Study design

The Healthy Life in an Urban Setting (HELIUS) study is a multi-ethnic cohort study conducted in Amsterdam in the Netherlands. It has been described in detail elsewhere [22,23]. In brief, baseline data collection took place from 2011 to 2015 and included residents of Amsterdam of various ethnic origins in the ages of 18 to 70. Participants from six ethnic groups–of Dutch, African Surinamese, South-Asian Surinamese, Turkish, Moroccan and Ghanaian ethnic descent–were randomly selected from the municipal population register, stratified by ethnicity to ensure roughly equal numbers from each group. Sociohistorical information on the ethnic minority groups included in the study is provided elsewhere [23]. Data were collected by questionnaire and in a physical examination, in which biological samples were also obtained. Participants who were unable to complete questionnaires in Dutch were provided questionnaires in English or Turkish or received assistance from a trained, ethnically matched interviewer.

For the current study, cross-sectional baseline data were analysed. From the total sample completing the baseline questionnaire (N = 23,942), we first excluded participants not belonging to the six ethnic groups listed above (n = 586). We also excluded all respondents of Ghanaian origin (n = 2,484), as they included only a small number of immigrant offspring. From the remaining 20,872 participants, we excluded those with missing scores on the depressive symptomatology (n = 183), on the acculturation measures (n = 455), on any of the three measures of socioeconomic position (n = 257), on Dutch language proficiency (n = 16) or on the discrimination scale (n = 57). Our final study sample thus consisted of 19,904 participants: 4,591 of Dutch origin, 3,213 of South-Asian Surinamese, 4,212 of African Surinamese, 3,800 of Turkish and 4,088 of Moroccan ethnic origin.

The Medical Ethics Committee of the Amsterdam Academic Medical Centre (AMC) approved the study protocols. Written informed consent was obtained from all participants.

Variables

Statistical analyses

Baseline sociodemographic characteristics and social conditions are reported for respondents from the host population, for immigrants and for immigrants’ offspring, differentiated by ethnicity. Our first aim was to assess the prevalence of depressed mood in immigrant offspring relative to that in the Dutch host population. We used logistic regression models, controlling for age and gender, and obtained average marginal effects (AME) for each ethnic minority group [34]. AMEs were estimated using the margins post estimation command in STATA (College Station, TX: StataCorp LP). Effects of all predictors in the model are calculated for each observation in the dataset, and then averaged. The resulting average marginal effect can be interpreted as the average increase in the probability of depressed mood over all values of the covariates [35]. As an example, an average marginal effect of .20 indicates a 20% higher probability of depressed mood as compared to the reference group. We prefer AMEs above ORs, given that our aim is to compare the contribution of social determinants to the probability of depressed mood across groups, in particular generations as well as ethnic groups. When making comparisons across groups, ORs can be misleading as an indicator of the size of the contribution of each predictor as a large change in odds ratios does not always correspond with a large change in predicted probability. In contrast, AMEs do allow for a fair comparison of effect sizes across groups [34]. Finally, since the age ranges among immigrants and immigrants’ offspring were very dissimilar, we did not directly compare the prevalence of depressed mood. Any differences between groups should be viewed with that in mind.

We also assessed the associations among immigrant offspring between indicators for social conditions (socioeconomic position, perceived discrimination, sociocultural conditions) and depressed mood, controlling for age and gender. Our second research aim was to analyse whether the distributions of social conditions accounted for the risks of depressed mood in immigrant offspring. To that end, we used logistic regression models to compare the risks of depression in the immigrant populations (both first generation and offspring) and the host population, adding the indicators for social conditions to the model both one by one, as well as a set (socioeconomic position, perceived discrimination, sociocultural conditions respectively). Because the host population had no values on the sociocultural indicators, we also created models in each immigrant and offspring group based on what we considered the most favourable sociocultural profiles–integrated ethnic identity, integrated cultural orientation and integrated social network. We used decreasing AME as an indicator of the relative importance of a specific social condition for the higher depression risk.

To evaluate whether the patterns observed in immigrants’ offspring resembled those in immigrants (our third research question), similar statistical analyses as described were performed on the immigrants in the sample.

Analyses were conducted using IBM SPSS Statistics for Windows (Armonk, NY: IBM Corporation) and STATA Statistical Software, Release 15 (College Station, TX: StataCorp LP). The level of statistical significance was set at p < .05.

Limitations of the study design

As this study used cross-sectional data, we cannot exclude the possibility that the social conditions of participants would be the consequence, rather than a cause, of depressed mood. E.g. a depressed mood might have hindered people to have contacts with other people, increasing the risk of becoming ‘marginalised’. Longitudinal data will be required to demonstrate whether the prevalence of depressed mood will decrease if social conditions for the second generation improve.

A second limitation is the use of self-reported data on depressed mood, based on the PHQ-9 among different ethnic groups. A recent analysis on the measurement invariability of the PHQ-9 showed that the PHQ-9 items were measuring the same underlying construct in the different ethnic groups included in our study [27]. That suggests that PHQ-9 is a good measure of depressive symptoms in those groups. The PHQ-9 outcome is not equivalent, though, to a diagnosis of clinical depression, implying that our results are not automatically generalisable to an outcome of clinical depression.

Third, we were not able to draw firm conclusions about whether the prevalence of depressed mood in immigrants’ offspring was lower than or similar to that in people of the first generation. The indirect comparison, whereby the host population was used as the reference group to both the immigrants and the immigrants’ offspring, led us to conclude that the risk in the offspring was slightly less elevated than that in the immigrants. However, in view of the minimal overlap in the age distributions of the two generations, that conclusion may be biased by an age effect. A direct comparison of prevalence in both generations might have been more reliable, but it yielded wide confidence intervals, probably related to the age distribution differences. To enable conclusions on changes in risk across generations, future research should include both first and second generations and ensure, if possible, substantial overlap in age range.

Interpretations and implications

In all four minority ethnic groups we studied, the prevalence of depressed mood in immigrants’ offspring was higher than that in the host population. In contrast with the ‘immigrant health paradox’, a similar pattern was observed for immigrants. This result also contrasts with several previous studies that have found comparable rates of depression in immigrant offspring and host populations. Studies based on the European Social Survey, for instance, found no elevated risk for immigrant offspring in Europe [1,8]. Those studies, however, combined data from ethnic minority populations into one category, thereby possibly masking differences between ethnic groups. A study in Australia, as a case in point, has indicated a higher prevalence of mental health problems in the offspring of immigrants of European descent, but not of other ethnic groups [14]. This highlights the importance of distinguishing by ethnic background in future studies.

The size of risk differed between ethnic groups in our own study, in fact, with offspring of Turkish descent showing the highest risk and those of African Surinamese descent the lowest. Since we found similar within-group risk patterns for immigrants and offspring, that could point to risk factors inherent in particular ethnic groups, such as culture or genetic profiles. The higher risk of depression in offspring of parents with depression is well documented [36], and it could be linked to genetic factors or to family interaction (communication, experience of affect). Such factors could partially underlie similarities in pattern across generations in a particular ethnic group. In addition, the differences in size of the risk between ethnic groups might reflect the prevalence of depressed mood in the country of origin [4]. The prevalence of this health problem has been shown to differ between regions of the world, with, e.g., a relatively low prevalence in Sub Saharan countries where the African Suriname originally come from [37].

This is not to say, of course, that a risk of depression in a particular ethnic group is determined entirely by attributional factors–unique and rather stable attributes of that group such as genetics and culture [22]. Quite the contrary, our study has revealed that the higher risk of depression for immigrants and their offspring in ethnic minority groups is strongly associated with the living conditions to which the groups are exposed in the host country, so-called post-migration factors. This is in line with previous studies that show the salience of such social conditions for immigrants themselves. To the best of our knowledge, we are the first to study the impact of multiple types of social conditions simultaneously on the burden of depressed mood in immigrants’ offspring. Similarly to the results in previous studies of immigrants, we found that the higher risk in the offspring was accounted for by social conditions to which they were subject, in particular an adverse socioeconomic position and the experience of discrimination. Should that be confirmed in future studies using longitudinal data, it will be an indication that improving the living conditions of immigrants’ offspring–and in particular their socioeconomic position and their exposure to discrimination–could be a promising strategy to reduce depression rates in the offspring generations.

Whereas in our study the greater risk of depression in immigrants’ offspring could be fully explained by adverse socioeconomic conditions and discrimination, a higher risk partially persisted in immigrants after those social conditions had been taken into account. A small part of their remaining higher risk was explained by sociocultural factors, whereas the analysis for immigrants’ offspring left no room for additional contributions by sociocultural factors. Important for this difference seems the higher percentage of first-generation respondents who were marginalised or separated–subgroups that showed higher depression risk than immigrants classified as integrated. Apart from the minor difference in risk levels, though, it is important to note that the distribution of immigrants’ offspring across acculturation categories was largely similar to that of the immigrants themselves. In addition, also the risk of depression associated with specific acculturation categories was largely similar across generations, implying that also for the immigrants’ offspring, the process of acculturation might be as stressful as it is for first generation immigrant.

Also worth highlighting, and in line with the results of a meta-analysis by Yoon and colleagues [12], is that our respondents who were classified as integrated–that is, showing an orientation towards the cultures of both the host country and the country of origin–had the lowest risks of depressed mood. They thus showed a lower risk than those classified as assimilated (integrated into Dutch society with a low orientation to the country of origin). This provides indications that it is beneficial for people from ethnic minority populations to stay attached to the original country’s culture and people. Whereas this has been put forward as an explanation for the relative good mental health of immigrants who fit the ‘healthy immigrant paradox’ [14], our results provide indications to suggest that this beneficial effect also holds for immigrant’s offspring.

In our statistical analysis, we dealt with social conditions as three separate categories–socioeconomic position, discrimination and sociocultural situation. In reality, of course, such conditions are interlinked, and they could have an interactive impact on mental health. People without paid jobs, for example, may have less contact with people from the host country, with a consequent influence on their acculturation strategy. Similarly, also experiences of discrimination might impact on the process of acculturation [20]. Socioeconomic position can also impact mood through cultural factors such as low cultural entitlement–a feeling of not ‘being a relevant and legitimate citizen who matters in society’ [5]. Moreover, multiple categories of social conditions may share the same underlying mechanisms, such as institutional racism [38]. For example, institutional racism may reduce the likelihood that people from ethnic minority populations will get a job, and thereby weaken their sense of affiliation with the host society’s culture. Unravelling the complex ways in which such conditions interact, and probably also have synergic effects, is crucial if the aim is to counter the high risk of depressed mood in ethnic minority populations through improving their social conditions. The results of our study suggest that the way in which these conditions interact might be different for immigrants and immigrants’ offspring, given, e.g. the fact that, in our multivariable model, sociocultural conditions did not contribute to the increased risk of immigrants’ offspring whereas it did for immigrants. As we found similar associations in the univariate model, this might indicate that sociocultural conditions impact on mental health in the immigrant’s offspring through socio-economic conditions. The differential impact of discrimination on the process of acculturation between generations is another example of how mechanisms might differ for immigrants and immigrants’ offspring [20].