Abstract Bovine brucellosis remains endemic in Bangladesh and poses significant public health and economic risks. Traditional risk factor analyses predominantly employ binary outcome classifications, failing to capture the spectrum of serological states from healthy through latent infection to active disease. In this study, a Bayesian mixed-effects multinomial regression was applied to identify risk factors associated with latent and disease states, and a linear mixed-effects model was used to quantify milk yield loss. A cross-sectional study was conducted across 17 large-scale dairy herds in Bangladesh (n = 2,696). Milk samples from 2,696 lactating cows were analyzed using indirect ELISA (iELISA). On the basis of Bayesian mixture model thresholds, the animals were classified as healthy (S/P% 82.0). Bayesian mixed-effects multinomial logistic regression with herd-level random intercepts was implemented in Stan to identify risk factors. Milk yield was further analyzed using a linear mixed-effects model to quantify the magnitude of production loss associated with brucellosis status. Retained placenta was the strongest predictor for both latent (RRR = 5.82) and disease (RRR = 11.81) states. High parity (3rd–5th calf) was significantly associated with active disease. Anestrous increased latent risk but not disease status. Higher milk yield was inversely associated with both states. The model showed excellent discrimination for disease classification (AUC = 0.90) and moderate discrimination for latent status (AUC = 0.74). Compared with healthy cows, latent and diseased cows produced 0.53 L/day and 1.01 L/day less milk, respectively. These findings demonstrate distinct epidemiological risk patterns and measurable production losses associated with different brucellosis states, informing targeted surveillance and culling strategies.
Alam et al. (Wed,) studied this question.