Abstract / Summary
Abstract Background Lung ultrasound (LUS) is an established diagnostic tool for pediatric pneumonia, yet its utility as radiation-free alternative to computed tomography (CT) for monitoring disease progression and differentiating complicated phenotypes, specifically necrotizing pneumonia (NP) and parapneumonic effusion (PPE), has not been prospectively validated in a single comparative cohort. This study aimed to evaluate the diagnostic and longitudinal performance of LUS, including a quantitative LUS score, in children with community-acquired pneumonia (CAP), CT-confirmed NP, and PPE, and to evaluate the diagnostic accuracy and longitudinal performance of LUS for identifying and monitoring complicated pediatric pneumonia, including its potential as a radiation-free follow-up tool. Methods This prospective single-center diagnostic accuracy study enrolled 257 children diagnosed with CAP ( n = 116), NP ( n = 54), and PPE ( n = 87). NP was confirmed by CT in all 54 cases; LUS was performed by trained operators at admission and on day five as a paired assessment against CT and chest radiograph findings. LUS qualitative features and quantitative scores were compared between groups and over time. Multivariable regression and ROC analysis were performed to assess diagnostic accuracy and identify predictors of clinical outcome. Results At admission, NP and PPE showed significantly higher LUS scores and more frequent absent lung sliding and heterogeneous hypoechoic consolidation than CAP ( p < 0.001). A LUS score cutoff of > 8 demonstrated strong diagnostic performance at both time points (admission AUC: 0.881, sensitivity 79.02%; Day 5 AUC: 0.861, specificity 93.86%), supporting its potential role as an early screening marker and as a radiation-free bedside monitoring tool during follow-up. Longitudinal LUS assessment demonstrated significant re-aeration by Day 5 ( p < 0.001), evidenced by an increased proportion of separated B-lines (35.0% to 54.9%) and a reduction in lung consolidation. On multivariable analysis, higher LUS scores independently predicted longer hospitalization on both admission (β = 0.221, p < 0.001) and Day 5 (β = 0.215, p < 0.001), and elevated D-dimer on admission (β = 0.334, p < 0.001) and Day 5 (β = 0.386, p < 0.001). Conclusion A LUS score > 8 showed promising diagnostic performance for identifying complicated pediatric pneumonia at admission and during Day 5 follow-up. Although internal bootstrap validation supported the stability of its performance, prospective validation in independent cohorts is required before implementation as a CT-sparing follow-up strategy.