Preterm birth and maternal country of birth in a French district with a multiethnic population
Dr J. Zeitlin, INSERM U149, 123 Boulevard Port Royal, 75014 Paris, France.
Objectives This analysis explores the association between preterm birth and maternal country of birth in a French district with a multiethnic population.
Design Prospective observational study.
Setting District of Seine-Saint-Denis in France
Population 48,746 singleton live births from a population-based birth register between October 1998 and December 2000.
Methods We compare preterm birth rates by mother's country of birth controlling for demograhic and obstetric factors as well as insurance coverage and timing of initiation of antenatal care.
Main outcome measures Overall preterm birth rates and preterm birth rates by timing of delivery (<33 weeks versus 33–36 weeks of gestation), mode of onset (spontaneous or indicated preterm birth) and the presence of hypertension in pregnancy.
Results Women born in Northern Africa, Southern Europe and South/East Asia did not have higher preterm birth rates than women born in continental France. Rates were significantly higher for women born in the overseas French districts in the Caribbean and Indian Ocean and Sub-Saharan Africa. Excess risk was greatest for early preterm births, medically indicated births and preterm births associated with hypertension.
Conclusions Patterns of preterm birth with relation to timing, mode of onset and medical complications among of Afro-Caribbean origin should be confirmed in future research.
Ethnic origin and country of birth have been associated with the risk of preterm birth in many settings.1–6 Understanding this association is important for improving perinatal outcomes in minority populations and gaining insight into the aetiology of preterm birth. This analysis explores the relationship between country of birth and preterm delivery in the French district of Seine-Saint-Denis, where almost half of women of childbearing age are born outside France. A study undertaken in the early 1990s by the maternal and child protection agency found that perinatal mortality was higher for women born outside of France, with the risk being almost doubled for women from the French overseas departments in the Caribbean and Indian Ocean and women from Sub-Saharan Africa and a 50% greater risk for women from Northern Africa and Southern Europe.7,8 After this study, the district put into place a routine information system on births that included information about mother's country of birth,9 making it possible to evaluate the association with preterm birth overall and for subgroups of preterm births defined with regard to timing of delivery (whether early or moderate preterm birth), the mode of onset of delivery (whether the preterm birth was spontaneous or indicated) and the medical complications associated with the preterm birth.
This analysis used data on 48,746 singleton live births that occurred in Seine-Saint-Denis from October 1998 to December 2000 to assess the association between mother's country of birth and preterm birth.
The Seine-Saint-Denis register includes all births from 22 weeks of gestation in the 20 maternity units in the district. Information on the mother, her medical and obstetric history, the current pregnancy, the delivery and the newborn(s) are recorded in the certificat du 8ème jour, a health certificate required by law for all live births in France. The administrative section of the certificate is completed by the mother, while data on the pregnancy and the delivery are filled in by the midwife in the delivery room using the mother's medical records. An experimental version of this certificate with additional data items, including the mother's country of birth, was introduced in the district between 1995 and 1998, with full coverage starting in October 1998. Investigators from the maternal and child services visit every maternity unit in the district to verify the completeness of ascertainment by comparing the list of certificates received with the birth registers in the delivery wards. A separate certificate also collects data on all stillbirths and is verified by the investigators in the same way. Although stillbirths are not included in this analysis, the Seine-Saint-Denis database includes all births from 22 weeks of gestation and over, which is important for the completeness of registration of births at early gestations.
The primary dependent variable is preterm birth, defined as a duration of gestation less than 37 completed weeks. Gestational age is based on clinical and ultrasound data in the first trimester. Ninety percent of women have an ultrasound in the first trimester. Gestational age was missing for only 121 cases (0.25% of the sample). Preterm births are divided into early (<33 weeks) and moderate (33 to 36 weeks) and into spontaneous (either preterm labour or premature rupture of membranes) and/or indicated (if a caesarean section was carried out before the onset of labour or if labour was induced for reasons other than premature rupture of membranes).10 Medical complications of pregnancy associated with preterm birth were ascertained from the certificate which includes specific questions on hypertension, pre-eclampsia and eclampsia, preterm labour, and whether growth restriction was detected by ultrasound and the reasons for antenatal hospitalisation. Small for gestational age preterm infants were identified using the 10th centile of French population reference standards which is only available for births over 27 weeks of gestation.11
The primary independent variable is mother's country of birth. In France, special authorisation from the French data protection authority, la commission nationale informatique et libértés, is necessary to create a database which includes information on country of origin. Questions that can be included are only mother's country of birth or current nationality; questions on ethnicity or race are not allowed. The variable used in the Seine-Saint-Denis database is country of birth, which is a proxy for ethnic origin for women who were born outside of continental France. The group of women born in France includes women of French origin and women of other ethnic origins whose parents or grandparents immigrated to France.
Mother's country of birth is coded into 11 groups of which six are included in this analysis: continental France, the French overseas districts (mainly women from the Caribbean islands of Guadeloupe, Martinique and La Réunion in Indian Ocean, whose populations are primarily African in origin), Southern Europe (Yugoslavia, Portugal, Spain, Italy, Greece), Northern Africa (Algeria, Morocco, Tunisia), Sub-Saharan Africa and South/East Asia (mainly women from Southern Asia—Pakistan, India, Bangladesh, Sri Lanka, but also including women from Eastern Asian countries). The number of women born in other countries of Europe, the Middle East, the Americas and Australia are too small to permit analysis. The country groupings are the same as those used for the study of perinatal mortality in 1989–1992.
Covariates include maternal age, parity, obstetric history (whether the mother had a previous preterm birth, growth restricted birth or previous perinatal death), antenatal care (timing of first antenatal visit, number of visits, number of ultrasounds and hospitalisation during pregnancy) and health insurance status. Health insurance status measures whether the woman is covered under the national health insurance plan; all women were eligible for medical coverage during pregnancy in this period through supplementary district health programs.
The population includes all 48,746 singleton live births that occurred in Seine-Saint-Denis from October 1998 to December 2000. From this total, we excluded 3825 births for which country of birth was unknown (7.9% of the sample) as well as women born outside of the six country of birth groups (n= 1348, 2.8% of sample). Unknown cases had a higher rate of preterm birth (7.5%) than those with data on country of birth (5.2%). This reflects the fact that emergency cases tended to have higher levels of missing data, especially when the delivery occurred at another maternity unit than initially intended.
To measure the influence that cases with unknown country of birth could have on our results, we carried out a sensitivity analysis using the final multivariate model. We imputed all cases with missing information on country of birth to each country of birth grouping in turn to see if it would significantly change the adjusted odds ratio for that group. This exercise evaluates the maximum impact that these cases could have on our estimates.
Other items have varying levels of completeness. Recording is complete for information that is systematically included in medical records, such as gestational age, birthweight and timing of first antenatal visit (due to a legal requirement that the pregnancy be declared at first visit so that the maximum family allowance can be collected). In contrast, maternal date of birth is filled in by the mother and is missing in 14% of the sample. Other items depend on the completeness of recording in the medical records. The percent missing is reported in descriptive tables. For multivariate analyses, missing responses are included as a separate category. Because data are missing for a variety of reasons, including in some cases because of an emergency at birth, these coefficients do not have a clear interpretation and are not reported.
The impact of country of birth on preterm delivery was estimated using logistic regression. Further analyses that distinguish between preterm births by timing, mode of onset and presence of hypertension use multinomial logistic regression. Equality of coefficients for the subgroups of preterm births was tested using the Wald statistic, constructed from the estimated covariance matrix. Statistical analysis was conducted using Stata Statistical Software (Release 7.0, Stata, College Station, Texas, USA).
Table 1 presents maternal characteristics and care during pregnancy and delivery by country of birth for all singleton live births. Missing cases are excluded from the calculation of percentages, but the percent missing is provided below each item in parentheses. Because of the large sample sizes, small differences between groups achieve statistical significance and, thus, almost all differences between the characteristics of women born in France and women born elsewhere are significant. For this reason, we have opted not to present significance levels in these tables.
Table 1. Maternal characteristics and care during pregnancy and delivery by country of birth, singleton live births.
|National health insurance|
|First antenatal visit|
|Less than 4 antenatal visits||3.5||5.6||5.9||6.5||9.2||6.0|
|No. of antenatal visits (mean)||8.3||8.2||8.0||8.0||7.2||7.8|
|No. of scans (mean)||3.4||3.5||3.4||3.4||3.4||3.3|
Mothers born in continental France were less likely to be under 20, were more likely to be having their first child and had fewer previous adverse obstetric outcomes than women born elsewhere. Women from Sub-Saharan Africa had the most distinct age and parity profile with a large number of women with four or more previous deliveries. Women born outside of continental France had lower insurance coverage and were more likely to have had inadequate antenatal care measured by late initiation of care and fewer than four visits. Women from Sub-Saharan Africa were least likely to have insurance and most likely to have had inadequate antenatal care, followed by women from South/East Asia. Despite these differences, however, a large majority of women from all groups received timely antenatal care, as measured by the time of first visit, and received care in line with overall norms for antenatal care in France, as measured by the number of visits and the number of ultrasound scans. Rates of hospitalisation during pregnancy were higher for women born outside of continental France, with the highest proportion among women from the French islands in the Caribbean and Indian Ocean, followed by women from Sub-Saharan Africa.
Preterm birth rates by country of birth are given in Table 2. Women from continental France had the lowest preterm birth rates (4.6%), but were closely followed by women from North Africa. Women from Southern Europe and South/East Asia, have singleton preterm birth rates of about 5%. None of these rates differed significantly from those for women born in continental France. In contrast, women from the overseas French districts as well as women born in Sub-Saharan Africa had significantly higher rates of preterm birth: 7.9% and 7.2%, respectively. Adjusting for demographic and obstetric factors as well as insurance coverage and timing of initiation of antenatal care attenuated the risk for women from Sub-Saharan Africa (OR from 1.6 to 1.3), but did not affect the other results.
Table 2. Association between preterm birth and mother's country of birth, singleton live births.
|French Caribbean/Indian Ocean||7.9||1.77 (1.47–2.13)||1.62 (1.35–1.96)|
|Southern Europe||5.0||1.09 (0.91–1.31)||1.04 (0.87–1.26)|
|Northern Africa||4.7||1.02 (0.90–1.15)||0.94 (0.83–1.07)|
|Sub-Saharan Africa||7.2||1.61 (1.43–1.81)||1.36 (1.20–1.55)|
|South/East Asia||5.1||1.12 (0.93–1.34)||1.04 (0.86–1.24)|
Sensitivity analyses, carried out to evaluate the impact of missing country of origin using this model, did not find important differences in odds ratios for any of the subgroups. The assumption that unknown cases were from French overseas districts or from Sub-Saharan Africa slightly attenuated odds ratios, although they remained significantly different from 1, whereas for the other groups OR were increased, but not to significant levels (Southern Europe, 1.12; Northern Africa, 1.02; and South/East Asia, 1.15).
Table 3 presents preterm birth rates by timing and onset of preterm birth. The excess risk for women from the French Carribean and Indian Ocean and from Sub-Saharan Africa was highest for early preterm versus later preterm births—rates over two times those observed for the other country of birth groups—as well as for indicated versus spontaneous preterm births. Differences are significant statistically, but less marked for preterm births between 33 and 36 weeks of gestational age and for spontaneous preterm births (OR of 1.4 and 1.5). These patterns persist after adjustment. The Wald test for equality of coefficients shows that the risks associated with country of birth differ by timing of the birth and mode of onset.
Table 3. Association between mother's country of birth and preterm births defined in relation to timing of preterm birth and onset, singleton live births.
|French Caribbean/Indian Ocean||2.3||2.67 (1.91–3.78)||5.5||1.52 (1.22–1.88)|
|Southern Europe||1.0||1.15 (0.77–1.71)||4.0||1.07 (0.88–1.32)|
|Northern Africa||1.0||1.12 (0.86–1.46)||3.7||0.99 (0.87–1.14)|
|Sub-Saharan Africa||2.1||2.43 (1.93–3.05)||5.1||1.39 (1.21–1.59)|
|South/East Asia||0.8||0.94 (0.61–1.44)||4.3||1.16 (0.95–1.41)|
|Test of equality between coefficients|| || ||P < 0.001|| |
| ||Onset of preterm birth|
|Spontaneous preterm birth %||Adjusted ORa (CI)||Indicated preterm birth %||Adjusted ORa (CI)|
|French Caribbean/Indian Ocean||4.6||1.34 (0.91–1.38)||2.9||2.91 (2.14–3.95)|
|Southern Europe||3.9||1.12 (0.91–1.38)||1.1||1.06 (0.73–1.56)|
|Northern Africa||3.3||0.96 (0.83–1.11)||1.2||1.15 (0.90–1.47)|
|Sub-Saharan Africa||4.7||1.38 (1.19–1.58)||2.3||2.26 (1.82–2.80)|
|South/East Asia||3.8||1.09 (0.89–1.35)||1.1||0.99 (0.67–1.46)|
|Test of quality between coefficients|| || ||P < 0.001|| |
| ||Preterm birth associated with hypertension|
|Yes (%)||Adjusted ORa (CI)||No (%)||Adjusted ORa (CI)|
|French Caribbean/Indian Ocean||2.5||3.98 (2.80–5.64)||5.1||1.25 (0.99–1.57)|
|Southern Europe||0.7||1.20 (0.74–1.93)||4.2||1.00 (0.82–1.23)|
|Northern Africa||0.5||0.67 (0.46–0.98)||4.1||0.99 (0.87–1.14)|
|Sub-Saharan Africa||1.9||2.55 (1.92–3.40)||5.1||1.18 (1.02–1.37)|
|South/East Asia||0.7||1.14 (0.71–1.83)||4.3||1.06 (0.68–1.29)|
|Test of equality between coefficients|| || ||P= 0.001|| |
The descriptive analysis of medical complications associated with preterm births found that women born in Sub-Saharan Africa and in the French Caribbean and Indian Ocean who experienced a preterm birth had significantly higher rates of hypertension than women born in continental France: 26.4% and 32.4%versus 12.2%, respectively. There were few other significant differences between continental France and the other countries of birth for other pregnancy complications associated with preterm birth (diagnosis of IUGR without hypertension, birthweight under the 10th centile, preterm labour and preterm premature rupture of membranes). The exceptions were that women born in Northern Africa were less likely to have infants with birthweights under the 10th centile than women born in France (14.5%vs 19.7%), while women from Sub-Saharan Africa had higher rates of SGA (25.1%). Preterm premature rupture of membranes was lower for women from Sub-Saharan Africa (26.2%vs 34.3%).
An additional analysis, presented in Table 3, explored in more depth the association between country of birth and preterm births complicated by hypertension. Preterm births associated with hypertension were more frequent among women from the French Caribbean and Indian Ocean and Sub-Saharan Africa than among women born elsewhere, as already noted in the descriptive analysis of pregnancy complications. Adjusting for covariates, these groups were at much higher risk of experiencing this type of preterm birth. However, even among preterm births not associated with hypertension, these groups still had a slight excess risk of preterm birth, although for women from the French Caribbean and Indian Ocean, this association was no longer significant.
Maternal country of birth was associated with an increased risk of preterm birth in Seine-Saint-Denis, but only for women from the overseas French districts in the Caribbean and Indian Ocean and from Sub-Saharan Africa. Women born in Northern Africa, Southern Europe and South/East Asia did not have a higher risk of preterm birth than women born in continental France. The singleton preterm birth rate observed for the group of women born in France was the same as the French national rate (4.7%).12 The excess risk for women from the Caribbean and Sub-Saharan Africa was particularly pronounced for preterm deliveries before 33 weeks of gestation, for medically indicated births and for preterm births associated with hypertension. These differences persisted after controlling for demographic and obstetric risk factors, health insurance coverage and trimester of initiation of antenatal care.
The Seine-Saint-Denis birth register is a strong source of information on preterm birth because it is population based and has high quality recording of gestational age. However, some data items are not recorded for all births. In particular, country of birth was missing for 7.9% of the sample. Sensitivity analyses found that excluding these missing cases did not create a sufficient bias to affect our conclusions. Another weakness of this database is the absence of information on socio-economic characteristics, such as education or occupational level. Socio-economic status is a known predictor of risk of preterm birth.13 The one available measure of socio-economic status is health insurance coverage.
Confounding by socio-economic factors cannot be the sole explanation for the variation in risk linked to mother's country of birth, however. The population with the highest risk of preterm birth—women born in the French districts of the Caribbean and Indian Ocean—has a higher socio-economic status than foreign-born populations, as measured by occupational status and education8 as well as insurance coverage. Conversely, women from North Africa and Southern and Eastern Asia, populations with lower levels of income, education and insurance coverage than women born in continental France, did not have an increased risk of preterm birth. Nonetheless, socio-economic status is lower among the other higher risk group: women born in Sub-Saharan Africa.
Our findings are supported by other research. Studies from the Netherlands, Sweden, Norway and the United Kingdom have found increases in the risk of preterm delivery of between 1.4 and 1.8 for black women from the Caribbean and Sub-Saharan Africa.1–4,14 In the United States, black women have almost two times the risk of white women.15 In contrast, some minority populations have not been found to be at higher risk of preterm birth, despite a greater prevalence of socio-economic risk factors. In particular, women from Northern Africa have not been found to have an increased risk of preterm birth and have heavier babies overall.1,16–18 Results for women from South and East Asia in European studies are not consistent,1,14,17 however, the geographic zone covered by this category includes a large group of countries and proportional representation from individual countries differs in Europe. These discrepancies underline the importance of defining homogenous groups for comparative analyses. In the United States, the risk of preterm birth has been shown to be lower among first generation immigrants from East Asia and Mexico than in the white non-Hispanic population.18,19
Identifying the explanations for the associations between ethnic origin and preterm birth is complex. Country of birth and ethnicity represent various and interconnected social, cultural, temporal, biological and geographical constructs.20,21 While immigrant and minority populations face poorer socio-economic conditions overall, other cultural factors may mitigate these. Researchers have suggested, for instance, that stronger social support networks contribute to lower preterm birth rates observed for immigrant Hispanic women in the United States.22 Alternatively, individuals in better health may be more likely to immigrate, creating a selection bias.19
Our results showing higher risks associated with early preterm birth indicated, preterm birth and preterm birth associated with hypertension suggest that the medical risk factors for preterm birth differ by country of birth. Women delivering preterm who were born in Sub-Saharan Africa and in the French islands in the Caribbean and Indian Ocean had a higher prevalence of preterm births associated with hypertension during pregnancy. The higher prevalence of hypertension may be one of the reasons for the higher rates of indicated preterm births and early preterm births. Hypertension does not, however, explain all the difference in preterm birth rates, as there is still a raised risk for preterm births not associated with hypertension.
A stronger association of some risk factors with early versus later preterm birth has been noted in other studies. For instance, maternal age, obstetric history and underprivileged social status were stronger predictors of preterm births before 33 weeks versus those between 34 and 36 weeks.23 This same effect is found in our analysis and could reflect a higher sensitivity of risk of early preterm births to social and biological factors.
Other European studies of preterm birth and ethnic origin have not analysed the characteristics of preterm birth or associated medical complications. Several American studies have reported a similar risk pattern related to early preterm birth among black women, with relative risks for preterm birth before 33 or 34 weeks of 3 versus 2 for moderate preterm births.5,24,25 Two studies have also reported greater relative risks between black and white women for indicated preterm birth compared with spontaneous preterm births.24,25 Other studies, however, did not find a difference in relative risks between blacks and whites for medically indicated births.26–28 It is noteworthy, given the large disparity in antenatal care between black and white women, that one of the studies to find a difference in risk for indicated versus spontaneous delivery was in a military population where all women had equal access to health services.25 Several American studies on hypertension in pregnancy have also found that black women have higher rates of hypertension during pregnancy,29 but these results are not consistent.30 One European study has identified a higher risk of hypertension in pregnancy among women of African origin.31
Further research on the timing, mode of onset and medical complications associated with preterm births in other European countries where access to health care is good and where similar associations have been found for foreign-born populations would be useful to confirm these findings for populations of Afro-Caribbean origin. This analysis also suggests that preterm birth is not always a principal explanation for higher perinatal mortality, as women from Northern Africa and Southern Europe, for whom perinatal mortality was higher in this district, did not have more preterm deliveries. This has also been reported in other European studies4,32 and orients research towards socio-cultural, linguistic and health system factors that contribute to the greater risks of perinatal mortality at term.33
The Seine-Saint-Denis experimental birth certificates are financed by the Conseil Général de la Seine-Saint-Denis. The authors would like to thank the maternal and child protection services for their collaboration.