The Effect of Birth in Secondary- or Tertiary-Level Hospitals in Finland on Mortality in Very Preterm Infants: A Birth-Register Study
OBJECTIVE. Our goal was to test the hypothesis that the level of the delivery hospital affects 1-year mortality of very preterm infants in Finland.
PATIENTS AND METHODS. This retrospective national medical birth-register study included 2291 very preterm infants (gestational age of <32 weeks at birth or birth weight of ≤1500 g) born in 14 level II (central) and 5 level III (university) hospitals in 2000–2003. The main outcome measures were adjusted total mortality (including stillbirths) and mortality of live-born infants until the age of 1 year.
RESULTS. Both the total 1-year mortality and the 1-year mortality of live-born infants were higher in level II hospitals compared with level III hospitals. Total mortality was higher in very preterm infants who were not born during office hours. In theory, delivery of all very preterm infants in level III instead of level II hospitals translates into an annual prevention of 69 of the 170 total deaths and prevention of 18 of the 45 deaths of live-born infants.
CONCLUSIONS. Resources in neonatal intensive care should be increased, especially during non–office hours, to have an equally distributed service through the 24-hour day. More efficient regionalization of very preterm deliveries may improve 1-year survival of very preterm infants in Finland.
Care of very preterm infants is extremely specialized and requires 24-hour coverage by highly trained and experienced medical and nursing staff, as well as expensive equipment. Regionalization of very preterm deliveries results in patient volumes adequate to provide staff with sufficient experience in dealing with such deliveries. A higher level of delivery hospital is associated with lower mortality rates in preterm infants.1–4 High patient volumes have also been associated with better outcomes in some studies.5,6 In addition, regionalization of deliveries has the advantage of avoiding transportation of neonates, which increases mortality and morbidity.7 The benefits and costs of regionalization of the deliveries of preterm infants, however, have been debated, and even the concept of deregionalization has been brought up.8
The Finnish National Medical Birth Register (MBR) at the National Research and Development Centre for Welfare and Health includes all live births and stillbirths of infants weighing ≥500 g or with a gestational age of ≥22 weeks. The coverage of the register is practically 100%, because the information on the few missing cases is received from the central population register (live births) and the Cause-of-Death Register at Statistics Finland (stillbirths and infant deaths). The systematically gathered national database enabled us to examine the effects of delivery-hospital level on perinatal and 1-year mortality. Antenatal maternal information was gathered during publicly funded antenatal care, which is free of charge to the customer and covers 99.8% of the pregnant mothers (according to national statistics).
One of the reasons for low perinatal mortality in Finland may be the generally adopted policy to regionalize the care of at-risk pregnancies, including those that are very preterm.9 Because health care practices are evolving rapidly, there is a need to update current referral policies and reevaluate the effects of regionalizing deliveries of very preterm infants. The aim of our study was to test the hypothesis that the level of delivery hospital affects 1-year mortality of very preterm infants in Finland.
PATIENTS AND METHODS
The study population consisted of very preterm infants, defined as infants with a gestational age of <32 weeks at birth or a birth weight of ≤1500 g, born in Finnish level II (central) and level III (university) hospitals from 2000 to 2003. Seventeen of the infants born in level II hospitals were transferred to level III hospitals during their first week of life. According to definitions proposed by the American Academy of Pediatrics Committee on Fetus and Newborn,10 all 5 university hospitals in Finland have a level III NICU. In addition to the university hospitals, very preterm infants were delivered in 14 hospitals with a level II NICU during the study period.
The data were derived from the Finnish national MBR, which was linked to the Cause-of-Death Register to access information on infant mortality. According to the register, 2572 infants were born at <32 weeks of gestation or at a birth weight of ≤1500 g. Of them, 56 outliers were excluded because there was a major disparity in their register data when cross-tabulating gestational age and birth weight. Another 3 infants were excluded because of an incomplete social security number, and 9 others were excluded because there was no information on their mortality or treatment. We also excluded infants who had lethal congenital anomalies (see Appendix), according to the Finnish Register on Congenital Malformations (n = 95), and infants who were born in level I hospitals (regional or local; n = 92), at home (n = 3), or in hospitals with <5 very preterm deliveries during the study period (n = 23). Hospitals with <5 very preterm deliveries were excluded because it is likely that the deliveries were unavoidable emergency situations and would bias the results. Altogether, 2291 infants were included, 2021 of whom were born live.
One-year mortality with and without stillborn infants was examined in relation to hospital type (level II or III). For statistical analyses, SAS 9.1.3 for Microsoft Windows (SAS Institute, Inc, Cary, NC) was applied. Logistic regression was used in all analyses. The potential number of lives saved if all the infants had been born in level III hospitals instead of level II hospitals was calculated. This was done for stillborn and live-born infants and separately for live-born infants. First, logistic regression was applied to estimate the parameters for all covariants. On the basis of these estimates and by hypothesizing that all the infants were born in level III hospitals, we calculated the number of infants saved, defined as the difference between the actual number of the observed deaths and the number of deaths predicted by the model.
The outcome data were adjusted for maternal hospitalization for hypertension during pregnancy, maternal age, smoking during pregnancy, primiparity, birth during non–office hours, intrauterine growth (categorized as small [SGA], appropriate [AGA], or large [LGA] birth weight for gestational age, defined as birth weight below −2 SDs, between −2 SDs and 2 SDs, and >2 SDs according to reference values from the Finnish population, respectively), birth weight, categorized gestational age, and gender. Marital status of the mother, previous induced abortions, previous extrauterine pregnancies, previous miscarriages, mode of delivery (cesarean section or other), multiple gestations, and the year of birth were excluded as covariants in the models, because they did not have a significant effect on the outcome. Smoking during pregnancy was included in the model because it was close to being significant and was considered a variable of general interest. Socioeconomic status of the mother based on maternal occupation was not used as a covariant, because the information could not be estimated reliably for young women of childbearing age. Births between 8 am and 4 pm Monday to Friday were included as “office-hour” births. Births during public holidays, on weekends, or from 4:01 pm to 7:59 am on weekdays were included in births “during non–office hours.”
The adjusted mortality indexes in different gestational-age classes in level II and III hospitals were calculated as the ratio of observed mortality divided by expected mortality. The expected mortality for each unit was produced by summing the individual predictions for 1-year mortality of logistic regression adjusted for gender, SGA, LGA, maternal age, maternal smoking, and maternal hospitalization for hypertension. In addition to these variables, classified gestational age was included in the adjustment of mortality indexes of hospitals arranged by the number of very preterm births.
The study protocol was approved by the National Research and Development Centre for Welfare and Health Ethics Committee.
The distribution of very preterm deliveries according to the hospital type is shown in Table 1. Very preterm deliveries occurred in 5 level III and 14 level II hospitals. Depending on the university hospital district, the proportion of very preterm infants born in level III hospitals varied from 55% to 95%. As seen in Table 1, only a few infants of ≤27 weeks' gestation were born in level II hospitals.
The total 1-year mortality including stillbirths was 21.6% (494 of 2291 infants). Of the live-born infants, 11.1% (224 of 2021) died during the first year of life. The median age for death of the live-born infants was 2 days (range: 0–336 days). The results of the logistic regression showed that the risk of total 1-year mortality was significantly increased by birth in a level II hospital, a low birth weight and a low gestational age, inappropriate fetal growth (SGA or LGA), male gender, decreasing maternal age, and birth during non–office hours. The independent effect of maternal smoking during pregnancy was of statistically borderline significance. Maternal hospitalization for hypertension and primiparity decreased the risk of the total mortality (Table 2). The 1-year mortality of live-born infants was significantly affected by the same variables with 2 exceptions; birth during non–office hours and maternal smoking had no significant effect (Table 2). In theory, delivery of all very preterm infants in level III instead of level II hospitals translated into an annual prevention of 69 of the 170 total deaths and prevention of 18 of the 45 deaths of live-born infants.
The very preterm infants were divided into subgroups of those born at 22 to 28 weeks' gestation and those born at 29 to 33 weeks' gestation. At 22 to 28 weeks' gestation, both the total 1-year mortality (odds ratio [OR]: 4.7; 95% confidence interval [CI]: 3.0–7.5) and the 1-year mortality of live-born infants (OR: 3.3; 95% CI: 1.8–6.1) were higher in level II hospitals compared with level III hospitals. At 29 to 33 gestational weeks, the total mortality was higher (OR: 3.3; 95% CI: 2.1–5.4), but the mortality of the live-born infants did not differ between level II and level III hospitals (OR: 1.2; 95% CI: 0.5–2.6). The distribution of 1-year mortality according to gestational age is shown in Figs 1 and 2. A birth during non–office hours was associated with higher total 1-year mortality (OR: 1.8; 95% CI: 1.3–2.3), but the association was not significant in live-born infants only.
The median number of very preterm deliveries was 7.5 (range: 1–29) per year in level II hospitals and 68.5 (range: 45–211) per year in level III hospitals. The association between very preterm birth rate and mortality is shown in Figs 3 and 4.
Our national population-based study showed that the 1-year mortality of Finnish very preterm infants was increased for infants born in level II hospitals compared with those born in level III university hospitals. In the subgroup of infants born at 22 to 28 gestational weeks, a significant difference was seen both in the total mortality (including stillbirths) and the mortality of live-born infants. The total mortality was higher in the subgroup of infants born at 29 to 33 weeks. Birth during non–office hours increased the total mortality.
The reliability and coverage of the data included in the registers is one of the most important limitations of register studies generally. We were able to avoid most of these limitations because the MBR covers all Finnish delivery hospitals and has an established way to collect an extensive amount of data. These register data have been shown to be reliable.11 Adjusting mortality for infant- and mother-related confounding variables enabled us to further improve the accuracy.
Besides prematurity, congenital anomalies are the main cause of death among infants. Excluding the infants with lethal anomalies (ie, those infants who would have died irrespective of the quality of the care) decreases the possibility that the differences in mortality are explained by differences in patient characteristics between hospitals. Infants with minor anomalies or anomalies with variable degrees were not excluded to avoid selection bias, because there were regional differences in reporting minor anomalies. There is no general agreement between different studies on the criteria for anomalies to be excluded.2–4,12–14
Our study included a national 4-year cohort of very preterm births, thus avoiding a selection bias typical of hospital-based studies. We analyzed our data by both excluding and including stillbirths. Therefore, we achieved better coverage of births at each gestational week compared with studies that do not include stillbirths.12 The relation between the time of fetal death and the time of birth, however, remain unknown in our study. Therefore, we do not know whether the larger proportion of stillborn infants in level II hospitals reflects a lower referral rate for dead fetuses or a difference in the level of obstetric care.
When compared with countries with equally low infant mortality (eg, Sweden), very preterm deliveries are more efficiently regionalized in the Finnish health care system. In Sweden, only 41% of singleton deliveries at 24 to 31 weeks of gestation occurred in level III hospitals in the 1990s,3 compared with 79% of the live-born infants in our study. However, there were significant regional differences in regionalization. Our study suggests that a higher degree of regionalization could be achieved because regionalization up to 95% was reached in 1 university hospital area. There will always be, however, a small proportion of unavoidable emergency deliveries outside level III hospitals. Such emergency deliveries comprise a larger proportion of very preterm deliveries in level II hospitals than in level III hospitals. There may be more potential disadvantages for the infant in these deliveries, such as missing the benefit of antenatal glucocorticoids, which leads to a selection bias that needs to be taken into consideration when interpreting our results.
Our findings parallel the results of a Swedish population-based study by Johansson et al.3 They found that the mortality of infants born at 24 to 27 gestational weeks was increased in general hospitals compared with university hospitals. Also, the authors of a Finnish study found increased mortality of extremely low birth weight infants born in level II hospitals compared with level III hospitals in the 1990s.1 It is interesting to note that Warner et al4 found the effect of birth-hospital level on mortality to be even greater in infants with birth weight between 1000 and 1499 g than in infants with birth weight between 500 and 999 g.
Previous studies have shown increased mortality in infants born at night12,13,15–17 and during holidays,15 but contradictory results have been reported for those born during weekends.13,14,18 In our study, the majority of preterm births occurred during non–office hours, which was associated with increased total mortality and a trend toward increased mortality in live-born infants. This suggests that resources should be increased especially during non–office hours to have an equally distributed service throughout the 24-hour day.
There are conflicting data about the effect of hospital size and very preterm patient volumes on infant mortality. Although some studies have not demonstrated an advantage of birth in hospitals with large volumes of very preterm patients,19,20 others have shown that very low birth weight infants have lower mortality rates if they are born in larger compared with smaller units.5,6 Our study shows an association between a higher rate of preterm births and a lower mortality. However, the high preterm birth rate in our study is also related to higher hospital level, because all Finnish hospitals with >44 very preterm deliveries per year are level III university hospitals, whereas all those with <30 very preterm births are level II hospitals.
The explanations for survival advantage in level III hospitals may be related to higher patient volumes, more experienced and highly trained staff, better coverage during non–office hours, and more up-to-date patient-management styles and equipment. There might also be differences in how actively infants with borderline viability are resuscitated. Antenatal transfer and resuscitation practices of preterm infants have been shown to differ in Australia21 and European countries,22 and obstetricians in level II hospitals have underestimated the chances of survival of extremely preterm infants.21
Although Finnish neonatal mortality rates remain very low, our study indicates that the deliveries of very preterm infants, especially those born at <29 gestational weeks, should be more efficiently referred to hospitals with a level III NICU to ensure the best chances for survival. Consistent national criteria for regionalization would enhance the development of more equal care of very preterm infants.
This national study provides information about the effects of regionalization on mortality of very preterm infants. In the future, the effect of birth-hospital level on the long-term outcome of very preterm infants should be evaluated, because morbidity might be a more sensitive outcome measure than mortality. Care of very preterm infants should also be evaluated from both health-outcome and economic perspectives to provide a basis for rational decision-making on the organization of perinatal care.
The data on malformations were derived from the Register of Congenital Malformations, which was linked to the Finnish National MBR by using the mother's and child's unique identification numbers. All congenital anomalies are reported prospectively to the register by the delivery hospitals and hospitals treating congenital anomalies. Information on congenital anomalies is also received from other health registers such as the Hospital-Discharge Register and MBR.
Anomalies that were considered lethal and resulted in the exclusion of the affected infants from the study population were trisomy 13 or 18, triploidy, double-outlet right ventricle, double-outlet left ventricle, Taussig-Bing anomaly, transposition of great arteries, hypoplasia of the heart, left-heart hypoplasia syndrome, hypoplasia of the left ventricle, univentricular heart, atresia of the pulmonary valve, anencephaly, bilateral agenesis of the kidneys, Potter syndrome, acrania, anencephaly, occipital encephalocele, holoprosencephaly, semilobar holoprosencephaly, cervical meningocele, large bilateral polycystic/microcystic kidneys, bilateral cystic dysplasia of the kidneys, bilateral severe hypoplasia of the lungs, bilateral rudimentary lungs, congenital tracheal stenosis, hypoplasia of the trachea, and agenesis of the diaphragm.
This study was supported by the Finnish Academy (Research Program on Health Services Research), the South-West Finnish Fund of Neonatal Research, Turku University Hospital EVO Funds, and the Turku University Hospital Foundation.
Members of the Performance, Effectiveness, and Cost of Treatment Episodes (PERFECT) Preterm Infant Study Group are Sture Andersson, MD, PhD, Mika Gissler, DrPhil, MSocSci, Mikko Hallman, MD, PhD, Jaana Leipälä, MD, PhD, Unto Häkkinen, PhD, Emmi Korvenranta, MSc (Econ), BM, Heikki Korvenranta, MD, PhD, Mikko Lavonius, MD, Liisa Lehtonen, MD, PhD, Miika Linna, PhD, Mikko Peltola, MSc, Liisi Rautava, BM, and Outi Tammela, MD, PhD.
We thank Turku University Hospital Foundation for providing a research office for Ms Rautava.
- Accepted August 10, 2006.
- Address correspondence to Liisi Rautava, Department of Pediatrics, Turku University Hospital, Kiinamyllynkatu 4-8, 20520 Turku, Finland. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Tommiska V, Heinonen K, Ikonen S, et al. A national short-term follow-up study of extremely low birth weight infants born in Finland in 1996–1997. Pediatrics.2001;107(1) . Available at: www.pediatrics.org/cgi/content/full/107/1/e2
- ↵Johansson S, Montgomery SM, Ekbom A, et al. Preterm delivery, level of care, and infant death in Sweden: a population-based study. Pediatrics.2004;113 :1230– 1235
- ↵Warner B, Musial MJ, Chenier T, Donovan E. The effect of birth hospital type on the outcome of very low birth weight infants. Pediatrics.2004;113 :35– 41
- ↵Bartels DB, Kreienbrock L, Dammann O, Wenzlaff P, Poets CF. Population based study on the outcome of small for gestational age newborns. Arch Dis Child Fetal Neonatal Ed.2005;90 :F53– F59
- ↵Stark AR; American Academy of Pediatrics, Committee on Fetus and Newborn. Levels of neonatal care [published correction appears in Pediatrics. 2005;115:1118]. Pediatrics.2004;114 :1341– 1347
- ↵Luo ZC, Karlberg J. Timing of birth and infant and early neonatal mortality in Sweden 1973–95: longitudinal birth register study. BMJ.2001;323 :1327– 1330
- ↵Luo ZC, Liu S, Wilkins R, Kramer MS; Fetal and Infant Health Study Group of the Canadian Perinatal Surveillance System. Risks of stillbirth and early neonatal death by day of week. CMAJ.2004;170 :337– 341
- ↵Field D, Draper ES. Survival and place of delivery following preterm birth: 1994–96. Arch Dis Child Fetal Neonatal Ed.1999;80 :F111—F114
- Copyright © 2007 by the American Academy of Pediatrics