Table 4 Characteristics of studies included in review relating unemployment status and mental health outcomes, 2004–2014

[78]

National population sample, USA

Cohort

1510

(1986–2002)

>25 y

Individual-level indicators

Employment status

Socioeconomic status

Psychological Well-being

Depressive symptoms (CES-D)

Job loss is linked with follow-up depressive symptoms and, occupational prestige significantly heightened this vulnerability. Unemployment status is significantly associated with depressive symptoms (r: 0.333, S.E.: 0.108)

Temporal order of exposures, confounders, and the outcome under consideration affected all participants at the same time producing stronger causal conclusions.

Difficult to distinguish truly involuntary job losses from health-related separations. Did not account for life course effects, the role of neighbourhood, or other such effects by which inequality may shape health.

[79]

National population sample, Australia

Cohort

7176

2001

20–55 y

Individual-level indicators

Employment status

Psychological Well-being

Mental health distress (MHI-5)

Negative correlation (r = −0.16) between unemployment and mental health across waves. Mental health is both a consequence of and risk factor for unemployment.

Uses a continuous measure of mental health symptoms.

It simultaneously investigates the bi-directional effects of unemployment and mental health.

The analyses was restricted to working age population (20 to 55 years at baseline).

The results and recommendations should not be generalized to other cohorts.

[80]

National population sample, Britain, UK

Cohort

14,686 (1991–2000)

≥16 y

Individual-level indicators

Employment status

Financial situation

Psychological Well-being

Mental health distress caseness (GHQ-12)

Job loss increased risk of distress for men (OR = 3.15; 95 % CI:2.50–3.98) and women (OR = 2.60; 95 % CI:1.97–3.43). Moving to paid work reduced risk of distress for men (OR = 0.52;95 % CI: 0.41–0.68) and for women (OR = 0.68;95 % CI: 0.69–1.40). Worse off unemployed men are more distressed (OR = 4.19; 95 % CI:3.20–5.50).

Temporal order of exposures, confounders, and the outcome under consideration affected all participants at the same time producing stronger causal conclusions.

Although, subjective financial difficulty was associated with psychological distress, whether it is causal or the consequence of negative affectivity is not clear.

[81]

National population sample, New Zealand

Cohort

15,095

(2004–2009)

15–60y

Individual-level indicators

Employment Status

Deprivation

Psychological Well-being

Mental health distress (Kessler-10 and SF-36)

Job loss decreased MH (SF-36) in 1.34 points (95 % CI −1.85 to −0.82) and increased mental distress in 0.50 points (95 % CI 0.34 to 0.67). Deprivation was associated with a 1.47 (95 % CI −1.67 to −1.28) decline in MH and a 0.57 unit (95 % CI 0.51 to 0.63) increase in mental distress.

Large sample over 5 years. Temporal order of exposures, confounders, and the outcome under consideration affected all participants at the same time producing stronger causal conclusions.

Those with poor mental health are more likely to subsequently become unemployed or experience more deprivation, so reverse causation might be possible.

[82]

National population sample, Britain, UK

Cohort

10,300

16–64y

(1991–2009)

Individual-level indicators

Employment Status

Psychological Well-being

Mental health distress (GHQ-12)

Moving from unemployment to employment was strongly associated with an improvement in mental health −2.1 [95 % CI −2.4 to −1.7], whereas becoming unemployed was detrimental 2.5 (95 % CI 2.2–2.7).

Annual data collected over a 19-year period. Temporal order of exposures, confounders, and the outcome under consideration affected all participants at the same time producing stronger causal conclusions.

Possible bias due to selection effects threat causal inference since those with poor mental health are more likely to subsequently become unemployed.

[83]

National population sample, Britain, UK

Cohort

10,264

16–65y

(1991–2007)

Individual-level indicators

Employment Status

Psychological Well-being

Mental health distress (GHQ-12)

Job loss significantly predicted poorer psychological well-being in comparison to those still employed (2.21; 95 %; CI: 1.99–2.43).

Annual data collected over a 16-year period. Temporal order of exposures, confounders, and the outcome under consideration affected all participants at the same time producing stronger causal conclusions.

Possible bias due to selection effects threat causal inference since those with poor mental health are more likely to subsequently become unemployed.

[84]

National population sample, Japan

Cohort

4800

(2007–2011)

20–40y

Individual-level indicators

Employment Status

Psychological Well-being

Mental health distress (MHI-5)

Job loss decreases mental health by 12.0 points (MHI-5) after controlling for other variables.

Temporal order of exposures, confounders, and the outcome under consideration affected all participants at the same time producing stronger causal conclusions.

Direction of causality even after controlling for individual heterogeneity, is difficult to distinguish.

[85]

National population sample, Australia

Cohort

5846

2007 > 15 y

Individual-level indicators

Employment status

Unemployment duration

Psychological Well-being

Mental health distress (MHI-5)

Baseline mental health status predicts overall time spent unemployed. 19.1 % of those with poor mental health experience subsequent unemployment compared with 14.6 % of those with better mental health.

Temporal order of exposures, confounders, and the outcome under consideration affected all participants at the same time producing stronger causal link.

The analysis was restricted to respondents aging 20–50 years at baseline. The results and recommendations should not be generalized to other cohorts.

[86]

National population sample, Australia

Cohort

21,280

(2001–2010) ≥16 y

Individual-level indicators

Employment status Number of unemployment spells

Psychological Well-being

Mental well-being (SF-36)

Compared to employed people, unemployed people show a 1.64 decrease (95 % CI −2.05 to −1.23, p < 0.001) in mental health, and those who had two or more spells of unemployment show a 2.56 decrease (95 % CI −3.93 to −1.19, p < 0.001).

Large sample. Temporal order of exposures, confounders, and the outcome under consideration affected all participants at the same time producing stronger causal conclusions.

Lack of data on voluntarily or involuntary job loss (due to illness) self-reported nature of the data on mental health.

[87]

National population, Sweden

Cross-sectional

20,538

(2008)

18–85 y

Individual-level indicators

Employment status

Psychological Well-being

Mental health distress (GHQ-12) psychosomatic symptoms

Unemployed people had reduced psychological well-being (OR = 2.11; 95 % CI: 1.79–2.50) and more psychosomatic symptoms (OR = 1.62; 95 % CI: 1.37–1.92) compared with individuals who were employed.

Large sample. The postal survey reduces the potential bias introduced by interviewer and respondents may answer sensitive questions more honestly.

Its cross-sectional design removes the possibility of causal inference.

Postal questionnaire surveys increase non-response sample bias.

[88]

Community sample, Scania, Sweden

Cross-sectional

5180

18–64 y

Individual-level indicators

Employment status

Psychosocial job quality

Psychological Well-being

Mental health distress (GHQ-12)

People facing job strain (OR = 3.01; 95 % CI:2.26–4.02) and unemployment (OR = 5.81; 95 % CI:4.33–7.79) have significantly higher odds ratios of psychological distress.

The postal survey reduces the potential bias introduced by interviewer and respondents may answer sensitive questions more honestly.

Its cross-sectional design removes the possibility of causal inference.

Postal questionnaire surveys increase non-response sample bias.

[89]

Regional sample, North West of England, UK

Cross-sectional

15,228 (2009)

Individual-level indicators

Employment status

Deprivation

Psychological Well-being

Life satisfaction

Mental well-being

Deprivation strongly linked to low LS and MWB. 17.1 % of the most deprived tertile have low LS compared to 8.9 % in the most affluent.

It identifies the characteristics of individuals most likely to suffer from poor Well-being

Its cross-sectional design removes the possibility of causal inference.

[90]

National population sample, Brazil

Cross-sectional

5000

(2003) >18 y

Individual-level indicators

Employment status

Common Mental Disorders

State of animus (World Health Survey)

Among women, level of education and unemployment were associated to feelings of depression and anxiety. Among males, feelings of depression were strongly associated with unemployment.

This study provides data on the negative effects of unemployment on depression and anxiety, which are important predictors of subsequent morbidity.

Its cross-sectional design removes the possibility of causal inference.

Generalizing findings may be reasonably limited to the uniqueness of the Brazilian welfare system.

[91]

Cross-national samples of older adults from 13 EU countries and USA

Cohort

15,055

(2006–2010)

50–64 y

Individual-level indicators

Employment status

Common Mental Disorders

Depressive symptoms (EURO-D and CESD)

Unemployment was associated with 4.78 % [95 % (CI): 0.823 to 8.74 %] increase in depressive symptoms in the USA and 3.35 % (95 % CI: 0.486 to 6.22 %) increase in Europe.

Bias due to selection and reverse causality was lessen because the study distinguished job loss due to plant closures, and used individual fixed effect models.

Used two measures for depressive symptoms Euro-D for Europe and CESD for USA. However, these were normalized. The analysis was restricted to older adults (50–64y).

[92]

Cross-national sample of

primary care patients from EU and Chile

Cohort

10,059

(2003–2004)

18–75 y

Individual-level indicators

Employment status

Common Mental Disorders

Depression caseness Composite International Diagnostic Interview (CIDI)

Job loss between baseline and 6 months compared to those employed at both times had an adj relative risk ratio for 12-month depression of 1.58 (95 % CI:0.76, 3.27). Participants with depression at baseline and 6 months compared to neither time had an OR for 6-month unemployment of 1.58 (95 % CI:0.97, 2.58).

It examines the interrelations between unemployment and clinical depression in both directions across different countries producing stronger causal conclusions.

No available data on whether employment is full time or part-time or underemployment.

If unemployed adults with depression are less likely to seek medical treatment they may be under-represented in the GP-based sampling frame.

[95]

Cross-national samples of European Countries

Cross-sectional

34,395

(2001–2009) > 18 y

Individual-level indicators

Employment status

Education level

Income and Occupation

Common Mental Disorders

Anxiety

Mood disorder

Unemployed showed the highest prevalence and increased risk of 12-month mental disorders. Mood disorders and anxiety were more prevalent among those receiving a low and a low-average income

Northern Ireland, Portugal and Belgium were the countries with the highest risk for mental disorders.

This study examines the associations between employment status and mental health in a European representative sample.

Specifies which countries are at higher risk for mental disorders.

Since data derives from different countries during a wide time interval (2001–2009) to determine the impact of the adverse economic conditions of the past few years was not possible. Participants from different countries have been exposed to different economic scenarios and the study was unable to evaluate the impact of this.

No causal inference can be made due to the cross-sectional nature of the study

[93]

National population sample, England, UK

Cross-sectional

5090 ≥16 y

Individual-level indicators

Employment status

Common Mental Disorders

CIS-R interview: Common Mental Disorders (CMD)

Risk of CMD was significantly greater in unemployed individuals; economically inactive; not working due to physical health reasons; unable to find a suitable job among others. Individuals unemployed for less than 1 year or more than 3 years had a higher risk of CMD.

Uses a well validated scale for detection of common mental disorders.

Its cross-sectional design removes the possibility of causal inference.

[94]

National population sample, Sweden

Cross-sectional

42,448 (2004)

18–84 y

Individual-level indicators

Employment status

Financial strain

Common Mental Disorders

Anxiety Depression (EQ-5D)

Unemployment (OR = 2.9; 95 %; CI:2.2–4.0), economic hardship (OR = 3.1; 95 %; CI:2.4–3.9 were strongly and independently related with anxiety/depression.

Large and population-based study that uses an internationally validated scale of quality of life that measures anxiety and depression.

No causal inference can be made due to the cross-sectional nature of the study.

[96]

National population of employees of collapsed major banks, Iceland

Cross-sectional

1880

(2009) >20 y

Individual-level indicators

Employment status (Downsizing)

Common Mental Disorders

Depression and anxiety symptoms (AOSH)

Downsizing, salary cut, and transfer to another department is associated with increased psychological distress

Nationwide sample and the inclusion of all employees of collapsed major banks in one country highly hit by the economic recession

No causal inference can be made due to the cross-sectional design and self-reported data.

This sample was drawn from the collapse of banks in Iceland, so generalizing findings to other countires may be limited

[97]

National population sample, Finland

Case–control

5859 cases

74,809 controls

Individual-level indicators

Employment status

Socioeconomic status

Substance-Disorders

Driving under the influence of drugs (DUID)

Low education, unemployment, disability pension, being divorced and living alone were the strongest individual predictors of DUID in all substance groups.

Large sample size, based on two registers ensuring good coverage and validity, increases reliability of the study.

It shows that disadvantaged social background is related to driving under the influence of drugs.

Impaired drivers were over-represented: the cases were suspected and apprehended of DUID by the police. Not all people driving under the influence are caught (fewer than 10 %). The direction of causality remains unclear.

[98]

National population sample, USA

Repeated Cross-sectional

405,000

(2002–2010) >18y

Individual-level indicators

Employment status

Substance-Disorders

alcohol use/abuse/dependence; illicit drug use/abuse/dependence and tobacco use

Unemployed people show higher prevalence of alcohol use, illicit drug use, tobacco use, alcohol abuse or dependence, and illicit drug abuse or dependence then employed. This was before, at the start of, and during the 2009–2010 period of high unemployment.

Nationally representative sample of US adults. Strong association between substance disorders and unemployment.

Cross-sectional data does not allow tests of causality among the reported associations. Possible bias due to validity of self-reports of sensitive behaviours

[99]

National population sample, USA

Cross-sectional

5307

(2009–2010) >18y

Individual-level indicators

Employment status

housing payment problems

Substance-Disorders

Alcohol Dependence Negative drinking consequences

Housing payment instability was associated with experiencing more negative drinking consequences and alcohol dependence symptoms.

Job loss was strongly associated with alcohol problems in univariate models, but no significant associations were observed in multivariate models.

Nationally representative sample of US adults. Strong association between alcohol drinking patterns and housing instability and unemployment.

Does not preclude the possibility of reverse causation (individuals with existing alcohol problems prior to the study)

[100]

Community sample of job-seekers, Germany

Cross-sectional

7906 (2008–09)

18–64 y

Individual-level indicators

Employment status

Duration of unemployment

Substance-Disorders Smoking, risky drinking, illicit drug use.

Self-rated health

52.4 % of the sample (53.4 % male, 33.5 years mean age) had 3 or more health risk factors.

84.8 % of the 18–24 year old long-term unemployed men were smokers. Substance use risk factors were highest among the 18–24 year olds All health risk factors were associated with lower self-rated health.

Very high proportions of individuals with health risk behaviours were found, and associations with self-rated health were confirmed in a sample of job-seeker individuals.

No causal inference can be made due to the cross-sectional design.

The validity of self-reports of sensitive behaviours, such as alcohol consumption.

Since the research focused job-seekers the sample included both unemployed and employed individuals. This sample was drawn from one area in Germany so generalizing findings may be limited

[101]

National population sample, South Korea

Ecological Study

1995–2005

Individual-level indicators

Employment status

Substance-Disorders

Alcohol-attributable mortality

Found an incidence of 20 times higher alcohol-attributable deaths rate of unemployed compared to those of non-manual workers during recession

Brings national evidence on the inequalities in the health effects of economic changes.

Did not consider accidental deaths caused directly by alcohol (eg falls). The real magnitude of social disparity in alcohol-attributable death rates may be even greater than that estimated. Social disparity in alcohol-attributable mortality cannot be said to be a result of the crisis because this was tested.

[102]

Community sample, emergency departments in Edmonton, Canada

Case–control

507 cases

200 controls

(1993–94) >16 y

Individual-level indicators

Employment status

Suicidal Behaviours

Parasuicide

There is an association between unemployment and parasuicide (OR = 12.0; 95 % CI:6.0–23.9)

Brings strong evidence on the influence of exposure to unemployment on parasuicidal behaviour in comparison to a control group.

There is low response rate for both cases and controls.

[103]

Cross-national samples from 21 countries worldwide

Cross-sectional

108,705

(2001–07) >18y

Individual-level indicators

Employment status

Educational level

Suicidal Behaviours

Suicidal ideation and attempts

(CIDI)

12-month prevalence of suicide ideation, plans and attempts are 2.0, 0.6 and 0.3 % respectively for developed countries and 2.1, 0.7 and 0.4 % for developing countries. Risk factors for suicidal behaviours in both developed and developing countries included being a woman, low educated, low income, and being unemployed (among others).

Large cross-national epidemiological survey database

No causal inference can be made due to the cross-sectional nature of the study.

Reported suicidal ideation and suicide attempts can be susceptible to recall bias or to reluctance on the part of respondents to disclose such a sensitive information.

[104]

National population sample, Australia

Cross-sectional

4697

(2007–2009)

15–64 y

Individual-level indicators

Employment status

Suicidal Behaviours

Death by Suicide

During 2001–10 economically inactive/unemployed males suicide at 4.62 times (RR = 4.62; 95 % CI: 4.10, 5.19; P < 0.001) the rate of employed men (RR = 1.00). Economically inactive/unemployed females had a suicide RR of 8.44 compared with employed females (95 % CI 7.38, 9.67; P < 0.001).

Best available national data and provides information on the employment status of individual suicide cases.

Possible under-reporting of suicide data and under-report of the long-term unemployed that have given up looking for work (i.e. discouraged job seekers).

Lack of available data on confounding factors.

[105]

National population sample, Spain

Cross-sectional

4583

(2001–2002)

>18y

Individual-level indicators

Employment status

Suicidal Behaviours

Suicidal ideation and attempts

(CIDI)

Being unemployed or having work disability were also associated with suicidal ideation in people aged 18–49. The prevalence of suicidal ideation and attempts found in this study is similar to the one found ten years ago, before the economic crisis

Representative sample of the national population in Spain during the economic recession. The data was collected in the same way as in the ESEMED study, making it possible to compare current figures with the prevalence found before the crisis.

No causal inference can be made due to the cross-sectional nature of the study.

Reported suicidal ideation and suicide attempts can be susceptible to recall bias or to reluctance on the part of respondents to disclose such a sensitive information.