Towards reducing variations in infant mortality and morbidity: a population-based approach

Stream 1

This stream was designed to focus on an analysis of stillbirth and early childhood mortality rates, with the aim of improving understanding of the different influences on these rates in the UK. The work relied heavily on both routine data and more specialist data collected for other purposes.

Stream 2

This stream focused on the contribution of the group of babies born at 32–36 weeks’ gestation to both early mortality and morbidity. This work was based on a cohort of babies born LMPT and a random set of term-born control infants.

Description of data sets utilised

In order to achieve this objective, analyses utilised data at a national level to explore health inequalities in cause-specific infant, neonatal and perinatal mortality. As this study was based on routinely collected data that were anonymised, there was no requirement for ethics approval. The study focused on the 12-year period of 1 January 1997 to 31 December 2008 and utilised several national data sets.

Measurement of socioeconomic deprivation

Previous government targets in England and Wales measured the relative deprivation gap by using a classification of socioeconomic group based on the father’s occupation. 12 This excluded both infants whose parents had never worked and those who were solely registered by the mother. This led to the exclusion of a significantly at-risk group. Consequently, this method of measurement of socioeconomic group is inadequate. Ideally, individual measures of socioeconomic deprivation would have been used. However, as the overall aim of this programme was to utilise routine data to assess and monitor socioeconomic inequalities, an area-level measure of deprivation was chosen to fulfil this role. Therefore, socioeconomic inequalities were measured by using the Index of Multiple Deprivation (IMD) for 200416 at the LSOA level.

This measure of multiple deprivation is made up of seven domain indices at the LSOA level, which relate to (1) income deprivation; (2) employment deprivation; (3) health deprivation and disability; (4) education; (5) skills and training deprivation; (6) barriers to housing and services; and (7) living environment deprivation and crime. LSOAs are the smallest areas for which these deprivation data are available; although some degree of heterogeneity exists within them, the small size of the areas (only 1500 residents) limits this.

All LSOAs in England were ranked by deprivation score. They were then weighted by its population of births (using the live-birth denominator data) and divided into 10 groups with approximately equal populations of births in each from 1 (least deprived) to 10 (most deprived). Thus, when calculating mortality rates, if stillbirth or neonatal mortality was the same for all deprivation groups, a similar number of neonatal deaths would be expected in each of the 10 groups. This approach offered the potential not only to monitor mortality but also to allow targeted interventions at an area level in the future.

Statistical analyses

The aim of these analyses was to explore the deprivation gap in neonatal and infant mortality over time, first at an all-cause level and then by specific cause to unpick the key causes of death that related to socioeconomic inequalities in mortality. In order to explore trends by socioeconomic deprivation neonatal mortality rates for each cause of death by deprivation decile and time period were calculated. Analyses were undertaken solely for singleton births in this study because, in relation to multiple births, a variety of factors (such as differential access to fertility treatment) affects the rate at which multiple births occur and, in addition, multiple births are associated with both a higher mortality rate and additional specific causes of death.

Deprivation gap in all-cause neonatal mortality over time

Between 1997 and 2007, a total of 18,524 neonatal deaths of singleton infants were notified to CMACE All-cause neonatal mortality decreased over time from 31.4 per 10,000 live births (1997–9) to 25.1 per 10,000 live births (2006–7), but this differed by deprivation group. In absolute terms, rates were significantly higher in the most deprived areas (Table 1). Between 1997–9 and 2006–7 rates decreased more in the least deprived decile than in the most deprived decile (most deprived decile, 7.4 fewer deaths per 100,00 births; least deprived decile, 5.6 fewer deaths per 10,000 births). However, the relative reduction in mortality over time was smaller in the most deprived decile (17%) than in the least deprived decile (26%), leading to a widening of the deprivation gap. Table 2 details the regression equation for all-cause mortality. In 1997–9 infants risk of neonatal death in the most deprived decile was twice that in the least deprived [mortality rate ratio 2.08, 95% confidence interval (CI) 1.92 to 2.27], and this gap widened significantly over time, reaching a peak of 2.68 in 2003–5 before slightly narrowing to 2.35 in 2006–7. Consequently, the percentage of excess deaths associated with deprivation increased over the time period from 32.3% (1997–9) to 51.0% (2003–5) and then decreased to 37.5% (2006–7). Consequently, there was no evidence of an overall reduction over the time period to achieve the relevant NHS targets.

Deprivation gap in specific-cause neonatal mortality over time

The next step was to explore these deaths in more detail by investigating the cause-specific mortality. Immaturity was the most common cause (45.4%), followed by congenital anomalies (24.5%), intrapartum events (10.8%), infection (9.8%), accidents and other specific causes (7.1%), and sudden infant deaths (3.0%). The number of deaths increased with increasing deprivation for each cause, although the magnitude of this increase varied by cause. With the exception of deaths due to immaturity of < 24 weeks’ gestation, neonatal mortality fell over time for all causes, with the greatest falls seen for immaturity of 24–27 and 28–36 weeks’ gestation.

For five of the eight causes of death [congenital anomalies, immaturity (< 24 weeks and 24–27 weeks), infection and accidents and other specific causes], there was a larger relative fall in mortality over time in the least deprived decile than in the most deprived decile, although this was only statistically significant for congenital anomalies (Table 3). For these causes and immaturity (28–36 weeks) the trend in the deprivation gap over time was similar to all-cause neonatal mortality, with an initial two- to threefold ratio deprivation gap in 1997–9 (neonatal mortality rate ratio range 1.70–2.98) which increased up to 2003–5 (range 2.17–4.14) followed by a slight narrowing in 2006–7 (range 1.72–3.16) (see Table 3). The widest deprivation gap was seen for death due immaturity of < 24 weeks’ gestation, with the risk of death in the most deprived decile in 1997–9 being threefold that in the least deprived decile; the increased risk of death rose to over fourfold in 2003–5 and fell again slightly to threefold in 2006–7. As the absolute mortality did not fall over this time period, this widening of the deprivation gap led to the deaths due to immaturity of < 24 weeks’ gestation representing a larger proportion of all deaths in 2006–7 (21.7%) than in 1997–9 (16.9%).

Rates of intrapartum death and sudden infant death showed fell more among the most deprived decile than in the least deprived, leading to a non-significant narrowing of the deprivation gap in mortality of these causes. However, these deaths constituted only 13.5% of deaths, and their impact on all-cause mortality was small. Deaths due to intrapartum events showed the narrowest deprivation gap (ranging from 1.15 to 1.37 across the time period). Consequently, the reduction in deaths if the rates seen in the least deprived decile were applied across all deciles was small, ranging from 6.8% to 14.8%. In contrast, the deprivation gap for sudden infant deaths was the widest seen for any specific cause in 1997–9 (mortality rate ratio 3.62). This showed a non-significant decrease over time to 2.32 in 2006–7. Sudden infant deaths would have been reduced by over half in 1997–9 if mortality rates for the least deprived decile were applied to the whole population, compared with a reduction of just over one-third in 2006–7.

Graphical approach to understanding the deprivation gap

In order to best convey the cause-specific socioeconomic inequalities in mortality, a graphical approach was chosen. Figure 1 demonstrates the percentage of the deprivation gap in all-cause neonatal mortality explained by each specific cause. The width of the bands is directly related to the proportion of the deprivation gap in neonatal mortality that each cause accounts for. This graph clearly demonstrates that deaths due to immaturity (babies of < 24, 24–27 and 28–36 weeks’ gestation) and congenital anomalies explain the majority of the deprivation gap in all-cause mortality. Thus, all deaths due to immaturity and congenital anomalies combined accounted for 77% of the deprivation gap in 1997–9, rising to a peak of 81.9% in 2003–5 and then declining again to 79% in 2006–7. This increase in the proportion of the deprivation gap explained by prematurity and congenital anomalies was due to a combination of a widening deprivation gap in mortality for these causes, the high proportion of deaths due to these causes and a lack of decline in mortality due to immaturity at < 24 weeks. The remaining causes (sudden infant death, intrapartum events, infection, accidents and other causes) account for only 20% of the deprivation gap. Their reduced impact is partially related to their relatively smaller contribution to overall mortality, but also the narrow deprivation gap in mortality for intrapartum deaths. The percentage of the gap explained by sudden infant deaths fell over time from 5% (1997–9) to 2.5% (2006–7).

FIGURE 1.

Proportion of the deprivation gap in all-cause neonatal mortality rates explained by each cause of death over time. Reproduced from Smith et al.,20 © 2010, British Medical Journal Publishing Group, under the terms of the Creative Commons Attribution Non-Commercial Licence (CC BY-NC 2.0), which permits use, distribution and reproduction in any medium, provided the original work is properly cited, the use is non-commercial and is otherwise in compliance with the licence.

Key findings

• As there was a decrease in neonatal mortality over the period 1997–2007, the relative deprivation gap (the ratio of mortality in the most deprived decile to that in the least deprived decile) increased, particularly for deaths related to congenital anomalies and immaturity.

• Almost 80% of the relative deprivation gap in all-cause mortality was explained by premature birth and congenital anomalies.

Limitations and strengths

Implications for policy and practice

Although targets were set to reduce inequalities in infant mortality by 2010 in the UK,8 the deprivation gap did not narrow. Cause-specific analyses provide more detailed information, highlighting the contribution of each causal group and the impact of interventions or changes in society over time.

In the absence of cause-specific analyses various identifiable actions were recommended in England and Wales to reduce the deprivation gap in infant mortality by the target 10% (Dr Marilena Korkodilos, Specialist Public Health Services at Public Health England, 2010, personal communication). These are, in order of magnitude of anticipated reduction:

1. increasing breastfeeding rates and reducing obesity in the routine and manual group

2. reducing rates of smoking during pregnancy

3. alleviating overcrowded housing

4. reducing teenage conceptions.

However, while these were all laudable aims, it seems clear from the work presented here that, unless interventions target specifically the risk of very premature birth and potentially lethal congenital abnormalities, the impact on the deprivation gap is likely to be minor. For example, even accounting for the higher rates of infant mortality due to sudden infant death compared with neonatal mortality, based on our findings the contribution to the total mortality gap is simply too small to have a significant impact. The situation is somewhat different for smoking, as there is evidence to suggest that smoking is an important factor in the aetiology of preterm birth and, as a consequence, infant mortality. 21 The potential impact of reducing obesity and teenage pregnancy is less clear, in terms of existing evidence. However, smoking, obesity and teenage pregnancy have all been the subject of longstanding public health campaigns of limited success and the UK suggested goals required major behavioural changes. Our lack of understanding about the everyday environmental influences on the risk of preterm birth and major congenital abnormalities appears to be a significant impediment to the development of a rational strategy for diminishing the influence of deprivation on measures of early childhood mortality. Research has previously demonstrated the importance of immaturity in the UK compared with other European countries22 and the March of Dimes has highlighted the problem in global terms. 13 Tackling the wide deprivation gap among those less born at than 24 weeks’ gestation is likely to be achievable not through further progress in neonatal care but only through prevention. The need for a greater understanding of the mechanistic link between deprivation and prematurity is a major research priority, which would then allow a focus on primary preventative strategies to reduce the rate of prematurity itself. Our lack of understanding about the influence of health inequality in relation to major congenital anomalies deserves no less attention.

These findings point to ways in which our understanding of the social influences on early childhood mortality rates in particular localities might be improved, for example:

• The annual cause-specific mortality should be measured and overall neonatal and infant mortality rates should be reviewed. Cause-specific analyses provide much greater insight into socioeconomic inequalities in neonatal mortality on a global level, thus facilitating each country’s/area’s understanding of its early childhood mortality rates and identifying appropriate interventions for prioritisation.

• The use of the father’s occupational class to assess socioeconomic group should be avoided, as this excludes single mothers from analyses, a significant at-risk group. Area-level deprivation measures offer an inexpensive and quick way to continue monitoring inequalities. However, the collection of, and timely access to, comprehensive individual-level information for neonatal deaths and denominator data would be a further improvement.

• Improved data linkage between large routine data sets such as NN4B would facilitate considerable improvement in the research related to socioeconomic inequalities in infant mortality.

• Greater focus should be placed on the influence of preterm birth and congenital anomalies in neonatal mortality.

• Strategies for reducing socioeconomic inequalities should focus on the prevention of preterm birth rather than improvements in neonatal care, particularly for babies born before 24 weeks’ gestation.

Recommendations for future research

• There is a great need for a greater understanding of the mechanistic link between deprivation and prematurity. In order to reduce socioeconomic inequalities in neonatal mortality rates, there needs to be a shift in focus from targeting risk factors that have a minimal effect on prematurity rates to major primary prevention strategies for preterm birth.

• The methods proposed here for monitoring inequalities in neonatal mortality provide a quick and straightforward way of monitoring socioeconomic inequalities in the future. However, the collection of, and timely access to, comprehensive individual-level information for neonatal deaths is required to confirm the findings at an individual level.

Description of data sets utilised

In order to achieve these objectives, analyses utilised national-level data to explore health inequalities in cause-specific stillbirths. As this study was based on routinely collected data that were anonymised, there was no requirement for ethics approval. The study focused on the 12-year period between 1 January 1997 and 31 December 2008, and utilised several national data sets.

Individual-level data on all singleton stillbirths (infants born at ≥ 24 weeks’ gestation and showing no signs of life at birth) born to mothers resident in England between 1 January 1997 and 31 December 2008 were obtained from the CMACE, which had collected neonatal mortality and stillbirth data as part of its national perinatal mortality surveillance work funded by the National Patient Safety Agency until 2012. Data included cause of death, gestation, date of birth, mother’s age, sex of the infant, birthweight, ethnic group, multiplicity and mother’s place of residence (LSOA). Only stillbirths with a valid LSOA were included as, otherwise, no deprivation score could be assigned, but this excluded only 1% of deaths.

Classification of causes of deaths

For the national data on stillbirth, a local CMACE co-ordinator initially classified deaths using the obstetric (Aberdeen) classification system32 for stillbirths (Table 4). A CMACE regional manager then checked them with reference to post-mortem and coroner’s reports when available. Finally, CMACE carried out central cross-validation checks to ensure consistency.

Several of the rarer classification groups of the obstetric (Aberdeen) classification system were combined. As so many deaths were in the unexplained antepartum deaths category, these were then divided on the basis of birthweight (≤ 10th percentile or > 10th percentile), resulting in nine categories (see Table 4). Similar to the situation for neonatal deaths, data on stillbirths, with information on cause of death according to the Aberdeen classification, were available for only 8 years (i.e. from 1 January 2000 to 31 December 2007).

Denominator data: live births

The ONS birth registrations were combined with CMACE data on stillbirths to obtain a denominator of all births, used for calculating stillbirth rates. As there is no gestational age information in these data, this limited the detailed analyses of prematurity. As discussed in the analyses of neonatal deaths, birth data were obtained on the number of live births by year of birth and LSOA of residence in order to calculate mortality rates by LSOA. This allowed exploration of trends over time.

Measurement of socioeconomic deprivation

The same methodology was used for measuring socioeconomic deprivation as in the analyses of neonatal deaths, utilising routine data to assess and monitor socioeconomic inequalities. The IMD for 200416 at the LSOA level enabled allocation of a deprivation score to all infants with a valid postcode. These data were obtained from the ONS.

As for the analyses of neonatal deaths, all LSOAs in England were ranked by deprivation score. The LSOAs were then weighted by their population of births (using all births as a denominator) and divided into 10 groups with approximately equal populations of births in each from 1 (least deprived) to 10 (most deprived). Thus, when calculating mortality rates, if stillbirth rates were the same for all deprivation groups, a similar number of stillbirths would be expected in each decile.

Data linkage

National stillbirth data were provided with LSOA codes. The IMD 200416 was then linked to the mortality data matching on LSOA code. These data were then linked to the LSOA-level ONS birth denominator data set by LSOA-level code.

Statistical analyses

First, analyses were undertaken at an all-cause level and then by specific cause to identify the key causes of death that related to socioeconomic inequalities in stillbirth. In order to explore trends by socioeconomic deprivation, stillbirth rates were calculated for each cause of death by deprivation decile and time period. Analyses were undertaken for singleton births only, as, in relation to multiple births, a variety of factors (such as differential access to fertility treatment) affect the rate at which multiple births occur and, in addition, it is known that multiple births are associated with both a higher mortality rate and additional specific causes of death.

Exploring the deprivation gap

As for the analyses of neonatal mortality, Poisson regression models were used to assess trends in mortality by deprivation decile over time,17 fitting separate models for all-cause mortality and each specific cause of death for stillbirths. As discussed, UK targets for reducing socioeconomic inequalities in infant mortality8 were based on the relative deprivation gap, to avoid the influence of the underlying prevalence. The relative deprivation gap was assessed by fitting a linear trend between deprivation decile and mortality and calculating the mortality rate ratio between the most deprived and least deprived deciles, which is similar in approach to the relative index of inequality. 18 Significant change in the relative deprivation gap over time was assessed by fitting a separate deprivation effect for each time period. Reductions in stillbirth over time were assessed by calculating the relative change (percentage reduction in stillbirth rate by deprivation decile).

Once again the absolute change in the outcome (in this case stillbirth) was calculated over time by deprivation decile to assess improvements in mortality. The delta method was used to calculate confidence limits. 19 Excess mortality associated with deprivation as a percentage was estimated by separately applying the stillbirth rate in the least deprived decile to the total population and dividing that by the total number of stillbirths observed. The proportion of the deprivation gap in all-cause stillbirth rates explained by each cause was calculated for each time period. Then for each specific cause, the stillbirth rate was estimated in the least deprived decile and the most deprived decile for each time period by using the regression models. The absolute difference in these two rates was then calculated and expressed as a proportion of the absolute difference in rates for all causes combined.

Deprivation gap in all-cause stillbirth rates over time

From 2000 to 2007, 21,472 singleton stillbirths were reported to CMACE; of these, LSOA was missing in 120 (0.6%) and cause of death was missing or unclassifiable in 919 (4.3%), leaving 20,433 for analyses. First, the overall rate of stillbirth over time and by deprivation was assessed (Table 5). The overall stillbirth rate was 4.4 per 1000 births and there was no evidence of a change in stillbirth rate over time (2000–3 rate, 4.4 per 1000; 2004–7, 4.4 per 1000; p = 0.80). The total number of stillbirths in each deprivation decile increased as deprivation increased, with the number in the most deprived decile approximately double that in the least deprived. Women from the most deprived decile were twice as likely to experience a stillbirth of any cause than those from the least deprived (rate ratio 2.1, 95% CI 2.0 to 2.2; p < 0.0001). There was no evidence that this changed over time.

Deprivation gap in cause-specific stillbirth rates

Looking at stillbirths by cause of death (Table 6) revealed that antepartum deaths of unknown cause accounted for the highest percentage of stillbirths [59.2%: 21.3% small for gestational age (SGA); 37.9% not SGA] followed by antepartum haemorrhage (13.0%); maternal disorders (9.1%); congenital anomalies (7.8%); pre-eclampsia (4.2%) and mechanical issues during labour (2.4%). The remaining 4.3% were due to miscellaneous or unclassified reasons and were excluded from the Poisson regression analyses.

There was no evidence of trends of increasing or decreasing rates of stillbirth over time for any specific cause (Table 7); however, the deprivation gap varied by cause. Mechanical issues during labour was the only specific cause for which there was no evidence of a deprivation gap [relative risk (RR) 1.2, 95% CI 0.9 to 1.5]. All other causes of stillbirth showed a significant deprivation gap, varying from a 1.7- to a 3.1-fold difference. The widest deprivation gap was seen for deaths due to antepartum haemorrhage; women from the most deprived decile were 3.1 (95% CI 2.8 to 3.5) times more likely to experience stillbirth of this cause than those from the least deprived decile. Wide deprivation gaps were also seen for deaths due to congenital anomalies (rate ratio 2.8, 95% CI 2.4 to 3.3) and maternal disorders such as hypertension (RR 2.2, 95% CI 1.9 to 2.5). The deprivation gap was wider for stillbirths of infants who were SGA (RR 2.5, 95% CI 2.3 to 2.7) than for those who were not SGA (RR 1.7, 95% CI 1.5 to 1.8; p = 0.26) (see Table 4). The percentage excess deaths of all causes related to deprivation was 33%, suggesting in total one-third more stillbirths were observed than would have been expected if the stillbirth rate for all deprivation groups were the same as that for the least deprived decile.

Graphical approach to understanding the deprivation gap

Again, in order to best illustrate the cause-specific socioeconomic inequalities in stillbirth, a graphical approach was chosen. Figure 2 demonstrates the percentage of the deprivation gap in all-cause stillbirth mortality explained by each specific cause estimated from the Poisson regression models. This figure highlights that those deaths due to unexplained antepartum events account for 50% of the deprivation gap. Although, overall, among unexplained antepartum deaths, the proportion of infants who were SGA were smaller than the proportion who were not (22.3% SGA and 39.6% not SGA), the SGA group explains more of the deprivation gap. This is because the deprivation gap was wider for stillbirth infants who were SGA (RR 2.5, 95% CI 2.3 to 2.7) than for those who were not SGA (RR 1.7, 95% CI 1.5 to 1.8). There was no evidence of a change in the proportion of the deprivation gap explained by any of the different causes over time, which can be seen by the lack of change in the gradient of the lines representing each specific cause. Mechanical causes are seen to constitute a very small, insignificant proportion of the deprivation gap.

FIGURE 2.

Proportion of the deprivation gap in all-cause stillbirth rates explained by each cause of death over time. Reproduced from Seaton et al. 33 © 2012, British Medical Journal Publishing Group, under the terms of the Creative Commons Attribution Non-Commercial Licence (CC BY-NC 2.0), which permits use, distribution and reproduction in any medium, provided the original work is properly cited, the use is non-commercial and is otherwise in compliance with the licence.

Key findings

• Rates of stillbirth were twice as high in the most deprived decile as in the least deprived.

• This wide gap remained constant over time and was not diminishing.

• There was a significant deprivation gap for most specific causes of stillbirth.

• Unexplained antepartum stillbirths accounted for 50% of the deprivation gap.

Limitations and strengths

Implications for policy and practice

Cause-specific analyses provide more detailed information, highlighting the contribution of each causal group to deprivation differentials and the impact of interventions or changes in society over time.

Recent reductions in the stillbirth rate in other high-income countries24 suggest that there exist modifiable risk factors and that by the introduction of targeted interventions, an improvement in stillbirth rates could be seen, in particular the early identification of close monitoring of fetal movements. 39 Maternal smoking may be targeted successfully to impact on the rate of stillbirths, but effective tools to reduce maternal obesity and rates of teenage pregnancy are currently lacking. 40 Smoking, obesity and teenage pregnancy have all been the subject of longstanding public health campaigns of limited success and the UK suggested goals require major behavioural changes.

Some practical measures could help in improving our understanding of the role of deprivation in relation to stillbirth:

• Cause-specific socioeconomic inequalities in stillbirth should be monitored annually.

• Avoid the use of the father’s occupational class to assess socioeconomic group, as this excludes single mothers from analyses, a significant at-risk group. Area-level deprivation measures offer an inexpensive quick way to continue monitoring inequalities.

• Improved data linkage between large routine data sets such as NN4B would facilitate considerable improvement in the research related to socioeconomic inequalities in infant mortality.

• The implementation of an improved, internationally adopted, classification system would enhance the ability to identify other modifiable risk factors. In addition, such a system would facilitate the implementation of appropriate targets and interventions and reduce the proportion of stillbirths assigned to unknown causes.

Recommendations for future research

• The lack of change in stillbirth rates in the UK and the persistent wide socioeconomic inequalities highlights the need for further research to understand this intractable problem.

• Assessment of the impact of recent professional recommendations on the classification of births after 24 weeks’ gestation known to have died earlier as ‘late fetal losses’.

Socioeconomic inequalities in deaths associated with congenital anomaly

Following the national work on understanding socioeconomic inequalities in neonatal mortality and stillbirth, it became apparent that a key area of concern was the widening gap in socioeconomic inequalities in neonatal mortality relating to congenital anomalies. Deaths as a result of a congenital anomaly accounted for the largest proportion of the deprivation gap in neonatal mortality due to a single cause, and also represented a significant proportion of the deprivation gap in stillbirths. Understanding how these inequalities relating to congenital anomalies arose seemed key to implementing effective public health interventions to reduce socioeconomic inequalities in infant and neonatal mortality.

Socioeconomic inequalities in congenital anomalies have been demonstrated in the rates of stillbirth and perinatal, neonatal and infant mortality. 20,30,41,42 Research had shown an increasing risk of non-chromosomal anomalies with increasing deprivation, in contrast to a decreasing risk of chromosomal anomalies. 43 This last finding was predominantly a result of the increased risk of chromosomal anomalies with increasing maternal age. However, the influence of socioeconomic deprivation along the pathway from antenatal detection to delivery and possible neonatal mortality was not fully understood because of the lack of clearly defined standardised data in the antenatal period. Countries that have introduced the use of prenatal diagnostic techniques and access to termination of pregnancy because of congenital anomaly have reported large reductions in neonatal mortality rates, in contrast to those countries with more restrictive policies on pregnancy termination. 44–47 Nevertheless, the impact of these secondary preventative measures might vary with socioeconomic deprivation in terms of access to, and timing of, antenatal detection services through to the provision of information, the interpretation of risk, and the consequent decision-making regarding continuation or termination of a pregnancy.

Evidence in this area is sparse. A systematic review of UK studies showed no evidence of social inequalities in the uptake of prenatal screening,48 whereas research in Northern Ireland,49 where there is no provision of termination services, showed inequalities in both the offer and the uptake of screening. Further research suggested that socioeconomic differentials in decision-making following antenatal detection are a result of maternal age differences. 50 The term ‘congenital anomaly’ covers a very wide spectrum from the relatively minor to those with an exceptionally poor prognostic outcome and it is secondary preventative measures targeted at the latter that have the potential to improve infant mortality rates.

A detailed analysis of regional data was undertaken to identify the underlying reasons behind the inequalities seen in the analyses of neonatal mortality and stillbirth. Data were used from a large population-based congenital anomaly register in England (EMSYCAR) covering about 10% of the births in England and Wales, for 1998–2007, to investigate socioeconomic inequalities in the risk of congenital anomalies with a poor prognosis from antenatal diagnosis to end of pregnancy. The impact of variations in rates of termination of pregnancy for congenital anomaly on rates of stillbirth, live birth and neonatal mortality associated with congenital anomaly were explored to aid understanding of the reasons for the widening socioeconomic inequalities in neonatal and infant mortality in England.

Description of data sets

Regional data: congenital anomalies incidence and mortality

The EMSYCAR currently covers about 74,000 births annually (about one-tenth of all births in England and Wales), with around 2200 reported cases per year. The region covered by the register can be seen in Figure 3. Data for Northamptonshire, which joined the register only in 2003, were excluded, leaving a geographical area with about 60,000 births annually.

FIGURE 3.

English counties covered by the EMSYCAR.

This register is population based and includes all structural and chromosomal congenital anomalies in fetuses and infants of mothers living within the region at the time of delivery. It includes live births, stillbirths (from 24 weeks’ gestation), spontaneous fetal loss (before 24 weeks’ gestation) and termination of pregnancy for fetal anomaly at any gestational age. The register uses multiple sources of case ascertainment from within the care pathway, including antenatal ultrasonography, antenatal screening, delivery reports, birth notifications, pathology, cytogenetics, clinical genetics and paediatric surgery. All reported anomalies are coded in accordance with the International Classification Of Diseases, Tenth Edition (ICD-10). Information on maternal age and ethnicity, mother’s postcode of residence at delivery, end date of pregnancy and gestation at delivery were available from data collected by the register. The study included fetuses with an anomaly with an end of pregnancy date between 1 January 1998 and 31 December 2007. These data were linked to the ONS birth registrations and the Confidential Enquiry into Maternal and Child Health (CEMACH) data by LSOA to look at morbidity and mortality. Ideally, analyses would be based on the mother’s postcode at conception, but this was not available from the register. Analyses were therefore based on the mother’s postcode at delivery. To assess the potential impact of changes in the mother’s residence through the pregnancy on the observed socioeconomic inequalities, deprivation at antenatal detection and delivery was assessed separately for those fetuses for which this information was available (i.e. fetuses detected in the antenatal period when the mother opted to continue with the pregnancy).

Socioeconomic inequalities following diagnosis of a congenital anomaly

Table 9 demonstrates that there was no evidence of a difference in the overall rate of registrations of the nine anomalies by deprivation (rate ratio comparing the most deprived decile with the least deprived decile 1.05, 95% CI 0.89 to 1.23). In 86% of cases, the anomaly was detected during the antenatal period and, similarly, there was no evidence that this varied with deprivation (rate ratio 0.99, 95% CI 0.84 to 1.17) or ethnicity (p = 0.913). However, termination of pregnancy when an anomaly was detected antenatally was substantially less common in the most deprived decile (63%) than in the least deprived decile (79%; rate ratio 0.80, 95% CI 0.65 to 0.97), and this finding was similar after adjusting for maternal age differences (rate ratio 0.79, 95% CI 0.65 to 0.97) but slightly attenuated after adjusting for differences in termination rates by ethnicity (rate ratio 0.86, 95% CI 0.70 to 1.05). This was because, although rates of termination were similar for mothers classified as white British (71%) and Indian (Asian or Asian British) (71%), they were considerably lower for Pakistani (Asian or Asian British) mothers (42%).

These socioeconomic variations in termination of pregnancy for congenital anomaly impacted greatly on socioeconomic inequalities in the rate of fetal loss, stillbirth and live birth associated with an anomaly and also subsequent neonatal death. Considering all of the selected anomalies combined, rate of stillbirth or fetal loss was 20% higher in the most deprived decile than in the least deprived decile while the rate of live birth was 61% higher and the rate of neonatal mortality was 98% higher. After adjusting for maternal age differences, socioeconomic inequality widened, with women of a similar age from the most deprived decile being 57% more likely to experience fetal loss or stillbirth, 85% more likely to give birth to a live infant and 123% more likely to have a baby who subsequently died in the neonatal period.

Exploring inequalities by type of anomaly

When looking at chromosomal and non-chromosomal anomalies separately (Table 10), the effect of deprivation differed, with women from the most deprived decile at increased risk of having a fetus with a non-chromosomal anomaly (rate ratio 1.41, 95% CI 1.17 to 1.70), but at reduced risk of a fetus with a chromosomal anomaly (rate ratio 0.52, 95% CI 0.39 to 0.69), compared with women from the least deprived decile. The greater risk of chromosomal anomalies was the result of differences in maternal age distribution between pregnant women from the most deprived and least deprived areas (women in the most deprived decile accounted for 25% of pregnant women > 35 years of age, whereas women in the least deprived decile accounted for 64% of this age group). As expected, the risk of chromosomal anomalies increased with increasing maternal age [women aged > 35 years were nearly five times more likely to carry a fetus with a chromosomal anomaly than younger mothers (rate ratio 4.96, 95% CI 4.12 to 5.98)]. In contrast, there was no evidence of a difference in the rates of non-chromosomal anomalies with maternal age (rate ratio 1.01, 95% CI 0.86 to 1.20). The age-adjusted RR for chromosomal anomalies showed no evidence of a difference in rate depending on deprivation level (rate ratio 0.85, 95% CI 0.63 to 1.15).

Despite these differences in registrations of chromosomal and non-chromosomal anomalies, the socioeconomic differences in antenatal detection rates and terminations of pregnancy were similar for both chromosomal and non-chromosomal anomalies. There was no evidence of a difference in antenatal detection rates with deprivation for either chromosomal (rate ratio 0.96, 95% CI 0.70 to 1.31) or non-chromosomal anomalies (RR 1.00, 95% CI 0.82 to 1.23). The socioeconomic differences in rates of termination of pregnancy seen for chromosomal and non-chromosomal anomalies when considered separately were similar to those seen for all anomalies combined.

In contrast, the rate of live birth associated with a non-chromosomal anomaly was 64% higher in the most deprived decile than in the least deprived decile and neonatal mortality was 130% higher. Socioeconomic inequalities for chromosomal anomalies appeared slightly narrower and were not statistically significant. The rate of live birth with a chromosomal anomaly was 50% higher in most deprived decile than in the least deprived decile (compared with 64% in the case of non-chromosomal anomalies) and neonatal mortality were 44% higher (compared with 130% in the case of non-chromosomal anomalies). However, adjusting for the wide maternal age differences increased the socioeconomic inequality for chromosomal anomalies to 121% increased risk for a live birth with an anomaly and to 104% for a neonatal mortality associated with an anomaly.

Key findings

• Rates of severe anomalies and method of detection were similar for all deprivation groups.

• Rates of termination after antenatal diagnosis of a congenital anomaly were lower in the most deprived areas than in the least deprived.

• This resulted in wide socioeconomic inequalities in live-born infants with a congenital anomaly and subsequent neonatal mortality.

Strengths and limitations

Implications for policy and practice

Although this work concentrated on a specific region of the UK, it is likely that these results are generalisable to the whole of the UK. The FASP aims to ensure consistent provision across the UK, although there may be some variation between centres in the antenatal detection rates of the selected anomalies and differences in uptake of screening programmes. This is unlikely to impact substantially on socioeconomic differences in behaviour following detection of a suspected anomaly. In England, Scotland and Wales secondary prevention of anomalies through access to termination of pregnancy is available to all. Internationally, this is not universally the case, as access to termination of pregnancy varies, but these findings may be mirrored in countries with a similar policy to that of England on the termination of pregnancy.

The use of prenatal diagnostic techniques and access to termination of pregnancy for fetal anomaly has reduced neonatal mortality internationally, but this research has demonstrated that these secondary prevention measures have had a knock-on effect of an increase in the socioeconomic inequality in mortality.

Going forward, comparisons of infant mortality rates between regions would be improved by taking into account of the variation in rates of termination of pregnancy for severe congenital anomaly since this will impact greatly on observed crude mortality rates. National monitoring of socioeconomic variations in congenital anomalies would facilitate evaluation of services and this may be possible following the introduction (in England) of the planned national congenital anomaly register.

Recommendations for future research

• It is vital that variations in congenital anomalies arising through the secondary prevention programmes based on screening and the uptake of termination of pregnancy do not detract from the importance of reducing inequalities in anomalies through primary prevention prior to conception as highlighted by Dolk. 53 Further research into the links between non-chromosomal anomalies and deprivation needs to be undertaken in order to identify primary prevention interventions.

• A decision to continue a pregnancy associated with a serious congenital anomaly should not be thought of as a flawed choice and may relate to societal and cultural norms. However, it is important that the reported socioeconomic variations in rates of termination do not arise from systematic differences in the delivery of services such as access to timely detection services, communication of mortality and morbidity risk by health professionals and access to a termination of pregnancy. Future research into the reasons underlying the socioeconomic variations in continuation of pregnancies associated with serious congenital anomalies is needed.

Description of data sets

Statistical analyses

Variation in preterm birth rates

There were 2,535,855 live births, 13,112 stillbirths and 2382 estimated late fetal deaths across the 147 PCTs. First, the live-birth rate was assessed by gestational age to assess how the rate of preterm live birth varied between PCTs. Rates of extremely preterm births (< 24 weeks’ gestation) varied widely between PCTs (Figure 4). Binomial regression models confirmed this pattern of variation between PCTs (SD between PCT variation < 24 weeks, 0.29; 24–27 weeks, 0.17; 28–32 weeks, 0.11) (see Figure 4), with the rate of extremely preterm birth being six times higher in the top 5% of PCTs than in the lowest 5% (90th percentile range 0.31 to 1.91 per 1000 births). In the older gestational age groups, this variation decreased, with a twofold and 1.5-fold difference at 24–27 and 28–32 weeks’ gestation, respectively (90th percentile range: 24–27 weeks, 2.48 to 5.26; 28–32 weeks, 10.24 to 15.75). This pattern was confirmed by binomial regression models with decreases in the between-PCT SD with increasing gestation (SD between PCT variation < 24 weeks, 0.38; 24–27 weeks, 0.18; 28–32 weeks, 0.11). After adjusting for socioeconomic deprivation, ethnicity and the age distribution of women in the PCT, the between-PCT variation was reduced for the more mature births but remained high in the extremely preterm group.

FIGURE 4.

Median, IQR and 90th percentile range of rate of preterm birth (live births and all births) by gestational age for PCTs (log-scale). Reproduced from Smith et al. 62 © 2013, BMJ Publishing Group Ltd and the Royal College of Paediatrics and Child Health, under the terms of the Creative Commons Attribution Non-Commercial Licence (CC BY-NC 3.0), which permits use, distribution and reproduction in any medium, provided the original work is properly cited, the use is non-commercial and is otherwise in compliance with the licence.

The next stage was to assess how the inclusion of fetal deaths and stillbirths affected the variation in the live-birth rate by gestational age. Including fetal deaths in the overall rate of very preterm birth (total births including fetal deaths) produced a reduction in the between-PCT variation in preterm births < 24 weeks’ gestation but little change at later gestations (see Figure 4). Binomial regression models confirmed this pattern of variation between PCTs (< 24 weeks, 0.29; 24–27 weeks, 0.17; 28–32 weeks, 0.11).

The proportion of births recorded as live showed an increase with increasing gestational age from 52.6% at < 24 weeks’ gestation to 73.9% at 24–27 weeks’ gestation and 91.9% at 28–32 weeks’ gestation, as would be expected as survival improves with increasing gestation (Figure 5). However, the variation at < 24 weeks’ gestation was wide with a 90% percentile range of 26.3% to 79.5%; that is, in the lower 5% of PCTs, around one-quarter of births were registered as live-born compared with over three-quarters in the upper 5% of PCTs. This variation reduced considerably with increasing gestation as seen in Figure 5 (90% percentile of range 24–27 weeks, 64.4% to 82.9%; 28–32 weeks, 88.5% to 95.4%). Binomial regression models again confirmed this pattern of decreasing between-PCT variation in the percentage of births registered as live-born with increasing gestation. Adjusting for PCT population characteristics resulted in little change in the between-PCT variation for all gestational age groups.

FIGURE 5.

Median, IQR and 90% percentile range for the percentage of preterm births registered as live-born by gestational age for PCTs (log-scale). Reproduced from Smith et al. 62 © 2013, BMJ Publishing Group Ltd and the Royal College of Paediatrics and Child Health, under the terms of the Creative Commons Attribution Non-Commercial Licence (CC BY-NC 3.0), which permits use, distribution and reproduction in any medium, provided the original work is properly cited, the use is non-commercial and is otherwise in compliance with the licence.

Impact of extremely preterm births on infant mortality rates

The overall infant mortality rate was 4.78 per 1000 live births (n = 12,083). This varied widely between PCTs from 2.34 to 8.93 per 1000 births. Despite making up only 1% of births, 19.5% of infant deaths arose from births < 24 weeks’ gestation. Between PCTs this varied considerably (90% percentile range 6.7% to 31.9%).

First, the impact of preterm births < 24 weeks on the actual rate of infant mortality was assessed. Comparing the overall infant mortality rate with the infant mortality rate excluding preterm births < 24 weeks’ gestation (Figure 6) showed a decrease in infant mortality of more than 1 death per 1000 births after exclusion of preterm births < 24 weeks’ gestation in 60 out of the 147 PCTs. Of these, 83% registered more than half of preterm births < 24 weeks’ gestation as live-born. Conversely, of those 87 PCTs which had a smaller decrease in infant mortality (≤ 1 death per 1000 births), only 43% had registered over half of preterm births < 24 weeks’ gestation as live-born.

FIGURE 6.

Overall infant mortality rate vs. infant mortality rate excluding infants of < 24 weeks’ gestation for PCT by percentage of births of < 24 weeks’ gestation registered as live-born. Lines indicate an absolute difference of one death and two deaths per 1000 live births. Reproduced from Smith et al. 62 © 2013, BMJ Publishing Group Ltd and the Royal College of Paediatrics and Child Health, under the terms of the Creative Commons Attribution Non-Commercial Licence (CC BY-NC 3.0), which permits use, distribution and reproduction in any medium, provided the original work is properly cited, the use is non-commercial and is otherwise in compliance with the licence.

Second, as infant mortality rates are often used to rank PCTs’ performance, the impact of births at < 24 weeks’ gestation on rankings of infant mortality rates was assessed. After excluding preterm births at < 24 weeks’ gestation (Figure 7), PCTs that registered fewer than half of preterm births of < 24 weeks’ gestation as live-born fell a median of 12 places whereas those that registered at least half of births as live-born improved by a median of 4 places.

FIGURE 7.

Rank of overall infant mortality rate vs. rank of infant mortality excluding infants born at < 24 weeks’ gestation for PCTs by percentage of births at < 24 weeks’ gestation registered as live-born. Lines refer to an absolute increase and decrease of 20 places in rankings, 1 = lowest mortality rate, 147 = highest mortality rate. Reproduced from Smith et al. 62 © 2013, BMJ Publishing Group Ltd and the Royal College of Paediatrics and Child Health, under the terms of the Creative Commons Attribution Non-Commercial Licence (CC BY-NC 3.0), which permits use, distribution and reproduction in any medium, provided the original work is properly cited, the use is non-commercial and is otherwise in compliance with the licence.

Key findings

• Wide between-PCT variation existed in extremely preterm birth (< 24 weeks’ gestation) rates.

• Consequently, the percentage of infant deaths arising from these births varied.

• Exclusion of births < 24 weeks’ gestation led to significant changes in infant mortality rankings of PCTs.

• Infant death rates in PCTs in England are influenced by variation in the registration of births for which viability is uncertain.

Limitations and strengths

The variation in the rate of live births at < 24 weeks’ gestation and the proportion of births registered as live-born may reflect true underlying differences in the extremely preterm birth rate and fetal death rate between PCTs. However, after adjusting for underlying aetiological differences such as socioeconomic deprivation, ethnicity and maternal age, wide variation between PCTs persisted in this early gestational age group. Furthermore, it was not possible to attribute such between-PCT variation to any reported approach to care or improved overall outcomes in particular hospitals around England. These are the first analyses exploring variation in the registration of < 24-week gestation births and deaths and their impact on infant mortality in the UK. By utilising national data collected in a validated, systematic way, it provides a large reliable data set to assess variation between regions across England.

Implications for policy and practice

This work demonstrates that registration differences impact on within-country comparisons, as well as those between countries. In order to make direct comparisons in mortality, be it at international, regional or unit level, detailed validation and standardisation is essential to ensure that ‘like-with-like’ comparisons are being made, distinguishing ‘real’ variation from ‘artefacts’ that arise from reporting and registration differences. A standardised approach to the collection, calculation and presentation of mortality rates would reduce artefactual differences.

Despite improvements in care, variations in outcomes persist. Differences in management strategies and decisions regarding viability by a range of health-care professionals present at delivery are likely to be responsible for the wide variations in delivery outcome of extremely preterm infants. Such variation in management is likely to reflect not policy at the commissioner level but rather the approach of individual hospitals and their clinical teams. In the UK, the World Health Organization (WHO) definition of a live birth based on signs of life irrespective of gestational age is generally accepted for registration purposes. However, there are practical difficulties in interpreting true signs of life as well as subjective differences in judgements about the best outcome for parents. In the USA, where the WHO definition is accepted, research shows that nearly one-third of physicians include gestational age in their personal criteria for defining a live birth with definitions ‘open to subjective interpretation’.

Decisions over registration greatly affect parents in the UK, as late fetal deaths (< 24 weeks of gestation) are not officially registered. Following the death of a live-born baby of any gestation or stillbirth after 24 weeks’ gestation, all parents are entitled to a birth and death certificate, to maternity/paternity leave and pay, to child benefit for 8 weeks and to means-tested assistance with funeral expenses, and mothers receive free prescriptions and dental care for 1 year; in addition, the case will be referred to a coroner for investigation. In comparison, following a late fetal loss, parents will be eligible for none of these benefits. This is likely to have financial impacts since extremely preterm birth is considerably higher in socioeconomically deprived areas. There may also be emotional impacts related to these babies not being formally registered as births and deaths. This issue was highlighted by a House of Commons debate exploring the introduction of birth certificates for stillbirths (from 24 weeks’ gestation), lobbied for by parents. 63

It would seem sensible that, when comparing localities, infant and neonatal mortality rates are calculated excluding births less than 24 weeks’ gestation to ensure that ‘like-with-like’ comparisons are being made. However, this would mean the exclusion of up to one-third of infant deaths in some areas and ultimately it would seem wise to work towards reducing the variation in the registration of extremely preterm infants in England.

Since 2013, the national perinatal mortality surveillance system in the UK has aimed to collect data on all deaths (including babies born dead) at 22 and 23 weeks’ gestation. This step should not only improve our ability to investigate within-country differences in early-life mortality rates but also allow for direct international comparisons, particularly across Europe, where the vast majority of countries use the WHO definition of birth and register all pregnancy outcomes from 22 weeks’ gestational age. Such an approach would seem sensible for all developed countries.

Recommendations for future research

• Research into understanding how this variation arises and standardising the implementation of guidelines regarding viability of infants of < 24 weeks’ gestational age is needed in order to tackle this variation and reduce its impact on both mortality rate comparisons and inequalities in parents’ access to benefits.

• Detailed understanding is needed of the potential emotional impact on parents of certifying babies at 22 and 23 weeks’ gestation as neonatal deaths or fetal losses so that guidelines can be introduced to ensure that all parents are treated in the most appropriate way at this difficult time.

Maternal infection

Maternal systemic, genital and urinary tract infection and inflammation have been causally linked to preterm birth. 74,75 The risk of infective complications increases with pre-labour premature rupture of the membranes. Maternal chorioamnionitis has been implicated as a factor contributing to the poor neurodevelopmental76 and neurological77 outcomes in affected neonates. However, the proportion of pregnancies complicated by infection at later preterm gestations is unknown.

Socioeconomic deprivation

Material or social deprivation refers to a variety of conditions experienced by people who lack certain resources in relation to others in the community. 18 Material, social and financial deprivation, educational disadvantage and poor access to health-care services are often closely linked. Areas of interest with respect to preterm birth have been the impact of living conditions and working conditions, including physical aspects of work. 78 These and other experiences that result in stress are thought to contribute to preterm delivery in the extremely preterm group, with rates of preterm birth in the most deprived women rising to almost twice those of the least deprived. 5 Poor dental health is a common problem in socioeconomically deprived populations. 79 Maternal oral health has been shown to be important for overall health and periodontal disease during pregnancy has been associated with adverse pregnancy outcomes. 80,81 Much of the research linking socioeconomic status with preterm birth has been carried out with respect to extreme prematurity. 5,82 The influence of socioeconomic deprivation in women delivering at later preterm gestations remains unexplored.

Lifestyle

Many aspects of lifestyle are closely linked to levels of material and financial deprivation. Lack of access to financial resources can limit the choice of lifestyle and ability to access education and health care, making it extremely difficult to separate material and behavioural issues. A number of elements of maternal lifestyle during pregnancy may have either positive or adverse effects on the fetal growth or development in utero and may influence rates of congenital anomaly. Although results of studies are conflicting, stress during pregnancy and, in particular, during the early stages of pregnancy is also thought to be an important risk factor for adverse pregnancy outcomes. 83 Such stress may be related either to pregnancy-specific anxieties84 or to life events. Stress responses are frequently linked to financial poverty as people living in poverty have fewer resources, both financial and psychological, with which to respond to major life events.

Other key issues potentially associated with preterm birth and adverse fetal development reflect lifestyle choices such as risk-taking behaviours including smoking,85 excessive alcohol intake,86,87 recreational drug use88 and dietary intake. Women are encouraged to eat a ‘healthy’ diet during pregnancy. There is evidence from research to suggest that a cholesterol-lowering, or Mediterranean-style, diet reduces the risk of preterm delivery. 89–92 Women at lower risk were those who ate fish at least twice a week, used olive or rapeseed oil, ate five or more portions of fruit and vegetables a day, ate meat at most twice a week and drank no more than two cups of coffee a day. 90 However, others have shown no such association. Maternal caffeine intake during pregnancy has, in itself, been associated with an increased risk of spontaneous abortion. 93,94 The UK Foods Standards Agency recently reassessed its recommendations, and suggested that caffeine intake should be no more than 200 mg per day (approximately two cups of coffee or tea) because of an increased risk of fetal growth restriction (FGR). 95 The impact of these factors of maternal lifestyle has not been assessed with respect to LMPT birth.

Antenatal corticosteroids

Research on the use of antenatal corticosteroids (ANSs) to prevent lung disease of prematurity has focused on very preterm deliveries and these drugs are recommended for use only up to 34 weeks’ gestation. Despite some evidence to suggest neonatal benefit even in term deliveries,96 and guidance suggesting that they may be considered in threatened preterm labour beyond 34 weeks, the use of ANSs in late preterm deliveries is rare.

Tocolytic agents

The use of tocolytic agents is closely linked to attempts to prolong pregnancy in order to allow administration of steroids at earlier gestations. The use of these medications beyond 34 weeks’ gestation is not recommended. 97

Pre-labour premature rupture of membranes

The Royal College of Obstetricians and Gynaecologists’ recommendations state that delivery should be considered when rupture of the membranes occurs at 34 weeks’ gestation or later. 98 This recommendation was based on the increased risk of development of chorioamnionitis with expectant management. Only a limited number of studies have addressed this question, and more research is required to determine the optimal time of delivery.

Maternal complications

Pregnancy-induced hypertension and pre-eclampsia are the most common complications of pregnancy. Delivery at 34 weeks’ gestation or later has been recommended in severe cases,99 but data suggest that women with mild disease are also more likely to be delivered in the late preterm period. 100

Previous delivery by caesarean section

Rates of caesarean section (CS) have been increasing in both the UK and the USA. Vaginal delivery after previous CS carries additional risks for both mother and baby101,102 and concerns about safety have led to a decline in both the number of obstetricians offering a trial of labour and the number of mothers taking up this option. Most obstetricians delivered such women at 36–37 weeks’ gestation, but no clear guidelines existed indicating the optimal time for delivery.

Intrauterine growth restriction

An estimated fetal weight below the 10th percentile for gestational age is associated with some chronic maternal medical conditions, placental disorders and fetal conditions such as chromosomal abnormalities and viral infections. Timing of delivery depends on both maternal and fetal condition; expectant management with close fetal monitoring is indicated and pregnancy is usually prolonged until at least 34 weeks unless monitoring indicates significant fetal compromise. However, the optimal timing for delivery for those who find continuous fetal monitoring reassuring has not been determined.

Oligohydramnios

Lower than expected volumes of amniotic fluid can occur because of a number of different pathologies, placing the fetus at increased risk. Close monitoring of the pregnancy is indicated, but the optimal management in the late preterm period has not been defined.

Multiple gestations

The incidence of twin and higher-order multiple pregnancies has increased, partly because of the increasing use of treatments for infertility. Multiple births are more likely to occur in the late preterm period because of higher rates of spontaneous preterm labour and complications of pregnancy. 104 According to data from the USA, the average gestational age at delivery of twins is 35.3 weeks and of triplets is 32.2 weeks. 105 There is evidence of increased risk of fetal death with increasing gestation in both monochorionic and dichorionic pregnancies. 106–108 At the time that this work was planned, the role of elective delivery in the LMPT period in multiple pregnancies was yet to be clarified.

Respiratory disease

As infants born at < 32 weeks’ gestation are at the highest risk of respiratory disease associated with prematurity, data reporting on rates and types of respiratory complications in more mature preterm infants were relatively sparse and there were no UK studies addressing this. Rubaltelli et al. 109 reported that the rate of any respiratory distress among infants born at 33–34 weeks’ gestation in Italian neonatal units (NNUs) was 20.6% and among those born at 35–36 weeks was 7.3%, compared with 0.6% in babies born at > 37 weeks. Wang et al. 110 compared outcomes of infants born at 35–36 weeks’ gestation between 1997 and 2000. In this study, 28.9% of late preterm infants experienced respiratory distress, compared with 5.3% of term-born infants. Similar results were reported by Escobar et al. 116 in 2006, with rates of 22.1%, 8.3% and 2.9% at 34, 35 and 36 weeks’ gestational age, respectively, in a retrospective review of the Kaiser Permanente cohort of babies in the USA. In this study, infants born at 35 weeks’ gestation were nine times more likely to be ventilated than those born at 38–40 weeks; infants born at 36 weeks were five times more likely to require ventilation. A retrospective single-centre study reported ventilation rates of 30%, 33% and 23%, at 34, 35 and 36 weeks’ gestational age, respectively. 117

Hypothermia

Temperature control and prevention of hypothermia form a part of routine newborn management. Wang et al. 110 found that 10% of late preterm infants required active management for hypothermia compared with term-born infants. Laptook and Jackson111 demonstrated increased susceptibility to cold stress in infants born at 34–37 weeks’ gestation. Temperature instability may be the presenting feature of other illness such as sepsis, but the frequency of, and factors associated with, hypothermia in LMPT infants in the UK has not been determined.

Hypoglycaemia

Hypoglycaemia is commonly encountered in neonatal care. The level of blood glucose that constitutes significant hypoglycaemia likely to impact on outcome is controversial for both term-born and preterm infants. 118,119 However, prevention of hypoglycaemia and its potential adverse neurodevelopmental effects is recognised as an important part of neonatal management. 120 Studies addressing hypoglycaemia in moderately and late preterm infants were few, but, in a retrospective review of records of more than 7000 infants, Wang et al. 110 observed hypoglycaemia three times as often in infants born at 35–36 weeks’ gestation as in term-born infants.

Feeding difficulties

Feeding represents one of the most basic skills that an infant must acquire in the neonatal period before discharge from hospital. Difficulties with feeding may influence length of hospital stay, early weight gain, glycaemic control and risk of infection. Sleepiness, difficulty with latching onto the breast and poor co-ordination of sucking and swallowing are all common causes of delay in establishing normal oral feeding in LMPT infants. 113 Wang et al. 110 found that 27% of late preterm infants, compared with 5% of term-born infants, required intravenous (i.v.) fluid administration. A substantial number, particularly those unable to feed enterally because of other conditions, may require a period of parenteral nutrition (PN). Although this may be necessary to meet nutritional requirements, it carries associated risks related to the PN itself and increased incidence of catheter-related infection with central venous line placement. As a result of the challenges of establishing breastfeeding in preterm infants, it is possible that mothers of such infants may be more likely to abandon breastfeeding than those delivering at term, although it is known that breast milk is the most appropriate nutrition for a newborn baby. Methods of feeding in LMPT babies at discharge from hospital and post discharge have not specifically been studied.

Jaundice

Prematurity places infants at an increased risk of hyperbilirubinaemia and the associated bilirubin encephalopathy (kernicterus) that can result from extremely high levels of serum bilirubin (SBR). It had been suggested that a trend towards earlier postnatal discharge of healthy newborn babies has led to an increased risk of severe jaundice secondary to postdischarge feeding difficulties and reduced monitoring in the first few days of life. In a study of hospital readmissions in the neonatal period,121 jaundice was one of the most common reasons for readmission; late preterm infants required readmission more frequently than did term-born infants. Bhutani et al. 112 studied cases reported to the Pilot Kernicterus Registry in the USA and found that late preterm babies were overrepresented in this registry, with breastfeeding being a significant risk factor.

Mortality

Worryingly, the observed risk of increased morbidity, as discussed above, seemed also to be associated with higher mortality rates in this group of infants than in those born at term. In 2000, Kramer122 assessed the contribution of ‘mild and moderate preterm birth’ to infant mortality and concluded that these infants had a high RR of death during infancy. Published infant mortality data for 2002 in the USA105 show that the mortality rate among babies born mild or moderately preterm is three times that of term-born infants. This was confirmed by Tomashek et al.,123 who found that early and late neonatal mortality were, respectively, six and three times higher, and infant mortality three times higher, among preterm infants than that of term-born infants. A substantial number of early deaths in both groups were related to congenital anomalies, but differences persisted even after exclusion of infants with anomalies.

Eligibility

Our original plan was for recruitment to take place over a 1-year period between September 2009 and August 2010. All mothers who were resident in a geographically defined area of Leicestershire and Nottinghamshire and delivered babies at 32⁺⁰–36⁺⁶ weeks’ gestation were eligible to participate in the study with their babies. There were no exclusion criteria. We also identified a group of babies born at term (≥ 37 weeks’ gestation) during the same time period and geographical region that would be recruited as a control group.

However, towards the end of the proposed 12-month recruitment period it became apparent that the target sample size of 800 singleton births in each group was unlikely to be met. An extension of the recruitment period was agreed with the ethics committee and recruitment continued until 31 December 2010.

Participating centres

We used a population-based design for the study to allow evaluation of the risks and outcomes within a regional population. The geographical area covered by the study comprised rural and urban populations, as well as a diverse range of ethnicities and a wide socioeconomic spectrum. This avoided the bias introduced in hospital-based studies, which often leads to recruitment of a study population with artificially elevated risk compared with the population as a whole. Within the chosen geographical area around Leicester and Nottingham, there were four hospitals and one community birth centre that provided maternity services to the population. Two hospitals in Nottingham provided all levels of neonatal care, from intensive care to low-dependency or ‘special’ care, with a total of approximately 10,000 deliveries annually. In Leicester, where there are also approximately 10,000 deliveries in total each year, one hospital provided both neonatal intensive and low-dependency care while the other provided low-dependency care only. In each of the two cities a single team of consultant neonatologists working across the two hospital sites provided medical management of neonatal care. Therefore, although infants were cared for in four separate geographical locations, for the purposes of delivery of neonatal care, these could be regarded as two neonatal centres. Both centres were regional referral units for women with complicated pregnancies and/or neonates with postnatal problems requiring higher-level care. The community birth centre within the region catered for around 250 deliveries per year. The deliveries here that were included were to low-risk women in whom there had been no complications of pregnancy. In addition, a proportion of the lowest-risk pregnancies were represented by deliveries at home.

Definition of geographical boundaries for recruitment

For the purposes of this study a geographical area covered by these hospitals was defined that maximised the number of participants in the study but minimised the number of residents who would deliver in other units not participating in the study (Figure 8). The region covered by the study was well known to the researchers and has previously been the setting for a number of successful population-based research studies.

FIGURE 8.

Map of area of residence for eligibility for recruitment into the Late And Moderately preterm Birth Study.

Data were obtained from the ONS for the hospital of birth for babies born in 2007 to mothers living in the postcode areas NG (Nottingham) and LE (Leicester). Based on these data, for each postcode sector (e.g. LE1 6) the percentage of births delivering at the five maternity units participating in the study was calculated. A cut-off point was then determined for the minimum percentage of births to occur in the five participating maternity units that would give at least 85% of all births in the total included area occurring in these units, that is to say that a cut-off point of 85% would mean that at least 85% of all births in each included postcode sector would deliver at one of the five participating delivery sites. The percentage of all births in the included area delivering to those five sites was actually much higher than this cut-off point as nearly all women living in close proximity to a participating hospital deliver there.

Based on these estimates, and the aim of including around 2000 singleton infants in the study, a cut-off point of 85% was decided on. This equated to 96.8% of births for the included postcode sectors actually occurring in the five participating hospitals, increasing to 99.2% when home births were included: that is, only 0.8% of births occurred in maternity units other than the five participating in the study.

Sample size

The sample size was based on detecting differences in rates of cognitive impairment and early rehospitalisation between the study and control groups of infants from singleton pregnancies. School-age follow-up of children in the EPICure study used a group of classroom controls,136 among whom the rate of cognitive impairment was 2%. There are no data for rates of cognitive impairment in moderately preterm infants in the UK. It was estimated that a rate of 5% cognitive impairment in our study group would be clinically relevant and justify targeted intervention in children’s health and education services. Oddie et al. 115 found that the rate of rehospitalisation in the first month of life in babies born at 35–37 weeks’ gestation was 6%, compared with 3% among term-born infants. A sample size estimation based on these figures predicted that recruitment of 800 singleton infants per group would allow detection of this difference in cognitive function with 90% power and in rehospitalisation with 80% power at the 5% significance level. To allow for potential non-consent and loss to follow-up, we decided to try to recruit 1000 singleton LMPT infants and 1000 singleton term-born infants.

In 2007, there were approximately 17,000 singleton births in the selected study area in Leicestershire and Nottinghamshire. Assuming that the proportion of births at 32⁺⁰–36⁺⁶ weeks’ gestation is 6–7%, it was anticipated there would be around 1000–1200 LMPT singleton births in 1 year. A minimum recruitment rate of 80% would give the required sample size.

Multiple births

To ensure inclusion into the study of sufficient term-born infants from multiple pregnancies it was decided to include all births at 32⁺⁰ weeks’ gestation or greater resulting from multiple pregnancies.

Selection of singleton control infants

Potential term-born control babies were identified using a pseudo-random sample of births obtained from ONS births data for 2007. For a random sample of 1000 births for the study area, ONS supplied information on the day of the week of birth, month of birth and place of delivery. A date and time of birth were then assigned to each of the births in this sample using known dates and time for births in Leicestershire obtained from the Child Health Registers for Leicester and Leicestershire and Rutland PCTs. These data were adjusted for day of the week to ensure an appropriate sampling of babies born at the weekend, as non-urgent induction of labour and CSs are more likely to be undertaken on a weekday.

This method overcomes the bias that may occur if more naive sampling methods are used, for example selecting the next term birth to occur at the maternity unit after a LMPT birth. 137,138 Such sampling methods would be likely to result in oversampling of high-risk term-born infants as these tend to be delivered in the sample places and at the same times as LMPT births; for example, no births would have been sampled from the midwifery-led unit.

Twelve births in the sample were excluded as they occurred at centres other than the five maternity units included in the Late And Moderately preterm Birth Study (LAMBS) or at home. Therefore, the 988 controls were used plus an additional 12 sampled from the 988 to make a pseudo-sample of 1000 controls. The list of potential control singleton births was then passed to the study midwives.

The birth to be included in the study was the first birth recorded at or after the time stated on the sample list. If there was uncertainty over which birth should be included because more than one birth occurred at the same time, then the NHS numbers of the mothers were examined and the infant included was the one born to the mother with the lowest final digit of NHS number. For example, if there were two eligible births to mothers with NHS numbers 2365892145 and 6523674352, then the mother with the latter number would be included. If the last digits were identical, then the next digit to the left would be used until there was a difference.

Timing of recruitment

We aimed to recruit both study and control infants during their initial hospital stay following delivery. The length of this stay varied considerably depending on the gestational age and clinical condition of the baby at birth and the condition of the mother. In all centres, it was common for term-born infants to be discharged home at 6–12 hours following delivery if there was no medical indication for either mother or baby to remain in hospital. Infants at the lower end of the gestational range required admission to a NNU and longer hospital stays.

At the time of recruitment, the mothers were either on the delivery suite or on a postnatal ward depending on the anticipated length of hospital stay. If any mother was too sick to give consent in the period following delivery, this process was delayed until the obstetrician caring for her was happy that she was in a fit condition to be approached for consent.

Identification of eligible mothers and infants

Clinical staff on the delivery suite of each participating hospital identified infants born at 32–36 weeks’ gestation and informed a member of the research team. The research midwives referred each day to the list of dates and times on which to base recruitment of the control infants in order to identify the next infant born at ≥ 37 weeks’ gestation eligible to be recruited to the control group.

Consent

The Derbyshire NHS Research Ethics Committee (REC) approved the study design and methods. A team of six midwives who worked collaboratively across the centres carried out recruitment to the study. A research midwife attempted to approach all eligible mothers within the first 24 hours after the birth to provide them with an information sheet and obtain written informed consent for their participation. Mothers who were discharged before this initial contact could be made to present the study were contacted by a member of the clinical staff and, when possible, a home visit by a research midwife was arranged. The midwife obtained informed consent at this visit.

Written consent was obtained from parents of study and control infants for the following:

• interview with a research midwife and completion of a maternal lifestyle questionnaire

• collection, retention and analysis of pregnancy, perinatal and neonatal data

• permission to send follow-up questionnaires at 6, 12, 18 and 24 months following the child’s birth.

Six months

The questionnaire at 6 months comprised items to assess their child’s use of health-care services, child-care costs and mothers’ general health (see Appendix 2). Questionnaires were posted to parents within 1 week of the date on which their child attained 6 months of age.

Twelve months

At 12 months of age, the same questionnaire was sent to obtain further longitudinal data relating to the use of health-care and social care services, with additional questions to determine whether or not the child had achieved key developmental milestones (see Appendix 3). Similar to the 6-month follow-up, questionnaires were mailed to parents within 1 week of the date on which their child reached 12 months of age and a second copy of the questionnaire was mailed 2–3 weeks later if a completed questionnaire had not been returned. However, at this follow-up, data were collected using a hierarchical mixed-mode approach. Parents who had not responded to two postal questionnaires were contacted by telephone, e-mail or text message to confirm that they had received the questionnaire and to offer the option to complete it as a telephone interview. Non-responders to postal questionnaires were therefore routinely offered the option of a telephone interview when contact could be made. At this time, parents were also given the opportunity to complete the 6-month questionnaire during the same telephone interview if this had not previously been returned. Up to three attempts were made to contact each family by telephone, e-mail or text message.

Two years

At 2 years, the child’s age at follow-up was corrected for prematurity. Data relating to health and neurodevelopmental outcomes were again collected using a hierarchical mixed-mode approach. Parents were mailed a questionnaire to complete 7–10 days prior to the day on which their child reached 2 years of age, corrected for prematurity. If the questionnaire was not received 10 days post 2Y-CA, parents were contacted by telephone, e-mail or text message to remind them to complete the questionnaire and to offer the option to complete it over the telephone at a time convenient to them. In addition, at this follow-up stage, a system for online completion of questionnaires was established to improve response rates. An identical format of the questionnaire was programmed online using SurveyMonkey^® (SurveyMonkey, Palo Alto, CA, USA) and, if this option was selected, parents were e-mailed a web link to the questionnaire to be completed at their convenience. If parents could not be contacted by telephone, text or e-mail and the questionnaire had not been received 2 weeks post 2Y-CA, a second postal questionnaire was sent. A final attempt was made 1–2 weeks later to contact parents by telephone, text or e-mail to offer a telephone interview or online completion. A maximum of three attempts were made to contact each family by telephone, e-mail or text message.

Live births

Of all mothers of live-born infants, 210 (15.6%) in the LMPT group and 328 (20.7%) in the term group either declined participation or were unable to be contacted by the clinical staff following discharge from hospital. A reason for non-participation was given by 118 (56%) of the LMPT group and by 191 (58%) of the term group. The most common reason in both groups was a lack of interest in the research, which was the reason given by 45% of LMPT mothers and 36% of mothers delivering at term.

Stillbirths

As might be expected, mothers of stillborn infants were less likely to agree to participate in the study: 55% of mothers of preterm stillborn infants declined to participate, as did 57% of mothers of stillborn infants born at term. Many of these mothers did not wish to hear about or engage in research after this distressing experience, although most indicated that they were willing to be contacted at a later date. Therefore, in most cases, a period of time was allowed to elapse before the research midwives attempted to approach these women. For those who wished to speak to a research midwife shortly after their delivery, every effort was made to offer support and, when necessary, practical advice at this time. Mothers who had given permission to contact them after discharge were telephoned at least 2 weeks after the delivery and were not approached again if they declined at this time.

Deaths

There were six deaths in the postnatal period prior to discharge home. Of these, five were in LMPT group and one was in the term-born group. There were two further deaths following discharge from hospital: one term-born infant died before the 6-month follow-up and a further preterm infant died between 6 months and 1 year.

Obstetric data

Data collection focused on exploration of the impact of key sociodemographic factors on the incidence of LMPT birth and on outcomes of newborn infants. We studied the distribution of pre-pregnancy health factors, including body mass index (BMI) at the time of booking for antenatal care, pre-existing chronic health problems, diabetes mellitus and hypertension. We also looked at previous obstetric history, including whether or not this was the first pregnancy and if the mother had experienced a previous preterm birth.

Maternal interview data

The maternal interview (see Appendix 1) involved collection of sociodemographic data from all mothers. This included information on mothers’ age, ethnicity, civil status, occupation during pregnancy, working patterns and conditions, education, socioeconomic status and lifestyle.

The highest maternal educational qualification was classified into four categories: (1) degree or higher; (2) A levels; (3) GCSE grades A–C; and (4) GCSE grades D–G or none of the above. Occupations were classified into eight analytic classes which were collapsed into managerial or professional (classes I and II); intermediate occupations (classes III–V); routine and semiroutine occupations (classes VI and VII); and unemployed, retired and students (class VIII).

Mothers’ occupational status was classified using the NS-SEC (ONS143), with mothers who reported that they looked after their family full time classified separately. Other economic information obtained comprised receipt of means-tested benefits, access to a car, home ownership, self-reported financial status and information about day-to-day financial hardship. These were combined into the composite working condition risk score described above.

Data collected relating to mothers’ health comprised self-reported chronic health conditions. Socioeconomic factors included maternal lifestyle and risk-taking behaviour including the use of tobacco, alcohol or recreational drugs.

Delivery and neonatal data

Data collection included mode of delivery, gestation, birthweight, gender, FGR and the occurrence of major chromosomal or structural congenital anomalies. Mode of delivery included spontaneous vaginal delivery; assisted vaginal delivery, which included all instrumental deliveries with either cephalic or breech presentation; and CSs and whether or not this was preceded by labour. Deliveries in which vaginal instrumental delivery was attempted, but the baby was ultimately delivered by a CS were classified as CS. FGR was defined using a customised weight percentile of calculator [Gestation Related Optimal Weight (GROW) Customised Centile Calculators, v5.16; Gestation Network LTD, Birmingham, UK].

There is no universally accepted definition of a major congenital anomaly. For the purposes of this study, major structural and chromosomal anomalies were classified by clinical consensus and with reference to the classifications used by the EMSYCAR.

Maternal age and ethnicity

The distribution of ages of mothers at the time of delivery was similar between groups, with the majority (57.3%) of mothers of singleton infants being between 25 and 35 years of age. Compared with the reference group (aged 25–29 years), much smaller proportions of teenage mothers and older mothers (> 35 years) delivered at LMPT gestations, but these differences did not reach statistical significance. No particular age band was associated with increased odds of delivering a singleton baby at LMPT gestation.

The odds of LMPT delivery were significantly increased in mothers of multiple births delivering at < 20 years of age, but the numbers of these women were extremely small, with only one in the term group.

The majority of the mothers of infants in both LMPT (71.6%) and term (77.5%) groups were of white British origin. Women of Asian and Asian British origin formed the next largest group, comprising 18.5% of mothers delivering LMPT and 14.5% of those delivering at ≥ 37 weeks’ gestation. The odds of these women delivering LMPT babies were significantly increased (OR 1.38, 95% CI 1.08 to 1.77; p = 0.01) compared with those of white origin. Odds of LMPT delivery were also increased in women of mixed racial origin, although the number of these women in the cohort were small (1.9% and 3.3% for LMPT and term births, respectively).

Maternal socioeconomic characteristics

Living as part of a couple was associated with significantly lower odds of delivering a singleton LMPT infant than was being a single parent (OR 0.69, 95% CI 0.54 to 0.88; p < 0.01). The number of mothers whose relationships were consanguineous was small, but similar in both groups of singleton births. Similar analysis was not possible for multiple pregnancies as there were only five mothers who were in consanguineous relationships.

With respect to maternal education, comparisons between levels of education were against the reference group comprising those women who had attained the highest level of educational qualification (degree or equivalent). When compared with this group, all those with lower levels of education had significantly increased odds of delivering at LMPT gestation. There was an inverse relationship between mothers’ highest education level and odds of LMPT birth, with the lowest educational group being at highest risk. A similar pattern was seen for mothers having multiple births, with those in the lowest educational group having significantly increased odds of delivering LMPT (OR 2.64, 95% CI 1.21 to 5.80; p = 0.015).

For occupational status during pregnancy, comparisons were made against the reference group of managerial and professional occupations. Compared with this group, all other groups had increased odds of delivering LMPT infants, and these differences were statistically significant in the case of those who had never worked or were unemployed (OR 1.66, 95% CI 1.23 to 2.24; p = 0.001), and in women who reported that they were looking after their family on a full-time basis (OR 1.40, 95% CI 1.08 to 1.81; p = 0.01). Within the group of multiple births, those who had never worked or were unemployed had increased odds of LMPT delivery (OR 2.94, 95% CI 1.10 to 7.89; p = 0.032)

Using the derived SES Index, as described previously, the odds of delivering LMPT were increased within the high-risk socioeconomic group compared with the low-risk group (OR 1.48, 95% CI 1.19 to 1.84; p < 0.001). Although the pattern was similar for multiple births, the difference between the high- and low-risk groups was of more marginal statistical significance (p = 0.044).

Maternal lifestyle

The majority of women in both study groups had a BMI within the normal range. Compared with those with booking weight within the normal range, however, those women whose weight fell within the obese range were less likely to deliver a LMPT baby (OR 0.68, 95% CI 0.52 to 0.89; p = 0.004). This was not the case for multiple births.

Smoking during pregnancy was associated with increased odds of delivery at LMPT gestation (OR 1.44, 95% CI 1.16 to 1.78; p = 0.001) in singleton but not in multiple pregnancies. The number of women who reported the use of recreational drugs was very small, but in this study this did not carry any increased risk of LMPT delivery. Alcohol use in pregnancy was also not associated with increased odds of LMPT delivery in either singleton or multiple pregnancies.

Previous preterm birth and pre-pregnancy health

Nineteen per cent of mothers of LMPT singleton infants had experienced a previous preterm delivery, compared with 5.2% of mothers delivering at term (OR 4.26, 95% CI 3.07 to 5.91). In the small number of multiple pregnancies in which the mother had previously had a preterm birth, there was a trend towards increased odds, but this did not reach statistical significance, and CIs were wide (OR 3.79, 95% CI 0.98 to 14.66). Women for whom this was the first pregnancy, regardless of whether this was a singleton or a multiple birth, were not at increased risk of delivering at LMPT gestations.

A large number of women delivering singletons reported having had chronic health problems during pregnancy (39.0% of LMPT singleton births and 31.2% of term births). Those reporting chronic health problems had increased odds of LMPT birth (OR 1.41, 95% CI 1.17 to 1.70; p < 0.001) compared with those who did not. Among mothers having multiple births, the proportion was similar for LMPT (33.0%) and term (32.6%) groups.

Specifically, pre-pregnancy diabetes and pre-pregnancy hypertension were associated with LMPT delivery in singleton pregnancies [OR 4.36, 95% CI 2.00 to 9.52 (p < 0.01), and OR 2.56, 95% CI 1.25 to 5.20, respectively]. The numbers of women having multiple pregnancies who had suffered from either of these conditions prior to becoming pregnant were too small to be suitable for analysis.

The characteristics of live-born singleton babies, including those with congenital anomalies, are shown in Table 14. The birth characteristics of 923 LMPT and 980 term-born singleton babies are compared. There were no statistically significant differences in the number of male babies or in the proportion of first-born infants. There were twice as many congenital anomalies in the LMPT group, but this difference did not reach statistical significance. LMPT babies were significantly more likely to have been affected by intrauterine growth restriction (OR 1.95, 95% CI 1.51 to 2.51; p = 0.001). Compared with spontaneous vaginal deliveries, LMPT babies were more likely to be born following pre-labour CSs (OR 2.17, 95% CI 1.67 to 2.81; p < 0.001) and significantly less likely to have an assisted vaginal delivery (OR 0.71, 95% CI 0.54 to 0.93; p = 0.014).

There were 207 LMPT and 275 live term-born infants from multiple pregnancies and their birth characteristics are shown in Table 15. As for singletons, there were no statistically significant differences in gender or in the proportion of first-born infants. Although a significant difference is seen for congenital anomalies, with LMPT having increased odds of being affected (OR 9.59, 95% CI 1.17 to 78.56; p = 0.035), this must be interpreted with caution because of the small numbers and very wide CIs. In contrast to singleton births, multiples were more likely to be delivered by CS after the onset of labour but, similar to singletons, the odds of assisted vaginal delivery in the LMPT group were reduced compared with spontaneous vaginal delivery (OR 0.40, 95% CI 0.21 to 0.75; p = 0.004).

Stillbirths

There were 16 recruited stillbirths in the LMPT group and three in the term group. The very small numbers of stillbirths occurring within the study period, and even smaller number of recruits in this group, means that meaningful analysis of the data collected for these mothers and their babies was not possible. However, the characteristics of the participating mothers and their stillborn infants are detailed in Table 16.

There were 15 singleton stillbirths in the LMPT group and two in the term group. There was one stillborn infant in each group that was from a twin pregnancy. There were no stillbirths in higher-order multiple births and in no case were both twins stillborn.

One mother consented to participate in the study following the birth of her baby that was known to have died in utero, but, when contacted at a later date to arrange an interview, she was too distressed to participate. Therefore, data for this case were available only from the mother’s maternity records and some data items were missing.

One stillbirth was the result of elective feticide in a twin pregnancy at 33 weeks’ gestation. This baby was included in the study, as the birth fulfilled the eligibility criteria. Documentation at the time of the feticide suggested that the decision for this course of action was taken because of antenatal detection of multiple fetal anomalies, but no further details were available about the nature of these anomalies.

Key findings

• There was an inverse relationship between mothers’ highest education level and odds of LMPT birth in both singleton and multiple pregnancies, with the lowest educational group being at highest risk.

• Mothers of Asian origin had increased odds of LMPT delivery than in those women whose ethnicity was white.

• Living as part of a couple during pregnancy was associated with lower odds of LMPT delivery than being single.

• Previous preterm delivery was associated with a fourfold increase in the odds of LMPT delivery compared with delivery at term.

• Maternal hypertension and diabetes originating prior to pregnancy were associated with two and four times increased odds of delivering at LMPT gestations, respectively, compared with term births.

• LMPT babies are more likely to be affected by intrauterine growth restriction than those born at term.

Strengths and limitations

The population-based nature of the study and inclusion of 80% of eligible mothers allowed recruitment from a socially and culturally diverse population. In addition, our study was able to recruit a large proportion of all births of twins and higher multiples. However, although study information was available in multiple languages for women from ethnic minorities, for the most part those with a limited understanding of English chose not to participate and this may have influenced our results with respect to analyses based on ethnicity.

We have identified an association between LMPT singleton delivery and previous preterm delivery, although from our data we are unable to determine whether this increased risk is related to previous LMPT birth or previous delivery at lower gestations.

In our study, information about specific pre-pregnancy diseases and any other chronic illnesses was ascertained by self-report at interview. It is well recognised that self-report may lead to either under- or over-reporting of conditions, although this effect is likely to occur in both study groups. No attempt has been made to determine the severity of these problems, but further subclassification of the conditions with respect to the likelihood of impact on pregnancy outcomes may be helpful.

Implications for practice

Although there is currently limited evidence for interventions to prevent preterm birth, our findings suggest that targeted counselling and/or care for a number of groups at risk of delivering at LMPT gestations may be appropriate. These include women who are likely to have difficulties accessing medical care, such as those with lower levels of education and those from ethnic minority communities. Women with longstanding chronic diseases may also benefit from enhanced medical surveillance or input to minimise the impact of their disease during pregnancy and optimise outcomes. However, the efficacy of targeted counselling and/or surveillance has not yet been tested and would require evaluation.

A number of sociodemographic factors were found to be associated with LMPT birth in both singletons and multiples. The impact and implications for these in singleton births are discussed in detail in a following section.

Future research

Further investigation of common and important pre-pregnancy medical conditions is warranted to clarify factors that place women at increased risk of LMPT delivery and determine whether targeted or general counselling, interventions and education can improve pregnancy outcomes for specific groups. Currently, information is limited about factors influencing spontaneous onset of labour at LMPT gestations and the role that maternal medical conditions play in this. There also remains a paucity of data about pregnancies in which pre-existing disease exerts an influence on the progress of pregnancy or well-being of the baby either directly, or by precipitating or exacerbating pregnancy-related complications. Exploration of this is necessary to guide obstetric decision-making in situations in which either expectant management or early delivery is an option, but the optimal course of action is uncertain.

Maternal complications

Women with systemic infection during pregnancy were no more likely to deliver at LMPT gestations than at term. However, when a raised CRP level was documented during labour, this was associated with odds of LMPT delivery that were 11 times greater than the odds of delivering at term (OR 11.85, 95% CI 3.62 to 38.79; p < 0.001). In contrast, women with a documented temperature of > 37.5 °C during labour were no more likely to deliver early. These results must be interpreted with caution. CRP level is generally regarded as a reasonably sensitive biochemical marker of the presence of infection, but is not measured in all women, so this finding is likely to reflect a group of women in whom infection was suspected. We do not have data for women in whom a CRP was measured, but was not found to be elevated, or for those in whom it was never measured. Modest fever of > 37.5 °C is less sensitive than a raised CRP level and can be caused by other, non-pathological, factors such as the ambient environmental temperature or the use of epidural anaesthesia in labour. Had data collection been limited to a more extreme value for mothers’ temperature, this may have produced different results.

Pre-eclampsia was also associated with increased odds of LMPT delivery, with affected women being four times more likely to delivery LMPT than at term (OR 4.16, 95% CI 2.84 to 6.10; p < 0.001). In addition, women who delivered LMPT infants were more likely than those delivering at term to have experienced rupture of the amniotic membranes > 24 hours prior to delivery. Gestational diabetes affected < 5% of women in each group and was not associated with increased odds of LMPT delivery compared with delivery at term.

Fetal complications

Fetal growth restriction and abnormal antenatal umbilical Doppler studies are closely related and are both likely to lead to decisions to deliver an affected infant prematurely. It is, therefore, not a surprising finding that, in this study, using univariable analysis, both were associated with increased odds of LMPT delivery [FGR: OR 1.99, 95% CI 1.55 to 2.56 (p < 0.001); abnormal Doppler: OR 13.4, 95% CI 3.17 to 56.80 (p < 0.001)]. Fetal growth restriction is likely to be the more reliable measure of the two, as it is impossible to know, with any degree of certainty, how many mothers in either group would have had abnormal studies if these had been performed. The uncertainty and the effect of small numbers are reflected in wide CIs for this measure.

Cardiotocographic abnormalities of any kind were common in both LMPT and term deliveries, with > 40% of all women in each group having some documented CTG abnormality during labour. However, abnormalities were more commonly documented in LMPT deliveries and this was associated with increased odds of LMPT delivery (OR 1.28, 95% CI 1.07 to 1.54; p < 0.007) compared with deliveries at term.

Meconium staining of the liquor was significantly less common in LMPT deliveries (OR 0.32, 95% CI 0.22 to 0.45; p < 0.001), which is an expected finding, as fetal distress associated with the passage of meconium in utero is principally a concern for women during labour at term, and in particular for post-term births (> 41 weeks’ gestation).

Umbilical cord prolapse was seen in only four (0.4%) LMPT deliveries and in no term deliveries, and so further analysis was not possible for this variable.

Key findings

• Pregnancies ending in LMPT delivery are more likely to have been affected by PROM, pre-eclampsia and FGR than are pregnancies that continue to term.

Implications for practice

The results of this study highlight for obstetric clinicians the increased likelihood of LMPT delivery in pregnancies complicated by pre-eclampsia, FGR and infection. Both expectant management and early delivery carry competing risks, with the risk of stillbirth at one end of the spectrum and the risk of neonatal morbidity at the other. As the optimal time for delivery has not yet been determined, options for management of these women at this stage of pregnancy should be approached with careful consideration of all factors.

Strengths and limitations

We have collected a wide range of obstetric information about mothers delivering at LMPT gestations and the in utero well-being of babies in order to determine factors that are associated with LMPT birth. We have shown that LMPT delivery is more common in pregnancies affected by complications. It was, however, beyond the scope of this study to explore the decision-making processes of obstetric teams that led to LMPT delivery in those in whom delivery was elective or semielective, and this is likely to play a crucial part in rates of LMPT birth and variation in these.

Future research

There is a need for further research into the management of complicated pregnancies at LMPT gestations and, in particular, to define optimum management of women who develop pre-eclampsia. The role of infection and inflammation in LMPT delivery and the value of clinical markers in guiding decisions to deliver require clarification.

There have been few studies of the outcomes of pregnancies affected by common complications with the aim of identifying the point at which the risk of in utero demise exceeds the risk of neonatal morbidity and thus justifies LMPT delivery. Exploration is needed to allow more detailed understanding of the obstetric decisions that lead to LMPT delivery.

Area-level socioeconomic risk

As in stream 1, area-level socioeconomic risk was calculated using the IMD 2010. 154 This measure of multiple deprivation is made up of seven domain indices at the LSOA level, which relate to income deprivation, employment deprivation, health deprivation and disability, education, skills and training deprivation, barriers to housing and services, and living environment deprivation and crime. LSOAs are the smallest areas for which these deprivation data are available; although some degree of heterogeneity will exist within them, the small size of the areas (only 1500 residents) limits this. IMD 2010 was linked to the LAMBS data constructed on LSOA of residence based on the mother’s postcode at the time of birth.

These data were obtained from the ONS and linked to the maternal data by LSOA. All LSOAs in England were ranked by deprivation score. They were then weighted by their population of births (using the live-birth denominator data) and divided into five groups, with approximately equal populations of births in each, from 1 (least deprived) to 5 (most deprived).

Individual-level socioeconomic risk

In order to operationalise the concept of socioeconomic risk, a multidimensional approach was taken, aiming to encapsulate five key areas of importance: (1) education; (2) occupation; (3) income; (4) wealth; and (5) social support. Indicators were then incorporated into the maternal interview to reflect these five aspects of socioeconomic risk.

Education was measured by collecting information on mother’s highest qualification and also age at leaving continuous education. Occupation was measured by both detailed questions on mother’s occupation during pregnancy and receipt of Jobseeker’s Allowance. Mother’s socio-occupational status was then classified using the NS-SEC (ONS143) based on the information collected in the maternal interview. Occupations were classified into eight analytic classes, which were collapsed into managerial or professional (classes I and II), intermediate occupations (classes III–V), routine and semiroutine occupations (classes VI and VII), and unemployed, retired and students (class VIII). Mothers who reported that they looked after their family full time were classified separately. Income was measured by the collection of occupation data but also was based on household access to a car, which is thought to reflect current income. It was decided not to ask for detailed information on personal income, as this was thought to reduce response rates. Wealth and long-term income was measured based on whether or not the mother was a home owner. Finally, social support was measured based on whether or not the mother was living with someone as a couple during the majority of the pregnancy.

Sociodemographic data

Demographic data were also collected in the maternal interview including mother’s age and ethnic group. Mother’s age was categorised into four groups: (1) < 20 years; (2) 20–29 years; (3) 30–39 years; and (4) ≥ 40 years. Ethnic group was based on groups used in the census: white; mixed; Asian or Asian British; black or black British; and Chinese or other ethnic group.

Statistical analyses

The aim of this work was to explore whether or not area-level socioeconomic deprivation was related to LMPT birth and then to try to unpick any inequalities by identifying whether or not individual socioeconomic-level factors explained observed variations.

First, univariable analyses were undertaken to explore the relationship with area-level deprivation and the five aspects of socioeconomic risk and rate of LMPT birth. Analyses were restricted to singleton births because differential access to fertility treatment may lead to a higher incidence of multiple births in less deprived areas. Following univariable analysis, multivariable unconditional logistic regression was used to examine the independent relationship between area- and individual-level socioeconomic risk factors and LMPT and to generate adjusted ORs with their 95% CIs. Variables likely to confound the relationship between socioeconomic position and LMPT were identified a priori based on evidence from the published literature and the results of univariable analysis (i.e. ethnicity and maternal age).

Model building using unconditional logistic regression was then undertaken based on LMPT as the outcome and including all five measures of individual socioeconomic risk, and area-level socioeconomic risk. Ethnicity and maternal age were also included, as there was a priori evidence that these factors were related to LMPT birth. Adjusted ORs were obtained for each risk factor and compared with the unadjusted ORs to assess the relative importance of the effect to LMPT birth. The individual effect of each variable on the fit of the data was assessed using the likelihood ratio test.

Finally, models were fitted to compare the effect of including the SES Index of combined socioeconomic risk with a model of the significant individual indicators to assess whether or not this combined score accounted for a greater amount of variation in LMPT births than the individual risk factors, as it reflected a multiplicative effect of the individual factors. These models were both adjusted for ethnicity and maternal age.

Key findings

• Women from the most deprived areas had a 49% higher risk of delivering LMPT than of delivering at term.

• After adjusting for individual-level socioeconomic factors, there was no significant association with area deprivation and LMPT birth.

• Not having access to a car had the strongest association with LMPT, with mothers at 37% increased odds of LMPT.

• Area- and individual-level socioeconomic factors did not explain the increased odds of Asian or British Asian women delivering an LMPT infant. Some of the apparent socioeconomic inequalities at the univariate level may be because of this higher rate of LMPT birth among Asian or Asian British women, who are more likely to be more deprived.

Strengths and limitations

Many research studies exploring socioeconomic inequalities in preterm birth are limited to area-level measures of deprivation. Here it has been possible to explore both area- and individual-level socioeconomic risk factors and assess the impact on LMPT birth with a rich source of data on these risk factors. Almost all respondents answered the socioeconomic questions in the maternal data set and so missing data were not a significant problem. Lack of access to a car showed the strongest relationship with LMPT birth, but there was strong correlation between all of the five individual socioeconomic risk factors. However, the SES Index, combining all five indicators of socioeconomic risk, did not account for any more of the variation in LMPT risk.

Data are available only for responders to the study. Study response has been shown to be related to socioeconomic status in a variety of studies, including the national census, and so the responders in this study are likely to be less deprived than all eligible study participants. Therefore, the socioeconomic effects seen here may be underestimated.

Implications

Socioeconomic inequalities exist in LMPT birth at both area and individual level. The effect of socioeconomic deprivation on very preterm birth rates appears to continue into LMPT gestations. This suggests that the burden of LMPT is greater in more deprived areas and is likely to lead to a greater burden of mortality, morbidity and adverse long-term outcomes in these areas owing to differential rates of LMPT birth.

In studies of socioeconomic deprivation, area measures are generally used as a proxy for individual risk. Consequently, as in this work and a recently published Italian study,155 there appears to be no residual influence of area-level deprivation after adjusting for individual measures, as there is often misclassification of deprivation at the area level. However, in studies from the USA and Canada, a consistent area-level deprivation effect has been shown even after accounting for individual-level behavioural and medical risk factors,156,157 which suggests that there may be geographical differences in access to good health care.

Interestingly, although all of the individual socioeconomic risk factors explored showed a relationship with LMPT birth at the univariable level, only lack of access to a car was significant after adjusting for other variables. This socioeconomic risk factor showed a stronger association than education, occupation, house ownership or cohabiting status. Lack of access to a car may reflect access to short-term income, but may also reflect ability to physically access services.

These data suggest that reducing the burden of disease as a result of LMPT birth will have to involve targeted, disease-specific preventative interventions in high-risk pregnancies. However, for low-risk groups, population-level public health action is needed to address risk factors for preterm birth associated with social deprivation.

Future research

The LAMBS collected rich and diverse data on socioeconomic risk, but also on lifestyle and health behaviours, stress and work conditions. The next stage of this research will be to understand how these factors are related to the observed socioeconomic inequalities seen here, exploring modifiable risk factors that may reduce the incidence of LMPT birth in the future. It is likely that the relationship between low socioeconomic group and LMPT may be explained by the clustering of lifestyle factors and health-related behaviours among socioeconomically deprived women. Both Danish and Swedish population-based studies have shown a reduction in the educational/socioeconomic status gradient in risk of preterm birth after adjustment for lifestyle factors including smoking. 158,159 Using data from the LAMBS we can extend these analyses to assess the impact of these risk factors on LMPT birth and assess whether or not the effects are similar to those seen at earlier gestations.

Mortality

Deaths were uncommon in the study group, but six babies died before discharge from hospital. Of these, five had been born at LMPT gestations and one at term, but the difference in the number of deaths between the two groups was not statistically significant. Four deaths occurred within the first week of life, one of which was of a term-born baby, and there were two deaths at more than 28 days of life. Table 21 shows the gestational ages of these infants at birth, the postnatal day of death and the cause of death.

Congenital anomalies

There were 23 babies in the LMPT group and nine babies in the term-born group with major congenital anomalies or chromosomal disorders. Table 22 shows the range of congenital anomalies seen and numbers of affected babies in both groups.

Resuscitation at birth

Compared with term-born babies, those born at LMPT gestations were more than twice as likely to require active resuscitation at birth (RR 2.37, 95% CI 1.82 to 3.08; p < 0.001), and were four times more likely to require endotracheal intubation (RR 4.02, 95% CI 1.85 to 8.72; p < 0.001).

Respiratory outcomes

The majority of babies in both the LMPT (84.2%) and term (98.4%) groups did not require any form of respiratory support. However, the risk of needing mechanical ventilation and/or non-invasive respiratory support in the LMPT group was more than twice that of infants born at term (RR 12.74, 95% CI 6.49 to 25.01; p < 0.001). Ventilatory support was required by 8.3% of LMPT infants, compared with < 1% of term-born infants. Non-invasive support was the maximum level of respiratory support given in 4% of LMPT infants and 0.1% of term-born infants. Oxygen therapy was required in 3.5% of LMPT infants, compared with only 0.7% of those born at term.

Nutritional outcomes

The proportion of babies requiring i.v. fluids was higher in the LMPT group than in the term-born group (RR 13.02, 95% CI 8.33 to 20.36; p < 0.01). The use of PN was also more common in LMPT infants (RR 11.43, 95% CI 3.51 to 37.21; p < 0.001) and the duration of i.v. fluids and nutrition use was also greater in the LMPT group. LMPT babies took longer than their term-born counterparts to attain full oral feeding by breast or bottle. Three infants in the LMPT group were fed via a NGT when they were discharged from hospital, compared with no infants in the term group.

With respect to breastfeeding, LMPT infants were less likely to receive any breast milk during their neonatal stay than infants born at term (64.2% vs. 73.8%; RR 0.87, 95% CI 0.82 to 0.92; p < 0.001). This proportion decreased during the neonatal stay in both groups, but fell more in the LMPT babies than in the term babies. The proportion of babies receiving any breast milk at the time of hospital discharge was 72.2% in the term group, compared with only 58% in the LMPT group. Exclusive breast milk feeding at discharge was seen in 65.1% of the term-born babies, but in fewer than half of the LMPT babies (65.1% vs. 39.3%; RR 0.60, 95% CI 0.55 to 0.66; p < 0.001).

Admission to neonatal unit and neonatal morbidity

Late and moderately preterm infants were almost 10 times more likely to be admitted to a NNU than babies in the term-born group (RR 9.82, 95% CI 7.27 to 13.27; p < 0.001). They were also at significantly higher risk of a number of common neonatal morbidities including hyperbilirubinaemia [that was of a level requiring management with phototherapy (RR 35.54, 95% CI 15.85 to 79.68; p < 0.001)], severe hypothermia (RR 8.09, 95% CI 4.34 to 15.07; p < 0.01) and severe hypoglycaemia (RR 8.17, 95% CI 3.93 to 16.98; p < 0.001). The incidence of neonatal encephalopathy of moderate or severe severity (grades II and III) was similar in the two groups.

Clinical investigations performed during the neonatal stay

In general, LMPT infants were more likely to undergo investigation than term-born babies. With respect to radiological examinations, LMPT babies were eight times more likely to have chest radiography performed (RR 8.52, 95% CI 5.52 to 13.16; p < 0.001) than term-born babies, and were also more likely to have had a cranial ultrasound scan performed (RR 5.19, 95% CI 2.88 to 9.37; p < 0.001). However, the number of MRI scans did not differ between the groups.

Screening for suspected infection was carried out in more than one-third of LMPT infants, compared with < 5% of those born at term (RR 7.19, 95% CI 5.38 to 9.60; p < 0.001). This yielded a positive culture in only 17 LMPT infants (5% of those tested). None of the infants born at ≥ 37 weeks’ gestation had a culture-positive infection.

Length of hospital stay

Three LMPT babies and one term-born baby died before discharge. For those infants surviving to discharge, the total length of hospital stay was greater in the infants born LMPT.

Resuscitation at birth

Compared with term-born multiples, LMPT multiples were more likely to require active resuscitation at birth (RR 1.81 95% CI 1.18 to 2.79; p = 0.007), but were no more likely to need endotracheal intubation.

Respiratory outcomes

As was the case with singleton births, high percentages of multiples in both groups did not receive any respiratory support (LMPT 84% and term 98.9%). Mechanical ventilation, non-invasive respiratory support and oxygen therapy were each required by only one baby in the term-born multiples group. However, 17 babies (8.5%) of LMPT multiples were ventilated and 11 (5.5%) and 4 (2%), respectively, needed non-invasive support and oxygen therapy. Among the multiple births, the risk of LMPT infants needing mechanical ventilation and/or non-invasive respiratory support was significantly higher than that of infants born at term (RR 19.18, 95% CI 4.62 to 79.70; p < 0.001). It should be noted that the small numbers for this analysis result in wide CIs.

Nutritional outcomes

Among infants from multiple births, the number of babies requiring any kind of nutritional support was small in both groups. No term-born infant required either PN or feeding by NGT, and only seven (3.5%) LMPT infants required PN and only one (0.5%) required feeding by NGT. More than one-quarter of LMPT infants required i.v. fluids, which was a significantly higher proportion than in the term-born group (RR 7.40, 95% CI 3.86 to 14.18; p < 0.001).

There were lower rates of breastfeeding among the LMPT multiple births than in the term-born group. A smaller percentage of LMPT than of term-born babies received any breast milk, but this difference did not reach statistical significance. However, a similar pattern of decline as seen in the singletons led to a significantly smaller proportion of LMPT multiples receiving breast milk at discharge (RR 0.83, 95% CI 0.71 to 0.97). Less than one-quarter of LMPT multiples were fed exclusively on the mother’s breast milk at discharge and, compared with term-born multiples, this was a highly significant difference (RR 0.60, 95% CI 0.46 to 0.80; p < 0.001).

Admission to neonatal unit and neonatal morbidity

Late and moderately preterm infants from multiple births were five times more likely to be admitted to a NNU than babies in the term-born group (RR 5.12, 95% CI 3.36 to 7.81; p < 0.001). Their risk of jaundice requiring phototherapy was also 10 times higher than that of term-born multiples (RR 10.62, 95% CI 3.80 to 29.63; p < 0.001), and they were twice as likely to have severe hypothermia (RR 2.54, 95% CI 1.03 to 6.27; p = 0.042). There was no increased risk for hypoglycaemia among LMPT multiples and neonatal encephalopathy was not seen in any baby from a multiple birth.

Clinical investigations performed during the neonatal stay

Late and moderately preterm infants from multiple births were more likely to under chest radiography (RR 5.75 95% CI 2.96 to 11.20; p < 0.001) and a cranial ultrasound (RR 5.94, 95% CI 1.71 to 20.58; p < 0.001) than term-born babies. The number of MRI scans performed was the same in both groups.

Screening for suspected infection was carried out in 35% of LMPT multiples, compared with only 5.8% of those born at term (RR 5.99, 95% 3.59 to 10.00; p < 0.001), but only two (1%) LMPT babies and one (36%) term-born baby were subsequently found to have positive cultures.

Length of hospital stay

All live-born infants from multiple births survived until discharge from hospital, but among these infants the average neonatal hospital stay was longer in the LMPT group than in the term-born group.

Key findings

• Infants born at LMPT gestations are at a significantly greater risk of neonatal morbidity than infants born at term.

• For most adverse outcomes, there is no difference in the effect between singletons and multiples born at LMPT gestations.

• Late and moderately preterm babies are less likely to receive maternal breast milk than those born at term.

• Non-spontaneous onset of labour at LMPT gestations is associated with more neonatal morbidity than delivery following spontaneous onset of labour.

• The effects of prematurity at LMPT gestations are compounded by effects relating to complications leading to medically indicated delivery.

Strengths and limitations

We have recruited a large population of LMPT infants and have collected detailed data for a wide range of outcomes. Data were collected either during the infants’ neonatal hospital stay or as soon as possible after discharge, minimising error and missing data. However, we were unable to recruit 100% of infants and it is possible that selection bias may have occurred if mothers of specific groups of infants declined to participate, for example if the baby was extremely sick or so well that very early discharge from hospital was possible. This may have led to both over- and under-estimation of effects.

Implications for practice

There is a need to examine current practice with regard to neonatal care for LMPT babies, as these babies have historically been cared for in a similar way to those born at term. Our findings and those of other researchers, in identifying increased risks of multiple common and important neonatal morbidities, should prompt consideration of whether or not current models of care are the most appropriate for this group. It may be that a greater level of surveillance following delivery or earlier intervention in some cases may lead to improved outcomes and shorter hospital stays, and this should be explored. Changes in clinical practice, if required in this large group of babies, are likely to be costly in terms of both resources and personnel, and careful planning would be needed to ensure that the highest risk groups are identified and targeted when possible.

Observation of poorer breastfeeding rates in LMPT babies is worrying, particularly as the longer duration of hospital stay in this group might be expected to allow greater support for mothers wishing to breastfeed their babies. As these babies appear to constitute a group at high risk of breastfeeding failure, targeted support in establishing and maintaining breastfeeding may lead to improvement in rates of breastfeeding at discharge.

High rates of obstetric intervention for delivery in LMPT babies and our finding of worse outcomes in this group suggested that greater caution may need to be exercised in determining the optimum timing for delivery in complicated pregnancies.

Future research

Having identified a higher risk of adverse neonatal outcomes in LMPT babies, future research is required to determine how much of this observed risk relates to immaturity and how much to the reasons for the early delivery. In this way, it may be possible to highlight certain groups for whom there may be scope to intervene and improve early outcomes. The identification of poorer breastfeeding rates in LMPT babies warrants urgent work to explore the factors influencing the initiation and maintenance of breastfeeding in this group and whether or not this can be improved with a different model of care. Ongoing work in this group is needed to determine whether problems in the neonatal period affect very large numbers of LMPT infants or whether they are confined to smaller, high-risk groups. Our findings of higher rates of morbidity in infants following medically indicated delivery rather than spontaneous onset of labour suggest that the latter explanation may be likely, with these infants forming one higher-risk group. Follow-up of the cohort will be crucial to indicate whether or not adverse neonatal outcomes translate into adverse outcomes in childhood and beyond.

Measures

Data relating to health, respiratory, neurodevelopmental, cognitive and behavioural outcomes were collected via parental reporting. A study questionnaire combining standard parent report measures and additional items to address health issues associated with preterm birth was developed. To assess general health, parents were asked to rate their children’s health in comparison with other children of the same age using a 4-point scale (excellent, good, fair or poor). Respiratory outcomes were assessed using forced-choice questions relating to the frequency of wheezing (never, occasionally, frequently or every day), type (preventer or reliever) and frequency of inhaler use (never, occasionally, frequently or every day), and prescription of steroids for wheezing (yes or no). Neurological outcomes were assessed using forced-choice questions relating to whether or not the child had experienced seizures over the last year (none, febrile only or neurological seizures) and prescription for anticonvulsant medication (yes or no).

To assess neurodevelopmental outcomes, parents were asked whether or not the child had been given a diagnosis of cerebral palsy (CP) and three forced-choice items were used to assess the child’s vision, hearing and neuromotor function. Responses were used to identify the severity of functional impairment (none, mild, moderate, severe) within each domain. Children were classified as having neurosensory impairment (NSI) if they had a moderate or severe impairment in any one of these three domains (irrespective of the diagnosis of CP). These questionnaire items were designed to map onto standard criteria for classifying neurodevelopmental outcomes at 2 years of age. 179

Cognitive development was assessed using the PARCA-R, a parent questionnaire of cognitive and language development. 180,181 Raw scores were summed to provide a subscale score for non-verbal cognition (NVC; range 0–34) and language development (range 0–124). These were then summed to provide a total parent report composite (PRC) score (range 0–158). The PARCA-R has excellent concurrent validity with gold standard developmental tests and PRC scores of < 49 have good diagnostic utility, with ≥ 80% sensitivity and specificity, for identifying very preterm children with developmental delay at 2 years of age. 180,181 Therefore, PRC scores of < 49 were used to identify children with suspected developmental delay in the present study. If fewer than four items were missing on the NVC scale, these items were replaced by the average NVC item score at the individual level. If more than five items were missing from the PARCA-R NVC scale, these children were excluded (n = 6). For 21 children whose first language was not English and whose parent was unable to complete the language section of the questionnaire, summary scores were not computed but cognitive impairment was classified using the NVC subscale alone, from which scores of < 22 were used to identify developmental delay corresponding with NVC scores of < 2.5th percentile of the term-born group. Classifications of cognitive impairment were combined with the NSI classification to determine the proportion of children with composite neurodevelopmental disability (defined as a moderate/severe impairment in vision, hearing, motor or cognitive function). 179

Behavioural outcomes were measured using the BITSEA,182 a norm-referenced parent report screener for social–emotional development and behavioural problems in children aged 1–3 years. The problem subscale comprises 31 items to assess problems such as aggression, defiance, hyperactivity, negative emotionality, anxiety and withdrawal, from which a summed total problem score is computed, with higher scores indicating greater problems. The competence subscale comprises 11 items to assess social–emotional abilities such as empathy, prosocial behaviours and compliance, from which a total competence score is computed, with lower scores indicating lesser competence. Behaviour problems (problem scores ≥ 75th percentile) and delayed social–emotional competence (competence scores ≤ 15th percentile) were identified using published age- and gender-specific norm-referenced cut-off points. 182 The questionnaire also comprises two questions to elicit level of parental concern about their child’s socioemotional and language development. The BITSEA has good test–retest reliability, inter-rater reliability and internal consistency, and positive screens have been shown to predict behaviour problems and psychiatric disorders at school age in both term-born and very preterm children. 182,183

Two-year follow-up rates

At 2Y-CA, seven (0.6%) LMPT children had died, the parents of eight (0.7%) had withdrawn them from the study and two (0.2%) children had moved into closed foster care precluding ongoing follow-up. Of the remaining 1113 LMPT children eligible for follow-up at 2 years, the parents of two (0.2%) refused to complete a study questionnaire, and we did not receive follow-up data from the parents of 460 (41.3%) children, despite numerous attempts to contact them. Thus, follow-up data were ultimately received for 651 (58.5%) infants born LMPT (Figure 10); excluding deaths, this equates to a follow-up rate of 58.0% of the total LMPT cohort.

FIGURE 10.

Follow-up rates and retention of the LAMBS cohort to 2Y-CA.

Of term-born infants, two (0.2%) had died, the parents of 16 (1.3%) children had withdrawn them from the study, and the parents of two (0.2%) had moved abroad, leaving 1235 children eligible for follow-up at 2Y-CA. Of these term-born infants, the parents of three (0.2%) infants refused to complete a study questionnaire, and we did not receive follow-up data from the parents of 461 (37.3%) children. The parents of 771 returned questionnaires, equating to a follow-up rate of 62.4% (see Figure 10) of eligible term-born children and, excluding deaths, 61.5% of the term-born cohort recruited. For all subsequent analyses, infants with major congenital anomalies were excluded; this equated to 638 responders in the LMPT group and 765 among the term-born controls.

Analysis of non-responders

Previous research has shown that non-responders to follow-up in preterm cohorts typically have higher medical and socioeconomic risk than those who respond to study questionnaires or formal evaluations. 185,186 To examine the effect of loss to follow-up on outcomes at 2 years, maternal and infant characteristics of non-responders were analysed for both LMPT and term-born children (see Table 28). Among infants born LMPT, all the maternal characteristics examined were significantly associated with non-response to follow-up at 2 years. Specifically, mothers who did not respond to follow-up were more likely to be younger (i.e. aged < 25 years), to have higher socioeconomic risk and poorer mental and general health, to be of non-white ethnic origin and not to speak English. The greatest risk was associated with socioeconomic deprivation, with those with medium- and high-risk SES Index scores being, respectively, 1.8 (95% CI 1.4 to 2.4) and 3.1 (95% CI 2.8 to 3.9) times more likely than those with low risk to be non-responders (see Table 28). Of all the infant characteristics examined, not receiving any breast milk at discharge was the only significant factor associated with non-response; this is likely to be a marker of higher social risk among those mothers not completing study questionnaires.

Analysis of factors associated with non-response in term-born controls produced similar results in terms of maternal characteristics, with non-responders more likely to be younger (aged < 25 years) or older mothers (aged ≥ 35 years), to have higher socioeconomic risk and to be of non-white ethic origin and not to speak English. However, there was no association with response to follow-up in terms of mothers’ self-reported general and mental health. Again, the greatest risk was associated with socioeconomic deprivation, with those with medium- and high-risk SES Index scores being, respectively, 1.7 (95% CI 1.3 to 2.1) and 2.8 (95% CI 2.3 to 3.4) times more likely than those with low risk to be non-responders. As in the LMPT group, not receiving any breast milk at discharge was significantly associated with non-response. In addition, mothers who gave birth to more than one baby and mothers of babies born at lower gestational age or of low birthweight were more likely to be non-responders (see Table 28).

Health, respiratory and neurological outcomes

Among those without congenital anomalies, health and respiratory outcomes for LMPT infants and term-born controls are shown in Table 29. RRs for adverse outcomes among LMPT infants are reported both unadjusted and adjusted for sex and SES Index (low, middle and high risk).

Overall, LMPT infants were at 2.4 times (95% CI 1.4 to 3.8 times) increased risk of poorer parent-reported general health than infants born at term. Notably, this increase was not evidenced in terms of ‘poor’ health, but instead there was an excess of LMPT infants reported as having ‘fair’ health. A similar pattern of results was evidenced for respiratory outcomes. Although LMPT infants did not have frequent wheeze more often than infants born at term (RR 0.89, 95% CI 0.45 to 1.75; p = 0.72), there was an excess of LMPT infants with occasional wheeze (29% vs. 20%) and a reduction in the proportion who were reported as never wheezing compared with infants born at term (69% vs. 78%). Similarly, although more LMPT infants had been prescribed an inhaler by 2 years of age (RR 1.38, 95% CI 1.10 to 1.73; p = 0.005), they did not have a higher rate of frequent inhaler use than term-born infants (RR 0.77 95% CI 0.39 to 1.55; p = 0.47); there was an increase only in occasional inhaler use, and LMPT infants had not been prescribed more corticosteroids than term-born controls, indicating no excess of severe respiratory problems. Similarly, there was no excess of neurological sequelae in LMPT infants: they were no more likely than term-born infants to have had seizures in the past year or to have been prescribed anticonvulsant medication (see Table 29). Adjustment for potential confounders of the association between outcomes and LMPT birth did not alter the statistical significance of results for any health, respiratory or neurological outcome (see Table 29).

Neurodevelopmental outcomes

Neurodevelopmental outcomes for LMPT and term-born children are shown in Table 30. RRs were adjusted for potential confounders (sex, corrected age and SES Index) when possible.

As in the term-born population, the prevalence of moderate/severe NSI was very low, with < 1% of LMPT children having functional hearing, visual or neuromotor impairment. When RRs could be computed, these showed no significant difference from term-born controls. When these were combined as a composite measure, overall there was a significantly increased risk for NSI among LMPT children compared with controls (1.6% vs. 0.3%; RR 6.00; 95% CI 1.32 to 27.28). Among the LMPT group, no parents reported that their child had received a diagnosis of CP compared with 0.5% of the term-born population (see Table 30).

Cognitive problems were more frequent, in general, and there was a significantly increased risk of cognitive impairment, which was present in 16% of LMPT children compared with 10% of term-born children (RR 1.56, 95% CI 1.18 to 2.06). There was also a significant difference in mean PARCA-R scores between LMPT children and controls (term: mean 94.5, SD 33.3; LMPT: mean 88.9, SD 36.0; unadjusted mean difference –5.61, 95% CI –9.28 to -1.95, p = 0.003; adjusted mean difference –4.16, 95% CI –7.70 to -0.63, p = 0.021). The distribution of PRC scores was similar in both the LMPT and term-born children, but there appeared to be a general shift to the left, resulting in an excess of LMPT children with scores below the cut-off point for cognitive impairment (Figure 11).

FIGURE 11.

Frequency distribution of PARCA-R PRC scores in LMPT and term-born children. Higher scores indicate better cognitive development. The vertical line indicates the cut-off point for identifying cognitive impairment.

When cognitive impairment was combined with NSI, LMPT children were at significantly increased risk of moderate/severe neurodevelopmental disability compared with term-born controls (RR 1.57, 95% CI 1.19 to 2.07). This was almost exclusively the result of cognitive impairment, with only two additional LMPT children and one additional control having neurodevelopmental disability over cognitive impairment alone (see Table 30). The adjusted RR for neurodevelopmental disability was 1.42 (95% CI 1.08 to 1.86).

Risk factors for neurodevelopmental disability in children born late and moderately preterm

In order to identify children at greatest risk of adverse outcomes, we explored the association of obstetric and neonatal factors with the presence of neurodevelopmental disability at 2Y-CA. As shown in Table 31, children whose mothers were of non-white ethnic origin, non-English speaking or of higher socioeconomic risk, or who smoked or took recreational drugs during pregnancy, or who had pre-pregnancy hypertension or pre-eclampsia were significantly more likely to have neurodevelopmental disability. Raised CRP levels during pregnancy were also marginally associated with disability (p < 0.1). LMPT children who were conceived via infertility treatment were less likely to have neurodevelopmental disability at 2 years. Of the neonatal factors analysed, being male, being resuscitated at birth and not having any breast milk at discharge were significantly associated with the risk of neurodevelopmental disability. These variables predominantly appear to be markers of high sociodemographic risk or hypertensive disease, either before or during pregnancy. The greatest impact was found for recreational drug use during pregnancy, pre-pregnancy hypertension and higher SES Index, each of which as associated with a three- to fourfold increased risk of disability.

Given the correlation between these factors, multivariable analyses were conducted to identify independent predictors of neurodevelopmental disability in LMPT children. As shown in Table 32, eight factors emerged as risk factors. Being male, being resuscitated at birth, having a mother who had pre-pregnancy hypertension or pre-eclampsia or who used recreational drugs during pregnancy was associated with a 1.5- to 2.5-fold increased risk of disability. Socioeconomic deprivation was also independently associated with disability, with risk of disability increasing with increasing deprivation. In addition, having a mother who did not speak English and not having had breast milk at discharge were also associated with disability.

Behavioural outcomes

Behavioural outcomes were assessed using the BITSEA parent report questionnaire from which both continuous and categorical outcomes were derived for the presence of behaviour problems and delays in socioemotional competence. Analyses are presented unadjusted and after adjustment for age, sex and socioeconomic status (Table 33).

Although there was a trend towards higher problem scores in LMPT children (p < 0.1), there was no statistically significant difference in the mean BITSEA problem scores. In contrast, LMPT children had significantly lower competence scores than term-born controls both before and after adjustment for confounders (adjusted mean difference –0.44, 95% CI –0.77 to –0.12; p = 0.008). When the BITSEA scores were compared with age- and sex-standardised norms, overall, LMPT children were significantly more likely to have a behaviour problem or delay, present in 37% of LMPT children, compared with 30% of controls (adjusted RR 1.17, 95% CI 1.01 to 1.35). However, although significantly more LMPT children had delays or deficits in socioemotional competence (adjusted RR 1.36, 1.12 to 1.65; p = 0.002), there was no excess of behavioural problems in this group. Despite the delays in socioemotional development observed among LMPT children, parents did not report more frequent concerns about their child’s behaviour or language development (see Table 33).

Key findings

• Parents report that children born LMPT have poorer health than children born at term and are more likely to be prescribed inhalers. These findings are indicative of an increase in minor health conditions and respiratory symptoms rather than severe health, respiratory or neurological outcomes.

• Children born LMPT have a six times higher risk of NSIs than children born at term. However, the absolute impact of these impairments is minimal compared with the increased risk of cognitive difficulties, which are present in 16% of children born LMPT.

• Late and moderately preterm children are more likely to have delayed socioemotional development at 2 years of age. Although parents do not report concern about these early delays/deficits, these may be indicative of longer-term peer relationship difficulties and mental health problems.

• The excess of socioemotional delays in the absence of behaviour problems may be indicative of a different behavioural profile following LMPT birth compared with children born very preterm.

• The significant increase observed in adverse health, cognitive and socioemotional outcomes is not accounted for by the association of LMPT birth with higher socioeconomic risk.

• Risk factors for poor neurodevelopmental disability in LMPT children include high socioeconomic risk, male sex, maternal hypertensive disease before or during pregnancy, antenatal recreational drug use and not feeding breast milk at discharge from hospital. These factors may be amenable to intervention for improving long-term outcomes in this population.

Strengths and limitations

Despite intensive efforts and resources directed at improving response rates to follow-up at 6, 12 and 24 months, this still remained at approximately 60%. Analysis showed that non-responders had higher socioeconomic and demographic risk than responders, and thus we may have underestimated the true prevalence of developmental problems in this study. Although the effect of selective dropout was similar in both groups, we intend to carry out multiple imputations to assess the impact of non-response bias on the prevalence of key adverse outcomes at 2Y-CA.

Given the size of the cohort, individual assessments were not possible at 2Y-CA and thus we opted to conduct follow-up using validated parent report measures. The utility of these for identifying children with poor longer-term outcomes requires investigation to determine the predictive validity of infant assessments in this population. The prospective nature of this study enabled us to collect important demographic, obstetric and neonatal information and thus we were able to adjust for the effects of socioeconomic deprivation on developmental outcomes at 2 years. Although multivariable analyses showed that NSI, cognitive impairment and delayed socioemotional competence were significantly increased after adjustment for our composite SES Index, there may be a significant interaction between preterm birth and socioeconomic factors. In particular, greater affluence and lower social risk may have a greater impact on outcomes in the LMPT population, potentially ‘protecting’ infants from the risks conferred by LMPT birth. As a priority, we intend to carry out further analyses to explore factors that mediate the relationship between LMPT birth and adverse developmental outcomes and the interaction of socioeconomic factors with the predictors of disability examined above. In addition, we intend to carry out further analyses to determine whether or not we can identify a subgroup of LMPT children at greatest risk of adverse outcomes in order to aid in decision-making around the time of birth and to help target follow-up and early intervention efforts.

Implications for practice

The results of the 2-year follow-up study may be used by obstetric practitioners in decision-making when considering the likely long-term outcomes of delivery of pregnancies at LMPT gestations. Information about developmental difficulties may also be used to counsel parents about the potential long-term outcomes for their child. This may aid in shared decision-making around delivery and also in raising awareness of the potential for long-term developmental problems, particularly in terms of cognitive and socioemotional development. Our patient and public involvement (PPI) work has identified that parents feel that community health-care workers lack specialist knowledge about the developmental and caregiving needs of babies born at LMPT gestations. Thus, the results of this study may be important in informing community paediatricians and health-care practitioners about outcomes for children born LMPT and for the need to carry out surveillance for developmental problems among these children.

The increased prevalence of cognitive and socioemotional difficulties at 2 years may warrant follow-up over the early years. Given the low prevalence of neurodevelopmental disorders and the size of the LMPT population, this may be best delivered by community health-care services rather than by neonatal services, as is currently the model of provision for children born very preterm. Low-cost alternatives to developmental assessments may be appropriate for this population, for which parent questionnaires may prove valuable. Enhanced antenatal surveillance for pregnant women at highest risk of pre-eclampsia and socioeconomic deprivation and ongoing developmental and educational support for mothers post discharge may also be beneficial for their infants who are born at LMPT gestations.

Future research

There is a need to assess the validity of the PARCA-R for predicting longer-term outcomes to determine its use as a clinical screening measure in children born LMPT. Longer-term follow-up of this cohort will be imperative to determine how early cognitive problems evolve over time and whether or not there is developmental plasticity in LMPT children. This will also enable an assessment of whether or not delays in early socioemotional development manifest as behavioural problems or peer relationship difficulties later in childhood.

Work to evaluate the efficacy of early parenting interventions for improving cognitive and socioemotional development in this population is needed. Given that we have identified an increase in problems in this group at 2 years, there is a need to develop, implement and evaluate potential follow-up schemes for families and children born LMPT that are feasible in such a large group of children. Detailed clinical studies are warranted to further explore and validate our findings relating to an observed excess of mild respiratory symptoms in this group of children. The impact of maternal mental health on behavioural outcomes in this population also requires exploration.

Measurement of resource use and costs

Relevant resource items were integrated into the perinatal and follow-up data collection instruments described previously. The neonatal and maternal data collection forms captured a comprehensive profile of resource use by each infant, encompassing length of stay by intensity of care, surgeries, investigations, procedures, drugs, consumables, transfers and post-mortem examinations until final hospital discharge or death (whichever was earliest). Resource inputs were valued using a combination of primary research, based on established accounting methods, and data collated from secondary national tariff sets. 211,212 All costs were expressed in pounds sterling and reflected values for the financial year 2010–11.

The total length of stay (total inpatient hospital days) was computed as the total number of hospital days until first discharge to home or death. This total incorporated any hospital stays following interhospital transfers that may have occurred. Postnatal costs for the mothers were based on the method of delivery and costs assigned using data from the NHS Reference Costs 2010–11. 212 Detailed information was available on the interventions and feeding received by infants in the NNU including number of days on oxygen, number of days on a ventilator, number of days of non-invasive respiratory support, number of days of PN and number of days of i.v. fluids. A clinician involved with the study (EMB) used this daily information on interventions and feeding to map the time spent in the NNU into days by level of neonatal care (special, high dependency or intensive). The cost of neonatal care was calculated for each infant by multiplying the length of stay in special care, high-dependency care or intensive care by the per diem cost of the corresponding level of care using data from the NHS Reference Costs 2010–11. 212 Non-routine investigations excluded from these per diem costs were identified by study clinicians and these were valued using a combination of primary and secondary costs. For situations in which these costs were not available from national tariffs, clinicians were asked to identify the staff and material inputs required for these investigations. Staff time was valued using the Unit Costs of Health and Social Care 2012. 211 The costs of surgeries were calculated by assignment of surgical procedures to relevant Healthcare Resource Group (HRG) codes and application of unit costs from national tariffs. 212 Transfers were recorded whenever an infant was transported from a general hospital to a specialist hospital for neonatal critical care, from a specialist centre to a general hospital, or from a specialist centre to a higher-level centre, and were valued using costs from the NHS Reference Costs 2010–11. 212 Post-mortem costs were based on data from secondary sources. 213 Table 34 contains for a list of resource values and their sources for the period from birth to initial hospital discharge.

As part of the battery of research instruments completed at the follow-up points (6 months, 1 year and 2 years), the main parent was asked to complete detailed postal questionnaires about their child’s resource utilisation over the previous period. The 6-month questionnaire covered the child’s resource utilisation between initial hospital discharge and 6 months; the 1-year questionnaire covered the child’s resource utilisation between 6 and 12 months; and the 2-year questionnaire covered the child’s resource utilisation between their first and second birthdays. The questionnaires were piloted to ascertain their acceptability, ease of comprehension and reliability and reminder letters were sent to parents to increase the response and completion rates. The data collected from the main parent included their child’s use of hospital inpatient, day care and outpatient services, community health and social care services, and medicines and drugs, and also recorded adaptations to the home, provision of special equipment and parental lost productivity attributable to the child’s health status, over the relevant time horizons. Resource inputs were valued using a combination of primary research, based on established accounting methods, and data collated from secondary national tariff sets211,212 (GBP, 2011 prices).

The use of hospital-based services over the period between initial hospital discharge and 2 years included inpatient admissions, use of hospital day-care services, paediatric outpatient department appointments, other outpatient department appointments and visits to accident and emergency units. Inpatient admissions over this time horizon were delineated by type and duration of neonatal or paediatric care and valued using per diem costs extracted from the NHS Reference Costs 2010–11. 212 Use of other hospital-based care was valued by applying unit costs extracted from national tariffs. 211 The use of community-based care included routine appointments (immunisations, weight checks and developmental checks), general practitioner (GP) or practice nurse appointments, physiotherapy contacts, walk-in health-care centre visits and telephone calls to NHS Direct. Costs for these community-based services were calculated by applying unit costs from national tariffs211 to resource volumes. A list of all medications prescribed was extracted from the resource use questionnaires and items were identified for inclusion in the economic analysis. This list included musculoskeletal, central nervous system, gastrointestinal, infections, skin, eye, cardiovascular, respiratory, ear, nose, oropharynx and endocrine system drugs, as well as nutrition and blood vitamins, food additives and borderline substances. NHS net prices per milligram for these medications were obtained from the British National Formulary for Children (BNFC). 215 Costs for individual infants were estimated based on their reported doses and frequencies if these were available, or otherwise on an assumed daily dose based on BNFC215 recommendations. Drugs cost were delineated in terms of different child age groups when the recommended daily dosage obtained from the BNFC215 varied by age (i.e. < 6 months, 6–12 months, 12–24 months). In estimating these costs, we assumed an average weight of 4.3 kg for children at 1 month, 7.7 kg for children at 6 months and 11.8 kg for children at 24 months. 216 Other costs captured in the study included adaptations to the home, provision of special equipment and parental lost productivity. Adaptations to the child’s home as a result of the child’s health status encompassed changing plugs, redecoration, heating alterations, sanitation, change of flooring, house extensions and purchase of humidifiers and equipment to ensure child safety. Cost estimates for these adaptations were informed by information obtained from the international electronic commerce company Amazon (Amazon.com Inc., Seattle, WA, USA; UK version). Adaptations such as house extensions and redecorations were guided by estimated prices provided by private building companies. Special equipment and aids required by the LAMBS infants over the course of follow-up fell into the following broad categories: breathing support equipment, feeding/breastfeeding equipment, chairs/mattresses and room ventilation equipment. Unit costs for these items were informed by information obtained from the international electronic commerce company Amazon (UK version). The costs to parents of taking time off work to care for the infant(s) were estimated by applying gender-specific median earnings data217 to occupational classifications derived from the self-reported employment information. A list of resource values and their sources for the period between initial hospital discharge and 24 months is provided in Table 35. All costs occurring beyond the first year after birth were discounted using the UK recommended discount rate of 3.5%. 220