Objective. To elucidate the role of level of care in combination with other perinatal risk factors for infant death in very preterm deliveries.
Design. Population-based cohort study.
Setting. Sweden, 1992–1998.
Subjects. Singleton infants (2285) born at 24 to 31 completed weeks of gestation to primiparous women.
Main Outcome Measure. Infant mortality.
Results. The rate of infant mortality increased from 5% among infants born at 31 weeks’ gestation to 56% among infants born at 24 weeks’ gestation. Compared with infants born at university hospitals, the unadjusted odds ratio (OR) of infant death was 0.70 (95% confidence interval [CI]: 0.54–0.90) among infants delivered at general hospitals. However, after adjustment, the OR of infant death shifted to 1.33 (95% CI: 0.88–2.02) for preterm births at general hospitals. This shift was primarily due to different gestational age distributions in regional and general hospitals. Among infants born at 24 to 27 weeks’ gestation, infant mortality rates were 23% (87 deaths) in university hospitals and 32% (73 deaths) in general hospitals, giving an adjusted OR of 2.00 for general versus university hospitals (95% CI: 1.15–3.49). The risk of death at 24 to 27 weeks’ gestation in general hospitals was increased specifically in pregnancies with placental complications.
Conclusion. Taking obstetric complications into account, there is an excess mortality risk among extremely preterm infants born at general hospitals.
The dramatic improvement in the prognosis of very preterm infants during the last decades1 has been attributed primarily to improvements of neonatal intensive care, including the introduction of surfactant and antenatal steroid therapy.2 However, extremely preterm infants still suffer high mortality and morbidity risks.3,4
The complex nature of intensive care for preterm infants demands highly qualified staffing as well as access to advanced technologies. In several studies from different countries, larger (level III) neonatal intensive care units have had lower mortality rates when compared with smaller (level II) units. However, these studies have some limitations. They categorized infants according to birth weight instead of gestational age,5–7 they were performed before recent improvements of neonatal practice,8,9 or they did not adjust for potential confounders such as obstetric complications.5,9
The aim of antenatal care is to identify pregnancy complications, which could be hazardous to the mother or the fetus, and to refer affected women to an appropriate specialist service. As a consequence, this strategy implies that complicated pregnancies resulting in very preterm birth should be overrepresented in university hospitals. However, unexpected preterm births still may occur in general hospitals. Both of these scenarios could influence neonatal outcome in the different hospital settings.
To elucidate the role of level of care for infant mortality, taking pregnancy complications and delivery characteristics into account, we performed a large population-based study including singleton infants born between 24 and 31 weeks of gestation from 1992 to 1998 in Sweden. The data presented herein suggest that level of care does influence the risk of infant death, but only among extremely preterm infants.
The study, approved by the ethical committee at the Karolinska Institute, was based on data from the general population-based Swedish Medical Birth Register. Starting at the first antenatal visit, information is collected prospectively in standardized antenatal, obstetric, and pediatric records, and copies are forwarded to the registry.10 Information about infant mortality, including time and cause of death, is obtained through individual linkage between the Birth Register and the Cause of Death Register.
Among primiparous women giving birth between 1992 and 1998 (n = 291 059), we identified 2374 live-born singleton infants delivered before 32 completed gestational weeks. Because there was no consensus on resuscitation of infants born at 22 to 23 weeks of gestation, 16 births at 22 completed weeks and 41 births at 23 weeks were excluded from additional analysis (infant mortality rates were 81% and 66%, respectively). We also excluded 32 accidental preterm births in 22 local maternity units without access to pediatric service. These local maternity units, of which several now are closed, were designed for a small number of normal pregnancies ending at term. In Sweden, there is a national consensus that very preterm infants should not be delivered in these units because of suboptimal conditions for treatment and postnatal transfer. Of the excluded infants born in maternity units, 16 were born at 24 to 27 gestational weeks and 16 were born at 28 to 31 weeks (infant mortality rates were 56% and 19%, respectively). Finally, by cross-tabulating gestational age, birth weight, birth length, head circumference, and neonatal diagnoses, we excluded an additional 32 infants for whom gestational age was either misclassified or very uncertain: 27 infants had complete records on birth weight (ranging from 2365 to 3939 g), birth length (ranging from 45 to 53 cm), and head circumference (ranging from 32 to 37 cm), and another 5 infants had recorded birth weights of ≥4120 g and missing data on birth length and/or head circumference. Because our data set was anonymized, we could not use the original medical records to obtain more correct information about gestational ages for these infants. The analyses were performed on the remaining 2253 live singleton infants, born at 24 to 31 completed weeks of gestation.
At the time of registration for antenatal care, maternal weight and length were registered, and body mass index was calculated as the weight in kilograms divided by the square of height in meters. Women were categorized as nonsmokers or smokers at registration to antenatal care and classified as either living with the infant’s father or not. The maternal age was registered at the time of delivery. Information about the mother’s country of birth was obtained through record linkage with the Immigration Register and categorized as a Nordic (Sweden, Denmark, Finland, Iceland, and Norway) or non-Nordic country of birth. The Swedish Medical Birth Register does not include information about maternal race. In our data set, 88% of mothers were born in the Nordic countries, and we assume that at least a similar proportion were white.
Diagnoses were coded according to the ninth or tenth edition of International Classification of Diseases. Pregnancy complications were grouped into 4 main groups: placental complications, hypertensive illnesses, diabetes, and other maternal diseases. Placental complications were defined as placenta previa, abruptio placenta, or unspecified antepartum hemorrhage. Hypertensive illnesses were classified as preeclampsia and other hypertensive illness. Diabetes was classified as pregestational and gestational diabetes. We also categorized women with regard to other chronic diseases (eg, systemic connective tissue diseases and renal diseases).
We obtained information about delivery characteristics: gestational age, mode of delivery, delivery hospital, birth weight for gestational age, infant gender, and fetal presentation. We also classified infants with regard to major malformations (n = 42; infant mortality rate: 55%) such as complex heart defects and renal agenesis.
During the study period, all pregnant women were offered ultrasonography before week 18, and >95% accepted this offer.11,12 Thus, gestational ages were determined mainly by ultrasound; otherwise, the date of the last menstrual period was used. Small for gestational age was defined as a birth weight <2 SD below the mean for gestational age according to the Swedish standard, and large for gestational age was defined as a birth weight >2 SD above the mean for gestational age.13 Infant mortality was defined as death during the first year of life, and neonatal mortality was defined as death during the first 4 weeks of life.
Delivery hospitals were categorized into general hospitals and university hospitals, according to a classification by the Swedish National Board of Health and Welfare.14 University hospitals are referral centers with full neonatal intensive care services. Neonatal surgery is provided, and neonatologists routinely attend high-risk deliveries. University hospitals typically receive referrals for 10 000 to 25 000 deliveries annually. General hospitals mainly care for inborn infants with 1000 to 5000 deliveries annually and provide neonatal intensive care before transfer to a university hospital. Major neonatal surgery is not provided in general hospitals, and high-risk deliveries are attended by neonatologists or pediatricians. Both university and general hospitals have similar obstetric and anesthesiologic services, and true “crash” cesarean sections are performed in both settings.
We estimated odds ratios (ORs) and 95% confidence intervals (CIs) before and after adjustment by using the logistic regression procedure (GENMOD) in the SAS 8 software package (SAS Institute, Cary, NC). As independent variables in the multivariate logistic regression models we included mode of delivery, hospital type, gestational age, birth weight for gestational age, infant gender, fetal presentation, placental complications, and maternal hypertensive illness. These variables were considered as risk factors or potential confounding factors a priori.
All variables were entered as independent categories. Gestational age was defined as a categorical variable divided into single completed gestational weeks.
We corrected for clustering within hospitals by using the generalized estimating equation method. Goodness of fit was assessed by the Hosmer and Lemeshow test, and there was no significant lack of fit in our regression model.
In all, 2253 infants were born before 32 completed weeks of gestation, of whom 267 (12.6%) died during the first year of life. The majority of deaths occurred during the neonatal period (n = 232).
Forty-one percent of the infants were delivered in university hospitals and 59% in general hospitals. Compared with general hospitals, extremely preterm infants were delivered more often in university hospitals (Table 1). No other differences between hospital types were found with regard to the included maternal, pregnancy, delivery, or neonatal characteristics (data are available on request).
In the univariate analysis, infants born in general hospitals had a reduced risk of infant mortality compared with infants born in university hospitals (Table 2). However, in the multivariate analysis, infants born in general hospitals faced a 33% increased risk of infant mortality. This shift was primarily due to different gestational age distributions in university and general hospitals.
The increased risk of infant death in general versus university hospitals was explained by increased neonatal mortality (adjusted OR: 1.41; 95% CI: 0.93–2.13), whereas the risk of postneonatal mortality was not increased in general hospitals (adjusted OR: 0.98; 95% CI: 0.41–2.36).
Infant mortality was inversely related to gestational age, rising from 5% among infants born at 31 weeks to 56% among infants born at 24 weeks. In the univariate analysis, infants born at 24 weeks had a 24-fold increased risk of infant mortality, compared with infants born at 31 weeks. In the multivariate analysis, this risk was increased further, which mainly was an effect of adjustments for hospital type and birth weight for gestational age. When we adjusted only for these 2 variables, the corresponding OR was 36.5 (95% CI: 19.3–68.7).
Compared with an infant with appropriate birth weight for gestational age, being either small or large for gestational age increased the mortality risk. Male infants had a higher mortality risk than female infants. Mode of delivery, fetal presentation at birth, placental complications, and maternal diseases were not independent risk factors for infant death. Preeclampsia was associated with a lower infant mortality risk before stratification by gestational age. Maternal characteristics (age, body mass index, smoking, cohabitation with infant’s father, and native country) were not associated with risk of infant death (data are available on request).
There was a significant interaction between gestational age and hospital type with regard to infant mortality (P = .049). Infant mortality rates were substantially higher in general hospitals compared with university hospitals up to 27 gestational weeks. For infants born at 27 weeks, the infant mortality rate was 18.6% in general hospitals and 11.3% in university hospitals, whereas the corresponding figures for infants born at 28 weeks were 10.4 and 9.5%, respectively. We therefore decided to stratify our cohort into 1 extremely preterm group (born at 24 to 27 gestational weeks) and 1 very preterm group (born at 28 to 31 gestational weeks) (Table 3).
For extremely preterm infants, infant mortality was 22% in university hospitals and 32% in general hospitals (Table 3). Compared with university hospitals, we found an increased risk of infant mortality among extremely preterm infants (24–27 weeks’ gestation) born in general hospitals, whereas there was no difference in infant mortality by hospital type among infants born at 28 to 31 weeks’ gestation. In both extremely and very preterm births, crude ORs were somewhat lower than adjusted ORs. For infants born at 24 to 27 weeks’ gestation, this was mainly an effect of adjusting for gestational age. When we adjusted solely for gestational age, the OR was 1.97 (95% CI: 1.12–3.47). For infants born at 28 to 31 weeks’ gestation, the shift was due to a combined effect of adjusting for gestational age and birth weight for gestational age. Adjusting for these 2 variables gave an OR of 0.83 (95% CI: 0.52–1.32). We also repeated the analyses of level of care and risk of infant mortality after excluding infants with major malformation (n = 42), with no change in risk estimates (data are available on request).
To explore the differences in infant mortality between hospital types further, we stratified births at 24 to 27 weeks’ gestation by pregnancy and delivery characteristics (Table 4). Compared with university hospitals, the excess infant mortality in general hospitals was particularly associated with placental complications (adjusted OR: 11.03; 95% CI: 5.46–22.29) and preeclampsia (adjusted OR: 3.23; 95% CI: 1.33–7.83).
At general hospitals, extremely preterm infants (24–27 weeks of gestation) suffered an excess mortality risk compared with those born at university hospitals. Besides level of care, 4 other factors were associated with an increased risk of infant mortality: decreasing gestational age, low and high birth weight for gestational age, and male gender. Mode of delivery was not associated with risk of infant mortality.
For infants born at 28 to 31 weeks of gestation, we found no difference in infant mortality by level of care. If anything, risk of infant mortality tended to be lower in general compared with university hospitals. Thus, our results differ from a recent US study,7 in which mortality among infants with birth weights from 1250 to 2000 g was found to be higher in community versus regional hospitals. However, the point estimates in both studies were nonsignificant, and CIs were overlapping, suggesting that the apparent difference in mortality risks may be a random effect. If a true difference exists, several factors may contribute, such as different study periods (1992–1998 vs 1992–1993) and study populations (singleton infants to primigravidas versus all singletons). In addition, pre- and postnatal transfer routines may not be the same in the 2 populations.
Among extremely preterm infants, our results raise the question of whether the increased risk of infant death in general hospitals was a matter of quality of care and whether additional centralization could improve survival. Immediate postnatal management is critical in extreme prematurity,15,16 and neonatal intensive care was possible to initiate in all hospitals. The higher number of mortality-associated placental complications in general hospitals may reflect unexpected obstetric emergencies in which antenatal transfer was contraindicated. Moreover, the survival of the extremely preterm infant delivered after placental abruption has, in itself, been found to be poor.17,18 However, a poorer outcome in general hospitals was not confined to situations of unforeseen obstetric complications, and less experience due to fewer cases may have contributed to worse outcomes in general hospitals. The assumption that quality-of-care differences exist is supported by the finding that the mortality risk among extremely preterm infants to preeclamptic mothers was higher in general hospitals compared with university hospitals. Possibly, a larger proportion of preeclamptic pregnancies could have been referred antenatally to a university setting.
Presently (2003), Sweden has 59 delivery units, but only 7 of them are affiliated with university hospitals. Although antenatal referrals are possible in the vast majority of cases, extremely preterm infants will be born unexpectedly in general hospitals. It therefore is important that well-functioning networks are created to decentralize experience in resuscitation and stabilization. Postnatal transports should be considered in collaboration with the accepting unit. If “mobile transport teams” were developed nationwide, they could travel to general hospitals to support local staff, ideally before delivery. Because timing also matters, such organization of transport may enable postnatal transfers on the first day of life.
This study has several methodological strengths. It was based on the Swedish Medical Birth Registry, which contains data on virtually all births in Sweden. The uniform health care system in Sweden provides a good model for research of level-of-care issues. The study population was well defined, including only singleton preterm infants born to primiparous women. Because virtually all pregnancies in Sweden are dated by an early ultrasound investigation, we could analyze mortality risks by gestational age, on which obstetric decisions are based. Data on pregnancy and obstetric complications were available, and we were able to adjust for maternal risk factors such as preeclampsia and placental complications. Unlike most previous research, our study was performed during a time period in which surfactant and prenatal steroids were generally implemented therapies.
Our study also has limitations. Our source of data did not allow the analyses of different clinical practices in different hospitals. However, the National Board of Health and Welfare and the Swedish Pediatric Society have proposed national policies regarding resuscitation of extremely preterm infants. According to a national survey, the professional community in Sweden agrees on and follows these policies regarding infants born after 24 weeks of gestation.19 More importantly, hospital data in the Swedish Medical Birth Registry only state the delivery hospital, and effects of postnatal transfers therefore could not be studied. In addition, given the fact that the proportion of the 24- to 27-week gestational age infants born in university hospitals increased from 60% in 1992% to 68% in 1998, it is possible that referral practices changed during the study period. Because the Swedish Medical Birth Register contains no data on fetal deaths before 28 gestational weeks, we could not address whether level of care influenced risk of early fetal death. Because obstetric and anesthesiologic services and the ability to perform “crash” cesarean deliveries are similar in both university and general hospitals, we find it unlikely that differences in fetal deaths due to the ability to respond to obstetric emergencies would have biased our results.
In the debate about organization of neonatal intensive care, our study supports the idea that extreme prematurity should be managed in university hospitals. Case-control studies and clinical audits within regional networks are 2 possibilities for additional study of clinical practices including referral routines, aiming at finding new and innovative ways to improve the prognosis for this vulnerable group of infants.
The study was supported by grants from the Swedish Research Council (71P-14158), the Swedish Council for Working Life and Social Research (2001-2247), the Samariten Foundation, and the Swedish Order of Freemasons.
- Received December 26, 2002.
- Accepted July 7, 2003.
- Reprint requests to (S.J.) Department of Neonatology, C2:02, Astrid Lindgren Children’s Hospital, Karolinska Hospital, SE-171 76 Stockholm, Sweden. Email:
- ↵Improved outcome into the 1990s for infants weighing 500-999 g at birth. The Victorian Infant Collaborative Study Group. Arch Dis Child Fetal Neonatal Ed. 1997;77 :F91– F94
- ↵Hack M, Taylor HG, Klein N, Mercuri-Minich N. Functional limitations and special health care needs of 10- to 14-year-old children weighing less than 750 grams at birth. Pediatrics. 2000;106 :554– 560
- ↵Cifuentes J, Bronstein J, Phibbs CS, Phibbs RH, Schmitt SK, Carlo WA. Mortality in low birth weight infants according to level of neonatal care at hospital of birth. Pediatrics. 2002;109 :745– 751
- ↵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
- ↵Field D, Hodges S, Mason E, Burton P. Survival and place of treatment after premature delivery. Arch Dis Child. 1991;66 (4 spec no):408– 410; discussion 410–411
- ↵Intensive Care of Newborn Children. Quality Requirements and Proposals of Organization. Stockholm, Sweden: National Board of Health and Welfare; 1997
- ↵Van Marter LJ, Allred EN, Pagano M, et al. Do clinical markers of barotrauma and oxygen toxicity explain interhospital variation in rates of chronic lung disease? The Neonatology Committee for the Developmental Network. Pediatrics. 2000;105 :1194– 1201
- ↵Ananth CV, Wilcox AJ. Placental abruption and perinatal mortality in the United States. Am J Epidemiol. 2001;153 :332– 337
- Copyright © 2004 by the American Academy of Pediatrics