We found that the risk factors associated with longer LOS differed for infants, depending on BPD severity (mild, moderate, or severe). We believe that this study makes a significant contribution to the literature because more individualized approaches to reducing LOS could be possible with this information. There were several notable findings in our study, including differences between the BPD groups. First, we found that the initial conditions of infants, such as GA at birth, birth weight, and CRIB II score, were independently associated with LOS, regardless of BPD severity. Second, maternal factors had little effect on the LOS, except for maternal hypertension in infants with severe BPD. Lastly, infants in the mild BPD group requiring surgical treatment during hospitalization in the NICU were independently associated with LOS, but the accompanying comorbidities were more important in infants in the severe BPD group.

Several studies have found that the incidence of BPD increases as the GA at birth and birth weight decrease [25]. Unlike previous studies, we divided the study population into three groups according to the severity of BPD and investigated them separately. We found that the severity of BPD increased, as GA at birth and birth weight of the infant decreased. Lower birth weight and premature birth were associated with an increased risk of developing severe BPD, consistent with a previous report by Han et al. [14]. Interestingly the CRIB II scores differed by group. As the CRIB II score assesses an infant’s initial illness at admission, the initial condition of the infant influences the severity of BPD.

In addition to the immature lungs of preterm infants, exposure to maternal oligohydramnios in the intrauterine period adversely affects fetal lung development and easily becomes dysplastic [26]. The highest rate of maternal oligohydramnios in the severe BPD group in our study is consistent with that reported in previous studies [27, 28] (Table 1).

There was little difference in the basic demographic information between the groups based on BPD severity. GA at birth and birth weight significantly impacted LOS, regardless of BPD severity. Using an anti-logarithmic conversion (eadj β), we could identify the factors common to all three BPD severity groups. The LOS increased in ELBW infants by 1.1–1.2 times compared to non-ELBW infants. The LOS of very preterm infants < 28 weeks of gestation at birth was also 1.1–1.2 times longer than that of infants ≥ 28 weeks of gestation at birth. This result conforms with a recent systematic review not limited to infants with BPD; GA and/or birth weight have also been identified as critical risk factors affecting the initial LOS [29]. The earlier the GA at birth, the more days are expected for maturation until the term-equivalent age; therefore, the LOS will become longer. Immaturity, represented by GA at birth, is a very strong factor and is known to be an important risk factor for major neonatal morbidities [30]. As LOS can be considered to reflect neonatal morbidities, we confirmed that it was affected by GA at birth and birth weight in all groups regardless of the severity of BPD.

Infants who were administered postnatal steroids had longer LOS than those not treated with steroids: the percent increase in LOS was 7.3%, 4.1%, and 11.6% for the mild, moderate, and severe BPD groups, respectively. Postnatal steroids can be administered to infants at various times, but administration of steroids soon after birth advantageously relieves the inflammatory cascade that leads to BPD development. Conversely, delaying treatment and using more selective criteria, such as the continued need for ventilatory support, can help identify infants most likely to develop BPD [31]. In addition, infants with more severe symptoms and signs who are diagnosed with severe BPD are less likely to benefit from steroid treatment and cannot be discharged earlier due to the continued need for ventilation [32]. This may explain the markedly prolonged LOS observed in the severe BPD group with regard to postnatal steroids.

Surgically treated PDA was independently associated with prolonged LOS in the mild BPD group. Similar results were recently published in South Korea; infants who underwent PDA ligation stayed longer in the NICU than those who did not [33]. The optimal timing for surgical ligation of PDA remains controversial. According to a recently published meta-analysis, early surgical ligation of PDA might have a better respiratory outcome and nutritional status than late surgical ligation [34]. Usually, surgical ligation is considered after medical treatment for PDA has failed or is contraindicated. Therefore, the timing of ligation is relatively later than that of medical treatment, and infants who require this surgery are treated for longer and have a longer LOS [35]. Infants in the mild BPD group were born at an older gestational age and needed fewer days to reach the term-equivalent age to be discharged than the others, and the timing of surgical ligation for PDA could be close to the term-equivalent age. In other words, surgically treated PDA might be a determinant of the prolonged LOS in the mild BPD group. Meanwhile, the requirement for surgical treatment for PDA was not a significant risk factor for increased LOS in infants in the moderate BPD group. As surgical ligation of the PDA is performed within two weeks in most cases [36], which is too early to discharge from the NICU for very preterm infants, born before 32 weeks of gestation. This factor increases the LOS in the NICU of infants with mild BPD. A few weeks after severe IVH, PHH may develop because of disturbances in Cerebrospinal fluid flow and absorption [37]. Additionally, white matter damage secondary to PHH is likely to be exacerbated by compression and ischemia due to increased intracranial pressure [38]. As preterm infants grow, their condition worsens or symptoms develop, and treatment is required, resulting in longer hospitalization [39].

Unlike the mild BPD group, the LOS of infants in the severe BPD group was affected by several neonatal outcomes, such as pulmonary hypertension, IVH, and sepsis. Higher incidences of neonatal outcomes in the severe BPD group were also found compared with those of mild and moderate BPD groups, results that are in line with a previous study [40]. These findings could reflect that infants in the severe BPD group had been affected by various comorbidities, easily. Although it was unclear whether combined comorbidities were directly associated with longer LOS, Hintz et al. reported the relationship between clinical comorbidity and longer LOS in NICU [41]. Therefore, combined comorbidities in the severe BPD group might have associations with prolonged LOS in NICU.

The risk level for preterm neonates can be determined based on their initial clinical condition owing to the immaturity of their structural and functional organs. Scoring systems developed in neonatal medicine for estimating mortality and morbidities can be applied to LOS predictions by accounting for the severity of illness in the first week of life beyond factors known at birth [42]. From our regression model, the CRIB II score can also help predict LOS even after adjusting for several confounding factors. Moreover, there were interesting dose–response relationships between the CRIB II score and LOS. Infants with severe initial illness were observed to require a longer LOS. Infants with a level 2 or 3 CRIB II score had a 1.1–1.2 times longer LOS than level 1 infants, while the most severely affected CRIB II level 4 infants had a 1.4–1.5 times longer LOS. The worse the initially evaluated CRIB II score, the longer the hospitalization period, which was observed in mild, moderate, and severe BPD groups. Considering these dose–response relationships, different cut-off points of the CRIB II score can predict longer hospitalization in each BPD group. Advanced neonatal care by skilled personnel is important to shorten hospitalization from resuscitation immediately after delivery to initial management after NICU admission.

The strengths of our study were that we used data from a population-based national cohort covering about 70% of VLBW infants in Korea and that the KNN uses a meticulous data collection system, without biases arising from differences between hospitals or individual neonatal units due to local discharge practices within units [16]. The KNN maintains a complete data-monitoring system to improve data quality [43]. However, this study has several limitations. Because the LOS values were not normally distributed, we log-transformed them and interpreted the results after the anti-logarithmic transformation of adj β. The second limitation was the use of registry data. Actual practices such as ventilator weaning protocol may inevitably differ across participating centers, and information about the exact size or hemodynamic significance of PDA could not be obtained. Third, longer LOS may have been underestimated. Infants transferred to the pediatric intensive care unit or general ward were excluded because the actual discharge date from the hospital was indistinguishable. Finally, along with reducing the socioeconomic burden, the overarching goal of reducing BPD severity for preterm infants should be considered to improve long-term neurodevelopmental outcomes for this cohort.

In conclusion, our study identified critical risk factors associated with extended initial hospitalization in preterm infants with mild, moderate, and severe BPD, respectively. Risk factors for longer LOS varied across different levels of severity of BPD. A unique approach is needed to reduce the LOS for infants in each BPD group, and this might be an important suggestion considering the socioeconomic burden. In addition, these risk factors may serve to predict the duration of initial hospitalization, which would be valuable to parents and families, clinicians, and other service providers.

In the future, shortening LOS could be possible by individualized approaches according to the BPD severity of the infants. For example, setting of management protocol through quality improvement (QI) programs including better neonatal resuscitation at birth, efforts to minimize accompanying comorbidities, and specific protocols for postnatal steroid use facilitates reducing LOS. These efforts, directed at reducing the LOS connected to the reduction of preterm morbidities including BPD, are ultimately aimed at having a better neurodevelopmental outcome.