The primary outcome measure was overall survival, which was defined as the number of infants admitted who were discharged alive. The secondary outcome measures were the four major adverse short-term morbidities: bronchopulmonary dysplasia (BPD), necrotizing enterocolitis (NEC), intraventricular haemorrhage (IVH), retinopathy of prematurity (ROP), and survival-free of major complications.

Overall mortality included all deaths that occurred during the first admission to our NICU until discharge from the hospital. All stillbirths and patients who died in the delivery room were excluded, and infants with congenital malformations were included in our calculations.

In accordance with the definition given by Jobe & Bancalari (8), BPD was defined as the requirement for supplemental oxygen at 36 weeks of postconceptional age or discharge to home, whichever came first. BPD was not graded according to severity (8). As we took part in the NIPPV (9) and SAIL (10) trials the definition of BPD changed in the second period of the study to respiratory support regardless of FiO₂ to maintain oxygen saturation ≥90% and to perform an oxygen reduction test (11–13). A completed course of antenatal steroids was considered as two doses with a 24-h interval, with the last dose administered more than 24 h before birth. NEC was defined according to Bell's criteria (14), where severe NEC requires surgical intervention. IVH was classified according to Papile et al. (15) and severe IVH was graded as IVH ≥grade III. ROP was graded in conformity with the international classification, and grades III–V were classified as severe ROP (16). Adverse short-term outcomes were defined as the development of any of the following adverse morbidities: BPD, severe NEC, severe IVH (grades III and IV), or severe ROP (grades III–V). The percentage of the cohort who survived without any of the four major short-term morbidities was summed up as survival free of major complications.

The nationwide registry was approved by the Ethics Committee of the Medical University of Vienna (EK 1828/2019). The study was approved by the Ethics Committee of the State of Vorarlberg (EK-2-7/2020), conducted in accordance with the Declaration of Helsinki and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for cohort studies. The authors disclose any potential conflicts of interest.

Data analyses were performed using SPSS software, version 21.0 for Windows (IBM; Armonk, New York, USA). Descriptive statistics are provided as percentages unless otherwise stated. Group differences were analysed using the chi-square test (χ2). Statistical significance was set at p < 0.05.

Overall, 25 very-preterm infants died (5.0%). For ELGAN, the mortality rate was 12%, and for those >28 weeks of GA, it was 1.7% (Table 2). The death rates in infants at 23, 24, 25, 26, and 27 weeks of GA were 20%, 26.1%, 26.7%, 4.9%, and 3.4%, respectively. The corresponding death rates for infants at 28, 29, 30, and 31 weeks of GA were 1.6%, 4.0%, 1.2%, and 0.8%, respectively (Table 1, Figure 1). Mortality decreased from 6.4% in P1 to 3.6% in P2 (p = .14) (Table 2).

Overall, the incidence rates were 11.2% for BPD, 4.0% for surgical NEC, 4.6% for severe IVH, and 2.6% for severe ROP. BPD rates significantly increased from P1 to P2 (7.2% to 15.1%, p = .04), whereas other morbidities showed non-significant changes from P1 (2.4% for NEC, 4.8% for IVH, 1.6% for ROP) to P2 (5.6% for NEC, 4.4% for IVH, 3.6% for ROP). For detailed information, see Figure 2.

The risk to be discharged with one or more of the predefined morbidities was 10.6% for one, 3.3% for two and 0.8% for three morbidities. No infants had all of the four morbidities at discharge. In ELGAN, the risk for at least one short-term morbidity ranged from 69.6% with 24 weeks of GA to 14.5% in those with 28 weeks and 5.8% in those with 31 weeks of GA.

Overall, the survival free of major complications rate was 79.0%. It ranged from 30.4% in infants with 24 weeks of GA to 94.2% in those with 31 weeks of GA (Figure 1), and showed no significant difference over time (82.4% for P1 to 75.7% for P2, p = .15).

Our study has some limitations. According to our register (ÖFGOR), all stillbirths and delivery room (DR) deaths are excluded which certainly influence the results, namely survival by underestimate of the actual mortality rate. When introducing the register in 2012 (5) we decided on a minimal dataset which did not include this parameter—as well as periventricular leucomalacia (PVL) and sepsis—because entering data is voluntary and not funded (2, 4, 5). Retrospectively, delivery room death can not be determined because live-born infants below 32 weeks of GA are not registered in national (29) or regional data sets (7). However, other studies (30, 31) also exclude DR deaths and a recent meta-analysis of 65 studies shows that this refers mainly to the ELGAN at the limit of viability (30): for infants of 25 weeks GA and above the percentage of survival of all live births and survival of infants admitted at the NICU does not show any notable differences. On the other hand, infants, where therapy was withheld after admission in the further course to the NICU, were also included in our death rates (4).

We did not adjust for risk factors because of the limited sample size. However, analyses conducted by ÖFGOR show that statistical risk predictors for death or adverse short-term outcome are not only low GA and low birth weight, but also missing or incomplete course of antenatal steroids, male sex, and multiple births (4).

To reduce the bias of different inclusion criteria, time periods, or countries with different cultures and health care systems (19, 32, 33) we describe our outcome with respect to national (4, 18), as well as to our neighboring countries Switzerland (17) and Germany (16, 21) and international data with similar timeframes and health care systems (19, 32–34). The long time period of our study may contribute to bias because patient care has changed. Certainly, the strategy in ventilatory support has undergone the biggest changes by new non-invasive measures like CPAP, NIPPV, less invasive surfactant application (LISA), or higher rates in non-invasive ventilation. Others remain constant like human milk feeding, or no probiotic supplementation. However, our study was not designed to quantify these effects and the number of included patients is far too limited to give any evidence. Of concern is the higher NEC rate in the ELGAN, when compared to others (16), which lead to several measures like an early implementation of human milk-based nutritional products in the latter time period of the study.

Finally, this study investigated outcome data at discharge but did not define long-term outcomes which are of great concern for patients, clinicians, and families, and affect the quality of life (12, 35).

The strength of this study is its prospective, population-based design over a long period of 14 years in a defined area. Considering the fact that data entry was based on voluntariness and no funding was available for individual centres (4), the inclusion rate seems adequate and is comparable to that of other studies: missing data in other population studies ranged from 3% in the Israel Neonatal Network to up to more than 40% in the Neonatal Research Network of Japan (19).

This is the first and only investigation of short-term outcomes in a regional setting in Austria. In this manuscript, we disclose our data and results to the public concerning benchmarks for very preterm infants and patient volumes which are strongly and repetitively discussed in health and politics (36, 37): some might argue that our group might not be representative. However, in our register, the mean birth weight, GA, sex or rate of completed antenatal steroids are the same as in the national studies for Austria and Switzerland (4, 17). In addition, we have a low rate of missing values, which reflects a good adherence to the register. Therefore, we are convinced that our cohort and subsequently our results are comparable to nationwide as well as international data.

Others might state that our patient volume is too low and the outcome might be better in a larger center (36–42). The effect of case load may be underestimated, if the results are not adjusted for the systematic referral of infants to the larger tertiary centers (40) which is not the case in our population. A recommended minimum provider volume does not exist (43) in Austria (3) nor in Germany (36, 42). The benchmark of a minimum of 30 preterm infants per year with <1,250 g birth weight was suspended by the Federal Social Court in 2015 in Germany (36) pending a final decision (42). Regionalization policies of neonatal care has been endorsed to improve neonatal outcome and geography plays a major role in how the neonatal care networks are set up and function (43). Although significant investments in infrastructure and other nationwide improvements were made in the European Union, Portugal is a country to demonstrate a significant decline in neonatal mortality coincident with closing smaller centers arguing strongly for the effect of regionalization on this outcome (41). While the positive effects of volume seem clear, the optimal threshold for discrimination low-volume from high-volume centers remains unclear (38, 41, 42). A study from Switzerland demonstrated that center variability in survival rates cannot be explained by baseline demographics or center case load (44). Providing slightly above 35 very preterm infants per year, we demonstrate one of the lowest mortality rates published in equal cohorts over an equal timeframe. Our results can be partially explained by the fact that all hospitals in our state are affiliated within a formal network with the same standard operating procedures, such as guidelines for antenatal maternal transfers or newborn resuscitation algorithms and the low rate of severe IVH may also reflect this good standard of care (36). Another explanation might be the mandatory regular attendance at simulation training programs for every team member and participation in the Video Apgar (22–26) and other trials (9, 10), which improved our DR management. Participating in such clinical studies means that we transparently expose our efforts to critical discussion within our institution and at other centers which may lead to standardization and improvement of our work.

Reducing morbidities among vulnerable infants remains a challenge for clinicians and investigators. Participation in (inter)national quality activities, multicenter registers, and clinical trials improves the quality of care and is increasingly considered an index of quality assurance. Precise data collection is an important component of guiding strategies for better outcomes. However, neither a continuous reduction in the viability limit nor an increase in survival rates, but rather the quality of life, should be the most important parameter in perinatal medicine.