RESUMEN
Bovine respiratory disease (BRD) is a frequent disease in feedlot cattle, but little is known on the role of pen housing conditions. The objective of this research is to use a retrospective analysis with data from 10 U.S. feedlots to determine potential associations between BRD risk during the first 45 days after arrival with pen-level management factors including the number of water sources, shared water sources, and shared fence lines. Generalized linear mixed models were used to evaluate associations between management factors, cattle demographics, and BRD incidence. The effect of shared water sources on BRD risk was modified by arrival weight and cohort size (p < 0.05). Cattle with two water sources had lower BRD morbidity (5.55% ± 0.98) compared to cattle with one water source (8.80% ± 1.50) when arrival weight was 227 kg to 272 kg, while there were few differences in heavier weight cattle. Cattle with two water sources had lower BRD morbidity (3.11% ± 0.56) compared to one water (5.50% ± 0.10) when cohort size was 100−175 head, but there were no BRD morbidity differences when bigger or smaller cohorts were evaluated. Shared fence lines and water sources were associated with BRD risk; however, no biologically meaningful results were identified. The number of water sources was associated with BRD risk, and effects were modified by cohort size and arrival weight.
RESUMEN
Bovine respiratory disease (BRD) is the leading cause of morbidity in feedlot cattle. The ability to accurately identify the expected BRD risk of cattle would allow managers to detect high-risk animals more frequently. Five classification models were built and evaluated towards predicting the expected BRD risk (high/low) of feedlot cattle within the first 45 days on feed (DOF) and incorporate an economic analysis to determine the potential health cost advantage when using a predictive model compared with standard methods. Retrospective data from 10 U.S. feedlots containing 1733 cohorts representing 188,188 cattle with known health outcomes were classified into high- (≥15% BRD morbidity) or low- (<15%) BRD risk in the first 45 DOF. Area under the curve was calculated from the test dataset for each model and ranged from 0.682 to 0.789. The economic performance for each model was dependent on the true proportion of high-risk cohorts in the population. The decision tree model displayed a greater potential economic advantage compared with standard procedures when the proportion of high-risk cohorts was ≤45%. Results illustrate that predictive models may be useful at delineating cattle as high or low risk for disease and may provide economic value relative to standard methods.
