Bayesian estimation of sensitivity and specificity of systematic thoracic ultrasound exam for diagnosis of bovine respiratory disease in pre-weaned calves

Among the different clinical presentations of bovine respiratory disease, active pneumonia, defined as an infection of the lower airway with signs of inflammation, is the most important to diagnose correctly so appropriate treatment can be initiated. Diagnostic tests that accurately identify cases of active pneumonia are lacking; however, thoracic ultrasonography (TUS) seems promising. The primary objective of this study was to estimate the accuracy of TUS compared to reference tests for the diagnosis of active pneumonia in pre-weaned calves, using a latent-class model method (LCM). The tests used for comparison were the Wisconsin Clinical Respiratory Scoring Chart (CRSC, positive if >= 5) and serum haptoglobin concentration (Hap, positive if >= 15 mg/dL). Secondary objectives were to assess the incremental value on TUS accuracy of combining TUS of the right cranial part and caudal parts, and to determine the accuracy of various thresholds for depth of consolidation (>= 0 cm, >= 1 cm, or >= 3 cm) for diagnosis of active pneumonia. One population of veal calves (n = 209) and one of dairy calves (n = 301) were enrolled. TUS, CRSC and Hap were all performed on each calf on the same day. TUS was performed by screening the mid to ventral portion of the lung caudal of the heart (the caudal sites), as well as the right parenchyma cranial to the heart (the cranial site). The maximal depth of consolidation (DEPTH) on TUS was recorded and noted separately for caudal and cranial sites. Different TUS cases were defined according to site and DEPTH. The accuracy of TUS was estimated by LCM for three tests conducted in two populations. Prevalence of active pneumonia was low (0.05) in both populations. In general, higher minimal consolidation depth thresholds led to increased TUS specificity (Sp) estimates, with minimal effects on TUS sensitivity (Se). With a TUS DEPTH threshold of >= 3 cm, adding TUS of the cranial site had little effect on accuracy. Using the >= 3 cm threshold with caudal sites only, posterior Se and Sp median estimates of 0.89 (95%BCI: 0.55, 1.0) and 0.95 (95%BCI: 0.92, 0.98), respectively, were obtained. In conclusion, in populations with low active pneumonia prevalence, adding TUS of the cranial site did not enhance the performance of the test. We suggest using a DEPTH threshold of >= 3 cm solely on caudal sites to detect active pneumonia.