Decision support system with semantic model to assess the risk of tail biting in pigs

Abstract

Tail biting is a multifactorial problem with important welfare as well as economic consequences. Different stakeholders in the pig production chain, such as farmers, consumers and policy makers are interested in reducing the level of tail biting, because tail biting may affect productivity, profit and animal welfare. To help assess the expected level of tail biting under specific husbandry conditions we constructed a computer-based decision support system called PIGTAIL, which is a relational database containing a tail biting model based on 133 statements derived from 61 scientific publications. A formalised procedure previously used for modelling animal welfare was used to construct the model. PIGTAIL contains 28 attributes that describe factors that have been associated in scientific research with different levels of tail biting in different housing conditions. Such factors include the provision of substrate, the docking of tails, sex and breed as well as feeding-, health- and climate-related factors. PIGTAIL takes a description of a housing and management system as input and produces a score for the overall risk of tail biting as output. This is a weighted average attribute score, where weighting factors are derived from different measures of tail biting performance, such as tail biting behaviour and tail wounds. The decision support system clarifies the reasoning steps involved in tail biting risk assessment and allows upgrading when new scientific information about tail biting becomes available.

Introduction

Tail biting is frequently observed in commercial pig production. It can be described as the sucking and chewing of the tail of another pig. It leads to damage of the tail and, if unchecked, can escalate into cannibalism. Such damage is painful for the victim pig, its growth rate is reduced, it may need veterinary treatment and the carcass may be downgraded at the slaughterhouse because of infection and spinal abscesses. Tail biting occurs to a greater or lesser extent in all countries and in all housing systems. Estimates of prevalence range from under 1% to over 10% at the national level, but individual farms may have figures as high as 20% and over 60% of farms have been reported to have experienced this problem (Chambers et al., 1995, Hunter et al., 1999). Considering the large numbers of pigs fattened annually, tail biting represents a major economic and welfare problem for the pig industry and, as yet, there is no solution.

The traditional management strategy to reduce the problem is tail docking, a procedure whereby the end of the piglet’s tail is removed without anaesthetics, leaving about 2 cm of tail stump. In the majority of countries this is still carried out as a routine procedure. It increases the long-term sensitivity of the tail stump resulting in a lower response threshold to being chewed on the tail (Simonsen et al., 1991). However, there is no difference in the frequency with which docked tails or undocked tails are seen in the mouth of another pig (Simonsen, 1995). A UK study reported that 3.1% of pigs with docked tails showed evidence of tail wounds due to tail biting (Hunter et al., 1999). Even if this is a reduction from the 9.2% of pigs with intact tails, it demonstrates that the tail docking procedure does not eliminate the problem. Thus, the common practice of tail docking is not only painful, but also only partially effective in controlling tail biting.

Many factors appear to affect tail biting and these have been studied for over 30 years (e.g. van Putten, 1968, Beattie et al., 2001, Schroder-Petersen and Simonsen, 2001, Guy et al., 2002). A supplementary approach to controlling tail biting could be the development of a tool that can help assess the risk of tail biting on a (existing or novel type of) farm or in a unit of pigs based on available scientific knowledge about the different risk factors.

Recently, a computer-based decision support system with a semantic model was developed for overall welfare assessment (Bracke et al., 2002a, Bracke et al., 2002b). The modelling of welfare followed certain procedural rules in order to safeguard the quality of the assessment. That paper explains how scientific knowledge is systematically incorporated into a list of welfare relevant factors (attributes), how the attributes are weighted and how the model applies to a predetermined domain of housing systems. The methodology used to construct a semantic model for welfare assessment may also be applicable to other multifactorial problems about which a substantial amount of scientific knowledge is available while not (yet) allowing the construction of stochastic models. Semantic models could improve the application of findings of applied animal behaviour research. Like welfare, tail biting is a multifactorial problem for which we constructed a decision support system called PIGTAIL. PIGTAIL was designed to assess tail biting risk based on presently available scientific knowledge and in such a way that the model could be upgraded when new scientific information would become available. PIGTAIL was not designed to make economic decisions or to assess welfare, although it could be incorporated as a module in economic and/or welfare models to assist in the assessment.

The aim of this paper is to describe PIGTAIL and the procedure used to construct its tail biting model. The subsequent paper will report a ‘validation’ of the model with an analysis of the literature.