You are here

Article Text

Article menu

Download PDFPDF

Other topics
Previous anxiety and depression as risk factors for early labour force exit
  1. Kara E Rudolph1,2,3,
  2. William W Eaton4
  1. 1School of Public Health, University of California, Berkeley, California, USA
  2. 2Center for Health and Community, University of California, San Francisco, California, USA
  3. 3Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
  4. 4Department of Mental Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
  1. Correspondence to Dr Kara E Rudolph, School of Public Health, University of California, 590 University Hall, Berkeley, CA 94720, USA; kara.rudolph{at}berkeley.edu

Abstract

Background Individuals with a history of mental illness have lower earnings than individuals without. A possible reason is that those with prior anxiety or depression may be more likely to exit the labour force prior to retirement age, but evidence has been mixed and limited. Our objective was to compare risk of early labour force exit between employed adults with a history of depression or anxiety versus those without, separately for men and women.

Methods We used data from the Baltimore Epidemiological Catchment Area Follow-up Cohort, which collected baseline data in 1981 and follow-up data 1993–1996 and 2004–2005. Cox proportional hazards models estimated the relative risk of labour force exit comparing those with versus without prior history of anxiety or depressive disorders.

Results Women with prior anxiety or depression are at 37% increased risk of dropping out of the labour force as compared to women without, controlling for age, socioeconomic status, race and marital status (HR: 1.37, 95% CI 1.04 to 1.79). Men with prior anxiety or depression are 18% more likely to subsequently drop out of the labour force as compared to men without, controlling for the above confounders as well as veteran status, but this association is not statistically significant (HR: 1.18, 95% CI 0.72 to 1.27).

Conclusions Prior anxiety or depression increases risk of early labour force exit for women. These findings may help explain previously reported lower earnings among female individuals with a history of mental illness and highlight the importance of considering anxiety and depressive disorders in policies supporting labour force participation.

  • MENTAL HEALTH
  • EPIDEMIOLOGY
  • EMPLOYMENT

http://dx.doi.org/10.1136/jech-2015-206202

Statistics from Altmetric.com

    Request Permissions

    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.

    Introduction

    Lifetime prevalence of anxiety and depressive disorders is higher in the USA than in any other country. The WHO World Mental Health Survey Initiative estimates that 21.4% of the US adult population has ever had a mood disorder and 31% has ever had an anxiety disorder.1 These disorders entail costs to individual and family life as well as to the economy. Depressive and anxiety disorders result in disability days2 ,3 as well as lost productivity while at work.4 In the USA, depressive disorders result in more than 500 million disability days per year and anxiety disorders result in more than 700 million disability days per year.5 For major depressive disorder alone, the annual cost of lost productivity in the USA is more than $36 billion.6

    These disorders not only result in lost income for the economy, they also result in lost income for the affected individuals. Individuals with history of mental illness have lower career attainment and lower career and past-year earnings than individuals without such a history.7 ,8 One hypothesised reason for these affected individuals earning less than their peers is that they may exit the labour force earlier.9 However, the evidence for this is limited and mixed.

    We know of few studies that have examined the link between previous anxiety or depression and subsequent labour force exit.10–13 Findings from these studies are mixed and share some limitations including measurement of depression using a symptom scale instead of a diagnostic schedule, and exclusion of participants who were lost to follow-up or died during the study period, which could compromise generalisability. In addition, these prior studies included middle-age to older-age adults, so it is unclear whether results apply to younger populations.

    Using data from the Epidemiological Catchment Area Baltimore Follow-up study (ECA), we compared risk of early labour force exit between adults with a history of depression and anxiety versus those without. Depression and anxiety were assessed using the Diagnostic Interview Schedule (DIS), which approximates a clinical diagnosis based on the Diagnostic Statistical Manual (DSM).14 ,15

    Methods

    The Baltimore ECA Follow-up cohort was designed to estimate the prevalence of mental disorders in the community, characterise their natural history and assess accessibility, coverage, and comprehensiveness of mental health services.16 The ECA's sampling strategy has been described previously.16 Briefly, it is a multistage, probabilistic sample of adults residing within the Eastern Baltimore Mental Health District, with an oversample of adults over age 65 (N=3481). The study includes four waves of data collection: wave 1 in 1981, wave 2 in 1982, wave 1993–1996 and wave 4 2004–2005. We used data from waves 1, 3 and 4, which occurred at approximately 10-year intervals. The study was approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board.

    Measures

    Exposure measure

    Our exposure of interest was lifetime history of either any anxiety disorder or depressive syndrome that occurred temporally prior to the outcome. ECA participants were administered the DIS to assess the presence of anxiety or depressive disorders/syndromes consistent with the DSM III at baseline. We defined any anxiety disorder as including obsessive-compulsive disorder, any phobia, panic disorder and generalised anxiety disorder. Depressive syndrome was defined as meeting DSM diagnostic criteria for major or minor depression without exclusions due to bereavement.17 Since the DIS is designed to estimate a psychiatric diagnosis (ie, illness present vs absent), lifetime history of depressive syndrome or anxiety disorder follows a Bernoulli (0/1) distribution. History of anxiety or depression was treated as a time-dependent exposure that could change over follow-up from unexposed to exposed. Once a participant reported any history, the exposure was valued at 1 for the duration of follow-up. If a participant reported no history at wave 1, her/his exposure was valued at 0 until report of either current or past anxiety or depression at follow-up waves 3 or 4, when the exposure would change to 1. Exposure assessment is depicted in figure 1.

    Figure 1
    Figure 1

    Epidemiological Catchment Area (ECA) data collection across follow-up: exposure assessment (history of anxiety disorder or depressive syndrome), time-varying covariate assessment (marital status), outcome assessment (labour force exit), and censoring due to death assessment.

    Outcome measure

    Our outcome of interest was first labour force exit prior to age 65 during follow-up, which we call early labour force exit. We defined exiting the labour force as changing from employed (full-time or part-time) to unemployed status and not seeking work. The US Census defines individuals as unemployed if they do not have a job but are looking for a job. Individuals that are not working and not seeking a job are defined as out of the labour force. For example, individuals who were retired, in school or otherwise not working and not looking for work were considered out of the labour force.

    Data on employment history was collected at waves 3 and 4 as part of the Life Chart Interview (LCI; figure 1). The LCI methodology has been described elsewhere.18 Briefly, this instrument collected yearly history from 1981 to 2005 from participants on life events (eg, marriage/separations, births, deaths, start/loss of job). The LCI was administered in waves 3 and 4 but asked about events occurring since the last interview, covering the entire follow-up time from 1981. Although the reliability of the LCI has not been extensively examined, employment history obtained using a similar instrument called the Life History Calendar was found to have test–retest reliability of 70–85% after 5 years and reliability of other life events, like marriage, of 90–100%.19

    Covariates

    Individual covariates included participant sex, age (years), black or white race, socioeconomic status, marital status and veteran status. A composite measure of socioeconomic status (SES) was used that is the mean household income percentile, educational attainment percentile and occupation status percentile (as measured by the Nam-Powers Occupational Status Score).20 These data were collected during the baseline household interview and are analysed in quartiles with those in the first quartile having the fewest resources and those in the fourth quartile having the most resources. Marital status was defined as currently married or living with a partner versus not, and was treated as a time-dependent covariate using baseline and LCI data spanning length of follow-up (figure 1). We initially planned on controlling for disability status, treating it as a time-dependent covariate indicating whether or not the participant reported any instrumental activities of daily living disability. However, models that included this covariate demonstrated worse fit as compared to models that did not include this covariate as assessed using Akaike information criterion. Consequently, disability status is excluded from the analyses.

    Statistical analysis

    The objective of our analysis was to examine whether having a history of anxiety disorder or depressive syndrome was associated with increased risk of subsequent labour force exit.

    Individuals who were lost to follow-up or died (as determined from the National Death Index) were censored at their age at last visit. We assume that censoring is non-informative, which means we assume it is unrelated to the conditional probability of dropping out of the labour force.

    Since participant death precludes observation of the event of interest, it can be thought of as a competing event. However, few employed participants at baseline died during follow-up, and this, coupled with a high percentage of censored individuals, made estimation of the mixture probabilities of the event of interest and the competing event unachievable. Thus, our analysis estimates the relative hazard of labour force exit in a hypothetical version of the ECA population in which those who died did not die.21 Although we do not know what the work history would have been for those individuals who died if they had not died, previous analysis of the Baltimore ECA Follow-up Cohort has shown that death over follow-up is not independently related to anxiety or depression.22

    Figure 2 shows inclusion/exclusion criteria. We limited our analyses to the 1172 individuals who contributed any employment history. We restricted our analysis further to 944 participants with employment data between ages 25 and 65, as we believe this age group to be at risk for exiting the labour force early. We also excluded four participants for whom temporality between labour force dropout and anxiety or depression could not be determined. Lastly, we excluded 28 participants who reported neither black nor white race. Our final sample size was N=912.

    Figure 2
    Figure 2

    Exclusion/inclusion criteria for analytic sample.

    Age 25 was defined as the origin of this survival analysis. Participants entered the analysis either at their age the first year they reported working during the ECA study period or at age 25 if they started working earlier.

    We first examine the association between a history of anxiety or depression and subsequent labour force exit with Kaplan–Meier estimates. We then use Cox proportional hazards models to estimate the relative risk of labour force exit comparing those with versus without prior history of anxiety or depressive disorders. We conduct the analyses separately for men and women, because literature suggests that men and women may have different relationships with anxiety and depression23–26 and with employment.27 We present two adjusted models. Both adjusted models account for age (via the time scale), gender (via stratification) and race. The fully adjusted model also controls for marital status, socioeconomic status and veteran status (for men only). Marital status, socioeconomic status and veteran status could function as confounding variables (each could affect risk of anxiety or depression and risk of labour force exit) as well as mediating variables (each could be affected by history of anxiety or depression and subsequently affect labour force exit). Adjusting for confounders that also may function as mediators would bias our estimates of the total effect to the extent that the total effect is mediated through such variables. Most variables (gender, age, race, socioeconomic status) did not have any missing data. One participant was missing marital status and seven participants were missing veteran status. We assessed the proportional hazards assumption by graphically examining scaled Schoenfeld residuals plotted against time. In addition, we statistically tested this. For men, there was no evidence of non-proportional hazards. For women, there was evidence of non-proportional hazards in the socioeconomic covariate. Consequently, we stratified our model by quartile of socioeconomic status. CIs are Wald CIs constructed on the log-transformed scale. We used R packages survival and KMsurv and R V.3.1.2.

    Results

    Of the N=912 Baltimore ECA participants included in our analysis, 53% were female and 43% reported lifetime history or anxiety or depression during follow-up. The mean age at entry was 33 years and the mean age at dropping out of the labour force or being censored was 50 years. Thirty-nine per cent of participants were black, and 39% reported being married or living with a partner. Most participants fell into the second or third SES quartiles (37% and 42%, respectively). Only 2% of participants reported a disability during follow-up. Thirty-seven per cent of men and 1% of women were veterans. Table 1 shows descriptive statistics separately for female and male participants by presence versus absence of history depression or anxiety.

    View this table:
    Table 1

    Baseline characteristics of study participants by history of depression or anxiety: Baltimore ECA Follow-up Study, 1981–2004

    Fifty per cent of working participants were followed for at least 16 years. The shortest follow-up time was 1 year and the longest was 26 years. A total of 317 participants dropped out of the labour force during follow-up (207 women and 110 men). The other 595 participants were censored. Of those who were censored, 15 died, 155 were lost to follow-up and the remaining 425 were event-free at age 65 or at the end of the study period in 2005.

    Figure 3 shows the association between labour force exit and prior anxiety or depression separately for men and women. For women, the curves begin to diverge shortly after entry at age 25 and widen until middle age. This divergence approaches statistical significance (log-rank test2=3.6, p=0.057). For men, the curves begin to diverge in the late 30s, but the gap between them remains narrower than for women and disappears altogether in the late 50s. For men, the divergence is not significant.

    Figure 3
    Figure 3

    Kaplan–Meier curves for remaining in the labour force for males (black) and females (grey) with (dashed lines) and without (solid lines) lifetime depression or anxiety.

    Figure 4 presents the relative hazard of labour force exit by prior anxiety or depression status separately for men and women. The unadjusted estimates, estimates adjusted for race and fully adjusted estimates are similar. For women, controlling for age, socioeconomic status, race and marital status, we estimate that prior anxiety or depression is significantly associated with a 37% increased risk of labour force dropout (HR: 1.37, 95% CI 1.05 to 1.79). For men, controlling for age, socioeconomic status, race, marital status and veteran status, we estimate that prior anxiety or depression is not significantly associated with an increased risk of labour force dropout (HR: 1.18, 95% CI 0.72 to 1.27). Table A1 in the online supplementary appendix presents coefficients and 95% CIs for each of the unadjusted, adjusted for race and fully adjusted models.

    Figure 4
    Figure 4

    Conditional HRs and 95% CIs comparing hazard of dropping out of the labour force for those with previous anxiety or depressive disorder compared to those without. Estimates are shown separately for men and women. 95% CIs for unadjusted estimates are shown as dotted-dashed lines. 95% CIs for estimates adjusted only for race are shown as dashed lines. 95% CIs for fully adjusted estimates are shown as solid lines. Fully adjusted models controlled for age, socioeconomic status, race and marital status. For men, fully adjusted models also controlled for veteran status.

    Discussion

    In a sample of Baltimore adults followed for more than 20 years, we found that a history of anxiety disorder or depressive syndrome increases risk of early labour force exit for women by nearly 40%. We found a slight increased risk of early labour force exit for men, but this risk was not statistically significant.

    Finding a significant relationship for women but not men corroborates previous research.10 ,13 ,28 This could be partially due to statistical power—more women than men report history of anxiety or depressive syndrome. It could also be due to historical societal expectations surrounding gender and work; research suggests that depressed men are more likely than depressed women to work.29 However, previous research has also suggested a greater effect of serious mental illness on past-year earnings for men versus women.7 This finding is not necessarily in conflict with ours as greater effects for men could reflect higher and more variable earnings7 among men. In addition, labour force outcomes may be seen at a lower threshold of severity for women versus men.

    Our use of depressive syndrome instead of major depressive disorder is supported by previous research, which found associations between subthreshold depression and subsequent labour force exit or unemployment but not major depressive disorder.10 This less severe form of depression has been described previously among ECA participants.17 The growing literature on its relationship with unemployment30 and early labour force exit adds to its importance.

    This study has several strengths. First, previous studies examining this question have included middle-aged to older-aged adults,10–13 but have stated the need to examine this question among younger samples.11 In addressing this research gap, our study demonstrates that risk for early labour force exit due to history of depression and anxiety is relevant not only for those close to retirement age but also for younger adults, suggesting that lost earnings potential may be greater than previously estimated, particularly for women. Second, the follow-up period of 24 years is longer than most previous studies, which include follow-up periods of a decade or less.10 ,11 ,13 Shorter follow-up periods may under-emphasise the potential for lost earnings. In addition, the availability of longitudinal data allows us to establish temporality between prior history of anxiety or depression and subsequent labour force exit. Temporality is of interest, because there is a larger literature on the cross-sectional association between labour force participation/employment and depression and anxiety8 ,29 ,31–33 as well as on the reverse relationship in which unemployment or inadequate employment increases risk of subsequent depression or anxiety.30 ,34 ,35 This latter relationship between lack of employment causing depression and anxiety may be part of a feedback loop, the documentation of which would be a valuable future contribution.

    We were interested in history of anxiety or depression instead of current anxiety and depression, because for the majority of people, depression and anxiety is either recurring or chronic.36 Using first report of history of depression or anxiety treats the variable as chronic condition instead of an acute, limited one. Future research contrasting labour force outcomes among those with acute, time-limited depression or anxiety episodes to those with chronic or recurring episodes would further understanding in this research arena.

    There are several limitations of this analysis related to sample size. First, the number of participants with history of anxiety or depression exiting the labour force was small: 37 men and 116 women. Thus, we may have been under-powered to find an association for men. These numbers become even smaller when stratified by type of anxiety and depressive disorder, precluding estimation of risk of labour force exit for each disorder.

    Another limitation is that our analytic approach assumes non-informative censoring, which may be violated if censoring is related to the conditional probability of dropping out of the labour force. In addition, we lack detailed, time-dependent information on socioeconomic status. Although we use a comprehensive measure of SES at baseline that includes wealth, occupational prestige and education level, it is plausible that this measure would change over time. If such changes were associated with history of anxiety and depression and precede changes in labour force status, then our estimates could be biased.

    Future research should examine possible mediation mechanisms by which anxiety and depression increase risk of subsequent labour force exit. Potential mediators could include variables like marital status, socioeconomic status and employment instability. We lacked information on time-dependent SES so could not consider it as a mediator, and although we had information on time-dependent marital status and employment, we were under-powered to test for mediation by these variables. Another potentially important mediator is receipt of appropriate treatment for anxiety and depression, as it is associated with increased employment and may prevent labour force exit.32 ,37 We lacked reliable, yearly data on treatment for anxiety and depression, so could not test this mediation hypothesis. Quantifying the extent to which receiving appropriate treatment mitigates the longitudinal relationship between depression and anxiety and labour force exit is an important area for future research.

    Conclusions

    In conclusion, history of anxiety or depression increases the risk of exiting the labour force early (prior to age 65). This increased risk is statistically significant for women and is present throughout much of adulthood. These findings may help explain previously reported lower earnings among female individuals with a history of mental illness,38 although research that examines the possible mediation of mental illness and lost earnings by early labour force exit is needed. In addition, our findings highlight the importance of considering anxiety and depressive disorders—even subthreshold—in policies supporting labour force participation.

    What is already known on this subject

    • Individuals with a history of mental illness have lower earnings than individuals without. A possible contributing reason is that those with prior anxiety or depression may be more likely to exit the labour force before retirement age, but few studies have examined the link between previous anxiety/depression and subsequent labour force exit. Those that have are limited by their measurement of depression and generalisability.

    What this study adds

    • Women with prior anxiety or depression have a statistically significant 37% increased risk of early labour force drop-out as compared to women without prior anxiety or depression, controlling for potential confounders. Men with prior anxiety or depression have an 18% increased risk of labour force drop-out, but this increase is not statistically significant. These findings may help explain previously reported lower earnings among female individuals with a history of mental illness and highlight the importance of considering anxiety and depressive disorders in policies supporting labour force participation.

    References

    1. ↵
      1. Kessler RC,
      2. Aguilar-Gaxiola S,
      3. Alonso J, et al
      . The global burden of mental disorders: an update from the WHO World Mental Health (WMH) surveys. Epidemiol Psichiatr Soc 2009;18:23–33. doi:10.1017/S1121189X00001421
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Kouzis AC,
      2. Eaton WW
      . Emotional disability days: prevalence and predictors. Am J Public Health 1994;84:1304–7. doi:10.2105/AJPH.84.8.1304
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      1. Judd LL,
      2. Paulus MP,
      3. Wells KB, et al
      . Socioeconomic burden of subsyndromal depressive symptoms and major depression in a sample of the general population. Am J Psychiatry 1996;153:1411–17. doi:10.1176/ajp.153.11.1411
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Stewart WF,
      2. Ricci JA,
      3. Chee E, et al
      . Cost of lost productive worktime among US workers with depression. JAMA 2003;289:3135–44. doi:10.1001/jama.289.23.3135
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Merikangas KR,
      2. Ames M,
      3. Cui L, et al
      . The impact of comorbidity of mental and physical conditions on role disability in the US adult household population. Arch Gen Psychiatry 2007;64:1180–8. doi:10.1001/archpsyc.64.10.1180
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Kessler RC,
      2. Akiskal HS,
      3. Ames M, et al
      . Prevalence and effects of mood disorders on work performance in a nationally representative sample of US workers. Am J Psychiatry 2006;163:1561–8. doi:10.1176/appi.ajp.163.9.1561
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      1. Kessler RC,
      2. Heeringa S,
      3. Lakoma MD, et al
      . Individual and societal effects of mental disorders on earnings in the United States: results from the national comorbidity survey replication. Am J Psychiatry 2008;165:703–11. doi:10.1176/appi.ajp.2008.08010126
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      1. Ettner SL,
      2. Frank RG,
      3. Kessler RC
      . The impact of psychiatric disorders on labor market outcomes. Ind Lab Relations Rev 1997;51:64–81. doi:10.1177/001979399705100105
      OpenUrlCrossRef
    9. ↵
      1. Schofield DJ,
      2. Shrestha RN,
      3. Percival R, et al
      . The personal and national costs of mental health conditions: impacts on income, taxes, government support payments due to lost labour force participation. BMC Psychiatry 2011;11:72. doi:10.1186/1471-244X-11-72
      OpenUrlCrossRefPubMed
    10. ↵
      1. Doshi JA,
      2. Cen L,
      3. Polsky D
      . Depression and retirement in late middle-aged us workers. Health Serv Res 2008;43:693–713. doi:10.1111/j.1475-6773.2007.00782.x
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Emptage NP,
      2. Sturm R,
      3. Robinson RL
      . Depression and comorbid pain as predictors of disability, employment, insurance status, and health care costs. Psychiatr Serv 2005;56:468–74. doi:10.1176/appi.ps.56.4.468
      OpenUrlCrossRefPubMed
    12. ↵
      1. Karpansalo M,
      2. Kauhanen J,
      3. Lakka T, et al
      . Depression and early retirement: prospective population based study in middle aged men. J Epidemiol Community Health 2005;59:70–4. doi:10.1136/jech.2003.010702
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Tian H,
      2. Sturm R
      . Datapoints: depression and leaving employment among older adult Americans. Psychiatr Serv 2004;55:10. doi:10.1176/appi.ps.55.1.10
      OpenUrlCrossRefPubMed
    14. ↵
      1. Anthony JC,
      2. Folstein M,
      3. Romanoski AJ, et al
      . Comparison of the lay diagnostic interview schedule and a standardized psychiatric diagnosis: experience in eastern Baltimore. Arch Gen Psychiatry 1985;42:667–75. doi:10.1001/archpsyc.1985.01790300029004
      OpenUrlCrossRefPubMedWeb of Science
    15. ↵
      1. Robins LN,
      2. Helzer JE,
      3. Croughan J, et al
      . National Institute of Mental Health Diagnostic Interview Schedule: its history, characteristics, and validity. Arch Gen Psych 1981;38:381–9. doi:10.1001/archpsyc.1981.01780290015001
      OpenUrlCrossRefPubMedWeb of Science
    16. ↵
      1. Eaton WW,
      2. Kessler LG
      . Epidemiologic field methods in psychiatry: the NIMH Epidemiologic Catchment Area Program. Orlando, FL: Academic Press, 2012.
    17. ↵
      1. Eaton WW,
      2. Muntaner C,
      3. Bovasso G, et al
      . Socioeconomic status and depressive syndrome: the role of inter- and intra-generational mobility, government assistance, and work environment. J Health Soc Behav 2001;42:277–94. doi:10.2307/3090215
      OpenUrlCrossRefPubMedWeb of Science
    18. ↵
      1. Lyketsos CG,
      2. Nestadt G,
      3. Cwi J, et al
      . The Life Chart Interview: a standardized method to describe the course of psychopathology. Int J Methods Psychiatr Res 1994;4:143–55.
      OpenUrlWeb of Science
    19. ↵
      1. Freedman D,
      2. Thornton A,
      3. Camburn D, et al
      . The life history calendar: a technique for collecting retrospective data. Sociol Methodol 1988;18:37–68. doi:10.2307/271044
      OpenUrlCrossRefPubMed
    20. ↵
      1. Nam CB,
      2. Powers MG
      . The socioeconomic approach to status measurement. Houston, TX: Cap and Gown Press, 1983.
    21. ↵
      1. Checkley W,
      2. Brower RG,
      3. Munoz A
      . Inference for mutually exclusive competing events through a mixture of generalized gamma distributions. Epidemiology 2010;21:557–65. doi:10.1097/EDE.0b013e3181e090ed
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      1. Badawi MA,
      2. Eaton WW,
      3. Myllyluoma J, et al
      . Psychopathology and attrition in the Baltimore ECA 15-year follow-up 1981–1996. Soc Psychiatry Psychiatr Epidemiol 1999;34:91–8. doi:10.1007/s001270050117
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      1. Kendler KS,
      2. Prescott CA,
      3. Myers J, et al
      . The structure of genetic and environmental risk factors for common psychiatric and substance use disorders in men and women. Arch Gen Psychiatry 2003;60:929–37. doi:10.1001/archpsyc.60.9.929
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      1. Krueger RF,
      2. McGue M,
      3. Iacono WG
      . The higher-order structure of common DSM mental disorders: internalization, externalization, and their connections to personality. Pers Ind Dif 2001;30:1245–59. doi:10.1016/S0191-8869(00)00106-9
      OpenUrlCrossRef
    25. ↵
      1. Trickett PK,
      2. McBride-Chang C
      . The developmental impact of different forms of child abuse and neglect. Dev Rev 1995;15:311–37. doi:10.1006/drev.1995.1012
      OpenUrlCrossRefWeb of Science
    26. ↵
      1. Cotter DA,
      2. Hermsen JM,
      3. Vanneman R
      . Gendered opportunities for work effects on employment in later life. Res Aging 2002;24:600–29. doi:10.1177/016402702237184
      OpenUrlAbstract/FREE Full Text
    27. ↵
      1. Horwath E,
      2. Cohen RS,
      3. Weissman MM
      . Epidemiology of depressive and anxiety disorders. In: M. T. Tsuang, M. Tohen (Eds.), Textbook in psychiatric epidemiology. New York, NY: Wiley-Liss, Inc, 2002:389–426.
    28. ↵
      1. Gutierrez-Lobos K,
      2. Wolfl G,
      3. Scherer M, et al
      . The gender gap in depression reconsidered: the influence of marital and employment status on the female/male ratio of treated incidence rates. Soc Psychiatry Psychiatr Epidemiol 2000;35:202–10. doi:10.1007/s001270050229
      OpenUrlCrossRefPubMedWeb of Science
    29. ↵
      1. Elinson L,
      2. Houck P,
      3. Marcus SC, et al
      . Depression and the ability to work. Psychiatr Serv 2004;55:29–34. doi:10.1176/appi.ps.55.1.29
      OpenUrlCrossRefPubMed
    30. ↵
      1. Dooley D,
      2. Catalano R,
      3. Wilson G
      . Depression and unemployment: panel findings from the Epidemiologic Catchment Area Study. Am J Community Psychol 1994;22:745–65. doi:10.1007/BF02521557
      OpenUrlCrossRefPubMedWeb of Science
    31. ↵
      1. Sturm R,
      2. Gresenz CR,
      3. Pacula RL, et al
      . Datapoints: labor force participation by persons with mental illness. Psychiatr Serv 1999;50:1407. doi:10.1176/ps.50.11.1407
      OpenUrlCrossRefPubMedWeb of Science
    32. ↵
      1. Waghorn G,
      2. Chant D,
      3. White P, et al
      . Disability, employment and work performance among people with ICD-10 anxiety disorders. Aust N Z J Psychiatry 2005;39:55–66. doi:10.1111/j.1440-1614.2005.01510.x
      OpenUrlAbstract/FREE Full Text
    33. ↵
      1. Marcotte DE,
      2. Redmon DP,
      3. Wilcox-Gok V
      . Prevalence and patterns of major depressive disorder in the United States labor force. J Ment Health Policy Econ 1999;2:123–31. doi:10.1002/(SICI)1099-176X(199909)2:3<123::AID-MHP55>3.0.CO;2-8
      OpenUrlCrossRefPubMed
    34. ↵
      1. Dooley D,
      2. Prause J,
      3. Ham-Rowbottom KA
      . Underemployment and depression: longitudinal relationships. J Health Soc Behav 2000;41:421–36. doi:10.2307/2676295
      OpenUrlCrossRefPubMedWeb of Science
    35. ↵
      1. Mossakowski KN
      . The influence of past unemployment duration on symptoms of depression among young women and men in the United States. Am J Public Health 2009;99:1826–32. doi:10.2105/AJPH.2008.152561
      OpenUrlCrossRefPubMedWeb of Science
    36. ↵
      American Psychological Association. Diagnostic and statistical manual of mental disorders,(DSM-5R). Arlington, VA: American Psychiatric Publishing, 2013.
    37. ↵
      1. Schoenbaum M,
      2. Unutzer J,
      3. McCaffrey D, et al
      . The effects of primary care depression treatment on patients’ clinical status and employment. Health Serv Res 2002;37:1145–58. doi:10.1111/1475-6773.01086
      OpenUrlCrossRefPubMedWeb of Science
    38. ↵
      1. Berndt ER,
      2. Koran LM,
      3. Finkelstein SN, et al
      . Lost human capital from early-onset chronic depression. Am J Psychiatry 2000;157:940–7. doi:10.1176/appi.ajp.157.6.940
      OpenUrlCrossRefPubMedWeb of Science

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • Contributors KER conceptualised and designed the study, performed the statistical analysis, interpreted results and drafted the manuscript. WWE was responsible for data acquisition and contributed to the study design, interpretation of the data and analysis, and critically revised the manuscript.

    • Funding KER was supported by the Robert Wood Johnson Foundation Health & Society Scholars Programme. WWE was supported by NIDA grant DA026652. The programme/funding agency had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.

    • Competing interests None declared.

    • Ethics approval Johns Hopkins Bloomberg School of Public Health Institutional Review Board.

    • Provenance and peer review Not commissioned; externally peer reviewed.