The overall results of the survey, showing the distribution of gait scores prior to slaughter, are given in . The figures in were calculated using the gait scores of a flock, weighted by the size of the flock as given by the number of birds placed as chicks. The minimum and maximum values in show that there was considerable variation in walking ability between flocks, but overall, 27.6 per cent of birds represented by this survey had a gait score of 3 or above.
The estimated percentage of birds in the survey population within each gait score category.
shows the percentage of birds within each gait score category broken down by the five companies and by the first and second visits. There was a large amount of variation in the distribution of gait score in flocks between the different companies. Company 4 notably produced only 8.5 per cent of birds with gait scores of 3 and above at the first visit compared with 22.7 to 29.7 per cent for the other companies, and only 0.6 per cent of birds with gait scores of 4 and above compared with 1.3 to 4.2 per cent for the other companies. also shows the deterioration of gait over time for companies 1, 4 and 5 where a minimum of four second visits were made, the figures for the second visit to company 3 representing only one flock.
The percentage of birds in each gait score category by producer and by first and second visit.
The average gait score for a flock was modelled in terms of the levels of the risk factors recorded by the veterinary assessors. In addition to the assessment of husbandry variables as linear predictors of average flock gait score, the age at which the birds were assessed, and a seasonal effect, were included in the model. It was necessary to correct for both age and season at the time of assessment, before the effects of different husbandry practices could be properly investigated. The seasonal effect was expected to be cyclical, so was modelled as a sinusoidal curve to identify any cyclical variation in average gait score over a 12 month cycle. Both Sine and Cosine terms for month (January
2, etc) were used in the modelling process allowing us to look for an annual maximum and a minimum effect of season on gait score. A number of variables in the model were centred [Text S1
] by subtracting the mean of the variable from each of its measurements.
The most parsimonious multilevel model we obtained of the relationship between the average gait score of a flock, and the risk factors that were recorded within the survey, is shown in . When company was included as a fixed effect, rather than a random effect, in the model the parameter estimates shown in were not meaningfully altered and there were no significant variable/company interactions. Similarly, when the veterinary assessors were included in the model as fixed effects the parameter estimates were not substantively altered. The parameter estimates in the final model are shown and also details of which of the variables are centred. The mean, minimum and maximum of the predictor variables are shown in , including all those variables which were centred.
The parameter estimates, their standard error and significance for the model of average flock gait score.
Mean, minimum and maximum values of the continuous predictor variables in the model of average flock gait score.
When all other variables were held constant, there was a seasonal pattern to the average gait score of the flocks ( and ) with the lowest (best) gait scores occurring in March and the highest (worst) in September. The age at which the birds were assessed was important in determining gait score, with every extra day, across the range of 28 to 56 days, leading to an average daily deterioration in score of 0.048. Although each flock was visited close to slaughter when gait is known to be poorest, within the survey as a whole we were able to evaluate the effect of age on locomotion problems throughout the growth period because of the wide range of age at slaughter. A post-thinning visit was associated with an increased average gait score of 0.25 over and above that due to the age at which the birds were assessed, probably due to the effect of the stress of the first thin, and/or the preponderance of larger, faster-growing male birds remaining in the flock after thinning.
The modelled seasonal change in average flock gait score.
We found that a number of fundamental husbandry practices were significantly associated with average flock gait score and these are detailed below.
A major influence was bird genotype. Broilers worldwide are predominantly of two types, from either one of two major international breeder companies, labelled here A and B. Birds from both genotypes are sometimes reared together within one flock. For every percentage increase in Breed A birds in a flock, from between 0 to 100 per cent, there was a 0.0024 improvement in flock gait score.
Whole wheat is sometimes fed to broilers as part of their diet, predominantly to improve digestive function. For every percentage increase in dietary wheat fed, from 0 to 30 per cent, as measured during their third week of life, there was a 0.017 per cent improvement in flock gait score.
Broilers are reared under a wide variety of artificial lighting regimes. For every 1 hour increase in the daily period of darkness, across the range of 0 to 8.5 hours, there was a 0.079 improvement in flock gait score.
There has been debate about the importance of stocking density as an influence on bird welfare and locomotion 
. Within limits, putting as many birds in a house as possible for each rearing cycle will improve profitability. For every 1 kg/m2
increase in stocking density as measured at the time of the flock assessment, across a range from 15.9 to 44.8 kg/m2
, there was a 0.013 deterioration in flock gait score.
Antibiotics are routinely used during different stages of broiler rearing and their use can be quite difficult to quantify accurately. In our study farmers were simply asked if a flock had received antibiotic. A reply of ‘yes’ in this context meant that a flock had received an extra antibiotic treatment in addition to that which would be part of normal rearing practice. For the flocks for which the farmer had answered ‘yes’ flock gait score was reduced by 0.17.
Broiler feed is pelleted to minimise wastage and to increase the amount of feed that a bird can consume within a given time. In our study, veterinary assessors made a subjective judgement of whether the quality of the pelleting was good or poor (dusty/broken). For flocks with poor pelleting, gait score was on average 0.15 improved.
The magnitude of a parameter estimate given in is not an indicator of the importance of the association of a variable with an increase or a reduction in gait score. This is because the magnitude of the parameter estimate is dependent upon the measurement scale used (e.g. kg or g, hour or day). Further, it is not possible to rank objectively the variables in terms of importance as subjective considerations such as practicality and the perceived cost of altering the husbandry practice have to be considered. For example, the results show that using 100% Breed A birds compared with 100% Breed B birds would reduce average flock gait score by 0.24. To achieve a similar reduction in average gait score by manipulating stocking density alone would require a reduction in stocking density of 18.5 kg/m2. Despite these considerations there is value in using our model in a predictive capacity to show the improvement in average flock gait score that might be achieved on a hypothetical farm if each relevant variable attained the ‘best’ value actually recorded within the survey. shows the potential improvement from a baseline flock average gait score of 2.5, the middle of the range of possible scores. The Figure shows that no, one variable was associated with a large shift in average flock gait towards the score of 0, a perfect gait, but that the age at which birds are slaughtered appears to be of primary importance, followed by the hours of darkness that the birds are allowed.
Figure 2 A modelling exercise reveals the extent to which average flock gait score might be improved (i.e. lowered from a notional average score of 2.5) by adoption of the ‘best’ management practice for each variable that we recorded within our (more ...)