Abstract
In order to support decision making on how to most effectively improve broiler welfare an innovative expert survey was conducted based on principles derived from semantic modelling. Twenty-seven experts, mainly broiler welfare scientists (n = 20; and 7 veterinarians), responded (response rate 38%) by giving welfare scores (GWS, scale 0–10) to 14 benchmarking housing systems (HSs), and explaining these overall scores by selecting, weighing and scoring main welfare parameters, including both input and output measures. Data exploration followed by REML (Linear Mixed Model) and ALM (Automatic Linear Modelling) analyses revealed 6 clusters of HSs, sorted from high to low welfare, i.e. mean GWS (with superscripts indicating significant differences): 1. (semi-natural backyard) Flock (8.8a); 2. Nature (7.7ab), Label Rouge II (7.4ab), Free range EU (7.2ab), Better Life (7.2ab); 3. Organic EU (7.0bc), Freedom Food (6.2bc); 4. Organic US (5.8bcd), Concepts NL (5.6abcdef), GAP 2 (4.9bcd); 5. Conventional EU (3.7de), Conventional US (2.9ef), Modern cage (2.9abcdef); 6. Battery cage (1.3f). Mean weighting factors (WF, scale 0–10) of frequently (n> = 15) scored parameters were: Lameness (8.8), Health status (8.6), Litter (8.3), Density (8.2), Air quality (8.1), Breed (8.0), Enrichment (7.0) and Outdoor (6.6). These did not differ significantly, and did not have much added value in explaining GWS. Effects of Role (Scientist/Vet), Gender (M/F) and Region (EU/non-EU) did not significantly affect GWS or WF, except that women provided higher WF than men (7.2 vs 6.4, p<0.001). The contribution of welfare components to overall welfare has been quantified in two ways: a) using the beta-coefficients of statistical regression (ALM) analyses, and b) using a semantic-modelling type (weighted average) calculation of overall scores (CalcWS) from parameter level scores (PLS) and WF. GWS and CalcWS were highly correlated (R = ~0.85). CalcWS identified Lameness, Health status, Density, Breed, Air quality and Litter as main parameters contributing to welfare. ALM showed that the main parameters which significantly explained the variance in GWS based on all PLS, were the output parameter Health status (with a beta-coefficient of 0.38), and the input parameters (stocking) Density (0.42), Litter (0.14) and Enrichment (0.27). The beta-coefficients indicated how much GWS would improve from 1 unit improvement in PLS for each parameter, thus the potential impact on GWS ranged from 1.4 welfare points for Litter to 4.2 points for Density. When all parameters were included, 81% of the variance in GWS was explained (77% for inputs alone; 39% for outputs alone). From this, it appears that experts use both input and output parameters to explain overall welfare, and that both are important. The major conventional systems and modern cages for broilers received low welfare scores (2.9–3.7), well below scores that may be considered acceptable (5.5). Also, several alternatives like GAP 2 (4.9), Concepts NL (5.6), Organic US (5.8) and Freedom Food (6.2) are unacceptable, or at risk of being unacceptable due to individual variation between experts and farms. Thus, this expert survey provides a preliminary semi-quantified decision-support tool to help determine how to most effectively improve broiler welfare in a wide range of HSs.
Highlights
Efforts to address positive changes in animal welfare should follow the principles underlying effective altruism [1, 2], namely to maximise efficiency in doing good
The contribution of welfare components to overall welfare has been quantified in two ways: a) using the beta-coefficients of statistical regression (ALM) analyses, and b) using a semantic-modelling type calculation of overall scores (CalcWS) from parameter level scores (PLS) and WF
Both input and output parameters were included in the survey in the current study because some scientists differ almost fundamentally regarding the issue of which types of parameters should be used for welfare assessment, for example that only animal-based output parameters are valid in principle [26] or that it is essential to use a mix of input and output measures [27]
Summary
A semantic modelling methodology has been developed to assess overall welfare, expressed as a score on a scale from 0 to 10, and used to prioritize welfare aspects in a systematic, formalised and science-based way [3, 4] This methodology, which involved empirical as well as expert validation surveys, has been applied to various farmed species, including pigs (Sow WELfare Model (SOWEL), [5], PIGTAIL on tail biting [6, 7] and RICHPIG on pig enrichment [8]), laying hens [9], dairy cattle [10], calves [11]) and farmed salmon [12,13,14]. No semantic model has been developed to assess the welfare of broiler chickens, and welfare scores for (attributes of) housing systems are lacking
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
