Activity Status

  • 17-18 March 2010 Project kickoff
  • Development of Decision Support functionality as part of WP4.
  • Demonstration of activities on this web-site
  • 1 April 2011 WP4 model and software development meeting
  • 26-27 May 2011 Project Meeting
  • 22 September 2011 WP4 meeting
  • 1 November 2011 WP4 modeling meeting
  • 2 February WP4 meeting
  • 24 February 2012 Midterm meeting
  • 14 May 2012 WP 4 meeting
  • 20 June 2012 Annual meeting
  • 6 July 2012 WP4 modeling meeting
  • 15 March 2013 WP4 modelling meeting
  • 15 September 2013 Annual meeting

Simple Vaccination Example - SimpleVac

This example was developed in corporation with Anna Garcia, Technical University of Denmark, National Food Institute.

The model deployed on this page is a simple example illustrating the use of (limited-memory) influence diagrams to support decision making on campylobacter vaccination of poultry.

The decision considered by the farmer is whether or not to and how to vaccinate a broiler. We assume he has two vaccines A and B to choose from (in addition to not vaccinating).

Figure 1: The SimpleVac model.

The farmer has to make a decision on whether or not and how to vaccinate a broilor. We assume that the farmer prior to the decision knowns something about the risk of campylobactor in the broiler, the effectiveness of each possible vaccination option, and the level of campylobacter in the broiler. This information relates, for instance, to properties of the farm such as history of campylobacter and preventive measures taken and seasonality. There are also factors unknown to the farmer at the time he makes the vaccination decision.

We assume the farmer is making his decision when the birds are two weeks of age and that the birds leave the farm at five weeks of age. The farmer wants to reduce the level of campylobacter when the birds leave the farm.

The different vaccination options none, vaccine A (oral delivery) or vaccine B (oral delivery) have different impact on the level of campylobactor at five weeks. If the farmer, for instance, decides not to vaccinate, then there is no impact and no reduction in level of campylobacter. If the farmer, on the other hand, decides to vaccinate using vaccine A and the effectiveness factors are high, then there will be a significant impact (reduction) in the level of campylobacter. The impact is assumed between 0 and two logs reduction in campylobacter level.

Figure 1 shows a graphical representation of the decision problem where

  • The box-shaped node is the decision on how to vaccinate: no vaccination, A oral (in water) or B oral (in water).

  • The oval-shaped nodes are random variables that describe the problem domain. There are nodes for the level of campylobacter before the vaccination decision at two weeks and at five weeks. One node represents the level of log-reduction at five weeks as a result of vaccination. The level of campylobacter is specified on a log scale from zero to six logs.

    The impact of the vaccination depends on the vaccination method and other factors that may be known or unknown at the time the vaccination decision is made.

    Finally, there are farm-specific risk factors that impact the level of campylobactor. Again, some risk factors may be known when the vaccination decision is made while others are unknown (to the farmer).

  • The two diamond-shaped are cost and reward functions relative to the decision. There is a cost associated with the vaccination and the reward (e.g., price of broiler) depends on the level of campylobactor at slaugthering.

    The cost function represents the cost associated with each vaccination option while the reward function represents the reward received by the farmer once the poultry is sold.

    The costs associated with A oral (in water) is 2 and with B oral (in water) is 5

    The assumption is that campylobactor free poultry can be sold at a higher price than poultry not free from campylobacter. We consider three levels of campylobactor 0-2, 2-4 and 4-6 logs. Under the assumption of the model there are three reward corresponding to the levels of campylobacter considered. The lowest reward has index 100 and each step adds an additional 15% to the reward producing rewards at 100, 115, 132.25.

Below there is a user interface for supporting the decision making under uncertainty made by the farmer. There are two observations to be entered prior to the decision. The risk factor observation has a causal influence on the level of campylobacter.

The farmer has to make a decision on the method of vaccination. The decision will cause a (uncertain) reduction in the level of campylobacter, which again impacts the campylobacter level at five weeks.

Finally, the expected costs and rewards (under the optimal strategy or selected strategy) are shown to guide the decision making.

Observations on risk factors and effectiveness factors related to the vaccination should be entered prior to making the decision on the method of vaccination. Select the appropriate values below.

Known Information at Decision (2 weeks) Unknown Campylobacter level (2 weeks)
Risk factors Logs level Probability Expected Utility
Observed level of Campylobacter
Effectiveness factors

Vaccination Options (2 weeks) Reduction in Campylobacter
Option Probability Expected Utility Logs reduction Probability Expected Utility

Select Vaccination Decision

Observations on risk, effectiveness factors and observed level of campylobacter at two weeks should be selected above deciding on the method of vaccination.

Campylobacter Level when leaving farm (5 weeks)

Logs Level Probability Probability Bar Expected Utility
The expected level of campylobacter is logs. The campylobacter level is computed as the weighted average using the middle points of the intervals listed in the table above. The expected value at five weeks is and the expect cost is . This means that the expected profit is .

Expected Cost and Reward

Expected Utility (EUR) Expected Utility (USD)
Cost + Reward

Placeholding text for USD-EUR exchange rate

Select scenario

Download current case (to store entered information on local disk)

Load a case file from local disk: