Statistical analyses:
To evaluate the strength and direction of the associations of the horse
and stud farm management risk factors with the seropositivity to BoDV by
IFAT in the group of 235 SP breeding horses sampled between 2011 and
2013, a multivariable logistic regression model was carried out, after
checking the independence of the stud farm factor. Odds ratios (ORs)
with their corresponding 95% confidence intervals (CIs) were used to
measure the associations. All analyses were performed with the software
Small STATA version 12.0.
Possible risk factors were included as individual variables. The
variables tested (n= 24) with their frequencies are listed in the
Supplementary Data. All variables showing a P value < 0.25 in
the univariable screening analysis of variables controlling for
important confounders (age) were selected for inclusion in a
multivariable model, which included those confounders. The multivariable
model was built using a stepwise backward variable selection procedure
in which variables with a P < 0.05 were retained in the model,
or where their removal resulted in a significant change in the effect of
other variables indicating confounding. Collinearity between variables
was checked and excluded. To improve the goodness of fit, biologically
plausible first-order interactions were determined between significant
and confounders variables. The possible multivariable significant models
were compared to each other by the Akaike Information Criterion (AIC)
and the Bayesian Information Criterion (BIC) (Dohoo et al., 2003). The
Hosmer-Lemeshow goodness-of-fit test (Hosmer, 2000) was used to evaluate
the fitting of the proposed model and the area under the receiver
operating characteristic (ROC) curve was also calculated.