Variability in Very Preterm Stillbirth and In-Hospital Mortality Across Europe
BACKGROUND AND OBJECTIVE: Stillbirth and in-hospital mortality rates associated with very preterm births (VPT) vary widely across Europe. International comparisons are complicated by a lack of standardized data collection and differences in definitions, registration, and reporting. This study aims to determine what proportion of the variation in stillbirth and in-hospital VPT mortality rates persists after adjusting for population demographics, case-mix, and timing of death.
METHODS: Standardized data collection for a geographically defined prospective cohort of VPTs (22+0–31+6 weeks gestation) across 16 regions in Europe. Crude and adjusted stillbirth and in-hospital mortality rates for VPT infants were calculated by time of death by using multinomial logistic regression models.
RESULTS: The stillbirth and in-hospital mortality rate for VPTs was 27.7% (range, 19.9%–35.9% by region). Adjusting for maternal and pregnancy characteristics had little impact on the variation. The addition of infant characteristics reduced the variation of mortality rates by approximately one-fifth (4.8% to 3.9%). The SD for deaths <12 hours after birth was reduced by one-quarter, but did not change after risk adjustment for deaths ≥12 hours after birth.
CONCLUSIONS: In terms of the regional variation in overall VPT mortality, over four-fifths of the variation could not be accounted for by maternal, pregnancy, and infant characteristics. Investigation of the timing of death showed that these characteristics only accounted for a small proportion of the variation in VPT deaths. These findings suggest that there may be an inequity in the quality of care provision and treatment of VPT infants across Europe.
- CI —
- confidence interval
- EPICE —
- Effective Perinatal Intensive Care in Europe
- GA —
- gestational age
- OR —
- odds ratio
- PPROM —
- preterm premature rupture of membranes
- SGA —
- small for gestational age
- VPT —
- very preterm birth
What’s Known on This Subject:
The mortality rates associated with very preterm births vary widely across Europe. Direct comparisons of international mortality rates for very preterm infants are complicated by a lack of standardized data collection and differences in definitions, registration, and reporting.
What This Study Adds:
Variations in stillbirth and in-hospital very preterm mortality rates were reduced by 20% after adjusting for maternal, pregnancy, and infant characteristics. Regional variation was not explained by the variation in the mortality of the earliest deaths.
The worldwide burden of preterm birth is increasing.1 However, since the early 1990s, many improvements in the care of these infants, including routine use of antenatal steroids for women at risk for preterm delivery and surfactant, have led to a dramatic reduction in the associated mortality of these infants, even for those born at <26 weeks’ gestation.2–6 Medical advances in subsequent years have led to a continued reduction in both the mortality and morbidity rates for very preterm infants.4,7,8 However, wide variations in stillbirth and in-hospital mortality rates for very preterm births (VPTs) have been observed across Europe, with up to twofold differences in these rates between regions.9–11 These comparisons raise questions about the quality and the equitability of access to perinatal and neonatal care in high-mortality regions or countries.
Using mortality rates as benchmarks for quality of care across countries is contingent on accounting for other sources of variation that affect these rates. Some of the variation can be explained by a lack of standardized data collection and reporting, specifically among the most preterm infants.12 Worldwide differences in definitions of stillbirths, births, and deaths, access to termination of pregnancy, and registration and reporting of very preterm infants make it difficult to carry out direct comparisons, despite using a standardized approach. In addition, perinatal management of infants at very low gestational ages (GAs) related to unit policy and treatment and ethical concerns around the withholding or withdrawal of care also differs across Europe and impacts on mortality rates.13–15 Finally, the broad diversity in the demographic characteristics of childbearing women in Europe may affect the patient case-mix.16,17
Using the Effective Perinatal Intensive Care in Europe (EPICE) standardized population-based VPT cohort from 19 regions across 11 European countries, this study aims to determine what proportion of the regional variation in stillbirth and in-hospital mortality rates in European regions persists after adjusting for reporting differences in outcomes among VPTs as well as population demographics and case-mix. We additionally investigate regional differences by the timing of death, because causes of death can differ for deaths which occur before, during, or immediately after delivery or on the neonatal unit before discharge.
The EPICE birth cohort comprises a geographically defined prospective cohort of outcomes for VPTs between 22+0 and 31+6 weeks of GA in 19 regions in 11 European countries (Table 1). Regions were selected with respect to geographic and organization diversity (all have at least 1 tertiary center), feasibility (preexisting data systems and expertise), and sample size considerations.
Regional data collection took place by using a standardized protocol between March and July 2011 for 12 months, except in France where data were collected for 6 months, during which there were ∼850 000 births in the EPICE regions. Data definitions were agreed on by the EPICE Consortium and data were collected from obstetric and neonatal medical records by the regional teams by using either the EPICE questionnaire or modified regional systems. Ascertainment was validated against birth registers within all hospitals by local teams. Determination of GA was based on the best estimate of the obstetric team caring for the pregnant women. When there were several estimates, GA was based on the following hierarchy: in vitro fertilization treatment, ultrasound based on earliest estimate, last menstrual period, fundal height measurement, and neonatal assessment at birth.
For this analysis, data were excluded for 3 regions with <150 cases (Table 1). Terminations of pregnancy and severe congenital anomalies (Supplemental Information) were also excluded due to the variation in policies regarding the treatment of such cases across the regions.
Outcomes were defined by the type and timing of death or survival to discharge from neonatal care by using the following categories: stillbirth, deaths <12 hours after birth, deaths ≥12 hours to 7 completed days after birth, deaths >7 days after birth, and survivors at discharge from neonatal care. Analysis by GA was performed by using 2-week GA bands to facilitate comparisons, namely 22+0–23+6 weeks, 24+0–25+6 weeks, 26+0–27+6 weeks, 28+0–29+6 weeks, and 30+0–31+6 weeks.
Maternal and pregnancy characteristics included in the analysis were maternal age, parity, multiple pregnancy, and pregnancy complications, defined as hypertensive diseases (preeclampsia, eclampsia, and HELLP syndrome), admission to hospital for antepartum hemorrhage after 20+0 weeks’ gestation, and preterm premature rupture of membranes (PPROM). Infant characteristics were GA at birth, birth weight, sex, multiplicity, and small for GA (SGA; <10th percentile).18
Ethics and data protection approvals were provided in each region to fulfill national legislation requirements.
Crude stillbirth, in-hospital mortality, and survival rates are presented using all births as the denominator, with differences between regions in the rates of mortality and of type of mortality tested by using the χ2 test. Maternal, pregnancy, and infant characteristics are presented as numbers and rates or means with SDs, and the statistical significance of any variation between the regions was tested by using the χ2 test or analysis of variance as appropriate.
The effects of maternal, pregnancy, and infant factors were investigated by using multinomial logistic regression models. The models simultaneously estimate the probability for each type of death (stillbirth, death <12 hours after birth, and death ≥12 hours after birth) while allowing the magnitude of the effect of each maternal, pregnancy, or infant factor to differ by type of death. This was undertaken first for maternal and pregnancy factors alone (overall mortality) and then for maternal, pregnancy, and infant factors (overall mortality and by timing of death) with associations presented as odds ratios (ORs) with 95% confidence intervals (CIs). Overall and for each type of death, the observed “unadjusted” mortality percentage was calculated for each region. In addition, a standardized mortality ratio was calculated overall and for each type of death for each region from the multinomial logistic model, namely, the ratio of observed to expected number of deaths was calculated by using the estimated risk-specific predicted probabilities from the whole cohort. An “adjusted” mortality percentage was calculated by multiplying the standardized mortality ratio by the appropriate observed death percentage across the whole cohort, both overall and for each type of death for each region. The variation between regions was quantified through the estimation of the SD of the mortality percentages for the regions. Absolute differences between the unadjusted and adjusted mortality percentages overall and for each type of death were then calculated for each region. This data analysis was generated by using SAS software, version 9 (SAS Institute, Inc, Cary, NC).
The EPICE study is comprised of 8888 infants that fulfilled the study criteria in the 16 regions included in this analysis. The overall stillbirth and in-hospital mortality rate for the cohort was 27.7%, with an almost twofold difference across the regions ranging from 19.9% in Stockholm, Sweden to 35.9% in Ile-de-France (Table 2) (P < .001). Individual components of the mortality rates showed even wider variation (overall P < .001), with stillbirth rates ranging from 11.5% in Stockholm to 24.2% in Ile-de-France; the high stillbirth rate in Ile-de-France largely explains their high overall mortality rate. The variation in hospital mortality for deaths <12 hours showed an over eightfold difference from 1.2% in the Northern region of the United Kingdom to 9.9% in Wielkopolska (Poland). Stockholm had the lowest rates of mortality for deaths occurring between 12 hours and 7 completed days after birth (0.7%). The highest rate for this category of death was 5.4% in Lisbon, Portugal. In-hospital mortality rates for deaths occurring >7 days after birth ranged from 2.3% in Estonia to 6.8% in Lisbon.
Wide variations were observed in maternal and pregnancy characteristics (Table 3), with the proportion of: mothers aged ≥35 years ranging from 15.0% in the Northern region of France to 45.4% in Lazio, Italy (P < .001); primiparous mothers ranging from 38.0% in Estonia to 67.6% in Northern Portugal (P < .001); and multiple pregnancies from 10.1% in Estonia to 22.1% in Eastern Denmark (P < .001). In terms of infant characteristics (Table 4) the mean GA varied between 28.0 weeks in Wielkopolska, Poland and Ile-de-France and 28.8 weeks in Estonia (P < .001); and the mean birth weight varied between 1052 g in Ile-de-France and 1234 g in Estonia (P < .001). Wide variations were also seen in the proportions of boy infants (P = .019) and SGA infants (P = .043).
Table 5 presents the proportion of the EPICE cohort at each band of GA (overall P < .001). The highest proportions in the lowest gestation band (22–23 weeks) were in Wielkopolska (13.7%) and Ile-de-France (13.2%) compared with only 2.8% in Estonia and 5.6% in Lisbon. Conversely, in the highest GA band (30–31 weeks), proportions varied from 31.4% in the Eastern region of Denmark to 45.2% in Estonia.
GA at birth has a major impact on the mortality rates of a region (Fig 1 A–E). Overall cohort survival by GA band was 4.2% at 22 to 23 weeks (survivors all 23 weeks), 44.6% at 24 to 25 weeks, 70.5% at 26 to 27 weeks, 84.9% at 28 to 29 weeks, and 91.0% at 30 to 31 weeks. Variation in stillbirth and in-hospital mortality rates was greatest at the lowest GA band (22–23 weeks) in terms of absolute mortality rates, but continued across all GAs in terms of the variation in relative rates, with a threefold range in stillbirth rates at 22 to 23 weeks GA (29.3% in Hesse, Germany to 85.7% in the Northern region of Portugal) and at 29 to 31 weeks GA (3.4% in Emilia Romagna, Italy to 9.7% in Ile-de-France). Detailed tables by GA group and region are provided as Supplemental Tables 7–11.
Adjusted ORs with 95% CIs are presented in Table 6 for maternal, pregnancy, and infant characteristics by outcome. Very preterm stillbirths and infants who died after ≥12 hours of neonatal care were significantly less likely to have a primiparous mother than very preterm survivors. Similarly, VPTs born to mothers with PPROM or mothers identified as having a hypertensive pathology (preeclampsia, eclampsia, or HELLP syndrome) had less risk of stillbirth or death after ≥12 hours of life. In addition to the expected inverse relationship between all poor outcomes and increasing GA, stillbirths were less likely to be from a multiple pregnancy compared with survivors and being SGA was associated with stillbirth or death regardless of timing. There were significant differences for all maternal, pregnancy, and infant characteristics presented across the 3 outcomes apart from maternal age.
Variation in the overall stillbirth and in-hospital mortality rates by region reduced little after adjusting for maternal and pregnancy factors from a SD of 4.8% for the distribution of unadjusted regional mortality rates (range, 20.0%–35.9%) to a SD of 4.7% (range, 20.5%–36.6%). However, additional adjustment for infant factors reduced the SD for the distribution of regional mortality rates to 3.9% (range, 20.0%–33.7%), largely due to GA. After adjusting for maternal, pregnancy, and infant factors, the SD of mortality rates by timing of death reduced by approximately one-tenth from 3.3% (range, 10.1%–24.3%) to 3.0% (range, 10.6%–22.0%) for stillbirths; by approximately one-quarter from 2.1% (range, 1.2%–9.9%) to 1.5% (range, 1.3%–7.8%) for deaths at <12 hours of age; and was 2.1% both before adjustment (range, 4.5%–12.2%) and after adjustment (range, 4.6%–11.4%) for deaths at ≥12 hours of age. The differences between regions all remained statistically significant (P < .001).
The impact of maternal, pregnancy, and infant characteristics on the timing of death for each region are shown in Fig 2, which illustrates the absolute differences in stillbirth and in-hospital mortality rates between the overall rates for the total cohort and the rates for each region. Adjustment for these characteristics changed the mortality estimates for Ile-de-France, Wielkopolska, and Estonia for stillbirths and deaths at <12 hours but had little effect on the UK region of Yorkshire and Humber for any mortality outcome. Deaths at ≥12 hours showed the smallest variation after risk adjustment for maternal, pregnancy, and infant characteristics.
This analysis identified wide variations in very preterm stillbirth and in-hospital mortality rates across Europe in the EPICE cohort from 2011 to 2012 by using standardized data collection procedures after accounting for population demographics, case mix, and timing of death and confirms previous findings.9 We show that these variations are evident for all outcomes, although variation is greatest (in relative terms) for the earliest deaths, predominantly those occurring <12 hours after birth. Although birth registration differences or policies for ascribing whether an infant is recorded as a stillbirth or neonatal death may account for some of the variation for the earliest GAs,12 this is not the case for births at later gestational ages (≥27 weeks), where GA plays a lesser role in the determination of potential viability. Nevertheless, because the highest levels of variation are found for the deaths that occur closest to birth, it suggests that unit and regional policy differences may account for a proportion of this difference, possibly representing the full spectrum of interventions for compassionate and palliative care across regions. This difference in policies may contribute to the variation seen in the later deaths because more proactive initiation of treatment may lead to increased survival in those infants who were previously thought to be at highest risk of mortality.19
To assess the full picture of outcomes associated with VPTs, it is important to investigate outcomes for the total cohort, including all stillbirths, intrapartum, and labor ward deaths as well as deaths on neonatal units to reduce the selection bias introduced by management policies.20 One example of this is the variation in the amount of infants admitted to a unit for comfort care when early death is anticipated, a policy adopted by the Polish region of Wielkopolska, whereas other regions may allow such deaths to occur in the delivery suite and, in some circumstances, they may fail to appear in official statistics. In many countries, decision-making about potential viability takes place in the maternity unit, reflecting differences in obstetric and perinatal practices in addition to health system factors and the impact of cultural attitudes.
A number of maternal and pregnancy characteristics are known to be associated with rates of very preterm delivery and the subsequent risk of mortality. In this analysis, we have noted both demographic variations in terms of maternal age, proportion of primiparous mothers, and pregnancy complications (hypertensive pathology and PPROM). However, despite variations in these characteristics across the regions, they had a minimal effect on the regional variation in VPT mortality rates.
Similarly, infant characteristics that are known to have an association with mortality were investigated, and variations in the proportion of VPTs that were boys, from multiple births, or had fetal growth restriction were noted across the regions. When infant characteristics were added into the risk adjustment model, the variation in overall stillbirth and in-hospital mortality rates across the study regions reduced by nearly one-quarter (23.3%), indicating the lack of homogeneity in the populations of infants in the different regions.
Investigation of the effect of adjusting for maternal, pregnancy, and infant characteristics by the timing of death shows that interregional variation persists at all 3 specified timings. This is contrary to the findings of the EXPRESS study in Sweden,11 which concluded that in the most extreme preterm infants, any regional differences were established <12 hours after birth, suggesting they could be related to variations in immediate perinatal practices. The continued wide variation in mortality for infants ≥12 hours of age in the EPICE cohort could indicate that there is variation in the quality, access, and provision of neonatal care across Europe. The profile of VPT is changing over time, with increasing numbers of earlier gestation infants, partly due to increasing iatrogenic early delivery as neonatal care has improved.21–23 The wide variation in mortality rates seen in the deaths at ≥12 hours of age may be due to the increasing complexity of these extremely preterm births.
Strengths of the Study
This study used a population-based prospective cohort study design with standardized data collection and definitions. Active validation of regional case ascertainment was carried out to ensure inclusion of all cases that fulfilled the study criteria, irrespective of local registration criteria for births and deaths. All terminations of pregnancy and major congenital anomalies were excluded to allow for other differences in the legislative practices between countries, although differences in perinatal management and policies for any other congenital anomaly may still have had an impact on our results. Our main analysis was carried out by using the timing of the death rather than admissions to the NICU, thus reducing the impact of the variations in unit and regional policies with respect to the treatment of this group.19
Limitations of the Study
There is no standardized method of ascertainment of socioeconomic status across Europe that can be derived from medical records24 and, therefore, no adjustment has been made for this particular population characteristic, which is known to be associated with preterm birth. However, although there is a doubling of the risk of VPT between the least and most deprived groups in populations,25 variations in the outcome of VPTs after admission to neonatal care do not appear to be explained by socioeconomic differences.26,27
Although the calculation of expected date of delivery for the vast majority of births is now assessed by ultrasound measurement, this is not always the case. In addition, there is variation in the dating formulae used for sonographic estimation of GA between regions,28 which could lead to systematic variation. In this study, we have used the best estimate for GA based on ultrasound as the gold standard where possible and have identified a hierarchy for the calculation of GA when a number of estimates were provided.
The aim of this study was to determine what proportion of the regional variations in stillbirth and in-hospital mortality rates in European regions persists after adjusting for population demographics and case mix. Over 75% of the regional variation could not be accounted for by these factors. Investigation of the timing of death showed that approximately one-tenth of the variation in the stillbirth rate was accounted for by maternal, pregnancy, and infant characteristics. However, one-quarter of the variation in the mortality rate for deaths occurring <12 hours after birth, but almost none of the variation in the mortality rate for deaths occurring ≥12 hours after birth, seems to be explained by these same characteristics. This finding suggests that there may be an inequity in the quality of care provision and treatment of very preterm infants across Europe.
EPICE Research Group: Flanders, Belgium: H. Cammu, E. Martens, G. Martens, and P. Van Reempts; Eastern region, Denmark: K. Boerch, A. Hasselager, L. Huusom, O. Pryds, and T. Weber; Estonia: L. Toome and H. Varendi; Burgundy, Ile-de France, and Northern region, France: P.Y. Ancel, B. Blondel, M. Bonet (Research Coordinator; INSERM), A. Burguet, J.B. Gouyon, P.H. Jarreau, P. Truffert, and J. Zeitlin (Project Leader; INSERM); Hesse, Germany: R.F. Maier, B. Misselwitz, and S. Schmidt; Saarland, Germany: Emilia Romagna, Italy: L. Gortner; D. Baronciani, G. Gargano; Lazio, Italy: R. Agostino, D. DiLallo, and F. Franco; Le Marche, Italy: V. Carnielli, M. Cuttini, and I. Croci; Eastern and Central regions, Netherlands: C. Hukkelhoven, M. Hulscher, L. Kollée, C. Koopman-Esseboom, and A. Van Heijst; Wielkopolska, Poland: J. Gadzinowski and J. Mazela; Lisbon and Tagus Valley, Portugal: L.M. Graça and Md.C. Machado; Northern region, Portugal: H. Barros, M.R.G. Carrapato, and T. Rodrigues; Stockholm, Sweden: A.K. Bonamy, M. Norman, and E. Wilson; East Midlands and Yorkshire and Humber, United Kingdom: E. Boyle, E.S. Draper, and B.N. Manktelow; and Northern region, United Kingdom: A.C. Fenton and D.W.A. Milligan.
- Accepted January 17, 2017.
- Address correspondence to Elizabeth S. Draper, PhD, Department of Health Sciences, College of Medicine, Biological Sciences and Psychology, University of Leicester, Centre for Medicine, University Rd, Leicester LE1 7RH. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: This work was supported by the European Union’s Seventh Framework Programme ([FP7/2007-2013]) under grant agreement 259882. Additional funding was received in the following regions: France (French Institute of Public Health Research/Institute of Public Health and its partners the French Health Ministry, the National Institutes of Health and Medical Research, the National Institute of Cancer, and the National Solidarity Fund for Autonomy; grant ANR-11-EQPX-0038 from the National Research Agency through the French Equipex Program of Investments in the Future; and the PremUp Foundation); Poland (2012–2015 allocation of funds for international projects from the Polish Ministry of Science and Higher Education); Sweden (Stockholm County Council [ALF project and Clinical Research Appointment (Anna-Karin Bonamy)] and by the Department of Neonatal Medicine, Karolinska University Hospital); and the United Kingdom (funding for the Neonatal Survey from Neonatal Networks for East Midlands and Yorkshire and Humber regions). The funders had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Blencowe H,
- Cousens S,
- Oestergaard MZ, et al
- Roberts D,
- Dalziel S
- Costeloe KL,
- Stacey F,
- Hennessy E et al
- Field DJ,
- Dorling JS,
- Manktelow BN,
- Draper ES
- McCall EM,
- Alderdice F,
- Halliday HL et al
- Draper ES,
- Zeitlin J,
- Fenton AC, et al; MOSAIC research group
- Zeitlin J,
- Draper ES,
- Kollée L, et al; MOSAIC research group
- Smith L,
- Draper ES,
- Manktelow BN,
- Pritchard C,
- Field DJ
- Draper ES
- Larroque B,
- Bréart G,
- Kaminski M, et al; Epipage study group
- Zeitlin J,
- Papiernik E,
- Bréart G,
- Draper E,
- Kollée L; MOSAIC Research Group
- Evans DJ,
- Levene MI
- Smith LK,
- Draper ES,
- Manktelow BN,
- Field DJ
- Copyright © 2017 by the American Academy of Pediatrics