Article Text
Abstract
Background The association between unemployment and poor mental health in general is explained by both causation and selection. The aim was to study whether experiencing unemployment was a risk factor for hospitalisation for depressive disorder specifically, and whether gender and immigrant status modified the hypothesised risk.
Methods A register-based prospective cohort study, 2000–2006, of persons aged 18–64 with a strong connection to the Swedish labour market. Outcome: hospital admission for a depressive episode; F32 in International Classification of Diseases, 10th revision. Exposure: employment status. Explanatory variables: gender and immigrant status. Confounders: age group, education and marital status. Cox regression models were used to estimate HRs with 95% CIs.
Results The cohort comprised 3 284 896 adults, 47.5% women. An excess relative risk for hospitalisation was found among those who became unemployed (HR=1.94, 95% CI 1.85 to 2.03). Foreign-born women who experienced unemployment had the highest relative risk (HR=3.47 95% CI 3.02 to 3.98).
Conclusions Among persons with a strong connection to the labour market experiencing unemployment, is a risk factor for hospitalisation for depressive disorders. Unemployed foreign-born women had the highest relative risk compared with all Swedish born, all foreign-born men and to employed foreign-born women.
- LONGITUDINAL STUDIES
- DEPRESSION
- GENDER
- MIGRATION
- UNEMPLOYMENT
Statistics from Altmetric.com
Introduction
This study will focus on hospitalisation for depressive disorders following unemployment. There is an abundance of studies of the association between unemployment and poor mental health and well-being in general1 but little is known about diagnosed depressive disorders specifically. There are two main hypotheses explaining the association between unemployment and poor mental health and well-being in general. One is that unemployment causes poor mental health (causation), the other that poor mental health causes unemployment (selection). Meta-analyses on the association between unemployment and poor mental health and well-being in general suggest that the average effect of causation is moderate and the selection effects are smaller.1 ,2 In terms of depressive disorders specifically, a selection effect has been found.3 Whether unemployment increases the likelihood of depressive disorders is less established. Studies of depressive disorders following unemployment have concluded that depressive symptoms (ie, symptoms that, if experienced together with other relevant depressive symptoms, would be classified as a depressive disorder) increased after termination of employment.4 ,5 However, none of these studies found evidence for an increased likelihood of depressive disorder.
Gender, age, education, socioeconomic position, heredity and personal characteristics modify the likelihood of poor mental health and low mental well-being in connection with unemployment.1 For depressive disorders specifically, the known risk factors include heredity, gender, marital status and socioeconomic position.6 However, the role of these factors in relation to the risk of being diagnosed with depressive disorder following unemployment has not been studied.
In the early days of research on the association between unemployment and poor mental health, researchers hypothesised that women would be less affected by unemployment than men, partly explained by the homemaker hypotheses—that women identified themselves with the homemaker.1 Differences in labour force participation rates between men and women are smaller in Sweden than in any other country,7 and in no other country are attitudes to women on the labour market more positive than in Sweden, according to the World Value Survey.8 In the Swedish setting the homemaker hypothesis has been adapted to suggest that migrant women from countries with more traditional gender roles than in Sweden have a lower likelihood of poor mental health following unemployment than Swedish-born women and Swedish and foreign-born men. Meta-analytic studies of gender differences in poor mental health and well-being during unemployment have found inconsistent results. Some studies have found women at higher likelihood of poor mental health; others have found higher likelihood among men.1 ,2 A cross-sectional study on 413 persons from Bosnia-Herzegovina who came to Sweden in 1993–1994 found that job occupancy was important for the mental health of men, but not of women.9
The homemaker hypothesis is disputed.1 Backhans and Hemmingsson10 propose that socioeconomic disadvantage leads to differential susceptibilities of mental health. According to this theory, unemployed persons who are socioeconomically disadvantaged will have a higher likelihood of poor mental health than more privileged persons who are unemployed, because the financial and socioeconomic vulnerability associated with unemployment increase the likelihood of poor mental health. The theory does not imply that the likelihood of poor mental health associated with unemployment is due to financial problems only. It suggests that job loss increases the vulnerability of poor mental health among socioeconomically disadvantaged groups. Foreign-born women in Sweden, on average, have a disadvantage compared with natives and foreign-born men,11 also on the labour market.12 If socioeconomic disadvantage leads to increased susceptibility to poor mental health, foreign-born women who experiences unemployment would have a higher likelihood of poor mental health than natives and foreign-born men who experiences unemployment.
To our knowledge, no study has had the statistical power to determine whether experiencing unemployment is a risk factor for hospitalisation for depressive disorder and no study has tested whether migrants are at greater risk.
The overall aim of this study was to assess whether to experience unemployment is a risk factor for hospitalisation for depressive disorder among persons with a strong connection to the labour market, and to evaluate whether gender and immigrant status modify the relative risk of it. We hypothesised that among persons with an initial strong connection to the labour market there was an increased risk of being hospitalised for depressive disorder following loss of employment. We also hypothesised that women and foreign-born persons had increased relative risks of being hospitalised for a depressive disorder following experiencing unemployment and that gender and immigrant status would interact, increasing the relative risks among foreign-born women.
The research questions were:
-
Among the persons with a strong connection to the labour market, is experiencing unemployment a risk factor for being hospitalised for depressive disorder?
-
If so, do gender and immigrant status modify the relative risk of it?
-
If gender and immigrant status do modify this relative risk, do they interact?
Methods
Design
The study had a prospective cohort design. The register-based cohort was created to compare risk differences of hospitalisation for depressive disorder following experience of unemployment. Time was measured in years. The cohort was closed. Follow-up started in January 2000, and ended by censoring in December 2006. Before start of follow-up, all participants were followed during a wash-out period for 3 years in order to detect hospitalisation for depressive disorders prior to unemployment.
Study population
For a schematic view of the selection process see figure 1. The study base was persons aged 18–64 years living in Sweden in January 2000. At the start of follow-up, all those included had a strong connection to the labour market, defined according to Lundin et al13: namely not being unemployed according to the Swedish employment service, not being sick-listed or taking parental-leave, not having a disability pension and having an annual income above Swedish-krona, the currency of Sweden (SEK) 67 100. Income criteria were adjusted for inflation as compared with the income criteria used in Lundin et al's definition. The number of persons fulfilling the study-base criteria during the start of follow-up in the year 2000 determined the study size. Only migrants who obtained resident permits prior to the start of the wash-out period in 1997 were included. Persons who had left Sweden according to the Population Registration System and persons who could be assumed to have left without informing the Swedish tax authorities were administratively censored from the year they left Sweden, according to a method described by Weitoft et al.14 Persons who died during the time of the study were censored from the year of death. Immigrants who had been given residence permits for refugee-reasons were excluded as studies have found them to be at a higher risk of poor mental health than other migrants.15
Measures to avoid section bias
In order to avoid selection bias we took the following measures. Prior to the start of follow-up, all participants were followed during a wash-out period for 3 years, 1997–1999 (as performed by Eliason and Storrie16). Participants who were hospitalised for depressive disorders during the wash-out period were excluded from the cohort. All participants had to have a strong connection to the labour market at the start of follow-up, as described above. All those who were hospitalised for depressive disorder prior to being unemployed, were censored throughout the study. Persons who were sick-listed for more than two-thirds of a year, on disability pension or parental leave were censored from the year they left the labour force. Censoring the sick-listed was carried out to ensure that those who became unemployed did not have a depressive disorder prior to unemployment. The outcome (described below under Variables) included first depressive episodes only; excluding recurrent depressive episodes.
Data sources
Register-based studies use data in official registers collected for generic purposes. Sweden has a high standard of official registers adapted for research purposes.17 As a means of identification, all included in the registers are assigned a personal identity number in the Population Registration System.18 Given ethical approval and a permit, registers can be linked for research purposes; after linkage, the data are anonymised. This study was performed with data from Statistics Sweden (http://www.scb.se/) and the National Board of Health and Welfare (http://www.socialstyrelsen.se/).
Variables
Outcome
The outcome was taken from the Swedish Hospital Discharge Register. The validity of the diagnoses in the register has been tested and found to have an overall high quality.18 For Swedish citizens and persons with a permanent residence permit, inpatient care in Sweden is almost free of charge.18 Hospitalisation for depressive disorder was defined as a hospital admission for a depressive episode; F32 in the International Classification of Diseases, 10th revision (ICD-10). This definition excludes recurrent depressive episodes and bipolar disorders as well as all other mood disorders. The outcome was coded binary: has or has not been hospitalised (reference category, from now on abbreviated to ref).
Explanatory variables
Gender: Men (ref) and women.
Employment status: Unemployment experience was defined according to the Swedish employment service. It was coded binary: employed (ref) or lost employment during follow-up. The participants were split into one exposed and one non-exposed groups, and treated as distinct groups throughout, as suggested by Clayton and Hills.19 The latter included those who had been employed (a criterion for being in the cohort) but had (voluntarily or involuntarily) lost the employment and were still able to work. Those who left the labour force were censored, according to the criteria stated above.
Immigrant status: The variable was coded binary: Swedish born (ref) or foreign born.
Country of origin by region: Countries were clustered as: Asia, South and Central America, sub-Saharan Africa, Middle East, Turkey and North Africa, former Yugoslavia, EU-countries joining since 2000,i EES-countries joining before 2004ii and others.
Employment status-gender-immigrant status: Combinations of the variables employment status, gender and immigrant status (for definitions see above) gave the following parameters: Employed men Swedish born (ref), Employed men foreign born, Unemployed men Swedish born, Unemployed men foreign born, Employed women Swedish born, Employed women foreign born, Unemployed women Swedish born and Unemployed women foreign born.
Confounders
Age at the start of follow-up in the year 2000 was coded into five groups: 18–24 (ref), 25–34, 35–44, 45–55 and 55–64 years.
Education was coded into two categories: low (all those with less than 11 years of schooling) or high (all those with 11 years or more years of schooling (ref)). Statistics Sweden converted education completed outside Sweden into equivalent levels of schooling in Sweden.
Marital status at the start of follow-up in 2000 was classified as married (ref) or not married.
Economic resources in each category male Swedish born, male foreign born, female Swedish born and female foreign born were measured by the gross individual median income with 90% confidence limits (90% CI) from paid employment together with all benefits based on social insurance, measured in Swedish Crowns (SEK, 100 SEK about € 11).
Statistical analyses
Incidence rates were calculated as the hospital episodes for depressive disorder per 1000 person-years. Cox regression models were used to estimate HRs. Five models were fitted: (1) age group and employment status, (2) age group, employment status and gender, (3) age group, employment status, gender and immigrant status, (4) age group, employment status, gender, immigrant status and education and (5) age group, employment status, gender, immigrant status, education and marital status. Best fit was tested with a step-wise procedure. Models were compared with a −2 log likelihood test.
Tests of statistical power, as well as graphical and statistical tests of fulfilment of the Cox regression assumption of proportional hazards, were performed as suggested by Allison.20 Results are presented as HRs with 95% CI. The analyses were conducted with the SAS software package V.9.2.
Results
The study base totalled 5 229 025. Of these, 1 944 129 (37.2%) were excluded due to not having a strong attachment to the labour market, having immigrated to Sweden post December 1999, refugee status or were hospitalised for depressive disorder.
At the start of follow-up in January 2000, the study population comprised 3 284 896 adults with a strong labour market attachment, of whom 1 560 646 (47.5%) were women and 273 427 (8.3%) were foreign born. For the excluded there were small gender differences too (51.7% women), the proportion of foreign born, however, were larger among the excluded (32.9% foreign born). For characteristics of the included see table 1.
The unadjusted incidence rate per 1000 person-years indicated differences in rates of hospitalisation for depressive disorder in total, by gender, by employment status during follow-up and by immigrant status (table 2). Women had a higher incidence than men had and employed had a lower incidence than those who lost their employment. Foreign-born persons had a higher rate than Swedish-born persons. Women who lost their employment during follow-up had the highest incidence of hospitalisation for depressive disorder.
In order to test whether the risk of hospitalisation for depressive disorder following experience of unemployment remained after adjusting for gender, immigrant status, education and social factors, we tested various models. The model with the best fit included employment status, gender, immigrant status, education, age and marital status. The three variables: children at home, area of residence in Sweden and for migrants, duration of stay in Sweden did not alter the model. Table 3 displays the results of the model including employment status, age group, gender, immigrant status, education and marital status. The model that included employment status and age group only was compared with the full model. In the full model, 10% of the effect of loss of employment on the relative risk of hospitalisation for depressive disorder was explained by gender, immigrant status, education and marital status. When testing the full model but stratified by gender the relative risk of hospitalisation for depressive disorder among those who lost their employment during follow-up was higher than that of the employed for men as well as women (see table 3). When testing the full model but stratified by immigrant status the relative risk of hospitalisation for depressive disorder among those who lost their employment during follow-up was higher than that of those who maintained employed for Swedish born (HR=1.99, CI 1.90 to 2.09) as well as foreign born (HR=1.59, CI 1.40 to 1.80; table not shown).
We created a model with the combination-variable employment status-gender-immigrant status and adjusted for age group, marital status and education, with the results presented in table 4. Adding the combined variable increased the model-fit significantly. We also tested several models including 2-way and 3-way interaction terms between employment status, gender and immigrant status as well as the main effects. Of these models the full model with a 3-way interaction term had the significantly best model-fit. But compared with the model with a combined variable, the model-fit was approximately the same. Employed Swedish-born men had the lowest relative risk of hospitalisation for depressive disorders. Foreign-born women who experienced unemployment had the highest. There was no significant difference between Swedish-born and foreign-born men following who experienced unemployment.
Discussion
The hypothesis that, among persons with a strong connection to the labour market, there was an increased risk of being hospitalised for depressive disorder following experience of unemployment was confirmed. Unemployed foreign-born women had the highest relative risk.
This study shows that among persons with a strong connection to the labour market, experiencing unemployment is a risk factor for hospitalisation for depressive disorder. This contributes to knowledge that experience of unemployment is a risk factor not only, as previously known, for poor mental health in general,1 but for hospitalisation for depressive disorder too. Our findings show that the risk seems to be present when analysing the whole study population. The risk was modified by gender and immigration status. Foreign-born women had the highest relative risk of hospitalisation when experiencing unemployment.
Our findings regarding the adjustment of the relative risk of hospitalisation for depressive disorder by age group, education and marital status can be compared with those in the meta-analysis by Paul and Moser1 concerning unemployment and poor mental health and well-being in general.
Limitations
Studying the risk of any measure of poor mental health following experience of unemployment raises the question of whether the risk could be due to selection, instead of causation. A number of measures controlled the risk of selection bias; however, there is possibly still selection not controlled. In Sweden, just as in most European countries, hospitalisation for any depressive disorder is rare.
Only persons with severe forms of depressive disorders receive inpatient care. This study uses psychiatric hospitalisation for depressive disorders as an outcome. The outcome is not a proxy for depressive disorders in general but a measure of hospitalisation due to depressive disorder. Owing to lack of information about the prevalence of depressive disorders it is difficult to estimate the proportion of persons with a depressive disorder that are hospitalised due to the disorder.
Compared with many other medical disciplines psychiatry has a stronger cultural component to it. Owing to this, it may be more difficult to transfer the symptoms of poor mental to a valid Diagnostic and Statistical Manual of Mental Disorders (DSM) or ICD diagnosis for migrants than for native patients.21 Utilisation of health services is culture bound too.22 This could possibly dilute the association found between immigrant status and hospitalisation for depressive disorder.
What might be a possible pathway from experience of unemployment to the higher relative risk of hospitalisation for depressive disorder among foreign-born women? Identification with the homemaker role has been hypothesised to protect women, especially foreign-born women in Sweden, from the mental health consequences of unemployment. This study does not support that theory. Foreign-born women had the lowest income; hence, the findings of the present study are more in line with the theory that disadvantaged groups are more vulnerable to the adverse mental health effects of unemployment.
During the follow-up, Sweden experienced an economic downturn, followed by steady growth. Hence, both good and bad economic times were present during the follow-up. Owing to the present economic crisis, the proportion of unemployed persons with a previously strong connection to the labour market will increase. According to this study, hospitalisation for depressive disorder may increase too, especially among foreign-born women.
The effect of the duration of unemployment was not tested and future studies should analyse whether there is a difference in the risk of hospitalisation for depressive disorder following unemployment for persons who get rehired quickly, compared with those who experience long-term unemployment. The finding that the relative risk was higher among foreign-born women was new to this study. Immigration status was defined as either Swedish born or foreign born and the latter is a rather unspecific category. Different subgroups of migrants have different access to the labour market and in future studies these groups should be tested separately.
In conclusion, this study shows that experiencing unemployment among persons with a strong connection to the labour market is a risk factor for hospitalisation for depressive disorders in Sweden. Foreign-born women had the highest relative risk.
What is already known on this subject
-
It is known that poor mental health can cause unemployment and that unemployment can cause poor mental health. In terms of depressive disorder, specifically, it is known that there is an association with unemployment.
What this study adds
-
This study adds that experiencing unemployment is a risk factor for hospitalisation for depressive disorder, especially for migrant women with a strong connection to the labour market.
-
Owing to the present economic crisis, hospitalisation for depressive disorder may increase.
Acknowledgments
Special thanks to Patrick Hort for language editing of the text.
References
Footnotes
-
Contributors A-CH designed the study, acquisitioned, prepared and carried out statistical analyses and interpretation of the data, and drafted the manuscript under supervision. DB, planned and performed statistical analyses, drafted the tables of the data and helped to draft the manuscript. JE and BB helped acquire data, participated in the study design and helped to draft the manuscript. SE and PI, main supervisor, conceived of the study, participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.
-
Funding This work was supported by FAS, The Swedish Council for Working Life and Research (FAS-2007-1961).
-
Competing interests None.
-
Ethics approval This study was approved by the Stockholm Regional Ethical Review Board (2008/732-31 and 2010/1983-32).
-
Provenance and peer review Not commissioned; externally peer reviewed.
-
↵i Bulgaria, Cyprus, Estonia, Latvia, Lithuania, Malta, Poland, Romania, Slovakia, Slovenia, the Czech Republic and Hungary.
-
↵ii Austria, Belgium, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Lichtenstein, Luxemburg, the Netherlands, Norway, Portugal, Spain and the UK.