Data - Bayesian Logistic Regression Model for bovine brucellosis cervical bursitis risk factors
Description
Bovine brucellosis is an endemic zoonosis prevalent in Brazil. Cervical bursitis is a brucellosis suggestive lesion found in bovine carcasses. This work aimed to study risk factors for brucellosis seropositivity and for Brucella abortus isolation in bovine carcasses with cervical bursitis from the states of Mato Grosso, Pará and Tocantins – Brazil, through Bayesian Logistic Regression Models (BLRM). Results of the diagnostic tests were the dependent variable in three distinct BLRM (model 1 serum aglutination test with 2-mercaptoethanol SAT/2ME, model 2 – rose Bengal RB and model 3 – isolation and identification Bac). Origin of the samples, sex and age of the animals were the explanatory variables. Females were 18.54 (95%CrI 4.727 – 85.98), 16.03 (95%CrI 4.198 – 71.57) and 5.796 (95%CrI 1.551 – 32.13) times more likely to be seropositive for brucellosis or had Brucella abortus isolated from the lesions than males, in models 1, 2 and 3 respectively. There was no risk associated with the age of the animals. The origin of the animals was a factor that reduced the risk for brucellosis in two models. The information provided in this study can support risk-based sampling for brucellosis vigilance and risk analysis.
Files
Steps to reproduce
The samples were collected by the Brazilian Federal Inspection Service (SIF) during the routine meat inspection in slaughterhouses in the states of MT, PA and TO – Fig.1, from 2010 to 2013. Once the lesions were found they were removed from the carcasses, then the blood was collected from the brachial vein, and allowed to clot at room temperature for serum collection. The paired samples were packaged in sterile plastic bags and tubes, respectivelly, identified and freezed -20°C until shipping to the Federal Laboratory of Agricultural Defense of Minas Gerais (LFDA-MG) to be processed. All samples were sent together with a form containing information about the date of collection, origin, breed, sex and age of the animals. Rose bengal (RB) and serum aglutination test with 2-mercaptoethanol (SAT/2ME) were performed according to Brazilian regulations (Brasil, 2017a). SAT/2ME cut-off was 1:25 (Brasil, 2017a). Isolation and identification tests (Bac) were carried out according to Alton et al. (1988) and Orzil et al. (2016). Bayesian logistic regresion models (BLRM) were condutect in OpenBUGS software, version 3.2.2 (Lunn et al., 2009; Ntzoufras, 2009). The results of the diagnostic tests were the dependent variable in three distinct models (model 1 - SAT/2ME, model 2 - RB and model 3 - Bac). The origin of the samples (state 1 PA, state 2 MT – reference, state 3 TO ), the sex (male – reference, female) and the age (<36 months – reference, ≥36 months) of the animals were used as the explanatory variables for the regression models. Linear regression on logit and likelihood function for each data point were as follow: logit(p[i]) = alpha + b.sex*sex[i] + b.age*age[i] + b.state1*state1[i] + b.state3*state3[i] Dpos[i] ~ dbern(p[i]) OpenBUGS run three chains for each model, with 100,000 iterations for each chain, being the first 5,000 iterations discarded as burn-in. Iterations were taken at every ten values generated by the models to minimize autocorrelation. Convergence of the chains was checked by visual inspection of OpenBUGS trace and history, and by Gelman-Rubin diagnostic plots. Deviance Criterion (DIC) and the number of parameters effectively estimated by the model (ρD), was used to check models’ goodness of fit to the data. Non-informative priors were specified for each parameter of interest (~dnorm (0.0, 1.0 E-4)). Posterior estimates of regresion coeficients were anti-log transformed (anti-loge) into odds ratios (OR) to allow better understanding and interpretation of independent variables (Noordhuizen et al., 2001). .