Feather pecking (FP) is a multifactorial problem in all housing systems.
Using a case-control design we wanted to determine a set of variables
which may be decisive for a flock to become case or control. At the same
time, we expected that the prevention of feather pecking not only
depends on certain factors, but also quantitatively on the extent of
compliance with recommendations. Data from three cross-sectional studies
were pooled, from which 165 flocks were allocated to cases (FP problem
flocks) or controls (no FP problem). Control flocks had at least 98% of
hens with a very good or nearly complete feather cover whereas in case
flocks 10% or more of the hens had highly damaged feathers or
featherless areas ≥ 5 cm² in at least one body region. From 32 potential
preventive factors that had been recorded in all data sets, 18 factors
passed univariable pre-selection and were presented to forward logistic
regression analysis. The resulting model for 137 flocks (due to missing
values) explained 41% of the variance, correctly classified 77% of cases
and comprised four variables with an effect size of f = 0.8. A
higher stocking density increased the likelihood of a ‘FP-problem’
whereas the presence of wooden perches and a littered veranda lowered
it. Unexpectedly, a higher drinking place/hen ratio also predicted a FP
problem. The results concerning wooden perches and drinking place/hen
ratio might be due to indirect effects and should be further
investigated. Non-FP flocks complied on average with 46.5% of
recommendations (from a list of 13 factors), which was greater than
FP-flocks (42.5%, P =  .036, U = 2537.500, n = 165, Mann-Whitney-U test, d_Cohen = 0.327).
We conclude that the number of fulfilled recommendations as well as the
combination of specific measures such as provision of a covered veranda
with dry litter or reduced stocking density are important to prevent
FP.