Ethnic differences in singleton preterm birth in England and Wales, 2006‐12: Analysis of national routinely collected data

Abstract Background Data recorded at birth and death registration in England and Wales have been routinely linked with data recorded at birth notification since 2006. These provide scope for detailed analyses on ethnic differences in preterm birth (PTB). Objectives We aimed to investigate ethnic differences in PTB and degree of prematurity in England and Wales, taking into account maternal sociodemographic characteristics and to further explore the contribution of mother's country of birth to these ethnic differences in PTB. Methods We analysed PTB and degree of prematurity by ethnic group, using routinely collected and linked data for all singleton live births in England and Wales, 2006‐2012. Logistic regression was used to adjust for mother's age, marital status/registration type, area deprivation and mother's country of birth. Results In the 4 634 932 births analysed, all minority ethnic groups except ‘Other White’ had significantly higher odds of PTB compared with White British babies (ORs between 1.04‐1.25); highest odds were in Black Caribbean, Indian, Bangladeshi and Pakistani groups. Ethnic differences in PTB tended to be greater at earlier gestational ages. In all ethnic groups, odds of PTB were lower for babies whose mothers were born outside the UK. Conclusions In England and Wales, Black Caribbean, Indian, Bangladeshi, Pakistani and Black African babies all have significantly increased odds of being born preterm compared with White British babies. Bangladeshis apart, these groups are particularly at risk of extremely PTB. In all ethnic groups, the odds of PTB are lower for babies whose mothers were born outside the UK. These ethnic differences do not appear to be wholly explained by area deprivation or other sociodemographic characteristics.


| BACKG ROU N D
Complications of preterm birth are the largest direct cause of neonatal deaths worldwide, accounting for 35% of 3.1 million neonatal deaths in 2010. 1 Babies born preterm who survive are also at higher risk of short-term and long-term morbidities, including neurodevelopmental impairments and respiratory and gastrointestinal complications, compared with babies born at term. 2 Much of the research on ethnic differences in preterm birth has been carried out in the United States. This has shown that compared with White mothers, the risks of preterm birth are consistently higher for Black mothers [3][4][5] but similar or lower for East Asian and Hispanic mothers. 2,4 Health disparities between ethnic groups have also been observed in England and Wales. [6][7][8] In particular, the risk of preterm birth, especially very preterm birth, is high among mothers of Caribbean and West African origin compared with White mothers. [9][10][11][12][13][14][15] South Asian groups have consistently high levels of low birthweight, whereas the risks of preterm birth in South Asian groups compared with the White group have been less consistent. [9][10][11][12][13][14][16][17][18][19][20] For example, some studies found higher risks, 11,14,16 some found similar risks 12,13,18 and others found mixed risks for specific Indian, Pakistani and Bangladeshi groups. 9,10,17,21 There have been relatively few studies from other high-income countries. [21][22][23][24][25][26] Although a number of socio-economic, obstetric and genetic explanations have been proposed to explain these differences, [1][2][3]16,23,27,28 the mechanisms or reasons behind them are not well understood.
Previous studies in the United Kingdom (UK) have some limitations. Studies based on regional data have provided important insights but have generally not included all ethnic groups, others have used parents' countries of birth or presented only crude rates because of small numbers. [9][10][11][12][13][14][16][17][18][19][20]29 National data about gestational age and ethnic group in England and Wales became available after 2005. 10,30 Data are now available for a longer period so larger numbers of births provide scope for more detailed analyses. The specific objectives of the study were as follows: (a) to investigate ethnic differences in preterm birth and degree of prematurity in England and Wales, taking into account maternal sociodemographic characteristics; and (b) to further explore the contribution of mother's country of birth to these ethnic differences in preterm birth.

| Data sources
This was a population-based study using routinely collected and Birth and death registration data are routinely checked by ONS.
However, ONS does not routinely exclude all implausible combinations of gestational age and birthweight. We therefore further checked and cleaned the data extract by excluding records with missing values or implausible combinations of gestational age and birthweight by sequentially removed births with: (a) a gestational age greater than or equal to 43 completed weeks; (b) a missing birthweight; (c) an implausible combination of birthweight and gestational age, defined as birthweight more than twice the interquartile range above or below the median birthweight of the sex-gestation-ethnic group-specific stratum of the study dataset.

| Ethnic group and other explanatory variables
Staff notifying a baby's birth to the birth notification system are asked to record the baby's ethnic group as reported by the mother based on the ethnic categories used in the 2001 Census in England and Wales. 10 These ethnic categories are based on three most common self-defined groups in the UK ('White', 'Asian' and 'Black') with several country or regional categories within each of them, a 'Mixed' group with several Mixed background categories, and a 'Chinese or other' group. 31 These ethnic group categories were then recoded for analy- Other explanatory variables included the baby's sex and year of birth, age of mother, area deprivation score of the mother's area of residence as coded according to the 2015 English Index of Multiple Deprivation (IMD) 32 and the 2014 Welsh IMD, 33 mother's country of birth, parents' marital status/registration type and mother's country of residence.

| Outcome measures
The primary outcome was preterm birth, defined as the live birth of a baby at less than 37 completed weeks of gestation. Our outcome measure was the percentage of live births that were preterm.

| Statistical analysis
Logistic regression models were used to explore ethnic differences in preterm birth using the White British group as the reference group.
We adjusted all models for the baby's sex and year of birth to account for differences between sexes and over time (models adjusted for these two variables only are referred to as the 'base model' hereafter).
Further adjustments were made (referred to as 'adjusted model') to account for IMD quintiles, age of mother (under 18, 18-19, 20-24, 25-29, 30-34, 35-39, 40 and over) and parents' marital status/registration type (married, joint registration/same address, joint registration/ different address, sole registration). As IMD scores are constructed differently in England and in Wales, we adjusted for the mother's country of residence (England vs Wales) in all models that included IMD.
Multinomial logistic regression was used to investigate the association between ethnic group and degree of prematurity, treating degree of prematurity as a categorical outcome variable and adjusting for the same covariates as above.
To further explore the contribution of mother's country of birth to ethnic differences in the odds of preterm birth, we then included mother's country of birth (inside versus outside the UK) in the adjusted model, tested the interaction between ethnic group and mother's country of birth, and present stratified results as appropriate. Finally, to assist in the interpretation of findings, we carried out additional exploratory analyses in which we adjusted for covariates (mother's country of birth, IMD quintiles, age of mother and parents' marital status/registration type) individually. All analyses were conducted in STATA version 13.

| Characteristics of the mothers and babies
Ethnicity was strongly associated with characteristics of mothers and babies ( The absolute risk of preterm birth was also strongly associated with these characteristics, having decreased over time and plateaued in 2010. The risk was highest in male babies, in the youngest and oldest groups of mothers, those living in the most deprived areas, mothers born in the UK and mothers who were sole registrants or jointly registered as living at different addresses from the baby's father; these results were confirmed using adjusted odds ratios (Table S1).

| Ethnic group and degree of prematurity
Compared with White babies, babies of other ethnic backgrounds were generally more likely to be born at earlier gestational ages.
Rates of extremely preterm birth were particularly high among babies of Black Caribbean and Black African origin (Table 1).
In the adjusted multinomial model, the differences compared with the White British group were greater at earlier gestational ages for all groups except Other White and Bangladeshi groups ( Figure 2).

The differences were most marked among Black African and Black
Caribbean babies for whom the ORs were around 1.5 for being born at 28-31 weeks and increased to around 2.6 for being born at 22-27 weeks. The differences were less marked for Indian and Pakistani babies.

| Baby's ethnic group, mother's country of birth and preterm birth
As we found strong interaction (P < .001) between ethnic group and whether the mother's country of birth was inside or outside the UK, we further stratified the study population into 18 subcategories by ethnic group (9 groups) and mother's country of birth, inside or outside UK, and compared the odds of preterm birth using White British babies born to mothers born in the UK as the reference group. Figure 3 and Table S2 show that consistently, within ethnic group, UK-born mothers had higher odds of preterm birth than non-UK-born mothers. The 'univariable' analysis adjusting for baby's sex, year of birth and each covariate individually also show that for all ethnic groups except Other White, adjusting for mother's country of birth increases the magnitude of the association between ethnicity and preterm birth (Table S3). Exploratory analysis describing the characteristics of the study population by ethnic group and mother's country of birth (Table S4) shows that for ethnic minority groups, in general, although non-UK-born mothers were more likely to live in the more deprived areas compared with UK-born mothers, they were also less likely to be younger mothers and more likely to be married or live at the same address with their partners.

Odds ratio
Adjusted odds ratio (95% confidence interval) for preterm birth Asian groups. Ethnic differences in preterm birth tend to be greater at earlier gestational ages. In all ethnic groups, the odds of preterm birth are lower for babies whose mothers were born outside the UK.

| Strengths of the study
The complete coverage of birth registration data, the large number of births and the low proportion with missing data mean that selection bias, detection and observer bias are minimized. The dataset also included sufficient numbers of births in the smaller ethnic groups, such as Black Caribbean and Bangladeshi groups, to enable disaggregated analyses.

| Limitations of the data
There is limited information available about the quality of ethnicity recording although it is possible that in some cases the mother may record her own ethnic group or that health care staff may record the mother or baby's ethnic group rather than asking the mother to report it. 10 Although some misclassification of ethnic group is possible, previous analysis suggests that this may predominantly affect the Mixed ethnic group 10 which was not the focus of our study. The availability of data about mother's country of birth linked to data about ethnicity enabled us to study country of birth and ethnicity separately, and this is a strength of our analysis.
We lack data on how gestational age was measured and the accuracy of the measurement, but the birth notification data that we used are regarded as the best source for national data on gestational age. 34 Birth attendants are asked to record data about gestational age based on last menstrual period in the birth notification system, but it seems probable that they were taken directly from woman's medical records and in the majority of cases, this would have been based on more accurate ultrasound dating. Information about clinical subtypes of preterm birth, that is spontaneous, preterm premature rupture of membranes and medically indicated (elective), is not available from the registration and birth notification systems. We also lack data about pregnancy history and other socio-economic or lifestyle factors that may be related to preterm birth 2 and differ between ethnic groups. IMD is based on the baby/mother's postcode of usual residence and may not capture the exposure during pregnancy for all women. However, IMD is an area-based measure of deprivation that is widely used in health research despite its limitations 35 and was used for this particular study because individual socio-economic status is available for only a 10% sample of births whereas IMD provides a measure of socio-economic disadvantage available for all births.
Although these national-level results are applicable to the populations of England and Wales, the different compositions of ethnic groups in other countries may limit international generalisability. In particular, we have used a combined 'Black African' group which consists of a diverse range of people originating from a whole continent within which there are wide differences in risks of preterm birth. 15 The generalisability to other parts of the world may also be affected by differences in health care access.

| Interpretation
Findings of higher odds of preterm birth in the Black groups are consistent with previous studies. [9][10][11][12][13][14][15] Area deprivation and marital status have been suggested to explain half of the excess risk of preterm birth in Afro-Caribbean but not African mothers. 12  Higher odds of preterm birth identified in all the three South Asian (Indian, Pakistani and Bangladeshi) groups in our study have been less well established. [9][10][11][12][13][14][16][17][18][19][20][21]24 Previous studies in which South Asian groups were aggregated showed inconsistent results with some showing higher risks 11,14,16,24 and others finding no increase in risks. 12,13,18 However, as South Asian groups are heterogeneous, results based on aggregated analysis may be misleading.
In studies that disaggregated South Asian groups, higher risks of preterm birth were sometimes only found in Indian 9,17 and/or Pakistani groups 10,21 but not all groups. 9,10,17,21 These disaggregated analyses typically have used groups with smaller numbers than ours and some differ from the national population in their ethnic composition.
It is plausible that higher odds of preterm birth in the South Asian groups are due to a higher prevalence of medical conditions that are associated with medically indicated preterm birth in these groups, but this is not something we were able to investigate. A UK study found being South Asian was a risk factor for medically indicated but not spontaneous delivery before 34 weeks. 18 Shorter gestational age and a non-significant increase in the risk of preterm birth have also been observed in Indo-Asian women in England compared with White British among women with chronic hypertension. 29 It is also plausible that South Asian fetuses, similar to Black fetuses, mature earlier in general compared with White fetuses. 11,16 Our finding that ethnic differences in the odds of preterm birth tend to be greater at earlier gestational ages has been less well studied as most studies only use one cut-off for preterm birth. However, our results are broadly consistent with findings from the few studies that have used different cut-off points for preterm birth to explore the changes in ethnic variation for Black groups. 12,[38][39][40][41] In particular, a study investigating North Birmingham women delivering singletons also showed that in Afro-Caribbean and African women, the risk of preterm birth was particularly high for delivery before 34 and 28 weeks. 12 However, the risks in South Asian women compared with White groups were not higher at earlier gestational ages in those studies. 12,40 We cannot provide an explanation of the ethnic differences in the increased odds of preterm birth at earlier gestational ages. It is worth exploring this phenomenon and potential mechanisms in other studies as very preterm births (births before 32 weeks' gestational age) account for the majority of neonatal deaths and serious morbidity. 27 Another important finding of our study is that within each ethnic group, babies of non-UK-born mothers had lower odds of preterm birth than those of UK-born mothers. The implications of this are likely to vary as the proportion of minority ethnic mothers born outside the UK varies. Differences in mother's country of birth by ethnic group, as seen in our study, may reflect differences in migration history, length of residence in the host country and degree of acculturation. 42 However, according to our exploratory analysis, in general, although non-UK-born mothers were more likely to live in the more deprived areas compared with UK-born mothers, which in turn would increase their odds of preterm birth, they were also less likely to be younger mothers and more likely to be married or live at the same address with their partners which are associated with lower odds for preterm birth. The finding of lower odds of preterm birth in non-UK-born mothers is broadly consistent with results from other studies that showed foreign-born mothers in ethnic minority groups had similar or lower risks of preterm birth compared with their US-born or UK-born counterparts. 15,22,43,44 The healthy migrant effect and the further loss of this effect after one generation have been suggested as a potential explanation. 5 Evidence also suggests that being able to maintain cultural links or live in areas with higher proportion of people from the same ethnic group may outweigh other hurdles and contribute to better well-being in general or pregnancy outcomes in particular in ethnic minority groups. 45,46 Unfortunately, we were unable to further explore these factors because of a lack of relevant data in our dataset.

| CON CLUS IONS
In England and Wales, Black Caribbean, Indian, Bangladeshi, Pakistani and Black African babies all have increased odds of being born preterm compared with White British babies. Bangladeshis apart, these groups are particularly at risk of extremely preterm birth. In all ethnic groups, the odds of preterm birth are lower for babies whose mothers were born outside the UK. These ethnic differences do not appear to be wholly explained by area deprivation or other sociodemographic characteristics; therefore, other factors should be explored, including the causes and subtypes of preterm birth, particularly the contribution of medically indicated preterm birth. The higher odds of preterm birth observed in UK-born compared with non-UK-born mothers in all ethnic groups may point to risk factors that increase with acculturation.

ACK N OWLED G EM ENTS
We thank Bliss, the Race Equality Foundation, Sands, and the Lullaby Trust for their helpful involvement in the Patient and Public Involvement (PPI) consultation for this project and in particular for their guidance on the ethnic groupings to be used in this study. This paper reports on an independent study which is funded by the NIHR Policy