OUP user menu

Cross-country variation in stillbirth and neonatal mortality in offspring of Turkish migrants in northern Europe

Sarah Fredsted Villadsen, Erika Sievers, Anne-Marie Nybo Andersen, Annett Arntzen, Marjorie Audard-Mariller, Guy Martens, Henry Ascher, Anders Hjern
DOI: http://dx.doi.org/10.1093/eurpub/ckq004 530-535 First published online: 24 February 2010


Background: Diverse early-life mortality outcomes have been documented in immigrant populations in northern Europe. A recent meta-analysis has suggested that national integration policy is a key factor in understanding this heterogeneous pattern. In this study, we investigated the variation of stillbirth and neonatal mortality between societies in northern Europe in one minority population, the Turkish. Method: Data on stillbirth and neonatal deaths in 239 387 births during 1990–2005, where the mother was of Turkish origin, was drawn from birth registries or surveys in nine northern European countries. Rates were compared with births from mothers who were born in the society of residence. Logistic regression was used to calculate odds ratios adjusted for year of birth of the offspring. Results: The risks for stillbirth were, or tended to be, elevated for Turkish mothers in all countries compared with the native population, with the highest risk in Austria (odds ratio (OR) 1.7; 95% confidence interval (CI) 1.4–2.1) and Switzerland (OR 1.6; 1.4–1.9). For neonatal mortality the results were heterogeneous, indicating no excess risk for Turkish-born children in the Netherlands, the UK and Norway, and elevated risks in Denmark (OR 1.3; 1.0–1.6), Switzerland (OR 1.3; 1.1–1.5), Austria (OR 1.4; 1.0–1.8) and Germany (OR 1.3; CI 1.2-1.5). Conclusion: This study suggests that preventable society-specific determinants are important for early-life mortality in Turkish migrants in Europe. An active integration policy is consistent with a favourable neonatal mortality outcome in continental Europe, but not with patterns in Scandinavia and the UK.

  • birth outcome
  • ethnicity
  • migration
  • stillbirth Turkish


Increased early-life mortality has been documented in ethnic minority populations in the United States and in several European countries.1–6 In the Netherlands, a doubled risk of death in early life has been described compared with the native population,7,8 and in the UK black and South Asian women have a considerable excess risk of stillbirth.1 Perinatal risk factors such as preterm birth, low, placental abruption and low birth weight and socio-demographic factors such as teenage pregnancies, grand multiparity, low maternal age and low socioeconomic status in the immigrant groups have been suggested to explain these patterns, although these risk factors seem to affect the different ethnic groups in a heterogeneous manner.1,7,8 There are, however, also reports of immigrant populations in Europe with similar or even better perinatal outcomes than the native population. In Norway, for instance, offspring of immigrant women from Vietnam have been reported to have lower perinatal mortality and offspring of North African immigrants an equivalent perinatal mortality compared with the native population.9

The composition of ethnic immigrant populations in the northern European countries is very heterogeneous, consisting of immigrants from different countries of origin with diverse motives for migration. One of the largest immigrant populations in northern Europe is the Turkish, with an estimated 3 million Turks currently living in the European Union, two-thirds of them in Germany.10 In the 1960s, Turkish emigration responded to labour recruitment in Western Europe in large numbers. In more recent decades, the labour migrants have been replaced by refugees and family migration with spouses from Turkey. In many of the host countries, Turkish migrants of predominantly rural origin tended to settle in clusters according to their localities of origin, supporting family as well as regional ties.11

Excess early-life mortality in the Turkish immigrant community has been repeatedly described in Germany12 and the Netherlands,7,8,13 although all reported infant mortality rates have been much lower than in Turkey, where it was estimated to be 24 per 1000 live born in 2006.14 Many studies of perinatal and neonatal health in immigrants in northern Europe focus on risk factors in the immigrant population in comparison with the native population, such as consanguinity, lifestyle and multiparity. A recent review of pregnancy outcomes in immigrant populations in Europe by Bollini et al.,15 however, suggests that integration policy in the resettlement country may have a more important influence on early-life mortality rates. The aim of this comparative study was to investigate disparities in stillbirth and neonatal mortality in Turkish immigrants compared with the native population between European countries, thereby looking for explanations arising from variations between countries rather than risk factors within the Turkish population.


This study was based on data on stillborn and neonatal deaths in Turkish immigrants from birth registries or comprehensive surveys of births in nine countries: Austria, the Flemish part of Belgium, Denmark, England and Wales, North Rhine-Westphalia in Germany, the Netherlands, Norway, Sweden and Switzerland. Table 1 presents some basic information about the data sources.

There were altogether 239 387 births classified as Turkish, with North Rhine-Westphalia contributing with >50% of these. The data were taken from different time periods during 1990–2005 to obtain sizeable populations of births of Turkish mothers in all countries. All data sources used in this study covered at least 99% of all births in the study setting. The results from the Netherlands13 and Germany12 used in the study are taken from previous publications, while the data from the remaining seven countries were collected and analysed specifically for this study.

View this table:
Table 1

Country-specific data sources and definitions

AustriaBelgium (Flanders)DenmarkEngland and WalesGermany (North Rhine- Westphalia)NetherlandsNorwaySwedenSwitzerland
Source of perinatal variablesNational RegisterNational Birth RegisterNational Medical Birth RegistryNational StatisticsPrevious Publication13 State registerPrevious publication13National Medical Birth RegistryNational Birth RegisterNational Statistics
Data ownerStatistik AustriaSPEStatistics DenmarkOffice of National StatisticsStatistical office, State of North Rhine- WestphaliaNational Institute of Public HealthNational Board of Health and WelfareNational Statistics/ Swiss federal statistical office
Data includes asylumseekers/ undocumented migrantsYesNoNoYesIn general yes, coverage potentially incompleteNoNoNoYes
Source of education variableNational registerNational registerPrevious publication13National registerNational register_
Definition of stillbirthAt least 500 g birth weight500 g or moreAt least 28 weeks gestationAt least 24 weeks gestationAt least 500 g birth weight, prior to 1 April 1994 at least 1000 gAt least 24 weeks gestationAt least 22 weeks gestationAt least 28 weeks gestationAt least 22 weeks gestation or a weight of more than 500 g
Number of births defined as Turkish13 666871716 26111 586128 78641 348312914 16227 480
Definition of TurkishChild nationalityMaternal country of birthMaternal country of birthMaternal country of birth yChild nationalityMaternal country of birthMaternal country of birthMaternal country of birthMaternal nationality
Years included2000–052000–051995–20001999–041990–971995–20001990–031992–051990–05


Country of birth as registered in the population registries was used to identify Turkish mothers in the three Scandinavian countries, England and Wales. Turkish nationality of the mother was used to identify the target population, including residents, asylum seekers as well as undocumented migrants in Switzerland and Flanders (Belgium), and, indirectly, by Turkish nationality of the child in North Rhine-Westphalia (Germany) and Austria. In the latter countries, attaining German or Austrian citizenship for the child required a parental minimum legal residency of 15 or 10 years, respectively. Asylum seekers and undocumented immigrants were also included in the population in England and Wales, but not included in the Scandinavian countries.

The gestational age or weight criteria for the stillbirth diagnosis differed between countries, as demonstrated in table 1. Neonatal mortality in all countries was defined as death of a live-born child within the first 28 days of life.

For six countries, some information on socio-demographic factors was available, and for five of these countries, we were able to obtain information about maternal education. The variable on educational level was grouped according to the International Standard Classification of Education (ISCED)16 into short (<10 years), intermediate (10–12 years) and long (13+ year) education.

Statistical methods

Mortality rates were calculated using the conventional definitions with a stillbirth rate calculated as number of stillbirths*1000 divided by all births, and likewise the neonatal mortality rate was calculated by the number of deaths during first 28 days of life in relation to all live births. Rates were estimated as an average over the study period.

The odds ratios (ORs) of stillbirth and neonatal mortality in offspring of Turkish migrants compared to the native population were calculated with logistic regression adjusted for year of birth of the offspring. Analyses were made separately for each country with SPSS 16.0 or SAS 9.2. The OR for neonatal mortality in the Netherlands was based on the data presented in the previously published Dutch study by Troe et al.,13 which did not allow for adjustment for year of birth.


Table 2 demonstrates the socio-demographic and perinatal indicators in the Turkish immigrant group and in the native populations in the six countries where this information was available. The average educational level of the Turkish mothers was considerably lower than the level of the native mothers. The percentage of high education in Turkish mothers ranged from 3.7% in Belgium to 13% in the Netherlands. Teenage pregnancies were more frequent, while being a single mother was less frequent in Turkish mothers compared with the native populations in all countries. More than 90% of children of Turkish mothers in all countries had Turkish fathers, with the highest occurrence in Denmark (97.5%).

View this table:
Table 2

Socio-demographic and perinatal characteristics of births of Turkish mothers and of Native mothers in six of the nine countries included in the study

Maternal education (%)
Maternal age (%)
Mother employed year before birth (%)16.555.238.880.332.479.4
Single mother (%)
Paternal country of birth (%)
    Other immigrant1.
Preterm birth (%)
Multiple births (%)
Birth weight
    Mean (g)328232953392348433973485337535403393354633433302
    <2500 g6.
Parity (%)

The variation in mean birth weight of offspring of Turkish migrants varied only marginally across the countries, with the lowest mean in Flanders on 3282 g and the highest in Sweden on 3393 g. The mean birth weight varied more between the native population of the six countries, being highest in Norway and Sweden (3540 and 3546 g, respectively) and lowest in Flanders (3295 g).

The stillbirth rates were higher in the Turkish than in the native population in all countries (table 3). When the risk of stillbirth was estimated we found that almost all countries displayed elevated odds ratios, being highest in Austria (OR 1.7) and Switzerland (OR 1.6) and the lowest in Norway (OR 1.1) and Denmark (OR 1.2) (table 3).

View this table:
Table 3

Stillbirth, early neonatal and neonatal mortality rates in offspring of Turkish mothers in nine European countries

StillbornStillbornDay 0–6Day 7–27Day 0–27
Maternal country of originYearsN1/1000 bornOR (95% CI)1/1000 live born1/1000 live bornOR (95% CI)
AustriaNative2000–05408 6953.812.30.71
Turkish2000–0513 6666.51.7 (1.4–2.1) (1.04–1.8)
Flanders/BelgiumNative2000–05238 2333.411.90.51
Turkish2000–0587173.91.3 (0.9–1.8) (0.8–1.7)
DenmarkNative1990–2003812 3054.113.00.71
Turkish1990–200316 2614.71.2 (0.9–1.4) (1.0–1.6)
England and WalesNative1999–20043 037 3485.112.60.81
KingdomTurkish1999–200411 5866.41.3 (1.0–1.6) (0.6–1.3)
NRWNative1990–971 296 7983.812.4n.i.1
GermanyTurkish1990–97128 7865.41.4 (1.3–1.5)3.4n.i.1.3* (1.2–1.5)
NetherlandsNative1995–2000935 858n.i2.90.81
Turkish1995–200041 348n.i.** (0.7–1.1)
NorwayNative1990–2003675 3875.512.10.61
Turkish1990–200331296.11.1 (0.7–1.7) (0.2–1.2)
SwedenNative1990–20041 344 2373.311.90.61
Turkish1990–200414 1625.11.5 (1.2–2.0) (0.8–1.5)
SwitzerlandNative1990–2005900 8753.613.00.61
Turkish1990–200527 4805.81.6 (1.4–1.9) (1.1–1.5)
  • n.i.=no information.

  • *:Day 0–6.

  • **:not adjusted for year of birth.

The neonatal mortality rates showed a tendency towards lower rates in the Turkish group in the Netherlands, the United Kingdom and Norway when compared to the native populations, while the rates of the Turkish group in the remaining countries were or tended to be higher. Increased risks of neonatal mortality in children of Turkish mothers (compared with the native populations) were found in Switzerland (OR 1.3), Austria (OR 1.4), Germany (OR 1.3) and Denmark (OR 1.3).


This study is, to the best of our knowledge, the first to compare early-life mortality outcomes in one migrant group between different countries in Europe. We found that Turkish mothers in all countries had an increased risk of stillbirth, whereas the relative risk of neonatal mortality of offspring of Turkish migrants compared with the native population varied between countries. These patterns suggest a similarity in risk profile for stillbirth in Turkish immigrants in all societies, while the factors determining the neonatal mortality seem to differ between resettlement countries.

Stillbirth and neonatal mortality is difficult to study in migrant populations due to small numbers; however, this study was based on large data sets contributing with state-of-the-art information on early-life mortality from the respective contributing countries. With the exception of Norway, that only contributed about 3000 births in Turkish migrants, all country data sets contained at least 11 000 births. Analyses were performed specifically for this study (except the results from the Netherlands), allowing the analytic design to be as similar as possible. When one compares crude rates between countries, one also needs to consider the importance of the different definitions of the outcome variables. The gestational and weight limits for stillbirth (see table 1), for instance, varies considerably which needs to be borne in mind when one compares national rates. It is also possible that different practices regarding the categorization of deaths during labour as stillborn or neonatal deaths may differ between countries.17 It seems quite unlikely, however, that this was an important source of bias for the relative estimates ‘within’ each country. Five countries excluded asylum seekers, but since the numbers of asylum seekers from Turkey, during the years of the study, were comparatively small,18 it seems unlikely that this was an important bias in this study.

Bolini et al.,15 have suggested that integration policy is an important determinant of birth outcomes in migrant populations. They claim that societies have adopted either a ‘passive’ attitude, in which immigrants are expected to fit in without support of specific policies, or an ‘active’ attitude, where steps are taken to acknowledge cultural differences and specific needs, complemented by special services, such as interpreters and special measures on the labour market to facilitate the entry of newly settled immigrants. It seems probable that integration policy influences societal risk factors with direct effects on the immigrant mother and her fetus, such as socioeconomic living conditions and access to the best-quality healthcare. More indirectly, it is also possible that integration policy can be seen as an indicator of attitudes towards immigrants in the general population, thus indicating different levels of discrimination in all aspects of society.

According to Bolini et al., Austria, Germany, Switzerland and the UK should be categorized as countries with a restrictive integration policy, while the Scandinavian countries, the Netherlands and Belgium have permissive integration policies.15 Analysing the results of the present study with this categorization of the studied countries in mind, restrictive integration policy is consistent with the high risk for stillbirth in Turkish mothers in Austria and Switzerland. When it comes to the neonatal mortality, the results indicate no excess risk for Turkish-born children in the Netherlands and Norway (countries with a permissive integration policy) and elevated risks in the following countries with a restrictive integration policy; Switzerland, Austria and Germany. UK, a country with a restrictive integration policy and low neonatal mortality rates, and Denmark and Sweden, countries with a permissive integration policy, did not fit into this pattern. It should be noted that the size of the Turkish immigrants populations were largest in Germany, Switzerland and the Netherlands, giving further weight to the patterns associated with integration policy in Continental Europe. As a direct consequence of a passive immigration policy, obtaining citizenship in Austria, Switzerland and Germany was quite difficult and demanded at least 10–15 years residency. This also influenced the definition of Turkish; being maternal or child nationality in these three countries with a passive immigration policy as opposed to maternal country of birth in the other countries, where the majority of Turkish natives were citizens of the resettlement country. This definition may imply that the Turkish populations in Austria, Switzerland and Germany tend to have been residents in Europe for shorter periods of time than the Turkish populations in the other countries in the study. Since Troe et al.,19 have demonstrated that birth outcomes of Turkish immigrants in the Netherlands, if anything, tend to get worse with longer residency, it does not seem likely that this possible difference in time of residency explains the high neonatal and stillbirth rates in these three countries with a passive immigration policy. This study thus presents some support for the hypothesis that integration policy can be interpreted as an indicator of societal influences on early-life mortality of immigrants. Patterns in the UK and Denmark, however, seem to indicate that more complex theories are needed to explain early-life mortality outside the Continental European context.

It seems probable that policies that determine birth outcomes in immigrant populations do this through mechanisms of a more general nature. There are many studies that demonstrate that social disadvantage, in general, increases the risk of poor birth outcomes. Picket et al.,20 found that the US Census tract measures of disadvantage, including proportion of unemployed males and median household income, predicted poor birth outcomes and other North American studies21 have linked poor birth outcomes to neighbourhood poverty and housing vacancy. Bambang et al.,22 report a similar association with deprivation for perinatal mortality in the UK in 1993–95. Interestingly, Giuldea et al.,23 in a study in Wales, found an association with deprivation for stillbirth but no such association for neonatal deaths, a pattern somewhat similar to the pattern of Turkish immigrants in several of the countries in the present study. Many policies, apart from integration policy, can decrease or increase the poverty of vulnerable populations such as immigrants. One may speculate that policies that deal with the labour market and housing, including policies on discrimination, might be worthwhile to examine in cross-country comparisons of health outcomes.

An alternative explanation to the variation of birth outcomes in Turkish immigrants, from that of cross-country differences in policy, would be that the backgrounds of the Turkish migrant populations differ between countries because of selective migration from the country of origin. In general, however, we find that the immigration pattern has been similar for Turkish immigrant in the different countries of destination included in the study.11 The immigration started as a response to labour recruitment in Western Europe and later predominately continued by family reunification and to a certain extent by Kurdish and Christian refugees. The immigrants came primarily from rural regions in Turkey, and they often settled in clusters in big cities in the host societies, and the localities of origin appear to have influenced the country of destination. The educational level and the employment rates of the Turkish mothers in the study were consistently lower than the native populations, more than 90% of the mothers in each country had a Turkish-born father for their child, and the birth weight of the children varied much less between countries than it did between the native populations. This all seems to support the claim that the variation of socioeconomic background in Turkey was quite limited between the migrant groups in the study, and thus not a likely reason for the cross-country variation in outcomes.

The similarity of the stillborn rates in Turkish immigrants between societies suggests that some immigrant-specific risk factors may have a similar impact in all societies. Ethnic differences in health beliefs and language barriers between immigrant patients and native healthcare systems have been described to be important for the use of and the quality of care received by migrant populations.24,26 For the Somali immigrant group studies from the Nordic countries,24,25 and a recent meta-analysis of Somali immigrants in six different countries,6 have shown that complications during labour and suboptimal factors, such as delay in seeking healthcare, refusal and excess of caesarean section, inadequate medication and miscommunication in the antenatal and perinatal care could be explanations for excess perinatal mortality in this group. Similar patterns of suboptimal care have been described for Turkish immigrants in Berlin, with delayed first antenatal visits, higher rates of pre- and postpartum anaemia and less use of epidural anaesthesia during labour.26 It seems probable that specific policies for improved quality of health care, through efforts to facilitate communication and adjust the content of the care to the special needs of immigrant women, would be of importance to improve birth outcomes in migrant mothers.15

Consanguinity between parents is a risk factor for stillbirth and perinatal death, particularly if prenatal screening and terminations of pregnancies in case of major malformations are not accepted or available.27 From Norway, it has been reported that 17% of second-generation Turkish children have consanguine parents.28 No such figures are available for the other countries but the similar background of the Turkish migrant populations makes it quite likely that consanguinity account for some of the early-life mortality in Turkish migrants in this study. Analyses of causes of deaths in Turkish infants in the Netherlands, however, suggests that the excess mortality attributed to malformations, potentially caused by consanguinity, is small.13

This study demonstrates that risk factors and health outcomes in a migrant population in one country are not always applicable in another societal context. Future studies need to exploit the diversity of migrant health outcomes in Europe to identify the role of society-specific factors, such as national immigration policies in these vulnerable populations.


This study has been sponsored by the Nordic School of Public Health in Göteborg and a grant for Nordic Networks from the Swedish Council for Working Life and Social Research.

Conflicts of interest: None declared.

Key points

  • Stillborn as well as neonatal mortality in Turkish immigrants’ rates varies considerably between countries, suggesting that preventable society-specific determinants are important for these outcomes.

  • An active integration policy is consistent with a favourable neonatal mortality outcome in continental Europe, but not with patterns in Scandinavia and the UK.


We are indebted to the enthusiastic support from the European Society of Social Paediatrics and Child Health (ESSOP) and its president, Professor Nick Spencer, for this research project. The data from England and Wales have been provided by the Office for National Statistics, the data from Austria by Statistik Austria, both complemented by analyses made by the authors.


View Abstract