Evaluation of Welfare Methods

The objective of WP7 is to assess the overall strengths and weaknesses of each defined housing system for laying hens, as well as the overall welfare impact of each housing system.

In selecting the parameters to consider, the five freedoms provided a useful checklist to ensure that a range of measures was chosen. The five freedoms are popularly used as a baseline for animal welfare assessment and indicate that to experience good welfare an animal should be free from:

  1. Injury, disease and pain
  2. Fear and (dis)stress
  3. physical discomfort
  4. hunger and thirst (its diet should provide normal health and vigour)
  5. and should be free to express normal behaviour.

The LayWel partners spent some time carefully selecting the range of science-based outcomes used in the database. These therefore form the basis of our assessment in this WP7. However, as the database includes only welfare indicators, we have used the literature to incorporate supplementary information on bird choice where possible.

The two main methods we have used for integrating welfare are:

1. Database analysis

Data characterisation and selection
The first step was to characterise the data available in the LayWel database. The LayWel project was not intended to support many new research projects and this, together with the short timescale, has meant that the data available to evaluate have primarily been those from recent and ongoing studies. These studies were not designed to provide a comprehensive and balanced dataset. Nonetheless the database contains records from almost 1.2 million laying hens in seven different European countries.

Table 7.4 indicates that substantial data are available for three systems: small furnished cages (FC-small) with group sizes of up to 15 hens per cage; multi-tiered aviaries where nestboxes are separate from the tiers (MT-NN); and single tier non-cage systems (ST-NC). Data from 19 or more flocks are also available for conventional cage (CC) and a free range system (ST-NC+FR). Most of the non-cage system data came from commercial flocks. Whereas such data appear to be more relevant for the poultry industry, the lack of control of variables can affect the outcomes. For example, changes in diet may not be controlled and these have been shown to be one of the risk factors for outbreaks of feather pecking in epidemiological evaluation of commercial flocks (Green and others, 2000). For some of our data analysis, combinations of system types have been used, providing data from many flocks.

Table 7.4 Data available by housing system

SystemNo. of flocksNo. of replicatesNo. of partnersCommercial flocksExperimental or trial flocks
Conventional cage (CC)19326514
MT-IN (Multi-tiered
integrated nest)
MT-NC (Multi-tiered
non cage)+CV+FR
MT-NN (Multi-tiered
non-integrated nest)
ST-NC (single tier-non cage)54545522
ST-NC+CV+FR 77170

FC denotes Furnished cage
CV denotes covered veranda (‘Winter garden’)
FR indicates unroofed outside area available for hens to range on

Overall analysis of database
As already indicated, the datasets were not designed to integrate for overall analysis. It is particularly important to appreciate that a line of data in the database may represent anything from the mean of a single replicate from a small experimental trial of relatively few hens up to the mean of six replicates of a large-scale commercial evaluation. It has not been feasible to weight the data accordingly because of differences in experimental designs. It is therefore important to be very cautious in drawing conclusions from comparisons of these data as a whole. Whilst in general there was agreement that overall statistical analysis was not appropriate, we have performed a limited amount of modelling, for mortality and plumage condition, for which there were the most data as they were recorded in almost all studies. The next most recorded variables were egg production and feed intake so a brief overview of these is included, although they are not key indicators of welfare.

Data were analysed using the GLM procedure in SPSS. A number of potentially important explanatory variables were confounded either partially (e.g. country with breed) or totally (e.g. hen colour with breed). Thus an iterative approach was taken involving addition and removal of explanatory variables until the best-fit model was obtained, assessed by the magnitude of the R-Squared value. Judgement was also required as to the number of levels at which to categorise explanatory variables. Housing system was categorised at 6 levels: conventional cages, single tier systems, multi-tier systems, furnished cages (small groups), furnished cages (medium groups) and furnished cages (large groups).

With a large and more balanced dataset it would be worth employing multi-level or hierarchical analysis techniques e.g. to examine the effects of breed within housing system within country. This was not possible with the data available.

2. Welfare risk assessment

The traffic light approach
The comparison of welfare across systems for The European Food Safety Authority took the approach of estimating the risk for welfare of important factors for each system (EFSA, 2005). In this LayWel study we have built on that by using a ‘traffic light’ approach to compare welfare outcomes and risks to good welfare across systems. This uses colour coding to indicate the probable risk for welfare based on data (principally from LayWel) that has been screened for quality, supported by background information from preference tests and expert opinion. This is a simplification of benchmarking as a means of assessing welfare (e.g. Huxley and others, 2004).

In essence our system uses green to denote the probability of good or satisfactory welfare or performance. In this report we shall use it to denote a low risk of poor welfare. Green coding may not always represent optimal welfare, but invariably the probability of at least good welfare, based on results available to date.

Orange denotes a medium risk of poor welfare. We also use it to indicate factors that are highly variable (within systems or between farms).

Red indicates a high risk of poor welfare. Often the housing system simply does not provide the facilities required - or the characteristics are such that there is very high risk of undesirable outcomes without extreme vigilance, as indicated by recent data from LayWel and other studies.

Please note that the risk of poor welfare is being indicated. Thus in many cases it is possible for flocks within housing systems where an indicator is coloured red or orange to actually achieve good welfare. The system is not labelling systems as ‘good’ or ‘bad’ in terms of welfare.

The colour coding enables a visual integration of relative welfare risk to the whole flock, but does not weight the different outcomes nor indicate the suffering of small numbers of individual hens. Duration, severity and proportion of flock affected are the main issues considered in colour coding. In the main, the welfare assessment is on a flock basis. The table of welfare risks and outcomes enables visual integration within and between systems.

Current best practice
For some key welfare indicators in each category of housing system we have also compiled a table of numerical data representing the best results, where available. These indicate what is currently achievable in terms of improved welfare with better design, maintenance, management, husbandry and genotypes in the various housing systems. These ‘best figures’ may coincide with industry best practice; particularly where data are available from commercial farms (several examples in the LayWel database). In order that the figures are representative, we have mainly selected data that have been published in peer-reviewed journals or those data from the LayWel database that represent relatively large and replicated studies. Alternatively we have given small ranges or limits indicating what has been achieved in well-designed but unpublished studies. Whereas the preferred format for such data is medians and ranges, means are more commonly reported and these have been used.

Following these summarising tables is a short list and discussion of some of the main welfare indicators based on the conclusions of the other WPs. We also discuss the strengths and weaknesses of each principal housing system.