Are all immigrant mothers really at risk of low birth weight and perinatal mortality? The crucial role of socio-economic status

We used data from linked birth and death certificates taken from the Belgian civil registration system. The data relates to all singleton babies born between 1st January 1998 and 31st December 2010, whose mothers were living in Belgium, including those of asylum seekers and undocumented residents. In Belgium, all live births and foetal deaths from 22 completed weeks of gestation or a birth weight >500 g must be registered. Medical data are registered in hospital by midwives and gynecologists, and socio-economic data are recorded at the civil registration service within 15 days of the birth as reported by the parent(s). In Belgium, two perinatal epidemiological centers, SPE for the Dutch-speaking region and CEpiP for the French-speaking region are responsible for the quality and the completeness of the data encoded in birth and death certificates. They collaborate with the maternity hospitals and civil registration services in order to improve the quality of the data and avoid missing data.[15].

The data collection was carried out by Statistics Belgium and was exempted by law from requiring ethical approval. The process of obtaining data from linked death and birth certificates from the Belgian civil registration system is regulated by the Belgian Commission for the Protection of Privacy. After approval of this Commission, Statistics Belgium provided the data for the period 01/01/1998 to 31/12/2010.

From the birth certificates we extracted the following data: gestational age at birth, birth weight, maternal age, parity, maternal nationality of origin and at delivery, maternal educational level, maternal and paternal employment status, and if the mother is living alone or not.

We computed perinatal mortality, defined as foetal deaths from 22 weeks of gestation until 7 days after birth. We considered a birth weight to be low (LBW) when below 2,500 g.

Maternal nationality is recorded as declared at the civil registration service. We defined nationality of origin as the nationality at birth and the current nationality as the nationality at the child’s birth. We classified nationalities into the six most represented groups in the country, namely Belgium, members of the European Union (EU 27), Turkey, Maghreb, sub-Saharan Africa, Eastern European countries not included in the EU 27 and Russia. In this article, we have not shown the nationality category “others” because of the small number of subjects it contains and their heterogeneity (2.3 %). We defined naturalisation as the adoption of the Belgian nationality by the mother before delivery. Finally, we created eleven categories of nationalities, which took into account the mother’s nationality of origin and whether she subsequently acquired the Belgian nationality.

The maternal level of instruction was organised into five categories: superior (university or higher education), upper secondary (completed secondary school), lower secondary (up to the third completed year of secondary school), lower primary (completed primary or less), and other. Data on paternal and maternal occupational status were combined to derive the number of parents employed (two, one, or none). For single mothers, the paternal status was not taken into account, so in this case the number of parents employed could only vary between zero and one. We defined the number of parents employed as “parental employment status”.

Parity was categorised into three groups (nulliparity, 1 or 2 births, and > 3 births). Age of the mother was also categorised into three groups (<20 years, 20 to 40 years, and ≥ 40 years).

We used logistic regression to estimate odds ratios (ORs) for the associations between maternal nationality and perinatal outcomes (LBW and perinatal mortality), adjusted for maternal age, parity, education and employment status. First, the crude associations between nationality and perinatal outcomes were estimated. Thus, a series of multivariate models were developed to assess the individual effects of education and employment status as potential confounding variables..In the first multivariable model we examined the influence of parity and maternal age only (Model 1); in Model 2 we added the parental employment status; in Model 3 we adjusted for parity, maternal age and maternal education; and finally, in Model 4, we included all variables, namely parity, maternal age, parental employment status and maternal education. In order to study the interaction of SES with the association between nationality and perinatal outcomes, we stratified our models by parental employment status and maternal education. We present the odds ratios derived from the logistic regressions and the p-values for the Wald χ ² test. The significance level was set at α = 0.05, and all of the analyses were performed using Stata 12 software.

Table 1 shows the distribution of births according to maternal nationality and socio-demographic characteristics. 1,363,621 singleton births between 1998 and 2008 were included in our study. Seventy-five percent of births were to Belgian mothers, with the other most represented subgroups being EU27 (7.2 %) and Maghreb (3.8 %). Forty per cent of sub-Saharan African mothers lived alone and in 58 % of their households none of the parents were employed. The highest proportion of young mothers (<20 years) was found among Turks and Eastern Europeans (7.6 % and 8.1 % respectively). Belgian mothers had the highest proportion of superior educational level (42.2 %), whereas Turkish, Maghrebi and Eastern European mothers were most likely to have a less than primary educational level (26.4 %, 23.8 % and 23.4 % respectively). The proportion of women in the category “other educational level” was significant for Eastern Europe (25.2 %), sub-Saharan Africa (17 %) and Turkey (17.8 %). Belgian mothers had the highest proportion of both parents being employed in the household (71.4 %), contrasting with Eastern European mothers who had the highest proportion of none of the parents being employed (61.4 %). Naturalized mothers had a higher level of education and proportion of both parents being employed in the household compared to non-naturalized mothers.

Table 2 shows the crude and adjusted associations between maternal nationality and perinatal outcomes (LBW and perinatal mortality). We observed distinct patterns for LBW and perinatal mortality. In terms of LBW (Table 2), except for mothers from sub-Saharan Africa and EU27 countries, all nationalities were protected compared to Belgians. Sub-Saharan African mothers had a significantly higher risk of LBW compared to Belgians. Adjusting for maternal age and parity didn’t change the results. However, adjustment for maternal education significantly decreased the ORs of LBW in all nationality groups, and, for sub-Saharan African and EU27 naturalized women, the risks became comparable to Belgians. Adjusting for the employment status also decreased the ORs in all nationality groups, conferring protection against LBW when compared with Belgians. The ORs of for immigrant mothers were similar to those who adopted the Belgian nationality.

In terms of perinatal mortality (Table 3), all migrant women had a significantly higher risk compared to Belgian mothers, with the exception of naturalised mothers originally from Turkey, Eastern Europe and EU 27. As with LBW, adjusting for maternal age and parity did not change the results. Adjustment for the employment status decreased the ORs to a certain extent, although mortality remained higher for migrant women (with the exception of Eastern Europe). Adjusting for maternal education significantly reduced the mortality risks for all the nationality groups, although sub-Saharan African women remained more at risk than Belgians. Finally, in the full model, only mothers from Maghreb had a slight excess mortality rate compared to Belgians albeit at the margins of significance. In contrast to LBW, immigrant mothers had higher risks of perinatal mortality than immigrant mothers who adopted the Belgian nationality.

In order to study whether the SES is an effect modifier of the association between nationality and perinatal outcomes, we have constructed our models stratified by level of instruction (less than primary versus superior level) and parental employment status (no parents employed versus 2 parents employed) (Table 4  and Table 5). In Table 4, we observe that among households where none of the parents are employed, the rates of LBW for women of all nationalities are significantly lower than for Belgians (p < 0.0001). With two parents employed, the picture changes with sub-Saharan African mothers having a significant excess of LBW, whereas Turkish and Eastern European mothers have similar rates to Belgians. Women from EU27 and Maghreb, on the other hand, remain protected against LBW. When analysing maternal education, we observe a similar pattern: with a superior level of education, sub-Saharan African and Turkish mothers have a significant excess of LBW compared to Belgians. Among women with a very low level of education all migrant groups were significantly protected against LBW compared to Belgians.

For perinatal mortality (Table 5), a pattern similar to that of LBW is observed, although less pronounced. With none of the parents employed, the rates of perinatal mortality for women of all nationalities (and particularly EU27 and Eastern Europe) are lower than for Belgian women. With two employed parents, the pattern changes with women from Maghreb, sub-Saharan Africa and EU27 having a significant excess of perinatal mortality. Among women with the lowest level of education the rates of perinatal mortality are not significantly different to Belgian mothers. In contrast, with a superior level of education, Maghrebi, sub-Saharan African and Turkish mothers have a significant excess of perinatal mortality compared to Belgian mothers. Again, the acquisition of the Belgian nationality significantly decreased the mortality odds for all the nationalities.

The association between migration, socioeconomic factors and birth outcomes is not uniform. This association varies notably depending on the considered SES indicator, on the maternal nationality and on the particular outcome. Our results suggest the presence of an outcome and migration-specific process in the way that: risk factors of LBW and perinatal mortality are different and vary depending of the nationality considered.

For instance, regarding the SES, maternal education and parental employment status exert different effects on perinatal outcomes depending on the immigrant group. Adjusting for maternal education decreases the risk of LBW and perinatal mortality as observed in the literature, [16, 17] but in our study, parental employment status seems to weigh more than maternal education for sub-Saharan African mothers. We need further research to explore this observation but this suggests that appropriate measurements of socioeconomic status depend on the population being studied. For instance, the quality of education—and the relative differences in opportunities it represents—may vary across groups [11]. Different measures may capture different aspects of relative or absolute socioeconomic advantage, which may vary in their importance among immigrant groups [11].

Some risk factors for adverse perinatal outcomes appear to be specific to certain migrant sub-groups. For instance, we observe that sub-Saharan African mothers have higher risks of LBW whereas other migrant groups have lower risks of LBW. Various studies have shown that LBW among Sub-Saharan African women is associated to a higher incidence of hypertension and genito-urinary infectionswhereas perinatal mortality among Maghrebi and Turkish women is associated to a higher prevalence foetal or newborn deaths caused by congenital anomalies [12, 18‐20].

Our observation are consistent with Urquia et al., who show that the healthy migrant effect is outcome-specific in the way that migrants are healthier with regards to preterm birth, hospitalisation and illness during pregnancy but not with respect to postpartum depression. Moreover, this effect is also ethnicity-specific, since it only applies to non-European immigrants [21]. Similarly, Gould et al. observed that factors conferring protection against LBW in the white population were not protective among Asian Indian population in the US [6].

A key finding of our study is that being an immigrant is more protective against LBW and perinatal mortality for women with lower SES than for women with higher SES. This confirms previous North-American studies where native-born mothers with a low education had a higher likelihood of adverse birth outcomes compared to foreign-born mothers with a similar education [22, 23]. In addition to the ‘epidemiological paradox’, another paradox has emerged from studies that describe a weak or flat social gradient in health among immigrants [7, 9, 24, 25], which could explain our results. Different hypotheses have been postulated regarding this absence of gradient: it may be due to the fact that health inequalities in the migrants’ countries of origin don’t follow the same social patterns as in high-income countries [24]. Cultural and social factors may be responsible for attenuating the gradients. Indeed, it has been shown that recent migrants tend to have protective health behaviours such as healthier diets, lower consumption of cigarettes and alcohol, and better social support with stronger family ties and social networks [21, 26]. A further possible explanation is related to the ‘healthy migrant effect’ whereby a minimum level of health is thought to be needed to migrate, meaning that only a ‘selection’ of healthy migrants arrive in the host country. The poorer and less healthy people might therefore not be represented, which leads to flattening of the social gradient in health.

It also needs to be considered that the SES indicators which we used in our study may not have the same significance for the Belgian population as for the immigrants. In fact, we tend to assume that mothers with the lowest educational level are at the bottom of the social hierarchy. However, in their countries of origin, women with this educational level may not have been at the bottom of the social ladder. Moreover, the same number of years in school between Belgian and migrant doesn’t necessarily give the same education or the same chances on labor market for migrant population. Additional measures of SES such as occupational status, health insurance, living conditions, social integration, and neighbourhood characteristics may provide better markers of socioeconomic differentiation among immigrants. It has been shown that the effects of area-level socioeconomic characteristics (such as housing quality, safety, and social cohesion) may have an important impact on birth outcomes [11].

In our study, we observe a favourable effect of adopting the Belgian nationality on birth outcomes, which replicates our previous findings from Brussels [12]. However, this effect is less pronounced for LBW compared with perinatal mortality. In Belgium the nationality can be acquired in three ways: by statement, marriage or naturalization. The statement and marriage are open to adults who meet a series of requirements (related length of residence, place of birth, attachments to Belgium). For naturalization, a principal residence, covered by a legal residence in Belgium for at least three years (2 years for refugees) is required. The procedure takes at least 18 months and does not depend on cultural or language knowledge, nor on medical conditions [27]. The current nationality of the mother is a good way to assess the effects of first generation migration whereas the effects of 2nd or 3rd generation can be better approached by the use of the nationality of origin.

It has been shown that birth outcomes may either improve or deteriorate with length of residence, depending on the migrant group [28]. A common framework for the interpretation of these patterns is provided by the ‘acculturation theory’, which describes a socio-cultural process in which members of one cultural group adopt the beliefs and behaviours of another group [29, 30]. For example, the public health literature suggests that although immigrants from Latin America may arrive in the US with relatively healthy behaviours, they gradually disappear during the process of acculturation with higher proportions of preterm birth and morbidity during pregnancy [31]. In our study, we might view the acquisition of the Belgian nationality as a proxy indicator of acculturation, which exerts, in contrast, a protective effect. LBW did not increase and perinatal mortality decreased among women who acquired the Belgian nationality. These findings are consistent with Cacciani et al.’s, who found that perinatal outcomes improved over time in some immigrant women [32]. The acquisition of the Belgian nationality is more likely to be granted to people who have a longer duration of stay in Belgium, higher education, legal status and better language skills; which is indirectly linked to better access to quality health care, a key determinant to reduce adverse perinatal outcomes [33, 34].