Background: A study was undertaken to investigate whether job insecurity predicts incident use of antidepressant medication and whether the association is modified by a history of prolonged unemployment.
Methods: A prospective follow-up study was performed in 5142 Danish employees, including 632 employees with and 4510 without a history of prolonged unemployment. Participants were drawn from a random 10% sample of the Danish population. Survey data on job insecurity were linked with register data on history of unemployment and dispensing of antidepressant medication between June 2000 and December 2003 retrieved from the Danish Medicinal Product Statistics. Respondents with major depression at baseline or antidepressant use in the 5 years preceding baseline were excluded.
Results: Job insecurity predicted use of antidepressants after adjustment for sex, age, cohabitation, socioeconomic position and alcohol consumption (OR 1.43, 95% CI 1.09 to 1.88). The effect was attenuated after further adjustment for baseline depressive symptoms (OR 1.15, 95% CI 0.87 to 1.52). A history of prolonged unemployment predicted use of antidepressants in both models (OR 1.62, 95% CI 1.14 to 2.30 and OR 1.49, 95% CI 1.04 to 2.13, respectively) Compared with participants with neither job insecurity nor unemployment history, the OR for the joint effect of job insecurity and history of prolonged unemployment was substantially higher (OR 1.79, 95% CI 1.15 to 2.79) than the OR for job insecurity (OR 1.02) and unemployment history (OR 1.10) alone in the fully adjusted model.
Conclusion: Job insecurity predicts incident use of antidepressants among Danish employees with a history of prolonged unemployment.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Unemployment has been found to be associated with the risk of depression in previous studies.1 2 3 Recently, researchers have started to investigate whether job insecurity (ie, the threat of unemployment) influences the development of clinical depression and depressive symptoms as well.4 This is a public health issue of growing concern in a globalised economy that in many countries is characterised by an erosion of secure long-term employment, a rise in frequencies of downsizing and a marked increase in temporary and contract-based employment.5 6 7 8 9
To date, only a few large-scale epidemiological studies have investigated the effect of job insecurity on the onset of depression using a longitudinal design.10 11 12 13 In these studies, job insecurity was associated with future depression, although the effect sometimes differed by sex.
In epidemiological studies, depression is usually assessed by self-reported symptoms on rating scales or by a clinical diagnostic interview. When the exposure of interest is measured by self-reporting, the use of rating scales might introduce bias due to common method variance (ie, a spurious association between the exposure and the outcome due to a specific response style).14 The use of clinical interviews, on the other hand, involves substantial costs and is therefore often not feasible.
A cost-effective alternative is the use of population-based registries of prescription and purchase of antidepressant medication. Recently, Finnish researchers have investigated associations between work environment factors and use of antidepressant medication which included a cross-sectional study on antidepressant use in temporary workers15 and a register-based prospective study on the effect of downsizing on use of antidepressants and other psychotropic drugs.16 However, to our knowledge, no study has yet analysed whether self-reported job insecurity increases the risk of antidepressant treatment. Moreover, no studies have yet investigated whether the effect of job insecurity is modified by previous unemployment. Employees with a history of unemployment might experience job insecurity as more threatening and consequently might show more pronounced negative health effects than employees without such a history. This reasoning is based on a diathesis stress model of depression which assumes that the effect of a current stressor is modified by underlying vulnerability factors.17 An interaction between social marginalisation and stressors in the aetiology of depression has been discussed, for example, in the classic studies on “social origins of depression” by Brown and Harris.18
We have studied the effect of job insecurity on the incident use of antidepressant medication by combining self-reported data on job insecurity with administrative data on antidepressant treatment and with a specific focus on a potential effect modification by unemployment history. We hypothesised that (1) job insecurity at baseline would predict incident antidepressant use during follow-up and (2) the effect of job insecurity on antidepressant use would be stronger in people with a history of prolonged unemployment than in those without such a history.
Study design and population
This is a 3.5-year follow-up study on the effect of job insecurity on incident use of antidepressant medication. Baseline data included register data on the history of prolonged unemployment (1 October 1996 to 1 October 1999), register data on the current or past use of antidepressants (1 January 1995 to 31 May 2000) and survey data on job insecurity (March or April 2000, depending on when the survey was filled out). Baseline data collection was completed by 31 May 2000. Follow-up data collection on use of antidepressants started on 1 June 2000 and continued until 31 December 2003. The study data set was created by linking survey data from the Danish Longitudinal Study on Work, Unemployment and Health with register data on prescribed and purchased antidepressants from the Danish Medical Product Statistics located at Statistics Denmark.19 The linkage procedure made use of the fact that each resident in Denmark is assigned a unique person identification number. For research purposes, Statistics Denmark is thus able to combine survey data with register data in an anonymous process on a person level.
The Danish Longitudinal Study on Work, Unemployment and Health was originally designed to investigate associations between psychosocial factors, unemployment, social marginalisation and health.20 Participants were drawn from the Danish Institute of Governmental Research Longitudinal Register21 located at Statistics Denmark. This register includes data on sociodemographic variables and employment history of a random 10% sample of the Danish population aged 15 years or older. Of these, a random sample cohort (RS cohort) of 11 082 Danish residents aged 40 or 50 years by 1 October 1999 was drawn. Because only a few of the study participants had a history of prolonged unemployment, a second sample (“marginalisation sample”, MS cohort) was drawn from the same study population. This second sample included 4145 Danish residents aged 37–56 years who had been unemployed at least 70% of the time between 1 October 1996 and 1 October 1999. Both cohorts received a postal questionnaire comprising variables on socioeconomic and demographic factors, working conditions, health behaviours and physical and mental health in March 2000. There were 7583 respondents in the RS cohort (68% response rate) and 2287 respondents in the MS cohort (55% response rate). Of the 7583 respondents in the RS cohort, 205 had been unemployed at least 70% of the time between October 1996 and October 1999 and these respondents were moved to the MS cohort for the purpose of analysis in the present paper.
Figure 1 shows the exclusion process for both cohorts. Individuals were excluded if at baseline they (1) were not employed; (2) were current or past users of antidepressants; (3) had a major depression according to the criteria of the Diagnostical and Statistical Manual Version IV (DSM-IV)22 as assessed by the Major Depression Inventory (MDI) in the survey;23 or (4) had one or more missing values on any variable included in the multivariate analyses. The exclusion process resulted in a final sample of 5142 participants (53% women), with 4510 participants in the RS cohort and 632 participants in the MS cohort. Figure 1 shows that excluded individuals had a higher likelihood of antidepressant use at follow-up than study participants.
The characteristics of the study participants are shown in table 1, stratified by history of prolonged unemployment. Participants with a history of prolonged unemployment were more likely to be women (72.5% vs 49.9%) and more likely to report job insecurity at baseline (53.3% vs 34.7%) than participants without such a history.
Definition and measurement of antidepressant use
Use of antidepressant medication was defined by dispensing of an antidepressant at a pharmacy, which is the only legal way of purchasing this type of medication in Denmark for the non-hospitalised population. Data were retrieved from the Danish Medicinal Product Statistics that contains information on all prescribed medication purchased at pharmacies in Denmark since 1995.24 We used the registrations for all types of antidepressants, coded N06A, according to the Anatomical Therapeutic Chemical (ATC) classification system.25
Current or past use of antidepressants was defined by an entry of N06A in the database between 1 January 1995 and 31 May 2000 (ie, the month after the baseline survey was completed). Incident use of antidepressants was defined by two conditions: (a) an entry of N06A in the database during the 3.5 years of follow-up which ranged from 1 June 2000 to 31 December 2003 and (b) no current or past use of antidepressants.
Definition and measurement of job insecurity
Job insecurity was measured with the question: “Are you worried about becoming unemployed?”. The response options were “not at all”, “a little bit”, “somewhat” and “very much”. Respondents who answered anything other than “not at all” were classified as experiencing job insecurity.
In addition, we combined the information on job insecurity and history of prolonged unemployment into a new variable with the following categories: (1) neither job insecurity nor history of prolonged unemployment; (2) job insecurity without a history of prolonged unemployment; (3) history of prolonged unemployment without job insecurity; and (4) both job insecurity and a history of prolonged unemployment.
Definition and measurement of covariates
As covariates we included sex, age, cohabitation, socioeconomic position (SEP), heavy alcohol consumption and depressive symptom score at baseline. These covariates were selected because they were associated with a risk of depression in earlier studies.26 27 28
SEP was assessed based on the International Standard Classification of Education (ISCED) codes29 obtained from a register at Statistics Denmark, resulting in three SEP groups: “low education” (ISCED codes 0, 1, 2; ie, pre-primary to lower secondary education); “medium education” (ISCED codes 3 and 4; ie, upper and post-secondary education); and “high education” (ISCED codes 5 and 6; ie, first and secondary stage of tertiary education).
Heavy alcohol consumption was defined as drinking more than two units per day for women or three units per day for men. A unit was defined as one small bottle of beer (33 cl), one glass of wine or one shot of liquor.
The depressive symptom score at baseline was assessed with the MDI, a 10-item questionnaire that allows the measurement of both depressive symptoms on a continuous scale and major depression according to DSM-IV criteria. The MDI has been validated against clinical assessments of depression,23 30 and has been used to determine the prevalence of depression in a representative sample of Danish residents.27
We also assessed the level of self-efficacy in order to investigate if this measure of resilience might protect against the hypothesised adverse effects of job insecurity and prolonged unemployment. We used a 10-item Danish translation of the Generalized Self-Efficacy Scale developed by Jerusalem and Schwarzer.31 Items were added up and the sum score was divided into low (reference group) vs high self-efficacy by median split.
All analyses were conducted with the statistical program package SAS 9.0. The effect of job insecurity at baseline and of a history of prolonged unemployment on risk of incident use of antidepressant medication during follow-up was estimated by OR and 95% CI using multivariate logistic regression models. To control for the effects of the covariates we used two models: model 1 was adjusted for sex, age, cohabitation, SEP and alcohol consumption and model 2 was further adjusted for the continuous depressive symptom score at baseline. We used the depressive symptom score in a separate model because it is conceptually debatable whether a high level of depressive symptoms at baseline is a confounder or an intermediate variable.
To analyse whether the joint effect of simultaneous exposure to job insecurity and a history of prolonged unemployment was particularly health hazardous we calculated ORs and 95% CI for the four possible combinations of these two exposures using the combination “neither job insecurity nor history of prolonged unemployment” as the reference category.
Finally, we applied multivariate logistic regression models to analyse whether self-efficacy had a protective effect against depression among those participants who had been exposed to both job insecurity and a history of prolonged unemployment.
Effect of job insecurity and history of prolonged unemployment on incident antidepressant use
During the 3.5 years of follow-up, 232 participants (4.5%) had incident use of antidepressant medication. This includes 188 participants (4.2%) in the RS cohort and 44 participants (7.0%) in the MS cohort.
Table 2 shows that job insecurity at baseline predicted antidepressant use at follow-up in model 1 (OR 1.43), but not after further adjustment for the depressive symptom score at baseline in model 2 (OR 1.15). A history of prolonged unemployment (ie, belonging to the MS cohort) predicted antidepressant use in both models (OR 1.62 and OR 1.49 in models 1 and 2, respectively). Effect sizes were similar for both men and women (data not shown in table). Among men, the ORs for job insecurity were 1.56 (95% CI 1.03 to 2.37) and 1.18 (95% CI 0.76 to 1.81) in models 1 and 2, respectively. Among women, the ORs for job insecurity were 1.38 (95% CI 0.96 to 1.98) and 1.15 (95% CI 0.79 to 1.67) in models 1 and 2, respectively. For history of prolonged unemployment, the OR among men was 1.80 (95% CI 0.92 to 3.49) in model 1 and 1.54 (95% CI 0.77 to 3.10) in model 2. Among women the ORs were 1.64 (95% CI 1.09 to 2.49) and 1.54 (95% CI 1.01 to 2.36) in models 1 and 2, respectively.
Effect modification by history of prolonged unemployment
Table 3 shows the OR for the four combinations of the absence and presence of job insecurity and history of prolonged unemployment. Compared with participants with neither job insecurity nor unemployment history, the ORs for the joint effect of job insecurity and history of prolonged unemployment were substantially higher (2.38 and 1.79 in models 1 and 2, respectively) than the ORs for job insecurity and unemployment history alone. The effect was similar for men and for women (data not shown).
Finally, we analysed whether high self-efficacy had a protective effect among participants who had been exposed to both job insecurity and a history of prolonged unemployment. This was not the case. Although the incident rate of antidepressant use was slightly lower among participants with high self-efficacy (7.7% vs 9.9%), the effect was not statistically significant (table 4).
Summary of findings with regard to the two study hypotheses
We had hypothesised that (1) job insecurity would predict the use of antidepressants and (2) the effect of job insecurity would be stronger among participants with a history of prolonged unemployment. In the data analyses we found that job insecurity was prospectively associated with use of antidepressants before (OR 1.43) but not after (OR 1.15) adjustment for depressive symptom score at baseline. Thus, our first hypothesis was only partly confirmed. Participants with both job insecurity and a history of prolonged unemployment (OR 1.79) had a substantially higher risk of using antidepressants than all other groups. Thus, our second hypothesis was confirmed.
Comparison with previous studies on job insecurity, depression and antidepressant use
Previous studies on job insecurity and depression have measured depression by either self-reported symptoms or clinical interviews.10 11 12 13 To our knowledge, the present study is the first one to investigate this association by assessing the use of antidepressant medication. The results from our study support the conclusion from these other studies that job insecurity might contribute to the risk of clinical depression. However, our results also show that the effect is attenuated when analyses are adjusted for baseline depressive symptoms (which might or might not be overadjustment, see discussion below) and that the effect of job insecurity is modified by a history of prolonged unemployment.
Studies on the associations between temporary work and experiences of downsizing with antidepressant use have recently been conducted in Finland. In a cross-sectional analysis, Virtanen and colleagues found that temporary workers were more likely to use antidepressants.15 Since temporary workers are probably more prone to job insecurity than regularly employed workers, this might indicate an effect of job insecurity on antidepressant use, although the cross-sectional nature of the study limits causal inference. In a register-based study, Kivimäki and colleagues found that employees in organisations that had undergone downsizing in the past were at increased risk of using psychotropic drugs, including antidepressants.16 The reported effect estimates (1.49 and 1.12 for men and women, respectively) in their study were similar to the ORs in the present study (1.56 and 1.38 for men and women, respectively, before adjustment for baseline depressive symptoms). However, because survey data were not available in the study by Kivimäki et al, it remains unclear to what extent heightened job insecurity or exposure to other psychosocial factors (eg, increased workload or distrust) could explain the higher medication rate in the downsized organisations.
Role of history of unemployment
Conceptually, a history of prolonged unemployment can be regarded as a vulnerability factor that enhances the adverse effects of job insecurity. Mechanisms might be both material and psychosocial. Compared with people without a history of prolonged unemployment, those with such a history might have exhausted more of their financial resources in the past and might therefore be more anxious about financial difficulties and downward social mobility when facing the threat of job loss. People with previous experiences of prolonged unemployment might also be more pessimistic regarding their chances to get re-employed if they lose their current job. In this context it is of interest to note that, in another prospective Danish study, job insecurity was a stronger predictor for a decline in general self-rated health when it was accompanied by self-reported poor chances on the labour market.32 One might speculate that the experience of prolonged unemployment has induced a state of “learned helplessness”—which is regarded as an important psychological risk factor for the development of depression33 34—and that this state was psychologically activated when people became exposed to job insecurity.
Role of depressive symptoms at baseline
In addition to excluding participants who either had a history of antidepressant use or had a DSM-IV Major Depression according to their MDI score, we also adjusted the analyses for the continuous depressive symptom score at baseline. Subclinical depressive symptoms would be a confounder if they had influenced both the self-report of job insecurity and the subsequent risk of antidepressant treatment. However, these depressive symptoms would be an intermediate step in a causal pathway if job insecurity prior to the baseline survey had caused the symptoms at baseline, and the symptoms subsequently had caused antidepressant treatment during follow-up. To account for these two possibilities we presented data in two models, with and without adjustment for depressive symptoms. If depressive symptoms are both a confounder and an intermediate step in the causal pathway, then the true effect estimate would be somewhere between the overestimated effect size in model 1 and the underestimated effect size in model 2.
Strengths and weaknesses of the study
The strengths of this study are the longitudinal design and the complete coverage of all legally purchased antidepressants in Denmark. Furthermore, we were able to exclude participants who had used antidepressant medication during the 5 years prior to baseline or had a major depression at baseline. By conducting analyses for participants with and without a history of prolonged unemployment, we were able to identify a strong effect modifier which has not been reported previously.
A weakness of the study is the modest response rate of 55% in the marginalisation sample which limits the generalisability of the findings. People with depression at baseline might have been less likely to respond and it is therefore possible that depression was more widespread among non-responders than among responders. However, this would not have biased the results because all baseline participants with an indication of a present or past depression were excluded from the analyses.
The majority of study participants who had been marginalised (ie, had experienced prolonged unemployment in the past) were not employed at baseline (1625 of 2492) and consequently were excluded from the analysis. We further excluded all participants with a present or past depression. Hence, the participants of the MS cohort were a highly selected group. Despite a history of prolonged unemployment, they had managed to become employed again and they had neither a history of depression nor depression at baseline. It can be speculated that these participants had a high level of resilience, maybe because they had more material, psychological and/or social resources at hand to deal with adversity. It can also be argued that people who had experienced prolonged unemployment in the past might have become habituated to job insecurity and unemployment and therefore would be less likely to develop depression when exposed to current job insecurity. We believe that this makes our study results more interesting because the results imply that, even among people who might have shown a high level of resilience in the past and who might have been habituated to job insecurity and unemployment, the combination of job insecurity with a history of prolonged unemployment substantially increased the risk of depression. We acknowledge, however, that the generalisibility of the study findings is limited and that there is a need to investigate the adverse effect of shorter and less marginalising spells of unemployment. We also acknowledge that it would be important to investigate whether people with high resilience are less likely to develop depression when exposed to job insecurity and prolonged unemployment. In this study we conducted an additional analysis regarding the role of self-efficacy as a measure of resilience, but could not detect a protective effect.
We did not have information on job insecurity during follow-up. This is a limitation because exposure to job insecurity has probably changed for a number of participants during follow-up, resulting in a non-differential misclassification of exposure and a bias of the results towards the null hypothesis.
We have assumed in this study that use of antidepressant medication is an indicator of clinical depression. There might have been some false positive cases because antidepressants are sometimes used to treat medical conditions other than depression such as anxiety, pain and sleep disturbances.35 36 However, a greater concern is the number of potential false negative cases—that is, participants who developed a clinical depression during follow-up but were not treated with antidepressant medication. A recent survey on the prevalence of clinical depression in Denmark has shown that most cases of depression are neither diagnosed nor treated.27 If this misclassification is non-differential, it would have resulted in an underestimation of the effect size. However, if this misclassification is differential, it could have biased our results towards or away from the null hypothesis. We are currently in the process of developing an analytical strategy for investigating misclassification of depression in register-based studies.37
Finally, one has to be cautious of generalising the study findings to countries other than Denmark. As Bartley, Ferrie and others have pointed out, the health consequences of job insecurity might be at least in part dependent on country-specific societal factors such as employment protection laws, availability and duration of unemployment benefits and labour market policies.5 32
Job insecurity predicted the incident use of antidepressants among employees with a history of prolonged unemployment. The mechanisms might be both material and psychosocial. However, limitations regarding the generalisability of the findings have to be taken into consideration.
What is already known on this subject
Some studies have shown that job insecurity is associated with a risk of depression measured by self-reported symptoms or a diagnostic interview.
It has not yet been shown if job insecurity predicts use of antidepressant medication.
Moreover, it is unknown if a history of unemployment modifies the association between job insecurity and risk of depression.
What this study adds
Job insecurity predicted the use of antidepressant medication. However, the association was strongly attenuated when analyses were adjusted for baseline depressive symptom score.
The association between job insecurity and use of antidepressants was substantially stronger in participants with a history of prolonged unemployment than in participants without such a history.
Funding The study was supported by grants of the Danish Working Environment Research Fund (grant numbers 24-2005-09, 2-2006-04, and 5-2006-04) and of the Ministry of Health and Prevention, Public Health Fund (grant number 2005-14033-8).
Competing interests None.
Ethics approval The study has been notified to and registered by the Danish Data Protection Agency (Datatilsynet). According to Danish law, studies that include data from questionnaires and from registers only do not need approval from the Danish National Committee on Biomedical Research Ethics (Den Centrale Videnskabsetiske Komité).
Provenance and Peer review Not commissioned; externally peer reviewed.