Article Text


Changes in care and outcome of very preterm babies in the Parisian region between 1998 and 2003
  1. Jennifer Zeitlin1,
  2. Pierre-Yves Ancel1,
  3. Dominique Delmas1,
  4. Gérard Bréart1,
  5. Emile Papiernik2,
  6. the EPIPAGE and the MOSAIC Ile-de-France Groups*
  1. 1INSERM, UMR S953, IFR 69, Research Unit on Perinatal Health and Women's Health, Paris, France
  2. 2Université Paris V Réné Descartes et Maternité de Port-Royal, Assistance-Publique Hôpitaux de Paris, Paris, France
  1. Correspondence to Jennifer Zeitlin, INSERM U953, Hôpital Saint Vincent de Paul, 82 av Denfert Rochereau, Paris 75014; jennifer.zeitlin{at}


Objective To assess evolution in the care and health of very preterm babies between 1998 and 2003 after implementation of a regionalisation policy in France.

Design Comparison of two population-based cohorts.

Setting The Parisian region.

Patients All live births at 24–31 weeks of gestation in 1997 (EPIPAGE study, n=488) and in 2003 (MOSAIC study, n=580).

Interventions Implementation of regionalised perinatal networks.

Main outcome measures In-hospital mortality and morbidity, including intraventricular haemorrhage (IVH) grade III and IV, cystic periventricular leucomalacia (PVL) and bronchopulmonary dysplasia (BPD).

Results Over this period, babies born in level III units rose from 67% to 77% and use of antenatal corticosteroids, indicated deliveries and surfactant increased. In-hospital mortality and IVH grades III/IV declined, ORs of 0.66 (95% CI 0.46 to 0.95) and 0.27 (95% CI 0.15 to 0.47), respectively, while PVL and BPD stayed constant. The rate of very preterm babies discharged alive per 1000 total births increased by 18%, but declined for babies with severe brain lesions.

Conclusions The authors found improvements in mortality and morbidity for very preterm babies and changes in their care over a 6-year period following reinforcement of regionalisation policies.

Statistics from


The survival of very preterm babies has improved markedly since the 1980s because of medical innovations and in particular antenatal steroids and surfactant use, as well as policies to regionalise perinatal care.1,,5 Since the mid-1990s, there have been no similar advances for the care of these babies and improvements in outcome may have slowed or stopped.6,,9 No decrease in mortality or morbidity was observed for very low birthweight babies in the USA between the mid-1990s and 19996 7 and one study in Delaware reported increased mortality between 1995 and 2000.10

One factor affecting the prognosis of very preterm babies is whether delivery occurs in a maternity unit with on-site neonatal intensive care (level III unit). Programmes to regionalise the delivery of very preterm babies in these units began in the 1980s in the USA and Canada. Regionalisation was implemented in the late 1990s in France and some other countries of Europe and, in 2004, regionalised neonatal networks were established in the UK.11,,14 In the USA deregionalisation during the 1990s kept this issue on the research and policy agenda.15 16

What is already known on this topic

  • Mortality and morbidity for very preterm babies declined over past decades, but recent data indicate that improvements may have slowed.

  • A wide body of research finds that delivery in tertiary centres (often labelled level III units) improves prognosis after very preterm birth.

  • Improving the regionalisation of care is a continuing goal in many countries and may make it possible to improve outcomes for very preterm babies.

What this study adds

  • The implementation of perinatal networks was associated with changes in the place of birth of very preterm babies and their care in the Parisian region.

  • Within a six-year period, mortality and grade III and IV intraventricular haemorrhages decreased significantly.

  • The rate of babies discharged home after very preterm birth per 1000 total births increased 18%, but decreased for babies with severe brain haemorrhaging.

In the early 1990s, most very preterm babies in France were born outside of level III units and transported after delivery to neonatal intensive care units.17 Practices began to change at this time and by the late 1990s about two-thirds of babies under 32 weeks were born in level III centres.18 19 In 1998, an official policy required all non-level III maternity units to sign conventions with a regional reference level III for the management of high-risk deliveries.20 Level III maternity units were defined as having an on-site neonatal intensive care unit (NICU) and at least one neonatologist present 24 hours a day. Other maternity units were level I maternity units with no neonatal ward or level II units with special care nurseries that could provide care for babies born at 32 weeks of gestation or later.

In this analysis we explore the evolution in obstetric and neonatal care as well as mortality and short-term morbidity for babies born between 24 and 31 weeks of gestation in the Parisian region between 1997, one year before the regionalisation policy, and in 2003, five years later.


This analysis combines data from two population-based cohort studies on very preterm babies in the Parisian region.

The first study, EPIPAGE, aimed to study long-term outcomes of very preterm babies and has followed its cohort up to 8 years.21 The study included births between 22 and 32 weeks of gestation in nine French regions in 1997, including the Parisian region (districts of Paris, Val-de-Marne, Hauts-de-Seine and Seine-Saint-Denis). For each birth, questionnaires were filled in at the maternity and neonatal units on the family, pregnancy, delivery and the health of the baby during hospitalisation. Completeness of inclusion was verified by randomly selecting maternity units and checking the inclusions sheets with the birth registers in the maternity hospitals.

The second study, MOSAIC (Models for OrganiSing Access to Intensive Care for Very Preterm Births), is a European collaboration to study the impact of the organisation of care for very preterm births on access to care and health outcomes.22,,24 The study included all births between 22 and 31 weeks of gestation born in 10 European regions. Babies were followed up to discharge home. The Ile-de-France region, which includes the Parisian region, was one of the MOSAIC regions. Data were collected from medical files in each neonatal unit where the baby was hospitalised. For deaths before admission to neonatal care, the questionnaire was completed using the medical records in the maternity unit. Completeness of inclusion was verified by corroborating lists of babies in the study with birth registries in every maternity unit.

For this analysis, we defined a common population from both studies which included babies born alive between 24 and 31 weeks of gestation between February and July in the Parisian region. Babies below 24 weeks were excluded because in most maternity units these babies were considered below the limit of viability and pregnant women at this gestation would not be transferred systematically to specialised centres for delivery.25

Variables with the same definition in both studies were maternal age, gestational age, birth weight, sex, plurality, selected pregnancy complications (preterm rupture of membranes or preterm labour, haemorrhage and hypertensive diseases), antenatal corticosteroids administration, level of care of the maternity of birth, onset of delivery by caesarean section and surfactant administration (although not when it was given or the indication). In both studies, gestational age was based on the best obstetric assessment, using information on last menstrual period and ultrasound measures. First trimester ultrasound scans are part of routine antenatal care in France. Small for gestational age was defined as birth weight under the 10th percentile of intrauterine curves using methods proposed by Gardosi adapted for the French population.26

Outcome variables were in-hospital mortality, defined as death after live birth in the maternity unit or death in a neonatal unit before discharge home, and neurological and respiratory morbidity. Neurological morbidity was defined as whether the baby had a diagnosis of intraventricular haemorrhage (IVH) grade III or IV using the classifications by Papile et al27 or cystic periventricular leucomalacia (PVL). This rate was computed for all babies that were admitted to neonatal intensive care. Cases with missing data were excluded from the analysis (19/470 for the EPIPAGE cohort, 9/551 for the MOSAIC cohort). Respiratory morbidity was bronchopulmonary dysplasia (BPD) defined as oxygen dependency or ventilation at 36 weeks of gestational age and computed for babies discharged home from hospital. There were no missing values for the EPIPAGE cohort and nine for the MOSAIC cohort.

We compared the characteristics and care of the two cohorts for all babies as well as by gestational age group, 24–27 weeks of gestational age and 28–31 weeks of gestational age. Less than 28 weeks is a conventional cut-off for extreme prematurity and is recommended by WHO.28 We further explored the evolution of medical practices in level III and in non-level III units. Outcomes were compared between the two periods as well as by individual week of gestation. For analyses by week of gestation, data on morbidity for babies of 24 and 25 weeks were combined because of the small numbers in each group. Odds ratios for the period effect were calculated using logistic regression adjusting for population characteristics (gestational age, birth weight, sex, multiple gestation, medical complications of pregnancy and maternal age) for the entire sample and within gestational age groups. We also ran a logistic model adding level of care of the maternity unit of birth and antenatal corticosteroid administration. Rates of very preterm birth and of very preterm babies discharged home from hospital were calculated using as a denominator the total births in the study area in the two time periods provided by the National Institute for Statistics and Economic Studies (INSEE) and compared with rate ratios and 95% CI. Analyses were carried out using Stata 8.0 SE (StataCorp LP, College Station, TX, USA).


In 1997, there were 488 very preterm live births between 24 and 31 weeks of gestation for 51 390 total live births in the Parisian region, a rate of 9.5 per 1000 live births. The rate in 2003 was slightly higher: 10.7 per 1000 (580 very preterm births and 54 383 total live births), corresponding to an increase of 12% (rate ratio 1.12 (95% CI 0.99 to 1.27)).

Table 1 presents the characteristics of the two cohorts of very preterm babies. Their characteristics were similar, with the exception of a slight decrease in gestational age, birth weight and multiples in the later cohort. Average maternal age increased over the two periods. The prevalence of selected complications of pregnancy was similar, except for haemorrhage, which was more frequent in the 1997 cohort.

Table 1

Comparison of maternal and infant characteristics in the two cohorts of live very preterm births by gestational age group

In 2003, very preterm babies were more likely to be born in level III units (76.6% vs 67.2%) as shown in table 2. They also were more likely to receive antenatal steroids, be delivered by caesarean before onset of labour and to receive surfactant. These changes were observed for the two gestational age groups, although for extremely preterm births the difference in antenatal corticosteroids and caesareans before onset of labour was not statistically significant.

Table 2

Evolution of place of delivery and selected medical practices for very preterm births between 1997 and 2003 by gestational age group and level of care

In level III units, antenatal corticosteroids use rose from 85.4% to 94.0% between the two periods as did caesareans before labour (38.9–48.0%). After 27 weeks of gestation, 97.5% of babies received corticosteroids in 2003 versus 86.7% in 1997. Under 28 weeks, the proportions receiving corticosteroids also rose, but the difference was not statistically significant. Surfactant administration rose in both types of units (table 2). Mortality declined from 19.1% to 14.8% (p=0.065). Once adjusted for population characteristics, the period effect of being born in 2003 compared with 1997 on mortality became significant: 0.66 (95% CI 0.46 to 0.95) (table 3). The incidence of grade III/IV IVH decreased from 11.3% to 4.1% or an adjusted period effect of 0.27 (95% CI 0.15 to 0.47). PVL and BPD did not change.

Table 3

Evolution in very preterm mortality and morbidity between 1997 and 2003

Figures 1 and 2 present the evolution of mortality and morbidity by gestational age. The greatest changes in mortality and IVH occurred between 25 and 28 weeks of gestation, although decreases are observed at all gestational ages, except 25 and 31 weeks. The period effect was similar after adjustment for level of care and antenatal corticosteroids (OR for mortality 0.66 (0.45 to 0.98) and for IVH 0.32 (0.18 to 0.58); data not shown in tables).

Figure 1

In-hospital mortality by gestational age in 1997 and 2003.

Figure 2

Morbidity by gestational age in 1997 and 2003. (A) Intraventricular haemorrhage grades III and IV. (B) Cystic periventricular leucomalacia. (C) Bronchopulmonary dysplasia.

There were 495 survivors for 55 024 total live births in 2003, a rate of 8.9 per 1000 live births, versus 395 survivors for 51 910 total births in 1997, a rate of 7.6 per 1000 live births (rate ratio of 1.18 (1.03 to 1.35)). In contrast, babies discharged from hospital with IVH grades III and IV and PVL decreased to 19 for 55 024 live births in 2003, a rate of 0.3 per 1000, from 34 for 51 910 live births in 1997, a rate of 0.6 per 1000 (RR 0.53 (0.30 to 0.92)).


After the national regionalisation policy was put into place in France in 1998, the proportion of babies born in level III units rose in the Parisian region from 67% in 1997 to 77% in 2003. There were significant modifications in the care of very preterm babies and a decrease in rates of mortality and of grade III–IV IVH. The two other measures of morbidity (PVL and BDP) stayed constant. Between the two periods, the rate of very preterm births per 1000 total births increased slightly.

The decreases in mortality or morbidity were not explained by evolutions in the characteristics of very preterm babies; the 2003 cohort had slightly lower gestational ages and birth weight and adjusted period effects were greater than crude period effects. The 12% increase observed in the very preterm live birth rate between the two periods may be partially due to more aggressive intervention and also be related to regionalisation efforts. More active intervention and more caesarean sections have been associated with an increase in the rate of very preterm birth.29

It is also unlikely that the differences in mortality and IVH resulted from variability in protocols or definitions in the two studies. In both studies, in-hospital mortality included all deaths in the labour ward or any neonatal unit before discharge home. Completeness of the inclusions was verified and almost all babies had cranial ultrasounds. Discussions with clinicians in participating hospitals provided no evidence for a change in screening protocols. While improvements in imaging techniques may affect detection of some white matter anomalies and bleeding in the brain, both cystic PVL and grade III and IV IVH are easily identifiable and unlikely to be affected by technological improvements. Furthermore, if there were an effect, it would probably result in greater detection of these anomalies.

The proportion of babies born in level III units rose 10 points to 77%. While this is a significant increase, there was already substantial regionalisation in the Parisian region in 1997.18 19 More babies received antenatal corticosteroids, were born by caesarean section and were administered surfactant, reflecting the greater proportion of babies delivered in level III units where these practices were more frequent, but also modifications in practices within level III units. Other studies have observed that regionalisation programmes were associated with an increasing survival advantage for babies born in level III units30 and this could be related in part to changes in the care provided in these units.

While our results suggest that increased regionalisation changed healthcare provision and health, as found in previous studies,31 32 we cannot rule out the impact of other unmeasured factors, such as improved neonatal or obstetric care. Nor could we identify the medical practices that had the greatest effect. The period effect was not modified by the inclusion of maternity of birth and antenatal corticosteroids in logistic regressions. As regionalisation policies succeed, they change more than just the site of birth of preterm babies, but affect transfer decisions, and thus the characteristics of the babies transferred, as well as the care received in level III centres. These multiple changes are difficult to capture in multivariable models.

Our study found a decline in rates of IVH, but not cystic PVL. One possible explanation is that available interventions have a greater impact on IVH. Antenatal corticosteroids decrease rates of IVH33 and place of birth, neonatal transport and the structural characteristics of neonatal intensive care units are also associated with IVH.32 34 Fewer studies focused on PVL, which may be less responsive to these interventions.35 A selection effect may also be at work: unlike IVH, PVL is often detected after several weeks and babies that died in the earlier period would not have been diagnosed with PVL.

Measuring the trade-offs between mortality and morbidity is important for evaluating the impact of regionalisation.36 While decreases in morbidity have accompanied increasing survival,1 4 6 37 in some populations morbidity increased3 38 39 or remained constant.2 40 41 Studies on the impact of mortality declines on longer-term outcomes, such as cerebral palsy, have also provided inconsistent results.36 We found that short-term morbidity did not rise and that IVH declined. The rate of very preterm babies surviving to discharge rose by 18%; these babies had similar rates of respiratory morbidity, but fewer diagnoses of severe cerebral lesions. We could not assess the impact of these evolutions on longer-term outcome and it remains uncertain whether the decrease in severe neurological lesions, a major risk factor for cerebral palsy and other impairments, would offset the expected rise in the number of children with handicaps related to the larger size of the population at risk.


Significant changes in care and outcome of very preterm babies followed the reinforcement of regionalisation programmes in the Parisian region. These changes occurred over a short timespan showing that within a similar sociomedical context, outcomes can differ substantially. This result is interesting for the interpretation of variations in very preterm outcomes between countries with similar levels of development and medical care provision.23 These results support a continued focus on improving access to level III perinatal centres for these babies.


View Abstract


  • * EPIPAGE Ile-de-France Group: Pierre-Yves Ancel, Gérard Breart, Michel Dehan, Monique Kaminski, Christiane Du Mazaubrun, Michel Vodovar, Marcel Voyer, Véronique Zupan-Simunek; MOSAIC Ile-de-France Group: Gérard Breart, Jean-Louis Chabernaud, Dominique Delmas, Pierre-Henri Jarreau, Emile Papernik, Jennifer Zeitlin.

  • Funding This MOSAIC project was partially funded by a grant from the European Commission Research Directorate (QLG4-CT-2001-01907). The EPIPAGE study was supported by grants from INSERM (National Institute of Health and Medical Research), the Directorate General for Health of the Ministry for Social Affairs, Merck-Sharp and Dhome-Chibret, Medical Research Foundation, HAS (French National Authority for Health) and “Hospital Program for Clinical Research 2001 number AOMO1117” of the French Ministry of Health.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Commission Nationale de l’Informatique et des Libertés (the French data protection agency).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.