Seminar on Risk Analysis in Veterinary Science:
March 30, 2001 from 9 to 16 at SVS (room M1)

The IBIS-project and SVS invited to a one-day seminar on Risk Analysis in Veterinary Science. The seminar was open to everybody interested, and 55 persons attended the seminar: list of participants (pdf file).

The purpose of the seminar was to present the basic principles of risk analysis, to communicate some experiences with risk analysis from the IBIS-project and to initiate a general discussion on the applications of risk analysis in veterinary science. It was our hope that the seminar will be of interest both for researchers and administrators with a vague prior knowledge of risk analysis and for researchers working actively in the field.

The seminar was organised by the IBIS-project and arranged by Larry Paisley, Torben Grubbe and Henrik Stryhn.
On this page we keep the final programme and links to all presentations (at the bottom line of abstracts).

Programme

Abstracts

Larry Paisley: What is risk analysis

This presentation will consist of a discussion of:

- the(a?)definition of risk analysis (RA),

- who does RA?

- why has RA become fashionable?

- types of RA,

- components of RA,

- RA definitions,

- steps in quantitative risk analysis.

Henrik Stryhn: The meaning of a quantitative risk estimate and its uncertainty
Quantitative risk estimation is typically based on simulation from models involving one or several (probability) distributions. Initially we explain the use of distributions, in particular we draw the distinction between system variability and lack of knowledge. By way of examples we next review some common choices of distributions (and their justifications) and ways to combine them in a model. This leads to a discussion of the assumptions of a quantitative risk assessment and of the interpretation of the calculated distributions of risk. (pdf file)

Larry Paisley: Simulation modelling of BSE surveillance
Monte Carlo simulation models were developed to evaluate targeted surveillance programs for BSE. The model assumptions, data, structure and results for the system once proposed in Denmark will be presented. (Powerpoint file)

Henrik Stryhn: Quantitative risk assessment - what can it (not) do for you?
This introduction to debate attempts to formulate some questions that can be used as guidelines for, when a problem is well suited for a quantitative risk assessment and when it is not. (pdf file)

Jens Strodl Andersen: Qualitative risk analysis - From a veterinary engineering perspective
(Powerpoint file)

Torben Grubbe: Qualitative risk assessment - Import of cattle to Denmark
The Danish Veterinary Institute for Virus Research has made a qualitative risk assessment, which examines the risk of introducing viral diseases into Denmark via import of cattle. The risk assessment is conducted in accordance with the agreement on the application of sanitary and phytosanitary measures (the SPS agreement) and the Office International des Epizooties (OIE) guidelines for import risk analysis. The risk assessment is initiated with a hazard identification, where the cattle viruses are identified dichotomously as potential hazards or not, based on disease occurrence and control in Denmark. Following the hazard identification risk assessments are performed on each of the potential hazards. Each risk assessment results in a qualitative risk estimate by considering the risk of introduction, the risk of spread and consequences following either of these events. The aim of this presentation is to give a methodological description of import risk assessment, and give a short introduction to the SPS agreement and the OIE guidelines.

Preben Willeberg: Risk analysis in the veterinary services
(Powerpoint file)

Helle Sommer: Risk assessment on Campylobacter jejuni in chicken products - The slaughterhouse part
In January 2001 a first draft of a quantitative risk assessment on Campylobacter jejuni in chicken products has been revealed. The risk assessment was ordered by the Danish Veterinary and Food Administration as part of a strategy to control pathogenic microorganisms after the principles for Food Safety Risk Analysis. The work has shown that it is realistic to expect that at least a fraction of the human exposure to Campylobacter originate from Campylobacter in chickens. The outcome of the risk modelling reveals that important factors for human exposure to Campylobacter are the broiler flock prevalence and hence, the prevalence in retail chickens, the Campylobacter concentration on positive chicken products, and the extent of cross-contamination in private kitchens during food handling.

To quantify the risk, two sub-models have been developed, one describing the transfer and spread of Campylobacter through a chicken slaughterhouse and another dealing with the transfer and spread of Campylobacter during food handling in private kitchens. Due to the limited time only the slaughterhouse part will be presented in more details.

The quantitative model deals with prevalence and concentration of Campylobacter on the skin surface of whole chilled and frozen chickens. Through the different processes in the slaughterhouse the concentration of a given chicken goes up and down and the status can change from negative to positive or visas versa. Prevalence data are available at the entrance of the slaughterhouse and concentration data are available at the entrance of the slaughterhouse plus after certain selected processes. From these data the initial state (prevalence and concentration) of the chickens represented by distributions are determined plus the changes of the state over the different processes in the slaughterhouse. Prior of determine these input distributions the scanty data material are being analysed for equality of variances and for equality of the means. Moreover is the variance components determined in order to separate uncertainty and variability for the initial input distribution and the distributions describing the changes in concentration over the different processes. By separating the variability and the uncertainty it is possible to optimise future data sampling. The separation of variability and uncertainty is also used in order to model the changes in concentration over the different processes in the slaughterhouse - the changes in concentration over a given process is represented by a distribution with a mean equal to the change in concentration before and after the process and a variance equal the variance component.

However, this way of modelling the changes of concentration causes the variance at the end of the slaughterhouse to become larger than what is observed. Cross-contamination within Campylobacter positive flocks is not accounted for in the present model. This homogeneous effect leads to a smaller variance estimate than modelled. Another way of modelling the changes in concentration should be examined. In this presentation alternative ways will be discussed. (Powerpoint file, Additional Powerpoint slide)

Tine Hald, David Vose, Henrik C. Wegener: Quantifying the contribution of animal-food sources to domestic and sporadic human salmonellosis
In order to get a better understanding of the mechanisms behind the dynamics in the occurrence of Salmonella infections in humans, a quantitative risk assessment model was developed. The model quantifies the importance of the major animal-food sources of domestic and sporadic cases of human salmonellosis. Data from the national Salmonella surveillance of animals, foods and humans in Denmark in 1999 was used for demonstration. For modelling purposes, we applied a statistical technique called Bayesian Monte Carlo, which combines Bayesian inference and Monte Carlo simulation.

First, the number of domestic and sporadic cases caused by different Salmonella types was estimated based on the observed data, i.e. the registered number of cases of human salmonellosis. The principle was then to compare the occurrence of these estimated number of cases with the prevalence of the Salmonella types isolated from the different animal-food sources, weighted by the amount of food source consumed. However, the number of people being infected by a particular Salmonella type in a particular food source may depend on additional factors related to the Salmonella type and food source in question. Therefore, we introduced a multiparameter prior, which accounted for the presumed but undefined differences between Salmonella serotypes and food sources with respect to causing human Salmonella infections. A Poisson likelihood function was used for the probability of observing the actual number of human cases given the prevalence in the various food sources and the amount of food consumed. The maximum likelihood estimators (MLEs) were determined for the unknown parameters and uniform priors were constructed centred around the MLEs. The joint posterior distribution was estimated by the Bayesian Monte Carlo technique.

Based on the final multidimensional model, the number of human cases attributable to each food source was estimated. The most important sources were found to be table eggs, where to approximately 54% (90% C.I.: 51.4 - 56.1%) of the cases could be attributed. This was followed by domestically produced pork and imported poultry, which comprised around 9% (90% C.I.: 8.1 - 9.7%) and 8% (90% C.I.: 6.5 - 9.9%) of the cases, respectively. Sources such as domestically produced beef, broiler meat and duck meat had only a minor impact (< 3% each). We believe that the described method may prove to be an alternative to the "traditional" stable-to-table risk assessment, which often involves making a large number of assumptions with very variable plausibility. (Powerpoint file)