Strengths and weaknesses
The results of this study are based upon an exhaustive nationwide dataset covering sociodemographic, socioeconomic, migration and mortality information for the total de jure population in the period 2001–2011. These variables are stemming from three data sources (the Census, the National Registry and the death certificates) which were individually linked by Statistics Belgium, hereby avoiding a numerator-denominator bias. We disposed of information on all legal residents in Belgium at the moment of the census (October 1st 2001), and whether or not they emigrate or die during follow-up. The data are left truncated, meaning that immigrants who arrived later than the census were not included in this study.
Additional migration information such as the reason for migration was not available in the dataset. We were able to include all important migrant groups in Belgium, thereby maximizing the population with migrant roots. In addition, as men and women have a different migration history [1, 2, 4, 6, 13, 20], we decided to focus on the interplay between migrant origin, gender, SEP and cause-specific mortality, which is a topic that is still understudied [26]. Men used to migrate for work purposes and therefore had to be in good health and were more likely to receive medical checks at the workplace. Women on the other hand migrated for family reasons and often did not work outside the home. This different trajectory may have implications both for their health and for their SEP, as we indeed observed in our results (see next paragraph). Another strength of this paper is that we included three different SEP indicators, all representing different forms of disadvantage during different periods in life [23], and clearly having a different meaning in explaining mortality inequalities for men and women, and by migrant origin or COD. However, these indicators also have limits, e.g.: employment status does not contain information on the type of job, the working regime or the job history. Similar for home ownership, which is a proxy for the economic wealth of the household, yet it does not include information on the type or quality of the home, or the household income. These additional indicators of SEP could also have been informative, but were not present in our dataset. Including inequalities in morbidity patterns could be informative as well about the pathways behind the observed mortality patterns but this information was not available. Similarly, due to the semi-closed nature of the dataset, we were not able to include exposures experienced during the life course, prior to the migration to Belgium. Neither did we obtain data on health behaviours or healthcare use, which are likely to be gender- and culture-specific [6, 29,28,29].
Reflections on the main findings of the study
While describing the general characteristics of the various groups, we observed that the migrant community in Belgium is a very diverse group, as was stated in previous research as well [2]. Large variations have been observed for all variables of interest. In general terms we could discern different profiles. Migrants from Dutch descent were quite similar to Belgians in terms of their socioeconomic profile, while French migrants had a somewhat lower SEP compared with Belgians. Migrants from Italian, Turkish and Moroccan descent were those migrant groups that belonged to the early migration waves. They generally were situated in the lower socioeconomic strata. Migrants from Spanish descent are a diverse group including both traditional labour migrants as well as a recent influx of highly educated immigrants. Migrants from SSA descent belonged to the most recent influx of immigrants, and were more likely to be higher educated, yet living in rented dwellings. Furthermore, high levels of SSA migrants were unemployment but looking for a job. The last group, the migrants from Eastern European descent were a mixed group of traditional labour immigrants and more recent immigrants that were characterized by quite similar educational levels as the native Belgians, but higher levels of living in rented dwellings and unemployment. Finally, some important differences were found between men and women, indicating a different migration history and dependent on the country of origin. The fact that men most often migrated for job issues while women later followed for reasons of family reunification [2, 6, 20] was reflected in our results, at least among the earliest migrant groups. In these groups, women were most often unemployed at home while men were more often at work.
The unadjusted cause-specific mortality analyses by migrant origin confirmed the migrant mortality paradox for most migrant groups and all studied COD, as was also observed in previous studies [1, 4,5,6,7,8, 30]. Even though immigrants were more often situated in the lower socioeconomic strata, they tended to show advantageous mortality patterns. Among the different migrant origins, largest mortality advantages were found for Turkish and Moroccan migrants, without and with adjustment for SEP. This could be explained by the fact that compared with native Belgians, these traditional migrant groups may maintain a healthier lifestyle with earlier reproductive behaviour, longer breastfeeding periods, lower levels of alcohol and tobacco consumption, and a Mediterranean diet with high fruit and vegetable consumption [1, 6, 8, 9, 13, 17, 20, 26, 31,32,33,34]. Migrants from less industrialized countries thus seem to profit from the double advantage of maintaining the healthy lifestyle of the home country combined with the high-quality and accessible healthcare system of the host country [1, 7, 12, 13]. In contrast, migrants from French and to some extent Eastern European descent were the exceptions showing higher all-cause and cause-specific mortality compared with the native population. Again, part of the explanation is likely to be found in a difference in lifestyle as previous research has shown that these groups have high levels of tobacco and alcohol consumption [35]. For some part, the mortality disadvantage among French and in particular Eastern European migrants can be explained by their lower SEP, as in the adjusted models some of their excess mortality disappeared.
In general, we can say that men showed more variation in mortality by migrant origin than women, which we expected based on the healthy migrant effect. Besides the French and to some extent Eastern European migrant men, the mortality differences were generally in favour of the migrant groups. The fact that more differences were observed among men could be related to the differential history of migration: men used to migrate for job reasons and therefore had to be in good health [6, 13, 20]. Moreover, as men were more often employed, they were perhaps more likely to receive medical checks at work [13]. Women on the other hand migrated for family reasons, did not have a job but instead took care of the household, which did not necessarily demand good health. However, a counterargument is that the type of jobs offered to migrant men and women are different, implying different health risks [26, 36]: where women were more likely to be employed (if employed) in domestic or caretaking services, men were more likely to work in risky sectors as mining, agriculture or construction. On the other hand, not having a job and being dependent of their family or partner makes them even more vulnerable as they have a double disadvantage of being both a migrant and a women [26, 36]. However, if this was true, we would perhaps have observed a larger impact on the mortality differences among women when adjusting for SEP, which wasn’t the case. Instead, SEP seemed to have a larger impact on the mortality patterns of men than of women.
Adjusting the mortality differences for SEP generally resulted in an increased mortality advantage, although there was quite some variation by gender, COD and migrant origin. Among Turkish and Moroccan migrants, adjusting for the indicators of SEP even increased their mortality advantage compared with the native Belgian population. This did not come as a surprise since they belonged to the early migration flows who were more often situated in the lower SE strata. Among the other traditional early migration group, the Italians, adjustment for educational attainment and employment status resulted in an improvement of the mortality patterns whereas adjustment for ownership had no effect or a reverse effect. From the descriptive analyses we can deduce that people with an Italian background had similar levels of home ownership but lower levels of high education and employment, which could explain these results. Some of the migrant groups (Turkish, Moroccan and SSA) are on average younger, which could be related to this, as ownership increases with age. Among migrants from Eastern European descent, the mortality disadvantage disappeared after adjustment for SEP. On the contrary, among French migrants, the mortality disadvantage remained significant among men while among women the excess mortality had disappeared. Exceptions to these findings were Spanish and SSA men, whose mortality patterns even deteriorated after accounting for education. Striking however, no such change was observed among Spanish and SSA women. SSA men had an initial mortality advantage as a result of the protective effect of their educational level, while SSA women did not have a mortality advantage due to their high levels of unemployment.
As mentioned before, we observed quite some variation by indicator of SEP. There is not one indicator that was most important for all groups, yet the impact of the indicators of SEP varied by COD, migrant origin and gender. This stresses the importance of including SEP, and more specifically including more than one indicator of SEP when assessing mortality inequalities in the population. We expected that ownership would be a more important SEP marker for women as it reflects the financial means at the household level, including the earnings from the husband [37]. Since women are less likely to be employed and if employed have lower wages than men, we assumed this to be an important issue, certainly among the traditional migrant groups where employment levels are rather low. Yet, this was not reflected in our results, which could be due to the fact that we combined first- and second-generation migrants. The only main finding when it comes to indicator of SEP was that among men, and especially Spanish and SSA men adjusting for education had a negative impact on their mortality patterns, which was due to their high educational levels.
