Diagnosing bronchopneumonia in calves. A latent class analysis of the accuracy of clinical examination, thoracic ultrasonography and bacterial culturing from bronchoalveolar lavage fluid

Accurate detection of bronchopneumonia in calves is essential for management under field conditions. This study aimed to evaluate the diagnostic accuracy and agreement of three methods for detecting bronchopneumonia: clinical examination including auscultation (CE), thoracic ultrasonographic examination (TUS) (with positivity threshold ≥1 cm consolidation depth), and bacterial culturing from bronchoalveolar lavage fluid (BAL) (with a dominant or pure culture of ≥10 colonies of Pasteurella multocida, Mannheimia haemolytica, and/or Histophilus somni). Bayesian latent class analysis and Cohen's kappa were used to analyze data from 155 Norwegian calves in 15 herds with enzootic pneumonia, including two subpopulations, dairy and fattening herds, with expected difference in disease prevalence. TUS had the highest estimated sensitivity (Se) (89.5% [95% Bayesian credible interval 74.0-98.3%]), closely followed by CE (83.1% [68.7-93.9]), while BAL had the lowest Se (43.3% [29.1-62.7]). Bacterial culturing from BAL samples had the highest specificity (Sp) (98.6% [92.9-99.9]), followed by CE and TUS which had similar Sp of 77.5% (63.5-93.4) and 76.4% (62.1-92.1), respectively. Despite BAL's high Sp, its invasiveness and long turnaround time make it more suitable for identifying agents and conducting antimicrobial susceptibility testing than detecting bronchopneumonia. A serial interpretation of CE and TUS resulted in a Se of 74.4% (58.5-88.7) and a Sp of 94.4% (89.2-98.8). In calves from herds with enzootic pneumonia, TUS reduces the number of false negatives to a minimum; however, CE is nearly as sensitive. Conversely, a serial interpretation of CE and TUS increased Sp and reduced false positives, potentially supporting more targeted decisions under field conditions.