Using individual lung ultrasonography at the group level to define treatment thresholds in large groups of calves for tailoring antimicrobial use: A simulation approach

Bronchopneumonia is a major reason for antimicrobial use in young calves, especially in large groups of commingled calves. The clinical diagnosis may be challenging, and lung ultrasonography (LUS) has emerged as an effective calf-side diagnostic test. However, little is known about the applicability of this test at the group level in deciding whether a large group of calves should be treated. The objective of this study was to use a simulation-based assessment of the practical use of LUS to determine the optimal combination for deciding when to treat a group of calves. Therefore, we determined different plausible scenarios. We assessed the performance of LUS as a group test when assessing a random number of calves (N = 12, 20, and 30) and defining the group positive test based on the number of calves with lung consolidation (k; from 1 to 6), the true prevalence of bronchopneumonia in the tested population (TP; from 0.05 to 0.3), and the prevalence of positive herds (HTP; from 0.1 to 0.7). Accuracy of LUS at the individual level (sensitivity (Se) and specificity (Sp)) was assessed using a literature review and modeled using a beta distribution. Two thousand iterations of Se and Sp were then obtained, accounting for the negative correlation between these two parameters (known as the threshold effect). Then, group accuracy parameters (Herd Sensitivity, HSe; Herd Specificity, HSp) were obtained based on Se, Sp, k, N, and TP. The misclassification cost term (MCT), which accounts for different plausible cost ratios (r) between false negative and false positive groups, was also obtained to assess the robustness of the findings. Finally, the group-positive and negative predictive values (HPPV and HNPV) were calculated. For each parameter, the median, 5th, and 95th percentiles of obtained distributions were used to present the results. The results indicate important variations of HSe and HSp depending on N, k, and TP. The HSe decreased with decreased N, increased k, or decreased prevalence, whereas the opposite was observed for HSp. The context of use (TP, HTP, N and k) also impacted the HPPV and HNPV. Interestingly, whereas the HNPV increased and HPPV decreased for increasing k for TP≤ 0.15, the opposite was observed where TP≥ 0.2. The MCT varied as complex interactions between k, N, and TP. This study shows that the context of application of LUS as a group-level test was very important to consider to tailor antimicrobial use in calves.