Introduction

Numerous Swedish studies have shown that immigrants from non-European countries as well as most of the European immigrant groups suffer more from mental health problems than the indigenous people. In these studies different measurements of low mental health have been used [1, 29, 35]. Register based comparisons between immigrants and native Swedes concerning suicide [10, 18] and psychiatric admission [17] have also shown that such events are more frequent among foreign born.

Several risk factors for mental illness, such as being female [16], old age [26], low socio-economic status (SES) [19, 22], poor social support [30], poor financial situation [28] and unemployment [8], have been identified. It has also been shown that immigrants in Sweden are more socially and economically disadvantaged than native Swedes and that some immigrant populations are particularly disadvantaged [42]. In this study we will investigate if immigrants’ higher prevalence of mental illness compared with native born Swedes can be explained by higher prevalence of risk factors of social or economical nature. The present study use data from the PART-study which was specifically designed to investigate risk factors for mental illness. We deploy two well validated scales to assess mental illness (Major Depression Inventory and WHO (ten) wellbeing index). Only individuals with Swedish citizenship were eligible for participation in the PART study. The respondents have thereby had substantial contact with the Swedish society and are less likely to suffer from post-traumatic stress syndrome as it tends to disappear over time [25, 40].

The aims of this study are: (1) to investigate the association between immigrant status and mental illness among immigrants who have attained Swedish citizenship. We hypothesise that immigrants will have a higher prevalence of mental illness. (2) To investigate the association between immigrant status and exposure to risk factors of social and economic nature. The hypothesis is that immigrants are more exposed to such risk factors than native Swedes. (3) If the supposed association between immigration status and mental illness exist our aim is to further investigate to what extent previously known risk factors can account for the association between immigrant status and mental illness. The third proposed hypothesis is that immigrants’ poorer social and economic living conditions can “explain” the association between immigration status and mental illness.

Method

All subjects in the following study have answered a questionnaire that was included in the PART-study’s first phase. The PART-study is population-based, aiming to study risk factors for mental illness. The questionnaire contains questions on demography, life events, social support, working conditions, unemployment, physical health and several screening instruments for mental illness. The questionnaire was sent to 19 744 randomly selected inhabitants between the age of 20–65 in Stockholm county with Swedish citizenship during the years 1998–2000. The response rate was 53% (n = 10,441).

Outcome variables

In the WHO (ten) wellbeing index, both cognitive and affective elements are included in order to adequately measure subjective wellbeing (SWB)[2]. The scores from all the 10 items were added up ranging from 10 to 40, where higher scores indicated lower SWB. Those participants who had a total score above 29 were classified as having low SWB, a cut-off point also used by Forsell [12]. Her study showed that among those that met this criterion 62% had a psychiatric diagnosis on axel 1, implying that low SWB and mental illness are two highly overlapping phenomena. The internal consistency was very good for both the Swedish born subjects (Cronbach alpha 0.91) as well as for subjects born outside Europe (Cronbach alpha 0.90).

Major Depression Inventory, MDI was used to screen for depression. The scale’s 10 items were summarised to a score between 0–50, with each item giving a score from 0 (‘not at all’) to 5 (‘all the time’) in accordance with the instructions given by the author of the scale [3]. In order to be classified as suffering from depression the respondents had to have a total score above 19. This cut off point is suggested to be appropriate to use in population studies [11]. The Cronbach alpha calculations conducted showed that MDI had an excellent internal reliability both for the Swedish participants (0.88) and participants born outside Europe (0.91).

Immigration status

The indigenous Swedes and the foreign-born subjects are categorized into four different groups. The classification scheme is based on how classification of immigrants usually is performed within the official Swedish statistics. Although, this classification procedure, we believe, can also serve as an approximation of the respondents’ cultural distance to the Swedish society. The Swedish group includes all respondents who were born in Sweden. The Scandinavian group consists of individuals born in Finland (the majority), Norway, Denmark and Iceland. Respondents included in the third group were born either in Europe or in westernised countries except the Scandinavian countries, i.e. USA, Canada, Australia and New Zealand (only 8 subjects in the survey were born in these four countries). Respondents born in Poland, Germany and Yugoslavia are the largest subgroups within the European born category. The final group is comprised of individuals born outside Europe, except those from the westernised countries mentioned above. Notable nationalities included in this group are Turks, Iranians, Iraqis, Chileans and Lebanese, but also many individuals originating from other parts of Asia, South America and Africa. When comparing our sample with the official demographic statistics for Stockholm County, this sample turned out to be representative [34]. 18 respondents (0.2%) did not state their birth country in the questionnaire and were thus excluded in the analyses.

Potential confounders

Age is categorized into the following three groups: 20–34, 35–49, and 50–65.

The marital status was classified as yes or no depending on whether the respondents had stated that they at present were living together with another adult person.

The education variable specifies the respondent’s total years of schooling. It consists of three different values: “1–9 years” “10–12 years” and “more than 12 years”.

Poverty/economic security is measured with the question: “If you suddenly encounter an unforeseen situation and you had to get hold of 14,000 SEK (approximately 1,750 US$) within a week, would you be able to manage that?” The four alternative answers are: “Yes”, “Yes, probably”, “Probably not” and “No”.

The respondents’ level of income is based on self reported classification into one of five possible categories according to the respondents’ annual household earnings without possible subsidies. The income variable’s five possible categories are: “Less than 100,000 SEK”, “between 100,000 and 149,000 SEK”, “between 150,000 and 199,000 SEK”, “between 200,000 and 299,000 SEK” and “More than 300,000 SEK”.

The labour market position variable categorises the respondents into seven different groups. The actively employed or self-employed are classified into one of the following three categories: “blue collar worker”, “white collar worker” and “self employed”. These classifications were done in accordance with the classification system used by Statistics Sweden [31]. To also include those subjects who were not employed at the time of the survey in the analysis, we also included the following four categories: “unemployed”, “early retirement”, “students” and “early retirement as a result of poor health”.

Six questions about the availability of social integration (AVSI) e.g.”There are persons in my surroundings from whom I easily can ask for things e.g., borrow tools or kitchen ware”, and “Apart from those at home, there are others I can turn to if I am in trouble” [14, 37] was used to assess the respondents’ social support network. Those participants belonging to the lowest 25% when the six items were added up were classified as having a poor social support network.

Statistical analysis

For statistical analysis the computer program SPSS 11.5 was employed. First, the prevalence of each potential confounding factor and the outcome variables in different immigrant groups were shown. Second, the prevalence of depression and SWB were shown in categories of the confounding factors. Third, the association between immigrant status and depression and SWB was investigated. Fourth, the associations between immigrant status and depression and SWB was investigated by adjusting for all independent variables in multivariate logistic regression models. The results are shown as odds ratios (OR) with 95% confidence intervals (CI). To test the robustness of the full models, two Hosmer–Lemeshow’s goodness-of-fit tests (H–L test) were performed. The full model applied on depression had according to the test an acceptable fit (0.36), while the full model applied on low SWB showed a poor ‘goodness- of- fit’ (0.028), indicating that the model did not adequately describe the associations. In this case the insertion of parameters representing interactions between the independent variables was regarded as the reasonable way to try to obtain a more adequate model [6]. Thus, all two-way interactions were tested. One of them turned out to be significant (Position on the labour market × Level of education) in relation to low SWB. When this interaction term was applied the immigrant groups’ odds ratios changed marginally (between 0.00 and 0.02), and the H–L test showed an acceptable model fit (0.23).

Results

In Table 1 it is shown that the different groups have a markedly different age structure. The Scandinavian and European born group consists of fewer younger individuals in comparison with the Swedish born group, while the outside Europe born group is considerably younger. This is consistent with the demographic characteristics found within the total population of Stockholm [33]. The foreign born groups have both a lower annual income and poorer economic security than the Swedish born group and the outside European born group has much less economic resources or means than all the other groups. The level of education between the groups is quite similar, with the exception of the Scandinavians who have a lower level of education. The four groups have approximately the same probability of living alone, while the labour market position variable shows that the outside Europeans are over represented in the “blue collar worker” and unemployment categories. In addition, having a poor social support network is more common among the participants born outside Europe. It is also more common to suffer from mental illness among the participants born outside Scandinavia (e.g., 7.3% of native Swedes suffer from depression vs. 22.2% of the non-Europeans). In our sample the non-Europeans have on average lived in Sweden 17.7 years (SD 7.3), while the average for Europeans is 28.7 (SD 13.3) respectively 33.4 years (SD 10.9) for the Scandinavians.

Table 1 The distribution of the explanatory and outcome variables in percentages
Full size table

Table 2 shows that all the explanatory variables contribute to lower the risk of depression and low SWB. To be female, have a low level of education, be a “blue collar” worker and not cohabite with an adult person is associated with poor mental status in both the measures used. A low annual household income, poor economic security, unemployment and disability pensions are in the sample strongly associated with mental illness.

Table 2 The distribution of the explanatory variables in relation to the outcome variables in percentages
Full size table

Table 3 shows the association between category of country of birth and low SWB with the Swedish born population as reference category. In model 1 it is shown that those born in Scandinavia do not have an increased risk compared with the Swedish born population while to have been born outside Scandinavia are significantly associated with low SWB. In model 2–6 the effect on the associations found in model 1 of each potential confounder is shown. Model 7 (full model) further shows that both the outside European and the European groups’ increased relative risk for low SWB disappeared when all the explanatory variables were considered (born outside Europe: OR 0.98 CI (95% 0.76–1.27) and born in Europe: (OR 1.12 CI (95% 0.82–1.53)). The association between region of birth and depression showed a similar pattern as for SWB. However, the outside European group had a significantly increased risk of depression even after adjusting for risk factors (OR 1.62 CI (1.22–2.14)).

Table 3 Odds ratios (OR) with 95% confidence intervals (CI) for low SWB and depression in different models, (main effects) for groups of individuals with different birth countries. The Swedish born group is the reference group
Full size table

Discussion

The study revealed three important findings. First, our analysis showed that non-Scandinavians more frequently meet the different criteria for depression and low SWB. Second, members of those immigrant groups were also more often exposed to socioeconomic risk factors known to be associated with mental illness. Third, socioeconomic risk factors explained most of the immigrant groups’ increased risk for mental illness (e.g. −78% for depression and −98% for SWB in the non-European group). However, even after adjustment for risk factors the risk of depression was still significantly increased in the non-European group.

The outcome instruments have all been widely used. They are constructed to be applicable in different cultural settings and must be regarded as having a good cross-cultural validity as far as it is possible to establish. Furthermore, the two used ‘caseness criterions’ appear to be appropriate to use also in population based samples [11, 12]. It needs to be emphasised that the classification of immigrants into three different groups is a crude approximation of the respondents’ cultural distance, as the groups by no means are culturally homogenous. Particularly heterogeneous is probably the outside European born group. The cultural distance to the Swedish society for these respondents varies. Even though, this crude proxy for cultural distance is a limitation for the study and it would be preferred to asses the subjects’ cultural distance on an individual level, we still believe that this approximation tell us something about the respondents’ cultural distance to the Swedish society.

To examine if the slightly unevenly distributed missing data was likely to influence the results an additional logistic regression analysis was performed by including the missing data as separate dummy-variables. There was no indication that missing data was a cause for concern as the odds ratios were almost identical to those presented in Table 3. Whether the incomplete data on the outcome variables might influence the associations are more difficult to assess. It should be noted, that missing values was not completely random and was associated with immigrant status.

A non-response analysis [24] showed that respondents of the PART-questionnaire were more likely to be of female gender, over 50 years of age, have high income, be well educated and originate from Sweden or other Nordic countries in comparison to the non-participants. The analysis also showed that the odds ratios associated with these factors for receiving a psychiatric diagnosis in in-patient care were strikingly similar among the participants and non-participants. It thus seems that data from the PART-study may rather accurately reflect the true associations between immigrant status and other variables in this paper.

In cross-sectional studies the directions of the causality between the variables are generally not possible, to determine. Thus explanatory variables such as poor financial situation, poor social network, living alone and low socioeconomic status may be consequences as well as causes of mental illness. Arguments regarding causal inference in cross-sectional surveys must usually rely on earlier research findings from prospective designs. Studies of this kind have determined that social causation is a much more potent explanation for mental illness than the social selection hypothesis [7, 15, 22]. Even so, it is not clear how well these findings are applicable to immigrants. Most of the immigrants arriving to Sweden will initially have a poor socioeconomic situation, especially those that originate from less developed countries. A reasonable argument would be that immigrants with good mental health will be much more successful in improving their SES than those with mental illness. Low SES can in this scenario be perceived as a good marker for mental illness attained during the pre-migration period. However, it can be argued that because migration usually involves a tremendous effort, individuals with mental health problems (at least of a neurotic character) would be less likely to execute such a project. In a Swedish study it was concluded that exposure to pre-migration violence had no or very marginal importance for mental illness among immigrants [36].

The increased risk of depression found in the univariate analysis was substantially reduced (−78%) when adjusting for the confounding factors. However, a small significantly increased risk remained. The confounding factors were probably to some extent mis-classified which may have led to underestimation of the role of those factors. There may also be residual confounding from factors not controlled for in the analyses, such as area of living [32].

It has been suggested that a high level of status incongruence is a crucial factor to why mental illness are more common among immigrants [21]. However, a rather crude performed test, where highly educated individuals with a ‘blue collar’ occupation or currently unemployed were compared with individuals with same labour market position but less educated showed that those groups had the same prevalence of depression. This indicates that the unexplained relative risk of depression among non-Europeans could not be attributed to status incongruence.

It is also unlikely that the unexplained risk of depression among non-Europeans is an effect of poor social-cultural adjustment [38] as the immigrants in our sample has lived in Sweden for a fairly long period of time and all respondents had enough knowledge in Swedish to complete the questionnaire. The fact that the most culturally distant immigrant group’s excess risk was higher than the other groups might suggest that individuals belonging to this group have had greater difficulties to adapt psychologically to the new cultural setting [3, 39]. It has moreover widely been reported that perceived ethnic discrimination has a negative effect on a person’s well-being [27, 41].

Our finding that migrants’ rates of mental illness, indicated by MDI, are not fully explained by economic and social disadvantages are in line with previous performed Swedish studies [1, 29, 35]. In no previous study has the entire excess risk associated with immigrant status been explained by social and economic factors as is the case with low SWB in this study. It is plausible that some of the discrepancies between our study and former studies can be attributed to the fact that the immigrant sample in the present study on average had lived in Sweden for a longer period of time. There is no clear-cut empirical evidence on how the amount of time spent in a new country is related to mental illness [4] p. 368, but as stated above post-traumatic stress syndromes tend to diminish with time.

Research from different scientific disciplines have pointed out that there exist substantial cultural variations in the way individuals perceive and express mental illness [5, 9, 13, 23]. Kirmayer et al. [20] has for example argued that psychologization of ones mental problems actually can be regarded as a western cultural-bound syndrome. Although research have indicated that the outcome measurements used in this study are valid in different cultural settings, it is possible that at least part of the difference in odds ratios between the measurements are a consequence of cultural specific idioms of perception and presentation.

Conclusion

The findings in this study suggest that the association between immigrant status and mental illness appears above all to be an effect of a higher prevalence of social and economic disadvantage.