimer le nombre de fermes commerciales de poulets et de dindons. Nous avons validé les tailles des populations de fermes en les comparants aux données de 2007 fournies par l'industrie canadienne de la volaille (prétendues complètes et exhaustives). Nos résultats ont démontré que l'approche de modélisation log-linéaire était un outil approprié pour estimer les tailles des populations de fermes de poulets et dindons au Canada. Pour chaque catégorie de volaille, la taille de la population de fermes de 2007 était incluse dans l'intervalle de confiance des tailles estimées des populations de fermes avec un niveau de confiance de 95 %. La modélisation log-linéaire de type capture-recapture pourrait être utile pour estimer le nombre de fermes en utilisant des données de surveillance en particulier lorsqu'il n'existe aucun registre exhaustif.(Traduit par les auteurs).

Adapting a scenario tree model for freedom from disease as surveillance progresses: the Canadian notifiable avian influenza model.

Scenario tree models with temporal discounting have been applied in four continents to support claims of freedom from animal disease. Recently, a second (new) model was developed for the same population and disease. This is a natural development because surveillance is a dynamic process that needs to adapt to changing circumstances - the difficulty is the justification for, documentation of, presentation of and the acceptance of the changes. Our objective was to propose a systematic approach to present changes to an existing scenario tree model for freedom from disease. We used the example of how we adapted the deterministic Canadian Notifiable Avian Influenza scenario tree model published in 2011 to a stochastic scenario tree model where the definition of sub-populations and the estimation of probability of introduction of the pathogen were modified. We found that the standardized approach by Vanderstichel et al. (2013) with modifications provided a systematic approach to make and present changes to an existing scenario tree model. We believe that the new 2013 CanNAISS scenario tree model is a better model than the 2011 model because the 2013 model included more surveillance data. In particular, the new data on Notifiable Avian Influenza in Canada from the last 5 years were used to improve input parameters and model structure.

Conceptual evaluation of population health surveillance programs: method and example.

Veterinary and public health surveillance programs can be evaluated to assess and improve the planning, implementation and effectiveness of these programs. Guidelines, protocols and methods have been developed for such evaluation. In general, they focus on a limited set of attributes (e.g., sensitivity and simplicity), that are assessed quantitatively whenever possible, otherwise qualitatively. Despite efforts at standardization, replication by different evaluators is difficult, making evaluation outcomes open to interpretation. This ultimately limits the usefulness of surveillance evaluations. At the same time, the growing demand to prove freedom from disease or pathogen, and the Sanitary and Phytosanitary Agreement and the International Health Regulations require stronger surveillance programs. We developed a method for evaluating veterinary and public health surveillance programs that is detailed, structured, transparent and based on surveillance concepts that are part of all types of surveillance programs. The proposed conceptual evaluation method comprises four steps: (1) text analysis, (2) extraction of the surveillance conceptual model, (3) comparison of the extracted surveillance conceptual model to a theoretical standard, and (4) validation interview with a surveillance program designer. This conceptual evaluation method was applied in 2005 to C-EnterNet, a new Canadian zoonotic disease surveillance program that encompasses laboratory based surveillance of enteric diseases in humans and active surveillance of the pathogens in food, water, and livestock. The theoretical standard used for evaluating C-EnterNet was a relevant existing structure called the "Population Health Surveillance Theory". Five out of 152 surveillance concepts were absent in the design of C-EnterNet. However, all of the surveillance concept relationships found in C-EnterNet were valid. The proposed method can be used to improve the design and documentation of surveillance programs. It complements existing surveillance evaluation methods. Conceptual evaluation is not a performance-oriented evaluation method and so it is particularly useful for surveillance programs with a valid conceptual framework but limited technical capacity and resources. Such programs would be penalized using existing performance-based evaluation methods. Applying conjointly the conceptual evaluation along with existing performance-oriented evaluation methods will better judge the worth of surveillance programs. We recommend developing a comprehensive surveillance evaluation framework for veterinary and public health surveillance programs that integrates all existing surveillance evaluation tools and provides an appropriate way to evaluate various types of surveillance programs.

A population health surveillance theory.

OBJECTIVES: Despite its extensive use, the term "Surveillance" often takes on various meanings in the scientific literature pertinent to public health and animal health. A critical appraisal of this literature also reveals ambiguities relating to the scope and necessary structural components underpinning the surveillance process. The authors hypothesized that these inconsistencies translate to real or perceived deficiencies in the conceptual framework of population health surveillance. This paper presents a population health surveillance theory framed upon an explicit conceptual system relative to health surveillance performed in human and animal populations. METHODS: The population health surveillance theory reflects the authors' system of thinking and was based on a creative process. RESULTS: POPULATION HEALTH SURVEILLANCE INCLUDES TWO BROAD COMPONENTS: one relating to the human organization (which includes expertise and the administrative program), and one relating to the system per se (which includes elements of design and method) and which can be viewed as a process. The population health surveillance process is made of five sequential interrelated steps: 1) a trigger or need, 2) problem formulation, 3) surveillance planning, 4) surveillance implementation, and 5) information communication and audit. CONCLUSIONS: The population health surveillance theory provides a systematic way of understanding, organizing and evaluating the population health surveillance process.

A scenario tree model for the Canadian Notifiable Avian Influenza Surveillance System and its application to estimation of probability of freedom and sample size determination.

In 2008, Canada designed and implemented the Canadian Notifiable Avian Influenza Surveillance System (CanNAISS) with six surveillance activities in a phased-in approach. CanNAISS was a surveillance system because it had more than one surveillance activity or component in 2008: passive surveillance; pre-slaughter surveillance; and voluntary enhanced notifiable avian influenza surveillance. Our objectives were to give a short overview of two active surveillance components in CanNAISS; describe the CanNAISS scenario tree model and its application to estimation of probability of populations being free of NAI virus infection and sample size determination. Our data from the pre-slaughter surveillance component included diagnostic test results from 6296 serum samples representing 601 commercial chicken and turkey farms collected from 25 August 2008 to 29 January 2009. In addition, we included data from a sub-population of farms with high biosecurity standards: 36,164 samples from 55 farms sampled repeatedly over the 24 months study period from January 2007 to December 2008. All submissions were negative for Notifiable Avian Influenza (NAI) virus infection. We developed the CanNAISS scenario tree model, so that it will estimate the surveillance component sensitivity and the probability of a population being free of NAI at the 0.01 farm-level and 0.3 within-farm-level prevalences. We propose that a general model, such as the CanNAISS scenario tree model, may have a broader application than more detailed models that require disease specific input parameters, such as relative risk estimates.