Study population, protocol and ethical approval
The concept and protocol of investigation, definition of diagnosis of preterm morbidity and mortality as well as specific disease severity and level of care were adopted from the references cited in the manuscript, during the data collection, analysis and preparation of manuscript in 2018-2020. This study followed the whole regional birth population survey in 2015, in which a complete birth data was prospectively collected from totally 107 level I-III hospitals providing obstetric care (6 municipal, 16 county and 85 township), and 8 level II and 4 level III hospitals equipped with neonatal wards and/or intensive care units (NICU) [11,20]. Non-medical abortions (especially unplanned pregnancy) were excluded. The above birth data was integrated with the data on all hospitalized preterm infants derived from the complete birth population in whole region, retrospectively retrieved from regional perinatal information database (Fig. 1). The study protocol was approved by the ethic committee of Children’s Hospital of Fudan University, and accepted by HWCH as well as all participating hospitals in Huai’an [11,20]. As no specific intervention was applied, informed consent from parents/guardians was waived. Preterm birth was defined as delivery at 25+0-36+6 weeks of GA, divided into extreme (EPT, 25+0-27+6 weeks), very (VPT, 28+0-31+6 weeks), moderate (MPT, 32+0-33+6 weeks) and late (LPT, 34+0-36+6 weeks) preterm groups [21,22]. Those of EPT below 25 weeks of GA were not included due to very few numbers and parental decision not to provide resuscitation at delivery. Incidences of EPT, VPT, MPT, LPT and total preterm births were presented as per thousand (‰) of the number of total births (including term and post-term births) of whole region.
Definitions of preterm morbidity and mortality
The diagnostic criteria of pregnancy, perinatal and neonatal co-morbidities and complications are presented in additional file 1 [23-34]. Definitions regarding vital statistics are based on the original survey [11,20], and the 10th revision of the international classification of diseases [35]. Briefly, GA was mainly determined by the date of last menstrual period and/or fetal sonography in early pregnancy, or postnatal assessment by new Ballard score when prenatal records were missing or incomplete [36]. Birth weight (BW) was measured at birth. Small for GA was defined as a BW < 10th percentile for gender and GA [37]. Birth defects (BD) were identified prenatally or within the first seven postnatal days (PND) [11,20]. Severity of neonatal underlying diseases were characterized as requiring intensive or critical care based on the treatment strength during hospitalization, such as degree of the disease severity, in need of non-invasive/ mechanical ventilation (NIV/MV), surfactant, vasopressor, or surgery, etc (also see additional file 1) [7,28,38]. Fetal death was deemed to be synonymous with stillbirth. Deaths at delivery room (DR) referred to those born alive but died during resuscitation, usually after parents’ requests for withdrawal or withholding [39]. Perinatal mortality included stillbirths and neonatal deaths within first seven PND. Neonatal mortality was defined as deaths of livebirths within 28 PND, including deaths at DR. All-death was defined by including stillbirths, neonatal deaths or, for EPT and VPT, within 44 weeks of adjusted post menstrual age. In-hospital death was defined as deaths during hospitalization or, for EPT and VPT, within 44 weeks of adjusted post menstrual age. For those hospitalized in NICU but had early withdrawal at parental request, the outcome was estimated by attending physicians based on discharge or follow-up record.
In birth population, incidences of perinatal and neonatal morbidities were presented as percentage (%) with the number of preterm total births and livebirths excluding deaths at DR as denominator, respectively. In hospitalized population, incidences of postnatal morbidities were presented as percentage (%) with the number of total preterm admissions as the denominator. Rates (%) of perinatal mortality, stillbirths, deaths at DR and all-death were divided by the number of total preterm births. Neonatal mortality rates (%) were divided by the number of preterm livebirths (including deaths at DR) [20]. In-hospital death rates (%) were divided by the number of preterm livebirths excluding deaths at DR in birth population, or the number of total preterm admissions in hospitalized population, respectively. The GA-specific mortality rates (%, where numerators and denominators were both limited within specific GA stratum) [7,40] were adopted and corrected by the number of total births or livebirths of whole region [41], and presented as per 1,000 (‰, for calculation see Table 2).
Representativeness of HWCH and risk factors of deaths
To estimate the representativeness of HWCH of whole region, whole regional preterm births and hospitalized preterm infants were divided into HWCH and non-HWCH groups, respectively. Furthermore, the HWCH admissions were divided into inborn and out-born to compare the efficacy of perinatal-neonatal care by hospital categories (Fig. 1). Risks of all-death for preterm births and in-hospital death for preterm admissions were analyzed to determine the role of HWCH in the whole region, taking the effects of perinatal interventions and the underlying neonatal morbidities into consideration.
Statistical analysis
EPIDATA database was used for datasheet recordings and subjected to SPSS software (V. 16.0, SPSS Inc. Chicago, IL) for statistical analysis. Continuous variables were presented as mean and standard deviation (SD) or median [interquartile ranges]. One-way analysis of variance or non-parametric Mann-Whitney test was used for comparison among groups. Categorical variables were presented as number and rate, using a two-tailed Pearson Chi-squared or Fisher’s exact test where appropriate. P < 0.05 was considered statistically significant. Death risks of preterm births and hospitalizations were analyzed by uni- and multi-variable (binary) logistic regressions. Values measuring relative risks were given as odds ratio (OR) with 95% confidence intervals (CI).