Background: Comparisons of national perinatal and neonatal mortality often neglect the underlying causes.
Objective: To assess effects of very-preterm births in the UK and Australia.
Setting: Two geographically defined populations: the former Trent Health Region of the UK and New South Wales (NSW)/the Australian Capital Territory (ACT), Australia.
Method: All births 22+0 to 31+6 weeks in 2000, 2001 and 2002 were identified by established surveys of perinatal care. Rates of birth and death were compared.
Results: The population of NSW/ACT was 35% higher and there were 66% more births than in Trent (273 495 vs 164 824). The proportion of liveborn infants between 22 and 31 weeks gestation was about 25% higher in Trent (NSW/ACT 2945, rate per 1000 live births 10.82 (95% CI 10.43 to 11.22); Trent 2208, rate per 1000 live births 13.47 (95% CI 12.92 to 14.05)). The proportion of these infants admitted to a neonatal unit was also higher in Trent (91.2% vs 94.4%; OR 1.63 (95% CI 1.30 to 2.05)). Unadjusted mortality in infants admitted to a neonatal unit was similar: NSW/ACT 332/2686 (12.4%); Trent 284/2085 (13.6%); unadjusted OR 1.12 (95% CI 0.94 to 1.33; p = 0.21).
Conclusions: The higher rates of very premature birth and more ready admission to neonatal intensive care for infants in the UK may help to explain why perinatal and neonatal mortality are higher there than in Australia. Efforts to understand why the rate of premature birth in the UK is so high should be a national priority.
Statistics from Altmetric.com
The various death rates related to pregnancy and childbirth reported by individual countries still command considerable public attention despite wide variations in the way they are derived.1 For developed countries, very-preterm deliveries contribute the majority of deaths identified in these statistics.2 However, although individual or groups of neonatal units regularly report their mortality and morbidity, two other factors that have a profound affect on published perinatal and neonatal mortality are often ignored.
The first is the rate of very-preterm delivery. If country A has a higher proportion of very-preterm births than country B, then one would expect a higher perinatal mortality in country A. The second is the extent to which very-preterm infants are actively treated and whether those who are not offered intensive care are excluded from officially published statistics.
However, exploration of these issues, free from confounding variables, has proved very difficult. In studies that have explored variation in obstetric practice, results have been inconsistent. It is likely that some of the inconsistencies reflect the difficulty in establishing genuine like-for-like comparisons in which all live births are considered and appropriately adjusted for disease severity.3–9
The aim of this study therefore was to explore the differences in very-preterm delivery outcome between two geographically defined populations: the former Trent Health Region in central England and New South Wales (NSW)/the Australian Capital Territory (ACT), Australia. In both these localities, detailed information was available about the outcome of all births <32 weeks gestation. These populations are known to have broadly similar socioeconomic and healthcare provision but significant differences in organisation.
The study population comprised all very-preterm infants born between 22+0 and 3+6 weeks of gestation to mothers resident in the Trent region of the UK and the State of NSW and the ACT, Australia during the years 2000, 2001 and 2002.
Data for the Trent region were collected as part of the Trent Neonatal Survey. This is an ongoing study of neonatal intensive care activity in the former Trent Health Region of the UK National Health Service. All of the perinatal services in Trent contribute to the Trent Neonatal Survey, and units in adjacent regions also permit data collection on Trent infants. The present survey was established in February 1990. The database holds information on all infants of 32 weeks gestation or less born to a Trent resident mother and admitted to a neonatal unit since that time. Data for the Trent Neonatal Survey are collected by a group of five part-time research nurses who visit each of the neonatal units on a regular basis and complete a standardised dataset about each infant. Information is obtained from the clinical records, discussions with staff, and, where appropriate, personal observation. Data on immature infants who died outside a neonatal unit were obtained from the Confidential Enquiry into Stillbirths and Deaths in Infancy (CESDI) for the region. This is a similarly longstanding national study (more recently renamed Confidential Enquiry into Maternal and Child Health (CEMACH)) which collates, from a variety of sources, data on all infants delivered after 22 weeks gestation either born dead or who die in the first year of life. Over the course of the study period, 77 infants met the study criteria but were cared for outside the boundaries of the Trent region. As a result full details of their course could not be obtained and they were excluded from the detailed analysis.
Data for NSW and the ACT were collected as part of the Neonatal Intensive Care Units Data Collection (NICUS) from the 10 neonatal intensive care units (NICUs) in the two states. This is a routinely collected prospective audit of infants admitted to an NICU during the first 28 days of life who meet one of the following criteria: gestational age <32 weeks; birth weight ⩽1500 g; assisted ventilation for 4 h or more started during the first 28 days of life; major surgery (opening of a body cavity) first performed during the first 28 days of life.
There were 76 well infants of 30–31 weeks gestation who were cared for in the nursery of the non-tertiary obstetric hospital where they were born and were not admitted to an NICU. A further 22 infants born in non-tertiary hospitals near the inland borders of NSW were preferentially transferred interstate (Queensland, Victoria, South Australia) for intensive care. Another ∼20 infants per year, originating from near the inland borders of NSW, were preferentially transferred interstate in utero for intensive care. Data for these infants are not collected in NICUS unless they are also admitted to a NSW/ACT NICU within 28 days of birth.
The present version of the data collection was established in January 1992. NICUS data are collected by a designated neonatal clinical nurse specialist situated in each NICU who completes a standardised dataset for each mother/infant dyad. Information is obtained from the clinical records, discussions with the staff and mother, and, where appropriate, personal observation. The NICUS data collection was validated for 1997 births.10 Data for infants who were stillborn or died before admission to a NICU were obtained from the NSW Midwives Data Collection and the ACT Maternal and Perinatal Data Collection. These two data collections started in 1990 and include all births ⩾20+0 weeks gestation and/or ⩾400 g birth weight.
Organisation of maternity and neonatal care
At the time of the study, maternity care in Trent was centred on 16 perinatal units all with delivery and neonatal facilities. Access to regular antenatal assessment was freely available. The three smallest hospitals (with annual birth rates around 1000) offered no neonatal intensive care beyond stabilisation. However, the remainder (annual birth rates 2000–6000) all aimed to offer neonatal intensive care to a percentage of their inborn infants. Although there were four recognised tertiary centres, their limited number of intensive care cots meant there was no strategy for these units to take in the bulk of high-risk cases from the remainder of the region. Very few deliveries occurred at home or in private clinics. As a result, paediatric expertise was immediately available for almost all deliveries.
At the time of the study, maternity care in NSW/ACT was centred on 74 obstetric hospitals, 56 public and 18 private (all in urban areas). Access to regular antenatal assessment was freely available. Sixty-three hospitals offered no neonatal intensive care beyond stabilisation. Tertiary perinatal care was confined to 10 hospitals (eight perinatal centres and two co-located children’s hospitals). For many logistic and human resource reasons, all were located within the Sydney Metropolitan area, Newcastle (150 km north of Sydney) and Canberra (300 km south of Sydney). However, there was a well-established network of antenatal referral and transport of high-risk mothers to tertiary obstetric hospitals in conjunction with a high-risk fetal–maternal advisory consultant service. There was also an experienced neonatal and paediatric emergency transport service with comprehensive medical and nursing expertise that provided advice to paediatricians and postnatal transfer of all critically ill newborns to NICUs.
World Health Organisation (WHO) and Central Intelligence Agency (CIA) statistics suggest that the UK and Australia are socioeconomically similar. In 2002, gross domestic production (GDP) per capita was slightly higher in Australia at US$26 900 vs US$25 500.11 Between 1997 and 2001, Australia consistently spent ∼20% more on health care as a percentage of GDP (8.5% rising to 9.2% of GDP, compared with 6.8% rising to 7.6% of GDP in the UK).12
The WHO’s preferred statistic for comparing health inequality (based on the probability of a child surviving until 2 years) and termed equality of child survival suggests that the UK had less health inequality in 2000 (0.999 vs 0.977 for Australia).13 This is possibly due to poor access to care for the Australian indigenous and rural/remote communities.11 13 14 The Trent population (census 2001) is 93% White or European, 4% Asian, 1% Black with other minorities and mixed race comprising the other 2%, and in Australia (Births Australia 2002), White or European, Asians and others comprise 92%, 7% and 1%, respectively.11 Teenage pregnancy rates, an associate of low socioeconomic standing, are difficult to assess because of a paucity of recent Australian data, although for 1995 Singh and Darrock15 documented 43.7 pregnancies per 1000 Australian women aged 15–19, compared with 46.9 in England and Wales.
In 2003, infant mortality in Australia was 9% lower than in the UK (4.83 vs 5.28 deaths per 1000 live births).11 Mortality in the first 5 years of life was also lower in Australia at this time: six male and five female deaths per 1000 children compared with seven male and six female deaths per 1000 children in the UK.16
For the purposes of this comparison, a standard set of tables was agreed on and then created for both regions. Rates for events were reported together with 95% CI.
Differences in binary outcomes between the two regions were quantified using estimated odds ratios (ORs) with 95% CI. As the continuous outcomes were not thought to come from approximately normal distributions, the observed outcomes in the two regions were investigated using the Mann–Whitney test.
Risk-adjusted mortality was achieved using a reduced version of the clinical risk index for babies (CRIB) II.17 Missing data of items used to calculate the score were substituted with a value that had a neutral effect on overall risk—for example, for base, excess missing values were assumed to be zero.
SAS V8.2 software was used for all analyses.
Census data indicated that the population of NSW/ACT was ∼35% higher than that of Trent; however, the number of births (all gestations) over the period of the study was 66% higher (273 495 vs 164 824). Official rates of perinatal mortality were 7.7, 7.8 and 7.2 per 1000 births in NSW/ACT for 2000, 2001 and 2002, respectively. The equivalent figures for Trent were 8.5, 8.1 and 8.7 per 1000 births. Official rates of neonatal mortality were 3.2, 3.4 and 2.9 per 1000 births in NSW/ACT for 2000, 2001 and 2002, respectively. The equivalent figures for Trent were 4.4, 3.6 and 4.2 per 1000 births. Total stillbirths (all gestations) during the period were 1434 in NSW/ACT and 945 in Trent. These represented rates of 5.24 (95% CI 4.97 to 5.51) and 5.73 (95% CI 5.37 to 6.11), respectively. Risk of stillbirth (all gestations) was significantly higher in Trent: OR 1.09; 95% CI 1.00 to 1.19; p = 0.033).
Table 1 details the study populations. The proportion of liveborn infants between 22 and 31 weeks gestation was significantly higher in Trent (NSW/ACT 2945, rate per 1000 live births 10.82 (95% CI 10.43 to 11.22); Trent 2208, rate per 1000 live births 13.47 (95% CI 12.92 to 14.05)). The proportion of these infants admitted to a neonatal unit was also higher in Trent (91.2% vs 94.4%; OR 1.63 (95% CI 1.30 to 2.05)).
The unadjusted mortality in the two populations admitted to a neonatal unit within the study regions showed no difference: NSW/ACT 332/2686 (12.4%), Trent 284/2085 (13.6%); unadjusted OR = 1.12 (95% CI 0.94 to 1.33; p = 0.21). The use of CRIB II to adjust for disease severity increased the OR, but the 95% CI still crossed zero.
Although there was a trend towards improved survival in the population of very preterm infants admitted to a neonatal unit in NSW/ACT, there did appear to be a greater degree of selection in terms of which liveborn infants 22–31 weeks gestation were admitted for neonatal care in NSW/ACT. This is shown by looking at the mortality for all live births in the gestation category where the trend (again non-significant) is in the opposite direction: Trent 407/2208 (18.4%); NSW/ACT 591/2945 (20.1%); OR 0.90 (95% CI 0.78 to 1.04; p = 0.14).
Table 2 shows the effect of maternal age. In both settings, we used mothers aged 20–24 as a reference group, but the risk of death in each setting did not appear to be related to maternal age.
Table 3 shows details of the infants admitted to a neonatal unit. Although the median gestations were identical, the distribution was significantly different, with 44% vs 41% <29 weeks gestation in NSW/ACT and Trent, respectively. In terms of the management of those infants admitted to a neonatal unit, NSW/ACT infants were more likely than Trent infants to be transported after birth (all types of transfer included) (60.7% vs 31.0%; OR 0.29 (95% CI 0.25 to 0.33)), to receive antenatal corticosteroids (88.2% vs 84.5%; OR 0.72 (95% CI 0.61 to 0.86)), and to have been treated with any continuous positive airway pressure (73.9% vs 53.1%; OR 0.40 (95% CI 0.35 to 0.46)). They were less likely to receive surfactant (48.4% vs 65.6%; OR 2.04 (95% CI 1.81 to 2.29)). There was no difference between the two regions with regard to emergency and elective caesarean section (57.5% vs 55.2%), mechanical ventilation rates (61.7% vs 62.5%), and oxygen administration at 28 days, for infants alive at 28 days (30.3% vs 32.5%) for NSW/ACT and Trent, respectively. The median length of stay for surviving infants was significantly different: 56 days (interquartile range (IQR) 42–78) in NSW/ACT and 43 days (IQR 30–64) for Trent.
This study provides an important insight into why comparison of national statistics for perinatal and neonatal mortality is fraught with difficulty. Perhaps in developing countries, where mature infants make up the bulk of the deaths, the success of programmes focused on maternal and infant health can be assessed by these statistics. However, in countries such as Australia and the UK it is the very-preterm population that is the major component of the figures. We have identified two inherent, and unexplained, influences that appear to have a major impact on the number of deaths arising from the gestational age group we have studied.
Firstly, the UK appears to have a particularly high rate of very-preterm delivery. We have noted this previously in comparisons with Denmark and France,2 18 but in this study the nature of the available data makes the point most graphically. The similar rates of delivery by caesarean section of infants at this gestation indicates that this excess is not due to differences in obstetric management but rather to a higher rate of spontaneous preterm labour. Given what is known about these two countries and the data provided here, this difference seems unlikely to be related to the availability of health care, an excess of very young mothers, or even genetic differences. In this study, we were not able to assess the influence of increasing numbers of “older mothers”, noted by others19 to carry a higher risk of preterm delivery. Artificial reproductive treatments are associated with higher rates of prematurity.20 We have no direct data, but the higher rate of multiple births in the NSW/ACT population suggests that any influence of this type was more likely to have been present in the NSW/ACT births. Other potential causes of the observed differences are exposure to poor diet, manual work and a range of factors related to deprivation, which lead to increased stress in the pregnant woman. Evidence is emerging that links each of the above to a potential biological mechanism that results in preterm birth such as the presence of organisms in the genital tract.21–24 Further research related to these associations will be important if we wish to prevent preterm birth rather than deal with its consequences.
What is already known on this topic
The UK has a high rate of neonatal and infant mortality compared with apparently similar countries.
Much attention has been focused on whether this results from either the organisation or effectiveness of UK perinatal services.
What this study adds
UK neonatal and infant mortality is adversely affected by a high rate of premature delivery compared with countries such as Australia.
In this comparison, Australian units were more selective in terms of the proportion of very preterm infants admitted for neonatal care.
The second issue that has a major effect on the neonatal population of any locality is the concept of viability. It seems that a greater proportion of the infants born alive in this gestation band in Trent were admitted to a neonatal unit. In the literature it is clear that opinion does vary about the appropriateness of offering intensive care at the lowest gestations25 and we will not rehearse those arguments here.26 However, in this comparison, decisions about who to admit appear to have influenced not only survival figures but also the workload for the neonatal service and the apparent “performance” of that service. It is clear from the data presented here that the performance of the perinatal services can be viewed differently depending on whether one considers all very preterm live births or simply those that are admitted to a neonatal unit. It is clear from the comparison of how infants were treated in the respective services that some major differences exist. It seems unlikely that these represent population differences, but instead genuine differences in some aspects of how very preterm infants are cared for in the two settings. Whether these differences have an important effect on outcome is difficult to judge from these sorts of data. Of those outcome measures that were available to us, only length of stay was clearly different, and we believe that this was more likely to represent an effect of the organisation of the Australian service, where long distance “transfers back” are perhaps undertaken more cautiously.
It is reasonable to ask whether we have dealt adequately with the challenges posed by comparisons of this type. We may have missed some infants because of “cross boundary flows” or because fetal losses at a very early gestation have not been properly recorded. However, both of the populations used in this study have longstanding surveillance mechanisms in place to assess these sources of error.
There are important implications beyond the fact that it is difficult to compare the performance of reproductive health services by using crude perinatal and neonatal mortality. The first is that any comparison of reproductive health services must be able to take account of underlying differences of the type identified in this study.
The second relates to health policy. In countries such as the UK where the rate of very preterm delivery is higher than otherwise similar countries, understanding this difference should be a priority. Consequently there is a need for developed countries to establish accurately their rates of premature delivery.27 The causes of excess preterm birth are clearly far wider than teenage pregnancy, which has been the focus of much attention in the UK. In taking this agenda forward, it is hard to ignore the well-established link between “deprivation” and preterm birth and more recent evidence linking increasing maternal age and use of assisted conception with prematurity.28 Understanding the biological links between social inequality, maternal age, assisted conception, and preterm birth represents a major challenge for developed societies.
We acknowledge the continuing help and collaboration of the hospitals delivering perinatal care in both Trent and adjacent regions. This study is one of the Trent Infant Mortality and Morbidity Studies, which are funded by the PCTs of the old Trent Health Region. In NSW/ACT, we acknowledge the continuing help and collaboration of the staff of all the hospitals delivering perinatal care. This study was performed on behalf of the Neonatal Intensive Care Units’ (NICUS) Group.
All authors were involved in aspects of planning and carrying out the study and all authors contributed to the final version of the paper. BM performed the analysis and provided interpretation of the findings. DF is the guarantor.
Competing interests: None.
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.