Study design and population

This cross-sectional population-based comparative study uses national birth certificate data from Canada and Spain. The study population is composed of singleton liveborn infants in Canada (2000–2005) and Spain (1998–2007) to women born in Spanish-speaking Latin American countries and native-born women (Canadian-born women in Canada and Spanish-born women in Spain) in the last decade available in national birth certificate data. The longer study period for Spain reflects availability of data; however, the midpoint is the same for both countries, thus maintaining comparability. Both registers collect information on maternal country of birth and other sociodemographic characteristics. To minimize bias due to potential outliers, we only included records with birthweights within 250–5999 grams and gestational age within 20–45 weeks.

Data

Information on birth outcomes in Canada was obtained from the linked live birth-death file of the Public Health Agency of Canada’s Canadian Perinatal Surveillance System. Approval for use of this data was obtained from the Agency and Statistics Canada. Ethics approval was not legally required since data were anonymized, analyses were conducted in restricted access Research Data Centers, and guidelines for release of results were adhered to. Likewise, this study only used anonymized Spanish birth certificate data compiled by the National Institute of Statistics and the review of an Ethics Research Committee was not legally required in Spain.

Secondary analyses were conducted with a subsample of Latin American women delivering in Ontario hospitals (1998–2007) and captured through a linkage between the official immigration register, the Citizenship and Immigration Canada database, and the Discharge Abstracts Database, which comprise all hospital births. Data collection methods and use of Ontario hospital data were approved by the St. Michael’s Hospital Research Ethics Board. These analyses were conducted at the Institute for Clinical Evaluative Sciences (ICES) in Toronto, Ontario. ICES is a prescribed entity under Section 45 of the Ontario Personal Health Information Protection Act (PHIPA) 2004. As such, it is permitted to receive (collect) and use personal health information without consent. ICES implements strict privacy and confidentiality practices and procedures that have been reviewed and approved by the Information and Privacy Commissioner of Ontario as required under the Act.[25]

Outcome measures were singleton mean birthweight at term (37–41 weeks gestation), low birthweight (< 2500 grams) and preterm birth (<37 weeks of gestation). In Canadian birth certificates, the information on gestational age is recorded by the attending physician, nurse or midwife. In Spain, it is recorded by the parents based on the information provided to them by the birth attendant. A Spanish validation study using records of a single hospital as the gold standard, found that the sensitivity and specificity of birth certificate data for low birthweight were 90% and 99%, respectively and 78% and 98% for preterm birth.[26] Most notably, the birthweight distributions exhibited an almost perfect overlap between the curves at 37 and higher gestation weeks.

Maternal country of birth included Canada, Spain, Argentina, Bolivia, Chile, Colombia, Cuba, Ecuador, El Salvador, Honduras, Mexico, Nicaragua, Paraguay, Peru, Uruguay, Venezuela and Rest of Latin America (Costa Rica, Dominican Republic, Guatemala and Panama). The latter category was created by collapsing countries contributing very few births in at least one of the two countries. Country of delivery included Canada and Spain.

Analytic strategy

For descriptive statistics, we conducted a side-by-side analysis based on common definitions of the outcomes and birth characteristics in the two national datasets. Due to the restricted access of each dataset, it was not possible to pool the datasets of the two countries at the individual level but we could do so in aggregate form.[27] In order to compare birth outcomes in Canada versus Spain, we calculated within each dataset the mean term birthweight and the proportion of low birthweight and preterm birth, within strata of covariates (maternal country of destination, infant sex, maternal age groups, marital status, number of previous births, and father born in the same country as the mother), along with their respective standard deviations and number of observations. These aggregate estimates of each country were then combined and analyzed using regression techniques for aggregate data[27] to obtain unadjusted and adjusted mean birthweight differences and odds ratios with 95% confidence intervals, using births occurred in Spain as the reference group. Regression analysis of continuous and categorical aggregate data produces point estimates and confidence intervals similar to those obtained with individual-level data.[27] To account for unmeasured potential variations in circumstances of health care between receiving countries (i.e., geographic setting, health care setting, type of health care providers, and differences in the medical management between comparison groups) we further adjusted for a comparability score[28] by assigning a weight of 1/8 to each observation of the reference group (i.e., women delivering in Spain). Regression analyses were adjusted for infant sex, maternal age groups, marital status, number of previous births, and father born in the same country as the mother, as these factors may act as potential confounders.

Finally, we conducted secondary analyses restricted to Ontario hospital data linked to the official immigration register (Citizenship and Immigration Canada Database) to assess whether time since arrival to Canada was associated with improved birth outcomes of Latin Americans in Ontario, as a potential alternative explanation to our findings.

Main findings

Our findings provide support to the hypothesis that selective migration may influence the more favourable birth outcomes of Latin American women delivering in Canada vs. in Spain. First, the share of singleton births to Latin American immigrant women was larger in Spain (4.56%) than in Canada (2.18%), despite the physical proximity of Latin America to Canada. Second, compared to Latin American women giving birth in Canada, those delivering in Spain were more likely to be teenagers, unmarried and not having delivered previously; these being correlates of adverse outcomes and factors associated with negative selection. Adjustment for these factors reduced the differences but did not eliminate them. Third and more important, birth outcomes were more favourable among Latin American women delivering in Canada vs. those delivering in Spain. Notably, mean birthweight at term was 59 to 212 grams higher among infants born to foreign-born women from different Latin American countries giving birth in Canada compared to those of their same-country counterparts giving birth in Spain. The magnitude of these differences is comparable to the effect of smoking on birthweight.[29,30]

Strengths and limitations

Strengths of our study include the use of birth certificate data at the national level, comparable study periods and similar definition of the outcomes, maternal country of birth and covariates for statistical adjustment. Restriction of the study to a distinct immigrant group in two different receiving countries may diminish confounding resulting from unmeasured factors, since for example Mexican-born women giving birth in Canada may share background and cultural characteristics with their Mexican counterparts giving birth in Spain.

Limitations are several. First, gestational age information in Spanish birth certificates may slightly overestimate preterm birth rates, and consequently, low birthweight rates, according to a validation study that linked about 9000 birth certificates with records of one Madrilean hospital.[26] If this study was representative of Spain as a whole, then our findings regarding lower preterm birth and low birthweight rates among Latin American women delivering in Canada than in Spain should be taken with caution, since the true rates in Spain would be slightly lower, thus weakening the observed associations. This potential bias, however, does not affect our comparisons regarding birthweight at term. In the Madrilean study, the birthweight distribution at ≥37 weeks based on birth certificates was shown to be virtually identical to that of hospital records,[26] suggesting that the ubiquitous pattern of higher mean birthweight at term found among women from every Latin American country giving birth in Canada compared to Spain is unlikely to be affected by artifacts. Misclassification of gestational age in birth certificates is known to affect preterm and post-term births but not term births.[31,32]

While low birthweight and preterm birth mainly reflect pathological processes interrupting the pregnancy or its normal development, mean birthweight at term is a measure of normal fetal growth in the population.[33] If there were no significant differences in low birthweight and preterm birth rates between Latin American women delivering in Canada versus those delivering in Spain, we can still conclude that selective migration is associated with improved fetal growth.

Second, although we controlled for some maternal characteristics associated with birth outcomes, birth certificates do not provide information on other potentially relevant information, such as pre-migration characteristics (e.g. education, occupation status, work experience, and social class in their home countries), health behaviors during pregnancy, and post-migration social position and health care use. To compensate, we further adjusted for comparability scores[28] designed to account for unmeasured characteristics of the study settings and the results remained statistically significant.

Third, the lack of information on the type of migrants and the motivations for migration hampers our ability to explore in greater depth the nature of selective migration in our main analyses. Our a priori hypothesis assumed a pattern based on legal and voluntary migration. While most immigrants fall within this group, many immigrants may also be refugees or unauthorized. Births to both refugees and unauthorized immigrant mothers are captured, however, in both Canadian and Spanish birth certificates but it was not possible to identify them to conduct sub-group analyses. In the Ontario sub-sample, however, information on refugee status was available but refugee status was not associated with the study outcomes. The Ontario cohort was restricted to legal immigrants to Canada and therefore could not shed light on whether birth outcomes are different between legal and undocumented migrants.

Finally, we could not provide a more complete picture of the potential effects of selective migration on birth outcomes by including data of Latin American women who did not migrate and gave birth in their home countries.

Significance of the study

To our knowledge, this is the first study focusing on the impact of selective migration on birth outcomes in relation to the immigrant admission context of the receiving countries. A recent study comparing preterm birth rates of Mexico-born women delivering in Mexico and in California did not find evidence of a healthy immigrant effect.[10] This study, however, could not differentiate between authorized and unauthorized migrants, who are numerous in California, and therefore whether the healthy migrant effect applies differently to these two groups. A study conducted in New York City suggested that undocumented foreign-born Latinas had higher low birthweight rates than documented foreign-born Latinas.[34]

Our findings suggest that characteristics of the immigrant-receiving countries may be associated with differential selective migration patterns, which in turn may translate into different health outcomes. More specifically, the more restrictive immigration admission context in Canada seems to produce a “cream skimming” effect, further selecting women with characteristics conducive to favourable birth outcomes beyond self-selection to migrate. Selective migration of Latin Americans may take place both in Spain and Canada, but our findings suggest that it is stronger in Canada. While self-selection to emigrate may result in better birth outcomes among those who emigrated compared to those who did not migrate, Canada’s more restricted admission policies place a second filter on those who self-select themselves for migration, thus boosting the selection effect.In addition, Spain’s Latin American immigrant population has grown since the 2000s partly because of high demand for low-skill temporary laborers, many of whom became permanent residents through the 2005 regularization program. In Canada, there has not been a similar strong pull factor attracting lower SES Latin American immigrants.

While the selection hypothesis places emphasis on pre-migration characteristics, a potential alternative explanation to our findings is that the observed differences may reflect a more favourable post-migration socioeconomic environment in Canada than in Spain. While most Latin American women giving birth in Spain were recent immigrants, since immigration to Spain is a recent phenomenon that achieved its peak during the study period,[35] immigration to Canada has been more stable, meaning that Latin Americans have spent a longer time from arrival to the time of giving birth there. If the higher birthweight of Latin Americans in Canada is a result of adaptation to the local environment and improved living conditions, we would expect an increase in mean birthweight with longer stay in Canada. However, our analyses based on the Ontario sub-sample do not support that hypothesis. Prolonged time from migration to delivery was not associated with better outcomes thus contributing to a relative advantage of Canada versus Spain. Indeed, preterm birth increased rather than decreased with longer stay in Canada, as observed in other immigrant studies,[36,37] which could have contributed to a dilution of the cross-national differences. A longer residence of Latin American women in Canada does not explain our results.

Differences in the proportion of refugees in Canada and in Spain are unlikely to contribute to explain the observed differences. In our sensitivity analyses restricted to Ontario data, we did not find an association between refugee status and birth outcomes. Refugees to Spain are very few[20] and unlikely to influence birth outcomes of Latin Americans in Spain.

Finally, the presence of unauthorized immigrants may play an important role in explaining our findings. Although the unauthorized immigrant population in Canada is small,[18] it constitutes a sizable proportion of the overall immigrant population in Spain[21,22] and an unknown proportion of Latin American women giving birth in Spain during the study period may have stayed without governmental permission. In Spain, unauthorized residents have rights to healthcare during pregnancy, partum and postpartum. It was estimated that unauthorized stay accounts for about 25% of female immigrant workers aged 20 to 40 years in Spain in 2008.[38] Unauthorized immigrants may be negatively selected in terms of human capital and may experience socioeconomic disadvantage, challenges with healthcare access, discrimination, and fear of deportation in the receiving country;[39,40] stressors which may contribute to poor birth outcomes and therefore eroding any potential advantage associated with positive self-selection.

In sum, our findings provide support for the selective migration hypothesis, particularly due to a stricter immigrant selection in Canada compared to Spain. While most research on immigrant health focuses on the characteristics of the immigrant populations in relation to their places of origin, this study highlights the importance of paying attention to the destination environments, particularly the combination of “push-pull’ factors may determine the health trajectories of newly arrived immigrants. Future studies focusing on the differences in health outcomes between authorized and unauthorized migrants may help clarify the roles of self- and state-selection and the healthy migrant effect.