The invention discloses a predictive marker, a predictive model and a predictive system for bronchial lung dysplasia of a premature infant. EUGR, caffeine treatment days, total oxygen uptake days and LMR are independent risk factors of BPD occurrence of the premature infant. The lung ultrasound prompts the occurrence rate of the severe alveolar interstitial syndrome on the third day after the premature infant is born, the lung ultrasound prompts the pleural line change on the 28th day after the premature infant is born, and the occurrence rate of the alveolar interstitial syndrome, the pulmonary metaplasia or the fragment symptom has a prediction effect on the occurrence of the BPD. The serum eNOS level of a child patient in a BPD group on the first day after birth is remarkably higher than that of a child patient in a non-BPD group and is in positive correlation with the severity of BPD, and when the serum eNOS level of the child patient on the first day after birth is larger than 6.48 ng / ml, the risk that a premature infant suffers from BPD is obviously increased. The BPD occurrence of the premature infant can be well predicted by constructing a regression model through combination of six indexes, namely the serum eNOS (1d) level, the severe alveolar interstitial syndrome (3d), EUGR, caffeine treatment days, the total oxygen uptake days and LMR6 indexes.