Abstract
BACKGROUND: Bronchopulmonary dysplasia (BPD) is a risk factor for late respiratory diseases (LRD). Lung ultrasound (LUS) is a useful, radiation-free tool for predicting BPD, but data linking LUS to late respiratory diseases is limited. OBJECTIVE: To assess the predictive value of the modified lung ultrasound score (mLUS) for late respiratory diseases in premature infants. METHODS: We retrospectively reviewed data from 140 premature infants born before 32 weeks of gestation, treated for atelectasis due to LRD in our ICU from December 2023 to January 2025. LRD was defined as a composite endpoint with operational criteria: (1) Worsening BPD (need for ≥ 10% FiO(2) increment or reinitiated mechanical ventilation, per 2018 NICHD definition); (2) Bronchiolitis (tachypnea/wheezing + positive RSV RT-PCR or LUS/X-ray peribronchial thickening); (3) Pneumonia (fever > 38°C + CRP > 20 mg/L + LUS/X-ray consolidation); (4) Asthma/reactive airway disease (RAW, ≥ 3 bronchodilator-responsive wheezing episodes); (5) Respiratory-related hospitalizations (unplanned admissions for respiratory symptoms). Diagnosis was based on ICD-10 codes, emergency notes, medication records, and imaging. Sample size calculation was performed with α = 0.05, power = 80%, confirming 140 cases and 48 controls met requirements. Missing data (< 3%) were handled via multiple imputation (5 datasets), with complete-case sensitivity analyses. RESULTS: Multivariable logistic regression (adjusted for gestational age, birth weight, mechanical ventilation duration) showed mLUS independently predicted LRD (adjusted OR: 1.25, 95% CI: 1.10-1.42, p < 0.001), as did traditional LUS (adjusted OR: 1.19, 95% CI: 1.04-1.36, p = 0.012). Model fit was good (Hosmer-Lemeshow χ (2) = 7.23, p = 0.495). The median mLUS score was 15 (IQR 9-21) in infants with LRD versus 11 (IQR 4-16) without (p < 0.001); traditional LUS was 12 (IQR 6-16) versus 7 (IQR 4-9) (p < 0.001). The mLUS had an area under the receiver operating characteristic (ROC) curve (AUC) of 0.789 (95% CI 0.731-0.847), sensitivity 78.5%, specificity 57.2% (cutoff: 14). DeLong test confirmed mLUS AUC was higher than traditional LUS (0.726, 95% CI 0.658-0.794; p = 0.032). Spearman correlation between scores was 0.82 (p < 0.001), and mLUS improved net reclassification improvement (NRI = 0.28, p = 0.002). Component analysis: mLUS AUC for worsening BPD was 0.812, bronchiolitis 0.765, respiratory hospitalizations 0.793; excluding asthma/RAW (sensitivity analysis) gave AUC = 0.795 (p = 0.891). CONCLUSION: The modified lung ultrasound score is significantly linked to advanced respiratory disease in premature infants, demonstrating important clinical relevance.