Multivariate techniques used in a cross-sectional study of feather pecking and vent pecking in laying hens in alternative systems

The aim of this cross-sectional study was to identify and quantify management associations with feather pecking and vent pecking in adult laying hens in alternative systems on commercial farms in Great Britain. The data were derived from a postal questionnaire carried out in 1998. Of 200 farmers, 56.1% reported that feather pecking had occurred in the last depopulated flock, 36.9% reported vent pecking and 33.3% had both problems. The two outcomes, feather pecking and vent pecking, were compared with management exposures using univariate statistics and associated risk factors where P less than or equal to0.05 were then tested in two logistic regression models. The final feather pecking model contained eight risk factors: <50% of the flock used the outdoor area on a sunny day, the diet was changed more than twice, inspections were done by one person, no loose litter left by the end of lay, the hen house temperature was <20 degreesC, lights were turned up when the flock was inspected and bell drinkers were used. An increased risk of vent pecking was noted where dim lighting was used to stimulate the use of nest boxes, the diet was changed more than twice during the laying period, hanging bell drinkers were present and egg laying started before 20 weeks of age. The factors from both logistic models were then tested by survival analysis. In the vent pecking Cox proportional hazards regression model, all four factors significant in the logistic regression model remained significant. The three risk factors for feather pecking that remained in the Cox proportional hazards regression model were: <50% of the flock used the outdoor area on a sunny day, inspections were done by one person, lights were turned up when the flock was inspected. The outcome of a stratified Kaplan-Meier analysis suggested that after eliminating the effects of the risk factors from the Cox proportional hazards model, the proportion of unaffected flocks could be increased dramatically. The results indicated that survival analysis identified similar, but not identical risk factors, as the logistic regression modelling. Survival analysis permitted the testing of the effect of removing these risks on the reduction in hazard. It also provided a useful alternative analytical technique for indicating which risk factors delayed the occurrence of an outbreak of vent or feather pecking.