International migration and caesarean birth: a systematic review and meta-analysis

Background Perinatal health disparities including disparities in caesarean births have been observed between migrant and non-migrant women and some literature suggests that non-medical factors may be implicated. A systematic review was conducted to determine if migrants in Western industrialized countries consistently have different rates of caesarean than receiving-country-born women and to identify the reasons that explain these differences. Methods Reports were identified by searching 12 literature databases (from inception to January 2012; no language limits) and the web, by bibliographic citation hand-searches and through key informants. Studies that compared caesarean rates between international migrants and non-migrants living in industrialized countries and that did not have a ‘fatal flaw’ according to the US Preventative Services Task Force criteria were included. Studies were summarized, analyzed descriptively and where possible, meta-analyzed. Results Seventy-six studies met inclusion criteria. Caesarean rates between migrants and non-migrants differed in 69% of studies. Meta-analyses revealed consistently higher overall caesarean rates for Sub-Saharan African, Somali and South Asian women; higher emergency rates for North African/West Asian and Latin American women; and lower overall rates for Eastern European and Vietnamese women. Evidence to explain the consistently different rates was limited. Frequently postulated risk factors for caesarean included: language/communication barriers, low SES, poor maternal health, GDM/high BMI, feto-pelvic disproportion, and inadequate prenatal care. Suggested protective factors included: a healthy immigrant effect, preference for a vaginal birth, a healthier lifestyle, younger mothers and the use of fewer interventions during childbirth. Conclusion Certain groups of international migrants consistently have different caesarean rates than receiving-country-born women. There is insufficient evidence to explain the observed differences.


Background
The use of medical interventions in birth has quickly risen in the last quarter century particularly in industrialized countries where the development of medical technology has advanced rapidly [1]. Most notable has been the dramatic rise in caesarean births with rates in the last 15 years in the US, Canada, Australia and parts of Europe reaching 25% and above (see http://www.oecd.org/els/healthsystems/oecdhealthdata2012-frequentlyrequesteddata.htm). While a caesarean birth can be a life-saving procedure it is associated with a significantly increased risk of maternal death from complications of anaesthesia, puerperal infection, and venous thromboembolism [2]. The World Health Organization therefore recommends that caesareans be performed only in medically necessary instances [3].
Migration flows are increasing globally [4] and migrants contribute substantially to the total number of births, exceeding in some Western countries one fifth of births [5,6]. The health of migrants, including perinatal health, has therefore become a research priority [7]. Migrants (immigrants, refugees, asylum-seekers, undocumented and others with temporary or irregular statuses) [8,9] face multiple and intersecting social determinants of health that may compound childbearing health risks, including lower social economic status, lack of support, high levels of stress related to migration and resettlement and barriers in accessing healthcare [10,11]. Refugees and asylum-seekers may also have suffered trauma and abuse [12,13]. Perinatal health disparities [14], including disparities in caesarean births have been observed between migrants and receivingcountry-born women and some literature suggests that non-medical factors such as communication barriers, support and/or care practices during labour and delivery, female genital cutting, or cultural preference may also be implicated in caesarean rate differences [15][16][17]. We conducted a systematic review to address the questions: 1) Do international migrant women in Western industrialized countries have consistently different rates of caesarean birth than receiving-country-born women? And 2) What are the mechanisms (medical and/or non-medical factors) that might explain differences?

Methods
The "Meta-analysis of Observational Studies in Epidemiology (MOOSE)" recommended guidelines for the publication of meta-analyses of observational studies were used to prepare this manuscript [18].
We searched for cross-sectional and cohort studies that provided comparisons of caesarean rates between migrant and non-migrant women. To be included in the review, studies must have examined migrants who crossed international borders and who were living in an industrialized country, defined using the Organization for Economic Cooperation and Development (OECD) member list; studies conducted in refugee camps or with internally displaced persons were excluded. No restrictions were applied based on migrant origin, status or length of time in receivingcountry. We searched the following electronic citation databases: Medline, Health Star, Embase, PsycInfo, CINAHL, Sociological Abstracts, Web of Science, Proquest Research Library, Proquest Dissertations and Theses, POPLINE, Global Health and PAIS. The search strategy was developed in conjunction with a McGill University Health Sciences librarian and can be found in Additional file 1. No language limitations were applied. Searches were conducted from inception of each database until January 2012.
We also hand-searched reference lists of identified literature, conducted web searches and contacted migration and health research experts for additional literature. Web sites searched included professional agency [(e.g., International Federation of Gynecology and Obstetrics (FIGO)] and government (Canada, US and Australia) health sites (see Additional file 2). Each site was searched by browsing through relevant indices predefined by each website and by using 'search boxes'. Terms included: "c(a)esarean, csection, mode of delivery, (im)migrant, (im)migration, foreign, country of birth". Requests for literature were made through ROAM (Reproductive Outcomes and Migration), an international research collaboration, involving over 30 researchers from 13 countries, including Canada, Australia and Europe (see http://migrationandreproductivehealth.org/?page_id=76).
We reviewed all abstracts (or full articles if necessary) to determine if inclusion criteria were met. Those included were subsequently fully reviewed and quality-assessed and data were extracted into a database in Excel © . Non-English language reports were reviewed by ROAM collaborators and if no researcher could read the language of the report, we reviewed the English version abstract only.
The US Preventative Services Task Force criteria were used to assess the quality of studies [19]. Studies were evaluated on the 1) comparability of the study groups (assembly and maintenance); 2) degree of participant loss; 3) clarity of their "migrant" and "non-migrant" classifications; 4) validity and reliability of the approaches used to measure the migrant, caesarean birth and other variables; and 5) attention given to potential confounding factors (i.e., age and parity). Studies with a "fatal flaw" (i.e., ≥ 50% of the original sample lost to analysis and/or comparison groups not adequately defined) were removed from further analysis. Studies were scored as either "good" or "fair"the former if the study met all quality criteria and the latter if it did not meet all criteria but had no fatal flaw. One reviewer assessed all studies and a second reviewer independently assessed 25% of the studies to confirm the assigned quality scores. Discrepancies were resolved through discussion.
A standardized coding form was developed to extract three types of data: 1) publication year and language; 2) study characteristics (e.g., study sample and data sources) and 3) results. The migration indicator (i.e., how migrants were labelled in the study) was recorded verbatim from the report and categorized by type as: "source country/region", "foreign-born/non-national", "ethnicity", "length of time" or "migration status". Sources of data were classified as: 1) population-based registry; 2) population-based hospital data; 3) population-based survey; or 4) hospital records/research study. Timeframe and geographical coverage (local, regional, or national) of data collection, host country and number of migrants and non-migrants were also recorded. Extracted results included frequencies of caesarean rates (overall, emergency and elective) and indications for caesarean for migrants and non-migrants separately. Explanations for caesarean rate differences between groups postulated in discussion sections were also extracted.
For each study we defined caesarean rates for migrant women as being "higher", "lower" or "not different" compared to non-migrant women. If a study had results reported for more than one migrant group comparison they were coded as: "higher" if migrant results in all comparison groups were higher or a combination of higher and not different; "lower" if all were lower or lower and not different; "mixed" if some were higher and some lower; and "not different" if all rates between the migrant and non-migrant groups were not statistically different.
We used Review Manager 5.1 © to perform metaanalyses to estimate the 'effect' of being a migrant on the outcomes of interest: overall caesarean and emergency and elective caesarean rates. The studies were assumed to be heterogeneous due to the range of host countries and migrant groups examined and changes in obstetrical practice over time. We therefore used a random effects model which regards an effect as variable across studies and produces more conservative effect estimates with wider confidence intervals [20]. We combined and analyzed migrant sub-groups defined using UN macro-regions (see Additional file 3) or country of origin, migration status, and length of time in receivingcountry. Sub-analyses by parity, receiving-country/region and time period of data collection were also performed. Due to the diversity across studies in variables selected to control for confounding we chose to use unadjusted results in meta-analyses. We calculated each study's effect size as an odds ratio (OR) using the Mantel-Haenszel model (event/total for migrants and receiving-country women respectively) [20]. Attention was given to ensure duplicate data from different reports were not included twice in analyses and only one comparison from each study was combined for metaanalyses to ensure the results from a comparison group were not counted more than once. A weighted summary effect was calculated for each analysis performed.
To investigate the robustness of results which showed migrant women to consistently have different rates of caesarean compared to receiving-country-born women, sensitivity analyses were performed by limiting the analyses to only population-based data. To assess whether results would differ when covariates were adjusted, we also performed analyses by using adjusted odds ratios and applying the generic inverse variance method to produce summary estimates. Heterogeneity for all metaanalyses was assessed using I 2 tests and we considered a value of 50% or more to indicate substantial heterogeneity [20]. We also examined confidence intervals among the studies in each forest plot to assess the degree of overlap; less overlap being interpreted as evidence of heterogeneity [20]. Publication bias was assessed by visual examination of funnel plots of the OR against the standard error of the log of the OR of studies within each migrant subcategory [20].

Results
The database search yielded over eight thousand citations (n=8565). Once duplicates and studies that did not meet the inclusion criteria were excluded, 87 reports remained. Seventeen additional reports were identified: seven through hand searches, nine from ROAM collaborators and one annual report published by the New South Wales (Australia) Department of Health [21] online. Three reports could not be reviewed because of language [22][23][24]. Twenty studies were removed because the comparison group included migrants, and/or the migrant group included native-born, and/or the assembly of the migrant and non-migrant groups were dissimilar. Seventy-six studies met the inclusion criteria and were assigned a quality score (see Table 1 for a complete description of included studies) [15][16][17]. Twentytwo studies were scored as good and 54 were scored as fair (see Table 2). The primary reasons for not meeting all of the quality criteria was a lack of attention to confounding and ambiguity on how study groups (migrants vs. non-migrants) were defined or confirmed.
Across the 76 studies, data were reported on 1,029,454 migrants (see Table 2). The most common source regions represented were Latin America & Caribbean (39%) followed by 'origin unspecified' (11%) and South Asia (7%) (see Table 2). Region and country of origin were the migration indicators most often used to report caesarean rates (57 of 76 studies). "Foreign-born" was used in 16 studies; ethnicity/race/religion/language, length of time and migration status (e.g., "refugees"), were also used in five, three and nine studies respectively.
The receiving countries included Europe (68%), Australia (11%), the US (11%), Canada (6%) and Israel (4%). The primary language of publication was English (82%). Data sources were population-based in 36% of studies, and coverage for the majority (65%) of studies was one city (in most cases one hospital). Data were gathered from 1956 to 2010; 70% from the 1990s or later. More than half of the studies were published from 2000 onward (see Table 2).

Caesarean Birth Rates: Migrants vs. Non-migrants
Comparison of caesarean rates overall between migrants and non-migrants revealed that in 69% of the studies there were differences in caesarean rates between the two groups (see Table 2). For emergency caesareans, more than half of the studies showed migrants to have higher rates or higher and no different compared to receiving-country-born whereas for elective caesareans there was more variation: 26% showed higher rates or higher and no different; 37% showed lower rates or lower and no different and 5% of studies had mixed results (see Table 2).        [31,33,46,64] than non-migrants see Figure 1; multiparous women were also at increased risk [OR=2.02 (95% CI=1.51, 2.71)] [31,46]. Analysis by type of caesarean suggests an excess risk of emergency caesareans for Sub-Saharan African women (all estimates above 1), although results were heterogeneous [17,31,33,64,82,88,91,94]. Risks for elective caesareans varied by receiving-country; three studies in France [17,31,94], one in Italy [88] and another in Australia [91] showed higher risks of a planned caesarean, while two other studies, one in Canada [64] and the other in Norway [82], found African women to have the same and lower risk respectively.

Meta-analyses
Somali women in North America and Australia were found to have higher risks for a caesarean compared to non-migrant women [OR=1.13 (95% CI=1.02, 1.26)] [32,45]. Significant heterogeneity prevented the calculation of a summary estimate for Somali women in Europe, though the tendency appeared similar to North America/Australia results.
Sub-analysis by parity shows primiparous Somali women to have greater risk for caesarean [OR=1.45 (95% CI=1.30, 1.62)] [32,45] (see Figure 2). Multiparous Somali women also appear at increased risk however there was considerable heterogeneity preventing calculation of a summary estimate [32,45]. There were insufficient data for meta-analysis by type of caesarean in Somali women.  [26,53,88]. Results in other receiving countries were too few or heterogeneous to combine. Latin American women showed greater risk for an emergency caesarean compared to non-migrant women [OR=1.59(95% CI=1.13, 2.25)] [15,64,82,88]. Results for elective caesareans were too heterogeneous for calculation of a summary estimate. * Texeira 2010 did not specify variables controlled so excluded; each study may have controlled for more than one variable so total does not equal 19 (or 100%). † Texeira 2010 did not specify area of coverage; Small 2008 contained regional and national data coverage and so is counted twice. ‡ Texeira 2010 did not specify year(s) of data; total adds to more than 74 since a report may have data from more than one time period.

Eastern Europe
Results of studies reporting comparisons of Eastern European women to receiving-country-born are too heterogeneous to calculate a summary point estimate but suggest these women have lower risks for caesarean, with all estimates lower or tending to be lower than one [26,46,53,82,88,97]. Analysis specifically of primiparous women confirmed a reduced risk [OR= 0.52, (95% CI=0. 43 [16,46].

Southern Europe
Comparisons of women from Italy, Portugal, Spain, and Greece migrating to other parts of Europe or Australia were heterogeneous although the risks of caesarean were consistently greater than that of receiving-country-born women [16,52,56,61,63,68,73].

Other
Analyses by other migration indicators (migration status and length of time in receiving-country) were possible for migrants in Canada only. These included an analysis of asylum-seekers who showed no difference in rates compared to Canadian-born [OR=0.93 (95% CI=0.74, 1.17)] [64,66,95]; and migrant women in Canada five years or less who showed a greater risk for caesarean [OR=1.14 (95% CI=1.06, 1.23)] [64,66,97].
Summary  [16,32,41] were consistent in showing higher rates of caesarean compared to non-migrant women. Inconsistent findings were found however when adjusted meta-analyses were performed for South Asian and Latin American migrants; South Asian women still tended towards higher rates but with a wide, non-significant confidence interval [OR=1.10 (95%CI=0.87, 1.38)] [16,95] and Latin American women showed no difference in emergency caesarean rates compared to non-migrant women [OR=1.01 (95% CI=0.62, 1.63)] [82,88]. Adjusted analyses for emergency caesareans in North-African/West Asian women showed a similar point estimate to the unadjusted analyses but with a wide, nonsignificant confidence interval [OR=1.16 (95% CI=0.86,1.56)] [17,88]. Adjusted results for Eastern European women were too heterogeneous, but did generate a similar estimate to the unadjusted results.

Assessment for publication bias
Visual inspection of the individual funnel plots for studies reporting overall caesarean rates for Sub Saharan African, Somali, South Asian, Eastern European, Vietnamese, Latin American and North African/West Asian migrants (and emergency caesareans for the two latter) all showed symmetry, suggesting no publication bias, although very small studies were generally lacking. Table 3 summarizes important risk factors/mechanisms for caesareans in migrants cited in included studies and for each factor the number of studies citing this factor/ mechanism. Combined the most commonly reported (in order of frequency) were: language/communication barriers, low social economic status (SES), poor maternal health (e.g., anaemia, STIs, TB, parasitic infections), gestational diabetes/high body mass index (BMI), fetopelvic disproportion, and lack of prenatal care.

Mechanisms/indications for caesareans among migrant women
Evidence to explain the consistently different rates of caesarean between Sub Saharan African, Somali, South Asian, Vietnamese, Eastern European, Latin American and North African/West Asian (emergency caesareans for the two latter) and non-migrant women, was limited. Variables adjusted (see Table 2) or stratified differed across studies and few studies compared indications for caesarean between migrant and non-migrant groups making it difficult to draw firm conclusions regarding explanatory risk factors or mechanisms.
Reports that provided comparisons of medical indications for caesareans between Sub-Saharan African/Somali and non-migrant women show these women to have more caesareans due to fetal distress [31,32] and failed induction [32]. Similarly one study found Sub-Saharan African women who were induced to be more likely to have a caesarean compared to French women who were induced [17], suggesting complications with labour that could be due to any number of different reasons.
Adjusted analyses for South Asian [15] and Latin American (emergency caesareans) [82,88] migrants show age, parity or medical complications (e.g., feto-pelvic disproportion, fetal distress) partially explain excess caesareans in these women, although it is not possible to isolate which of these factors is important. For Latin American women, studies which show consistently higher rates of caesarean irrespective of adjustment of covariates or stratification (e.g., birthweight, hospital type) suggest different contributing factors [15,27,53,65,95]. Seeking social status [55]or cultural preference [15,46,53,95] is one suggested cause; higher rates of pre-eclampsia might also have a role [46]. Stratified analyses by neighbourhood (English-speaking and non-English-speaking) have shown higher rates of caesarean for Latinas compared to USborn when living in non-English-speaking neighbourhoods, leading to the suggestion that communication or cultural barriers might be explanatory mechanisms [27].
Vietnamese [15] and Eastern European [53,87,96] migrants seem to have protective factors that explain their consistently lower rates of caesarean. Proposed protective factors include a preference for a vaginal birth [15,87]; the healthy-immigrant effect (i.e., in which immigrants are healthier than the native population due to immigration selection criteria which excludes individuals with significant health problems) [87,93]; a healthier lifestyle (no smoking, drinking alcohol, or drug abuse; low BMI) [87,96]; young maternal age [87,96]; social support [16] and the use of fewer interventions during labour and birth [87,96]. In Southern Europe (Italy, Spain, Portugal), lower rates among migrants are also thought to be due to a preference for caesareans among non-migrant women [30,50] and because healthcare professionals are more concerned about litigation [25,36,86] from non-migrants.
Filipino migrants were explicitly discussed in a number of studies and hypotheses for their higher caesarean rates included maternal preference [15,39] and fetopelvic disproportion due to interracial marriage resulting in large birthweight babies [15,28,39,49,52]. More caesareans due to dystocia and feto-pelvic disproportion have been shown, however, this was deemed to not be due to interracial marriage (by comparing Filipino women in mixed marriages to Filipino women in non-mixed marriages) [28,39]. Filipino women who gave birth by caesarean were also found to experience a greater number of medical and obstetric complications (e.g., GDM, anaemia, viral diseases) compared to Australian-born women who delivered by caesarean [28].

Discussion
Our review of 76 studies comparing the rates of caesarean births between migrant and non-migrant women living in OECD countries, show that women from Sub-Saharan Africa, Somalia and South Asia consistently have an excess of caesareans compared to receivingcountry-born women while Eastern European and Vietnamese women have lower overall caesarean rates. North African/West Asian and Latin American women have higher emergency caesarean rates.
The literature provides inadequate empirical evidence to explain differences in caesarean rates observed. Overall it appears that a combination of factors and mechanisms are likely to be involved. The most frequently postulated risk factors for caesarean risk in migrant populations include: language/communication barriers, low SES, poor maternal health, gestational diabetes/high BMI, feto-pelvic disproportion, and lack of prenatal care. There were no studies identified examining determinants of caesareans in migrants specifically. However, gestational diabetes mellitus (GDM) and feto-pelvic disproportion are known to complicate delivery and increase the risk of caesarean [104,105] and these are more common in some migrant women [28,39,106]. Reasons for increased risk of GDM in migrants are unknown but might be associated with a genetic predisposition, or physiological response to dietary changes; the involvement of a stress response has also been proposed. Feto-pelvic disproportion might be the result of short stature, a large birthweight baby or childhood malnutrition [107].  A BMI above normal is associated with a number of adverse reproductive outcomes, including GDM, macrosomia, prolonged labour, pre-eclampsia and caesarean [108,109], although the exact mechanisms have not been identified. Excessive weight among childbearing women also varies by ethno-cultural background, with African women representing one of the groups with the highest rates of obesity [94,110].
Inadequate prenatal care is more common among migrants [111]. Language and no health insurance are often cited barriers to healthcare [11,112]. Women who do not receive prenatal care, or who cannot communicate their medical history, may have undiagnosed and untreated medical conditions and they do not benefit from other preventative interventions (e.g., advice about prenatal nutrition) [113]. Medical issues that may not be identified include parasitic infections, anaemia and TB which are known to affect some childbearing migrant women [114][115][116], although their association with caesarean birth is unclear.
Poverty, unemployment and low social status are significant concerns for migrants [117,118]. Low SES may indirectly be protective against caesarean since women of low SES are less likely to receive care in a private facility and there may be less concern of litigation among healthcare professionals caring for these women. However, low SES may be a barrier to accessing prenatal care, and is known to be independently associated with poorer health. Qualitative studies suggest that a lack of information and/or support during pregnancy and birth due to marginalization, communication barriers, and/or cultural insensitivity (e.g., genital cutting), may result in anxiety, fear and disempowerment [119,120] and lead to unnecessary caesareans.
Confirmation and further understanding of social support as a protective mechanism could have positive implications for reducing caesarean rates among migrants. Additional support (in addition to routine care) including emotional support as well as practical assistance provided during pregnancy has been shown to reduce the risk of caesarean birth for women in socially disadvantaged situations [121]. Continuous support during labour is also associated with a reduced risk of caesarean [122]. This strategy may be particularly beneficial for reducing emergency caesareans, which overall appear to be more common among migrants.
Our findings also suggest that receiving-country is an important variable to consider, particularly for certain groups (e.g., North African, Latin American, East Asian). The degree to which variation in caesarean outcomes across countries represents effects of policies and/or healthcare delivery [123] or other particularities of receiving countries (e.g., cultural factors) versus differences in the migrant populations resettling in each country, remains unknown. More precise definitions of migrants and more complete individual level migration-related data, including source country, length of time in host country, receiving-country language ability (at the time of pregnancy and birth), and migration status (e.g., refugee or economic immigrant) [124] would allow for better interpretation of results particularly since results from Canadian studies have shown that length of residence post-migration (< 5 years) was a significant predictor of more caesarean births. Confirmation of these findings in other OECD countries would be informative.
Research to establish evidence for risk factors associated with caesarean birth in migrants and to deconstruct the pathways (e.g., genetic or physiological, psychological stress, delivery of maternity care) by which they lead to disparities in mode of birth outcomes, is needed. Pathways are likely multi-factorial and complex. Future work using a combination of quantitative and qualitative approaches may be valuable in more fully expounding the processes. Moreover, greater emphasis on differences in caesarean indications/ mechanisms would be more informative than simply comparing caesarean rates between migrant and non-migrant women since rates may not be higher, but there may still be disparities in risk factors for caesarean births.

Strengths and limitations
There are limitations to this review. The web searches, although extensive, did not include all of the government and professional agency websites from all OECD countries. The majority of included studies were rated as 'fair' quality for not controlling for confounding or due to some ambiguity in their definitions of the study groups. The heterogeneity due to variation in the migrant populations studied or how source countries were grouped to represent regions, made it challenging to combine data for metaanalysis. US studies were largely missing from metaanalyses due to heterogeneity (for Sub-Saharan African women) or lack of data for the source regions analyzed (for Eastern European, Vietnamese, South Asian, Latin American and North-African/West Asian women). This might be problematic since other systematic reviews comparing the perinatal health of migrants to non-migrants [125,126] had varied outcomes between US and European regions. Nonetheless, a broad range of studies was included and analyzed; the database searches were exhaustive with no time or language limitations and only three reports could not be reviewed due to language. Rigorous methods for reviewing, extracting and analyzing data were applied, optimizing the quality of the results generated. The consistency of the results with population-based data offers confidence in the robustness of the findings.
A major strength of this review is that it is the first of which we are aware to systematically examine hypotheses put forward to explain differences in caesarean rates between migrant and non-migrant women and to assess these in light of the available empirical evidence.

Conclusion
Sub-Saharan African, Somali, and South Asian migrants consistently have higher caesarean rates while Eastern-European and Vietnamese migrants have lower overall caesarean rates compared to receiving-country-born women. North-African/West Asian and Latin American migrant women have higher emergency caesarean rates. To date there is inadequate empirical evidence to explain observed differences in caesarean rates; more focused research is urgently needed.