Accuracy and reliability of bedside thoracic ultrasound in detecting pulmonary pathology in a heterogeneous pediatric intensive care unit population

Purpose: This study was designed to access sensitivity and specificity of detection of lung abnormalities by the ultrasound (US) done by PICU providers of varying levels of experience compared to CXR and to assess the inter-rater reliability in the interpretation of the USG findings. Methods: Up to three US examinations were performed on patients meeting eligibility criteria. US examinations were reported by the operator and remotely by an expert reader. Both operator and readers interpretation were correlated with CXR read by an independent pulmonologist. Results: One hundred and thirty-five US examinations were performed on 91 patients over 9 months. Overall agreement between the operator and reader of the US was 0.53 (0.38-0.68). The agreement was highest with an expert-expert pair (0.75) and lowest with a novice-expert pair (0.27). Sensitivity and specificity of thoracic US to detect pulmonary abnormalities showed a high sensitivity by the operator (82.5%) compared to the reader (63.4%). Specificity was 25% and 42.8%, respectively. US was overall highly sensitive to detect pneumonia (96.4%) with a 100% PPV, but only modest for bronchiolitis. Conclusions: Lungs US is a rapid and sensitive bedside tool to assess lung consolidation in children in ICU. It, however, has low negative predictive values, and negative US examinations cannot rule out lung pathology.