Bayesian estimation of herd-level prevalence and risk factors associated with BoHV-1 infection in cattle herds in the State of Paraiba, Brazil

A cross-sectional study was carried out to estimate the animal- and herd-level prevalence of bovine herpesvirus 1 (BoHV-1) infection in cattle in the State of Paraiba, and to identify risk factors associated with herd-level infection. The state was divided into three sampling strata, and for each stratum, the prevalence of herds infected with BoHV-1 was estimated through a two-stage sampling survey carried out from September 2012 to January 2013. In total, 2443 animals were sampled from 478 herds. A virus-neutralization test was used for BoHV-1 antibody detection. A Bayesian latent-class model was used to describe the data, taking into account imperfect diagnostic test characteristics and the non-independence of test results from animals within the same herd, and using a dynamic within-model risk factor selection method based on indicator variable selection. The adjusted herd-level prevalence was estimated to be 84% (95% CI: 80-88%) for the State of Paraiba, and the animal-level prevalence was estimated to be 73% (95% CI: 66-84%). Only five of the available risk factors were used by the model, with the three most influential being disposal of aborted foetuses (3.78, 95% CI: 1.11-13.85), sharing resources with other farms (3.0, 95% CI: 1.1-8,6), and a herd size of > 23 animals (2.5, 95% CI: 1.1-6.0). Our findings suggest that the animal- and herd-level seroprevalence of BoHV-1 infection in the State of Paraiba is high. While some risk factors such as herd size and sharing resources were identified as risk factors for BoHV-1 infection, these risk factors are initially likely to be of only minor relevance in a control programme due to the extremely high prevalence of infected farms. However, the results are relevant to the risk of reintroduction of disease on farms that have previously eradicated the disease.