Improving Perinatal Regionalization by Predicting Neonatal Intensive Care Requirements of Preterm Infants: An EPIPAGE-Based Cohort Study
OBJECTIVE. Perinatal regionalization has been organized into 3 ascending levels of care, fitting increasing degrees of pathology. Current recommendations specify that very premature infants be referred prenatally to level III facilities, yet not all very preterm neonates require level III intensive care. The objective of our study was to determine the antenatal factors that, in association with gestational age, predict the need for neonatal intensive care in preterm infants, to match the size of birth with the level of care required.
METHODS. Data were analyzed from a cohort of very preterm infants born in nine French regions in 1997. We defined the need for neonatal intensive care as follows: (1) the requirement for specialized management (mechanical ventilation for >48 hours, high frequency oscillation, or inhaled nitric oxide) or (2) poor outcome (transfer to a level III facility within the first 2 days of life or early neonatal death). Triplet pregnancies and pregnancies marked by fetal malformations or intensive care requirements for the mother before delivery were excluded.
RESULTS. We focused our study on 1262 neonates aged 30, 31 and 32 weeks’ gestation, where the need for intensive care was 42.8%, 33.2%, and 22.8%, respectively. Multivariate analysis showed that the risk factors for intensive care requirement with low gestational age were twin pregnancies, maternal hypertension, antepartum hemorrhage, infection, and male gender. Antenatal steroid therapy and premature rupture of membranes were protective factors against intensive care requirement.
CONCLUSION. Infants <31 weeks' gestation should be referred to level III facilities. From 31 weeks’ gestation, some infants can be safely handled in level IIb facilities. However, the quality of perinatal regionalization may only be fully assessed by long-term follow-up.
Providing adequate care to preterm neonates is an important issue for perinatal medicine. The incidence of preterm deliveries has increased in all developed countries with quite similar trends.1 In France, very preterm deliveries accounted for 1.9% of all pregnancies in 2003.2 This increased incidence results from a variety of medical factors that also modify neonate adaptation to extrauterine life. Very preterm deliveries are also linked to social factors.3
To provide prompt and appropriate treatment to these premature infants, perinatal regionalization in France and other countries has organized care facilities in 3 ascending levels4, 5 corresponding with increasing degrees of pathology. These levels of care were well defined by the American Academy of Pediatrics.6 Level I facilities offer neonatal care to healthy term newborns. Level II facilities have 2 subgroups: level IIa can care for preterm neonates >32 weeks' gestation and weighing >1500 g, and level IIb offers short-term mechanical ventilation for infants with respiratory distress syndrome. Finally, level III facilities are NICUs capable of handling very preterm infants and infants needing ventilatory support for a long period of time, high frequency ventilatory oscillation, or nitric oxide inhalation. Infants of all gestational ages (GAs) with severe malformations or morbidities requiring surgical management should also be referred to level III maternity units. It is noteworthy that these perinatal care facilities are defined by their highest level of care, although they also provide less complex care.
Because care provided to very preterm infants is essential from birth to avoid poor outcome,7–9 it is mandatory that births be planned in facilities fitting the infant's level of pathology and care requirements.10 Perinatal recommendations were based previously on birth weight. Nowadays, routine early fetal ultrasonography in France and other countries enables GA to be precisely defined. Recommendations for management of preterm infants now stress GA over birth weight. French perinatal regionalization suggests that births before 33 weeks' gestation, as high-risk births, be prenatally referred to level III facilities by in utero transfers.11
However, not all preterm neonates <33 weeks' gestation require intensive care. Thus, some very preterm deliveries could be managed in level IIb facilities closer to the parents' home, avoiding pointless in utero transfers, saturation of level III facilities, and higher costs of centers providing advanced technologies and very specialized staffs.
Indeed, GA alone does not provide sufficient information for prediction of the infant's state at birth, because many other factors modify infant adaptation and outcome.10 Thus, the objective of our study was to determine the antenatal factors other than GA that are linked to the need for neonatal intensive care (NIC) in premature infants. To this end, we used the data of the EPIPAGE population-based cohort study.5, 12, 13
Data were analyzed from the French EPIPAGE (epidemiologic study of very preterm infants) cohort. The French EPIPAGE cohort included all preterm infants aged 22 to 32 weeks' gestation and a sample of preterm infants aged 33 to 34 weeks' gestation born in 1997 in 9 French regions. The aim of the EPIPAGE study was to assess mortality trends and neurologic outcomes of preterm infants at 5 years of age. The methodology of this study has been described in previous articles.12, 13
We defined the study population as mother-infant pairs. These pairs were included in our study if the infants were of a GA at which <50% of the neonates required NIC. Below this limit, GA by itself indicates the need for antenatal transfer to a level III unit. Only infants who were living at birth and born in a maternity unit were included in this analysis.
Pregnancies needing specialized monitoring or management in level III facilities (ie, amnioinfusion, amnioreduction, and intrauterine fetal transfusion), triplet pregnancies, mothers suffering from diseases requiring care provided only in level III units (ie, dialyzed mothers and mothers suffering from drepanocytic anemia or from heart or lung insufficiency), infants with congenital malformations, and infants referred to a level III facility after 2 days of life were excluded from this analysis.
Because the objective of our study was to aid decision-making in the referral of pregnant women at risk of preterm delivery, both mother and infant were excluded when one met exclusion criteria. In the case of twin pregnancies, the sickest infant (ie, the one most likely to have led to an in utero transfer) was retained in the analysis.
Defining the Need for NIC
We defined the need for NIC as meeting 1 of the 2 following criteria: (1) the use of specialized management (mechanical ventilation for >2 days, ventilation by high-frequency ventilatory oscillation, or nitric oxide for refractory hypoxemia) or (2) poor outcome (early neonatal death or transfer from a level IIb to a level III neonatal unit during the first 2 days of life for more intensive care).
Factors studied that were linked to the fetus included GA,14, 15 gender,16 intrauterine growth restriction ([IUGR] defined as birth weight beneath the third percentile on Yudkin's curves17 not associated with maternal hypertension),12 and fetal distress (defined as abnormal Doppler or fetal heart rate before labor onset). Factors linked to the mother were severe hypertension18 (defined as maternal hypertension associated with fetal IUGR or preeclamptic toxemia),19 antepartum hemorrhage,10 smoking,20 and history of preterm delivery or early neonatal death. Finally, factors linked to the pregnancy included pregnancy induced by assisted reproductive technology, twin pregnancy,21 antenatal steroid therapy,22 premature rupture of membranes (PROM),23 oligohydramnios or hydramnios, and infection (defined as probable infection suspected immediately after birth).
We first performed a univariate analysis to identify factors associated with the need for NIC, using χ2 tests. Then, known risk factors and variables for which P ≤ .05 were selected for multivariate analysis. Multivariate associations were analyzed with logistic regression models. We looked for significant interactions among the factors entered in the multivariate models. We adjusted our models for the region and the care level of the maternity unit at birth. The results of the logistic models were expressed as odds ratios (OR) with 95% confidence intervals (CI). The fits of the logistic models were assessed with the Hosmer and Lemeshow goodness-of-fit test, with P <.05 taken as evidence of a statistically significant difference between observed and predicted need for NIC.
We then determined the frequency of the need for NIC among infants aged 31 and 32 weeks' gestation, according to the factors significantly associated with the need for NIC in the multivariate analysis. All of the analyses were performed with SAS software, version 9.1 (SAS Institute, Cary, NC).
Description of the Cohort
Between January 1, 1997, and December 31, 1997, 3537 very preterm neonates were enrolled in the EPIPAGE study. According to our definition of requirement for NIC, >50% of infants aged ≤29 weeks' gestation required intensive care (Fig 1). Therefore, 1740 mother-infant pairs were selected for this study. All of these infants were aged 30 to 32 weeks' gestation, were living at birth, and received care in a delivery room and a neonatal unit.
Of the mother-infant pairs selected, 247 were excluded because of antenatal need for care in a level III facility (n = 58), triplet pregnancy (n = 71), infant congenital malformations (n = 49), severe maternal diseases (n = 21), or neonatal transfer after 2 days of life to a level III unit (n = 48). Of twin pregnancies, 231 infants were excluded to retain the sickest infant in the analysis or because the cotwin had met exclusion criteria. The remaining 1262 mother-infant pairs were included in this analysis.
The general characteristics of the study population are summarized in Table 1. Very few data were missing for the studied factors with the exception of tobacco use, for which 19% of the data were missing.
More than 70% of the preterm infants had antenatal steroid therapy. PROM was encountered in ∼30% of pregnancies, maternal hypertension in 20%, and maternal diabetes in 4%. The prevalence of the different diseases studied did not differ with GA, with the exception of oligohydramnios and hydramnios, maternal diabetes, and maternal hypertension, which were significantly more frequent in the youngest preterm infants.
Prevalence of Intensive Care Requirement and Associated Factors
There was an inversely proportional linear trend for the need for intensive care with increasing GA: 42.8%, 33.2%, and 22.8% at 30, 31, and 32 weeks' GA, respectively (Fig 1). In the univariate analysis, factors associated with the need for NIC are detailed in Table 1 and their ORs presented in Table 2. The need for NIC was significantly increased in the youngest preterm infants, twin pregnancies, and males and in pregnancies marked by fetal distress syndrome, maternal severe hypertension, antepartum hemorrhage, and infection. Conversely, the need for NIC was significantly decreased with antenatal steroid therapy, PROM, and maternal smoking.
Multivariate analysis performed on the whole population showed that GA remained significantly linked to the need for NIC. The risk factors and protective factors were the same as in the univariate analysis with the exception of fetal distress syndrome, which was no longer associated with an increased NIC requirement in the adjusted model (Table 3). According to the multivariate analysis, among single infants aged 31 and 32 weeks' GA, the need for NIC was calculated as 14.8% and 14.4%, respectively, for infants with antenatal steroid therapy but without severe hypertension, antepartum hemorrhage, or infection.
Infants born of smoking women (n = 356) had a significantly decreased risk of intensive care requirement compared with infants born of nonsmokers (24.8% vs 32.4%; P = .01), but this protective effect disappeared when only infants with antenatal steroid therapy were taken into account (22.1% vs 27.1%; P = .15). Tobacco use had complex interactions with several medical factors. Risk factors for requirement of NIC among smokers are presented in Table 4. Among infants of smoking mothers, the need for NIC was significantly increased for the youngest preterm infants, twin pregnancies, male infants, and IUGR. This need was significantly decreased with PROM (Table 4).
Between 30 and 32 weeks' gestation, GA is an important factor predicting the need for NIC, yet it is not the only factor. Twin pregnancies, male fetuses, severe maternal hypertension, antepartum hemorrhage, and infectious disease were also linked with significant increases in the need for NIC. In contrast, antenatal steroid therapy and PROM were associated with significant decreases in intensive care requirement. Moreover, the requirement for NIC was not significantly higher among infants prenatally transferred to a level III facility.
Antenatal factors linked with respiratory distress syndrome or mortality in preterm infants have been considered in previous articles.10, 21 However, to the best of our knowledge this is the first study to determine antenatal factors that, in association with GA, modify the need for NIC.
Among medical factors under physician control, prenatal steroid therapy significantly prevents the need for NIC for very preterm neonates. In 1997, 30% of mothers who delivered between 30 and 32 weeks' gestation did not receive steroids, yet 33.8% (n = 115) of these women were admitted to the hospital ≥24 hours before delivery.
PROM seems to protect preterm infants from poor short-term outcome. However, women presenting with PROM are carefully monitored, which could at least partially explain the protective effect.
The results of the univariate analysis suggest a decreased intensive care requirement for infants born of smoking mothers. Moreover, it seems that antenatal steroid therapy, a well-known protective factor, could have decreased efficacy in smokers.20 In contrast, IUGR in smokers is associated with an increased risk of intensive care requirement. Nevertheless, data concerning maternal smoking was lacking for 19% of the included pregnancies, so these analyses concerned only 1022 mother-infant pairs. This particular effect of tobacco use on preterm infants' immediate outcomes is not well understood and needs to be confirmed by additional studies. Multivariate analysis was performed separately for smokers. Finally, tobacco use may not be currently considered in the decision-making process to refer a pregnant woman to a level III unit.
The importance of perinatal regionalization has been emphasized in many studies. The concept of “critical mass”24 guarantees expertise for preterm infants requiring specialized care. According to this principle, the sickest preterm infants should be born in facilities possessing the most advanced technologies and most experienced staffs. Yet the number of high-technology centers depends on the health care resources of the country. Therefore, the prenatal referral of infants to a level III NICU should be closely linked with their needs to avoid saturation of level III maternity wards2 and unnecessary expense.
In this study, 31% of infants born in a level III facility after in utero transfer needed intensive care. To improve the prediction of intensive care requirement at birth, it is important to focus on the complications and medical situations that increase the risk. Infants presenting with such medical conditions should then be referred prenatally to a level III facility.
A recent policy statement presented by the American Committee on Fetus and Newborn6 uses both GA and birth weight as tools to define care provided at each neonatal facility level. Their definition of a level II facility is one that can care for infants aged >32 weeks' gestation or weighing >1500 g. However, our results showed that some neonates aged ≤32 weeks' gestation never require intensive care. In our study, only 14.6% (95% CI: 12.3–16.8) of single infants aged 31 and 32 weeks' gestation with antenatal steroid therapy but without severe hypertension, antepartum hemorrhage, or infection required intensive care, and the remainder could be handled in level IIb maternity units. Indeed, preterm infants aged 33 weeks' gestation are currently handled in level IIb units, and their requirement for intensive care reaches 13.1% (95% CI : 12.7–15.2). Moreover, in France, the requirements for physician presence and skills are the same in level IIb and III facilities.25 Therefore, it is important for perinatal regionalization to take physician and staff abilities into account so as to maintain skills in the appropriate level of care. Reassignment of perinatal transfers for pregnancies at 31 or 32 weeks' gestation to level IIb facilities could participate to this purpose. It could also benefit health resources by providing care at a decreased distance from home. However, it should be kept in mind that ∼14.6% of preterm infants at these GAs might require transfer to a level III unit after birth.
Currently, no clear definition of the need for NIC exists. The definition proposed in this study complies with physicians' current practice but may need to be revised according to perinatal care improvements and modifications in neonatal pathologies. In this definition, we excluded infants referred to level III facilities after 2 days of life, because it would have been difficult to distinguish the effects of antenatal factors from those of initial management.
Although we sought to determine antenatal factors associated with the need for NIC, IUGR and infectious disease were defined based on postnatal criteria, because prenatal information on these factors lacked precision. IUGR was, thus, defined based on birth weight criteria and infectious disease as probable infection, definitions commonly used in clinical practice. Infection and IUGR were not significantly associated with the need for NIC. Because these factors have been described in the literature to significantly increase respiratory distress syndrome,23, 26 they were included in the multivariate analysis.
In the multivariate analysis, associations between the factors and the need for NIC were adjusted for the level of the maternity unit at birth, because the intensity of the same complication and the quality of neonatal management may vary from one level to another. Because complications are also included in the models, an overadjustment can be feared. However, this would only weaken the association between a factor and the need for NIC. Medical practices may also experience regional variations. Thus, factors entered in the multivariate models were adjusted for the region of birth.
Our study was based on data from preterm infants born in 1997. In the years since, no major therapeutic modifications in neonatology have been introduced. The EPIPAGE study was conducted before the texts organizing French perinatal regionalization were promulgated.4 However, even in 1997, high-risk births were routinely referred to level III facilities by in utero transfers, because 62% of births before 33 weeks' gestation took place in such maternity wards.8 Finally, the quality of care in perinatal regionalization should focus not only on immediate outcomes of preterm neonates in level III and IIb facilities but also on long-term follow-up that needs to be performed both in level III and IIb facilities.
Short-term outcomes for preterm infants can be estimated before birth, allowing physicians to match the site of birth to the level of care required. Such predictions not only allow referral of sick infants to level III NICUs but also allow healthy preterm infants to be born and completely handled in level IIb facilities, optimizing the medical services available for very preterm neonates.
The EPIPAGE Study Group is composed of: Institut National de la Santé et de la Recherche Médicale U149: B. Larroque (national coordinator), G. Bréart, P.Y. Ancel, B. Blondel, M. Dehan, M. Garel, M. Kaminski, F. Maillard, C. du Mazaubrun, P. Missy, F. Sehili, K. Supernant; Alsace: M. Durant, J. Matis, J. Messer, A Treisser (Hôpital de Hautepierre, Strasbourg); Franche-Comté: A. Burguet, L. Abraham-Lerat, A. Menget, P. Roth, J-P. Schaal, G. Thiriez (Centre Hospitalier Universitaire St Jacques, Besançon); Haute-Normandie: C. Lévêque, S. Marret, L. Marpeau (Hôpital Charles Nicolle, Rouen); Languedoc-Roussillon: P. Boulot, J-C. Picaud (Hôpital Arnaud de Villeneuve, Montpellier), A-M. Donadio, B. Ledésert (Observatoire Regional de la Sante Montpellier); Lorraine: M. André, J-L. Boutroy, J. Fresson, J.M. Hascoët (Maternité Régionale, Nancy); Midi-Pyrénées: C. Arnaud, S. Bourdet-Loubère, H. Grandjean (Institut National de la Santé et de la Recherche Médicale U558, Toulouse), M. Rolland (Hôpital des enfants, Toulouse); Nord-Pas-de-Calais: C. Leignel, P. Lequien, V. Pierrat, F. Puech, D. Subtil, P. Truffert (Hôpital Jeanne de Flandre, Lille); Pays de la Loire: G. Boog, V. Rouger-Bureau, J-C. Rozé (Hôpital MèreEnfants, Nantes); Paris-Petite-Couronne: P-Y. Ancel, G. Bréart, M. Kaminski, C. du Mazaubrun (Institut National de la Santé et de la Recherche Médicale U149, Paris), M. Dehan, V. Zupan (Hôpital Antoine Béclère, Clamart), M. Vodovar, and M. Voyer (Institut de Puériculture, Paris).
We are indebted to Dr Antoine Burguet for his comments on drafts of the article and to Dr Beatrice Larroque for national coordination.
Funding for this study was obtained from INSERM (French National Institute of Health and Medical Research), Merck-Sharp, Dohme-Chibret, La Fondation de la Recherche Medicale (The Medical Research Foundation), and La Direction Generale de la Sante du Ministere des Affaires Sociales (The General Directorate for Health From the French Ministry for Social Affairs).
- Accepted January 3, 2006.
- Address correspondence to Rachel Vieux, Maternité Régionale A. Pinard –54000, 140 Rue du Dr Heydenreich, BP 4213, Nancy 54042, France. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Blondel B, Breart G, du Mazaubrun C, et al. The perinatal situation in France. Trends between 1981 and 1995. Gynecol Obstet Biol Reprod.1997;26 :770– 780
- ↵Blondel B, Supernant K, du Mazaubrun C, Breart G. Enquete nationale perinatale 2003. Rapport 2005. Available at: www.sante.gouv.fr/htm/dossiers/perinat03/sommaire.htm. Accessed May 21, 2006
- ↵Decrets n°98–899 et 98–900 relatifs aux conditions techniques de fonctionnement auxquelles doivent satisfaire les etablissements de sante pour etre autorises a pratiquer les activites d'obstetrique, de neonatologie ou de reanimation neonatale. Journal Officiel.1998;235 . Available at: www.legifrance.gouv.fr/html/plan/plan-site.html. Accessed May 21, 2006
- ↵Committee on Fetus and Newborn. Levels of neonatal care. Pediatrics.2004;114 :1341– 1347
- ↵Verloove-Vanhorick SP, Verwey RA, Ebeling MC, Brand R, Ruys JH. Mortality in very preterm and very low birth weight infants according to place of birth and level of care: results of a national collaborative survey of preterm and very low birth weight infants in The Netherlands. Pediatrics.1988;81 :404– 411
- ↵Yu VY, Downe L, Astbury J, Bajuk B. Perinatal factors and adverse outcome in extremely low birthweight infants. Arch Dis Child.1986;61 :554– 558.
- ↵Larroque B, Breart G, Kaminski M, et al. Epipage study group. Survival of very preterm infants: Epipage, a population based cohort study. Arch Dis Child Fetal Neonatal Ed.2004;89 :F139– F144
- ↵Larroque B, et le Groupe EPIPAGE. EPIPAGE: epidemiologic study of very premature infants. Protocol of the survey [in French]. Arch Pediatr.2000;7 :339s– 342s
- ↵Hoffman EL, Bennett FC. Birth weight less than 800 grams: changing outcomes and influences of gender and gestation number. Pediatrics.1990;86 :27– 34
- ↵Burguet A, Kaminski M, Truffert P, Menget A, Hascoet JM, Larroque B, and the EPIPAGE study group. Does smoking in pregnancy modify the impact of antenatal steroids en neonatal respiratory distress syndrome? Results of the EPIPAGE study. Arch Dis Child Fetal Neonatal Ed.2005;90 :F41– F45
- ↵Shinwell ES, Blickstein I, Lusky A, Reichman B. Effect of birth order on neonatal morbidity and mortality among very low birthweight twins: a population based study. Arch Dis Child Fetal Neonatal Ed.2004;89 :F145– F148
- ↵Crowley P. Prophylactic corticosteroids for preterm birth. Cochrane Database Syst Rev.2000;(2):CD000065
- ↵Circulaire DH/EO 3 n°99–402 du 8 juillet 1999. Available at: www.legifrance.gouv.fr/html/plan/plan-site.html. Accessed May 21, 2006
- Copyright © 2006 by the American Academy of Pediatrics