Preterm birth is related to individual social factors, including maternal education, occupation and ethnicity 1 2 , as well as neighbourhood social characteristics 1 3 . The influence of these factors seems to be greater for very preterm births (<32 weeks) than for moderately preterm births (32-36 weeks) 4 5 6 7 . Rates of preterm birth have increased steadily in recent decades in Europe 6 7 8 9 10 and social disparities have persisted over time 6 7 9 .

Mechanisms underlying the association between the risk of very preterm birth and social factors are not well understood 11 . Attempts to identify distinct social risk factors for subtypes of preterm births, including those associated with spontaneous very preterm labour 12 , specific maternal complications of pregnancy 13 14 15 or multiple pregnancies 16 17 , have yielded contradictory findings. Preterm birth has been also associated with a greater burden of neonatal care 18 , but little research has been carried out on whether social factors are related to very preterm infant outcomes before discharge from the neonatal unit 19 .

Wide variations in very preterm births rates exist in Europe (i.e. from 0.9% to 1.4% of total live births in France and England) 20 . Research suggests that the relationship between socio-economic deprivation and high risk births differ slightly across developed countries. For instance, the increased risk of very preterm birth in more deprived compared to less deprived neighbourhoods varied between 1.6 7 and 1.9 9 in the UK, and a two-fold risk for low birth-weight births(<2500 grams) in the US 21 . These two countries are known to have high rates of very preterm births when compared to other developed countries 20 . Cross-national comparisons make it possible to investigate whether the impact of social factors is more acute in contexts where preterm birth rates are higher. More generally, these comparisons enable exploration of variations in the social gradient in risk of preterm birth that may depend on policy or population factors 6 22 .

Cross-national comparisons of the impact of social factors on the risks and outcomes of very preterm birth have been limited because of the difficulties of defining comparable populations as well as the lack of socioeconomic information in studies on these infants. International variations in procedures for recording births and deaths at early gestational ages 23 as well as the assessment of gestational age make it difficult to compare populations of very preterm births unless a common inclusion protocol is used. Second, most data on very preterm births come from cohort studies 24 25 26 or hospital registers 27 and these do not usually include a comparison group of term births which is needed to assess the risk of preterm birth associated with social factors. Finally these studies are often based on data abstracted from medical records which contain limited information on individual social characteristics. While area-based measures offer a solution in this situation, common indicators are required; area-based measures may not be comparable because of the use of different indicators or scores of social disadvantage.

We investigated the influence of neighbourhood socio-economic status on the risk of very preterm singleton birth in two French and English regions with different very preterm rates and health systems using data from the population-based MOSAIC study collected with a common protocol and standardized inclusion and validation criteria. A second aim was to study whether the clinical characteristics of pregnancy, very preterm delivery, in-hospital mortality and short term infant health outcomes varied by neighbourhood of residence.

Table 1 shows median and interquartile range values for neighbourhood deprivation characteristics in the census population of babies “0” years of age. Sensitivity analyses comparing the socio-economic distribution of the census population of infants to birth registers yielded similar quartile cut-offs in both regions. Mothers of very preterm singletons were older in the Ile-de-France region than in Trent (Table 2). Gestational age, sex, birthweight and multiplicity were similar across the two regions. The distribution of pregnancy complications and very preterm infant outcomes differed between regions, except for hypertension during pregnancy, caesarean section before labour and in-hospital mortality.

a) Hypertension during pregnancy: pregnancy-induced hypertension, pre-eclampsia, eclampsia and HELLP syndrome.

b) In-hospital mortality includes death during labour, delivery or hospitalization after birth.

c) Includes only infants discharge alive.

d) Oxygen dependence or ventilation, including nasal continuous positive airway pressure, at 36 weeks of gestational age.

e) Brain lesions includes intraventricular haemorrhage and periventricular leukomalacia.

f) Long hospital stay: longer hospital stay than 75^th percentile for gestational age.

In both regions, the percentage of very preterm singletons increased with a higher rate of unemployment, a higher proportion of manual workers and non-home owners or a low educational level when compared to infants in the census (Table 3). More than 30% of the very preterm infants lived in the most deprived quartile compared with around 15% in the least deprived quartile for all four deprivation indicators, with statistically significant odds ratios ranging between 2.5 and 1.5 (test for trend of OR: p<.001) for most deprived compared to least deprived quartiles. There was no evidence of a difference in the magnitude of the effect of deprivation between the two areas (P values for test for interactions ranged between 0.1 and 0.9).

Test for trend of OR: p<.001 for singletons in Trent and Ile-de-France for all neighbourhood characteristics.

Table 4 compares pregnancy and delivery characteristics and infant outcomes for singleton very preterm births by unemployment rate quartiles. Older women tended to live in least deprived neighbourhoods in the Trent region; but no differences were observed in Ile-de-France. There was no evidence of a relationship between neighbourhood unemployment rate and gestational age, small for gestational age, indicated preterm delivery, haemorrhage during pregnancy or preterm labour. Hypertension during pregnancy decreased with neighbourhood deprivation in Trent but increased in Ile-de-France. The percentage of women with preterm premature rupture of membranes tended to be lower in more deprived neighbourhoods in both regions. There was no evidence of variation with neighbourhood deprivation for any of the very preterm outcomes (in-hospital mortality, bronchopulmonary dysplasia, brain lesions, ventilation and length of hospital stay) except for breastfeeding. Results were similar for the three other neighbourhood socio-economic indicators.

a) p value for bivariable analysis; b) adjusted for maternal age for pregnancy complications and caesarean section; p value adjusted for gestational age, small for gestational age and sex for infant outcomes. Small number of infants with brain lesions did not allowed calculation of aOR; c) <10th percentile of birthweight standards from the MOSAIC cohort; d) Hypertension during pregnancy: pregnancy-induced hypertension, pre-eclampsia, eclampsia and HELLP syndrome; e) In-hospital mortality includes death during labour, delivery or hospitalization after birth; f) Includes only infants discharge alive; g) Oxygen dependence or ventilation, including nasal continuous positive airway pressure, at 36 weeks of gestational age; h) Brain lesions includes intraventricular haemorrhage and periventricular leukomalacia; i) Long hospital stay: longer hospital stay than 75^th percentile for gestational age.

After adjustment for maternal age, gestational age, small for gestational age and sex, breastfeeding at discharge was lower in neighbourhoods with the highest unemployment rates (Trent: OR=0.2 (95% CI: 0.1-0.4); Ile-de-France: OR=0.4 (95% CI: 0.2-0.8)). The interaction between hypertension during pregnancy and region was not significant after adjustment. (interaction test p<0.07) (Additional file 1).

Neighbourhood socio-economic factors had a similar impact on the risk of very preterm singleton birth in two European regions with different health care systems and rates of very preterm birth. The risk of very preterm birth was doubled in the most deprived neighbourhoods compared to the least deprived. Consistent neighbourhood socio-economic gradients were not observed for any of the specific pregnancy or delivery characteristics for singleton very preterm births in the two regions and infant outcomes did not differ, with the exception of breastfeeding at discharge which was lower for infants living in more deprived neighbourhoods. These results support the hypothesis that the association between very preterm birth and socio-economic factors does not result from one single mechanism, but represents multiple disadvantages; these could include poor antenatal care, maternal stress, racism, poor health behaviours and biological factors, such as genetic predisposition, immune system changes or endocrine changes 11 30 .

The strengths of our study are its use of a common protocol to abstract data from medical records and to ensure completeness of inclusions in a geographically-based cohort of very preterm infants. To our knowledge this study is the first to look at direct socio-economic comparisons of very preterm births between countries using standardised data collection. Nonetheless, the definitions of some data items were difficult to standardise in terms of their severity or thresholds for inclusion, in particular some pregnancy complications such as maternal haemorrhage, and this may explain some of the variability in rates between the regions. Regional level differences in definitions should not, however, affect the within region association with neighbourhood socio-economic factors.

Comparable data across countries on maternal socio-economic characteristics at the individual level were not routinely available from medical records in either region as is often the case in studies based on medical records or registers. Therefore, we used census data to assigned socioeconomic position to very preterm infants according to the place of residence of the mother. Census data also allowed us to obtain a comparison group of infants which was not available in our cohort of very preterm births and to study the distribution of births by neighbourhood socio-economic characteristics. Our study shows that this methodology can be used to assess socioeconomic inequalities in a cross-national context. This methodology could be used to study other health issues at an international level for which a comparison population, i.e. in terms of age or sex, can be obtained from the census. Census data have some limitations, however. Disadvantaged families may be undercounted in the census, which would lead to an overestimate of the impact of deprivation on very preterm birth. We verified that this was not a problem in our analysis by validating the deprivation quartiles from our census population with other data sources on births in both regions. Other limits of our comparison group (census infants) are that we were unable to exclude very preterm infants and multiples from the census population, as gestational age and multiplicity are not available in the census data; but these represent a small proportion of all live births and would thus have a very minor impact on the birth population deprivation quartiles. Infants who died in the first days of life will most likely not be included in the census, although these infants are a very small proportion of total births.

Our results were robust despite differences in the definitions used to report census data on socio-economic characteristics in England and France. The magnitudes of the risk of very preterm births for singletons in most deprived versus least deprived neighbourhoods were quite similar for Trent and Ile-de-France, and slightly higher than those reported previously in Europe 7 9 using deprivation scores. The measures chosen include components of validated area-level composite deprivation scores in the UK (the Carstairs score and the Townsend Index) and in France 31 , although the scores do not include the same social characteristics (education is not included in the UK scores and home ownership is not included in the French score). Our study computed quartiles within each region, because of the differences in census definitions and since baseline values of our neighbourhood variables did not allow us to define common thresholds. The literature on neighbourhood socio-economic deprivation relies primarily on relative rather than absolute levels of deprivation. Use of relative deprivation measures may be less valid when inequality between regions is very different; however in the Trent and Ile-de-France, there were similar relative gaps between the lowest and the highest quartiles for the four socio-economic characteristics used. These various methodological challenges lead us to analyze the four variables separately and not attempt to develop a common deprivation score. However, given the consistency of our results across the indicators, a composite indicator looking at multi-dimensional deprivation would very likely give a similar or stronger effect. Further work would be needed, however, to validate this approach.

We found no systematic differences between pregnancy and delivery characteristics of very preterm births and neighbourhood of residence in the two regions. This finding corroborates previous studies which found no socio-demographic gradient for spontaneous delivery or PPROM for very preterm births in Australia 12 and no association of preterm birth with maternal complications (i.e. hypertension, preeclampsia and vaginal bleeding) during pregnancy in England 32 . In contrast, spontaneous very preterm deliveries in deprived neighbourhoods, particularly those before 29 weeks, have been associated with more fetal and maternal infections in the UK 15 . We found opposite associations between socio-economic characteristics and hypertension during pregnancy: lower rates among women living in deprived neighbourhoods in Trent and higher rates in Ile-de-France. Hypertensive complications have been reported to be more frequent among socially disadvantaged women 13 14 . In Ile-de-France, higher rates may be partly explained by higher proportions of women from sub-Saharan Africa who have been found to be at risk of hypertensive diseases during pregnancy and who live in poorer neighbourhoods 33 . It is not clear why rates were lower in Trent, but the distribution of maternal age might be an explanation if older women who are at higher risk of hypertension morbidity during pregnancy live in more affluent neighbourhoods. Another hypothesis could be under diagnosis of hypertensive diseases in women with inadequate antenatal care who live in less affluent neighbourhoods, such as young women.

Very preterm outcomes varied between the two regions, but they did not vary by neighbourhood socio-economic characteristics within each region, with the exception of breastfeeding. Similar results have been shown recently for survival, respiratory support, length of stay among very preterm infants 19 and the risk of cerebral palsy among very low birth-weight infants 34 . This suggests that once admitted to hospital equitable care is provided. Differences in breastfeeding at discharge may indicate, however, that socio-economic factors affect some aspects of care in the neonatal intensive care unit, in particular where parental involvement is required. Other studies have found that mothers of lower social class are less likely to breastfeed their high risk infants 35 36 37 . These mothers might experience greater difficulties establishing and maintaining lactation which requires acquiring knowledge about milk expression and storage procedures, having access to milk pumps and time for frequent expression and transport of milk to the neonatal unit. This difference also suggests that there may be differences in care after discharge and in longer term outcomes.

Despite the absence of an association between neighbourhood socio-economic status and infant mortality and morbidity, the total burden of mortality and respiratory and neurological morbidity will be greater in deprived areas because of their higher rates of preterm birth. In addition, the total burden of these socio-economic differences may be greater in England, where rates of very preterm births are higher, than in France 38 . The neonatal care load may also be greater in hospitals in deprived catchment areas.

In conclusion, trends in the risk of very preterm birth according to area-level socio-economic characteristics were robust in two European regions with different health care systems and rates of very preterm birth leading to a greater health burden in deprived areas despite the absence of socio-economic disparities in mortality and short-term morbidity. Planning for neonatal care provision, social support services for families and follow-up programmes needs to consider the socio-spatial distribution of very preterm children. Extending the methodology used in this paper to other countries with access to area-based census data would make it possible to test whether these trends are similar in other settings.

The authors declare that they have no competing interests.

Analysis and interpretation of the data were carried out by MB, LS and JZ. MB wrote the first draft of the paper and all authors participated in revisions and approved the final manuscript. All authors read and approved the final manuscript.