Skip Navigation LinksHome > August 2013 - Volume 55 - Issue 8 > Relationship Between Long Working Hours and Depression: A 3-...
Journal of Occupational & Environmental Medicine:
doi: 10.1097/JOM.0b013e31829b27fa
Original Articles

Relationship Between Long Working Hours and Depression: A 3-Year Longitudinal Study of Clerical Workers

Amagasa, Takashi MD, MPH; Nakayama, Takeo MD, PhD

Free Access
Continued Medical Education
Article Outline
Collapse Box

Author Information

From the Department of Health Informatics (Drs Amagasa and Nakayama), Kyoto University School of Public Health, Sakyo-ku, Kyoto, Japan; and Department of Psychiatry (Dr Amagasa), Yoyogi Hospital, Sendagaya, Shibuya-ku, Tokyo, Japan.

Address correspondence to: Takashi Amagasa, MD, MPH, Department of Health Informatics, Kyoto University School of Public Health, Konoe-cho, Yoshida, Sakyo-ku, Kyoto 606-8501, Japan (amagasa_t@tokyo-kinikai.com).

Authors Amagasa and Nakayama have no relationships/conditions/circumstances that present potential conflict of interest.

The JOEM editorial board and planners have no financial interest related to this research.

Collapse Box

Abstract

Objective: To clarify how long working hours affect the likelihood of current and future depression.

Methods: Using data from four repeated measurements collected from 218 clerical workers, four models associating work-related factors to the depressive mood scale were established. The final model was constructed after comparing and testing the goodness-of-fit index using structural equation modeling. Multiple logistic regression analysis was also performed.

Results: The final model showed the best fit (normed fit index = 0.908; goodness-of-fit index = 0.936; root-mean-square error of approximation = 0.018). Its standardized total effect indicated that long working hours affected depression at the time of evaluation and 1 to 3 years later. The odds ratio for depression risk was 14.7 in employees who were not long-hours overworked according to the initial survey but who were long-hours overworked according to the second survey.

Conclusions: Long working hours increase current and future risks of depression.

Back to Top | Article Outline
Learning Objectives

* Review previous knowledge on work-related factors affecting depression risk.

* Summarize the new findings on the association between long working hours and depression in clerical workers, both at initial evaluation and follow-up.

* Discuss the implications for potential steps to reduce the risk of major depression in employees.

Depressive state/major depressive disorder is a major occupational health issue in developed countries.1,2 On the basis of Karasek's job demand-control model,3 numerous studies have evaluated the relationship between high job demand and/or low job control and various states of mental and physical health.4–11 A growing body of evidence indicates that high job demand and low job control increase the risk of depressive state/major depressive disorder.6–12 Many studies have assessed the relationship between long working hours and various states of mental and physical health, and several have suggested a negative effect.13–18 Nevertheless, the results have not been consistent, and few studies have evaluated the long-term effects.13,15,19,20

Recent studies of British public servants observed study subjects for 5.3 and 5.8 years to clarify the risks of new depressive symptoms and new major depressive episodes, respectively. Both studies found that those with long working hours (more than 11 hours per day) showed an increased risk of new depressive symptoms/major depressive episodes compared with those who worked 7 to 8 hours per day.21,22 Nevertheless, our understanding of the long-term effects of long working hours is still limited because no one has exposed the factors involved in this process or documented the presence/absence of depressive symptoms at each time point during follow-up.15 Moreover, although we have been able to reveal that job demand is an intermediate variable for long working hours and depression,23 it has not been treated as such by other related studies, or by those in which the risk of being long-hours overworked (LHO) (defined as long working hours in addition to high job demand; hereafter, LHO), rather than long working hours, has been examined with regard to its effect on depression.23 In addition, working hours have not been considered a factor subject to time-dependent variation.15

The purpose of this prospective study was to clarify how long working hours affect future depressive states and investigate how individuals who are LHO exhibit an increased risk of future depression. This was done through repeated measurement of the variables at all four time points, spanning the 3 years of follow-up.

Back to Top | Article Outline

MATERIALS AND METHODS

Research Design and Study Populations
Research Design

A longitudinal study was conducted using unlinkable anonymous data collected from self-administered questionnaires. We obtained approval from the office administrator regarding the use of data for research purposes.

Back to Top | Article Outline
Research Participants and Data Anonymization

The original survey was administered at a business with branches located throughout a certain prefecture in the Tokyo metropolitan area. Participants included a total of 225 clerical full-time day-shift workers. The first author (T.A.) was consulted by the safety and health committee of the main office regarding measures to improve the mental health of employees. A baseline self-administered survey on the mental health status of employees was conducted in December 1999. The general affairs department distributed the questionnaire to branch managers, who instructed the employees to fill out and submit their responses. After filling out the questionnaire, the employee personally sealed it in an envelope, which was collected by a supervisor. The branch manager gathered and submitted the questionnaires to the survey office of the general affairs department in the main office. The survey office entered and double-checked the data, which were then anonymized. Digitized files were sent to the first author (T.A.), who statistically analyzed and interpreted the data and produced a report. The original aggregated questionnaires were destroyed. Data were obtained from 218 individuals (response rate 96.9%), excluding those on medical or administrative leave, after a survey period of 1 week. The instructions on the employee questionnaire clearly stated that “research results would be reported such that individuals/businesses cannot be identified and consent to participate will be assumed upon filling out and submitting the questionnaire.” Approval was also obtained from the health committee of the business. The first author retained the anonymized and digitized files. Details of the survey are provided in a separate report.24 Survey procedures were identical for three follow-up surveys (times 2 to 4), which were conducted at 1-year intervals after the 1999 baseline survey (time 1). In this study, the data obtained from all four surveys were analyzed.

Back to Top | Article Outline
Model Measurements
Number of Hours Worked

At Business A, the number of hours worked per week was measured on a scale previously used in a large cohort study of Japanese businesses.25 Subjects were asked: “Approximately how many hours on average do you work in a week? If you worked on holidays/weekends or overtime, please include those hours regardless of whether or not you were paid.” Participants could select one of eight options, ranging from “less than 40 hours” (option 1) to “100 hours or more” (option 8) in 10-hour intervals.

Back to Top | Article Outline
Job Demand and Job Control

At Business A, job demand, job control, and support were measured using a Japanese translation of a measurement scale (Job Content Questionnaire) on the basis of the extended Karasek model,6–12 which has been described previously.25,26 The reliability values (Cronbach α) for job demand (eg, “I feel that I have too much work”; 1 = “usually,” 2 = “sometimes,” and 3 = “never”) and job control (eg, “Can you determine the content of your work and methods used to complete tasks?”; 1 = “never,” 2 = “not usually,” 3 = “to some degree,” and 4 = “always”) were 0.88 and 0.51, respectively.

Back to Top | Article Outline
Depression

The General Well-being Scale was used as an index of depression.27 The reliability, validity, and acceptability of the Japanese version of this 18-item depressive mood scale were established. Depressive symptoms were measured by a subscale (total score of 0 to 30) (eg, “Have you been affected by nervousness or ‘nerves' this month?”; 1 = “severely” and 6 = “not at all”). The reliability (Cronbach α) of this scale was 0.84.

Back to Top | Article Outline
Statistical Analysis
Structural Equation Modeling (SEM)

Spearman's rank correlation coefficient between two variables was obtained for the target variables investigated. On the basis of previous results23 and the study by de Lange et al,11 covariance structure analysis28,29 was performed on several competing models regarding working hours, job demand, job control, and depressive mood. Using SEM, the goodness of fit of multiple models can be compared for two or more goodness-of-fit indices by determining the factor structure in several variables at once.11,28,29 In this study, analyses were only performed for data without missing values, and the following goodness-of-fit indices and criteria were established: goodness-of-fit index more than 0.90, root-mean-square error of approximation less than 0.05, and normed fit index more than 0.90. After obtaining the model with the best fit, SEM was repeated. The path with the largest significance probability was removed on the basis of the results of a test between two variables in the model at one of the four time points (at baseline and 1, 2, and 3 years later). A final model was constructed that satisfied the criteria for the goodness-of-fit indices, and the test results for the path coefficients were all statistically significant. From the standardized total effect of the final model, the effects of long working hours on depression during the evaluation period and potential for future depression were analyzed.

Back to Top | Article Outline
Multiple Logistic Regression Analysis

Multiple logistic regression analysis was performed to obtain associations between exposure factors and depression as odds ratios (OR) with two cross-sectional models and three longitudinal models. From the hours worked per week and job demand measured in each of the four surveys, data were binarized into LHO (score of 1), in which the number of hours worked per week was at least 60 and the job demand was considered high, and “other” (score of 0). We chose 60 hours per week as the cutoff point for the following reason: In 2005, the Industrial Safety and Health Act was amended in Japan. Since this amendment, when the amount of work beyond the standard weekly 40 hours exceeds 80 hours per month (which corresponds to 60 hours per week), and fatigue is observed or there is a health concern as a result, the business is required to provide face-to-face guidance by a physician upon the employee's request.30 The following three variables were binarized as 1 or 0: job demand (cutoff = the average score), job control (cutoff = the average score), and depression (cutoff = the upper 5th percentile for depressive mood). This cutoff for depression was not based on clinical diagnosis but rather on the estimated prevalence of major depressive episodes at the Japanese workplace (approximately 3% to 5%31–33). In addition, we performed a sensitivity analysis by setting the upper 3rd and 10th percentiles as the cutoff points for depressive mood scoring. Covariates were sex, age, and job control, although the explanatory and outcome variables were LHO status and depression, respectively. Cases with an endpoint determination of de novo “depression” during the target period and individuals who scored 1 for depression in 1999 (time 1) were excluded. The significance level of α for the tests was set as 0.05. All data analyses were performed using Microsoft Excel 2007, Amos Ver17.0, and SPSS Statistics 17.0.34

Back to Top | Article Outline
Competing Structural Models

On the basis of the previous study,11 tests were conducted on the following four causation models: within-wave, normal, reversed, and reciprocal. Regarding the number of hours worked per week, job demand, job control, and depressive mood, paths were drawn to the same factors 1, 2, and 3 years after baseline for all models. The models are described as follows:

1. Within-wave model. Paths were drawn from the number of hours worked per week to job demand, from job demand to depressive mood, and from job control to depressive mood for each of the four surveys.

2. Normal causation model. In addition to the within-wave model, paths were drawn from job demand at time 1 to depression at times 2 to 4, from job demand at time 2 to depression at times 3 to 4, and job demand at time 3 to depressive mood at time 4.

3. Reversed causation model. In addition to the within-wave model, paths were further drawn from depressive mood at times 1 to 3 to job demand and job control at times 2 to 4.

4. Reciprocal model. Paths in both the normal causation and reversed causation models were drawn.

Back to Top | Article Outline
Testing Logistic Models
Cross-Sectional (C) Models

C1. The relationship between LHO status and depression was determined for the 210 participants using the variables measured at time 1.

C2. This analysis was then performed for the 155 participants for whom data were available for SEM.

Back to Top | Article Outline
Longitudinal (L) Models

L1. Analyses were performed using LHO status at times 1 to 3 as the predictor variable and de novo depression at times 2 to 4 as the outcome variable.

L2. Given the effects of change in LHO status (a time-dependent exposure variable) across the follow-up period,15 we performed analyses, using LHO status at time 1 and times 2 to 4 as the predictor variable, and de novo depression at times 2 to 4 as the outcome variable.

L3. Analyses were performed using LHO status at times 1 to 3 as the predictor variable and de novo depression as the outcome variable, and LHO status at times 1 to 4 as the predictor variables and de novo depression as the outcome variable.

Back to Top | Article Outline
Ethical Considerations

This study was approved by the Ethics Committee at the Kyoto University Graduate School and Faculty of Medicine (No. E-673).

Back to Top | Article Outline

RESULTS

Baseline Characteristics

Baseline characteristics of the participants (N = 218 [171 men and 46 women], mean age 42.3 years [standard deviation 10.2]) were previously reported and are summarized as follows. Regarding the number of years of continuous employment, 23.5%, 50%, and 16.2% of participants had 5 years or less, at least 5 years and less than 20 years, and at least 20 years, respectively. The one-way commute was less than 1 hour for 84.7% of participants. The proportion of married participants was 71.3%. Regarding the actual number of days off per month, 48.1% responded having 4 days or less. At least 90% of participants responded that they worked overtime at least 14 days a month. For the number of hours worked per week, the modal response was “at least 60 hours but less than 80 hours” (101 participants; 47.2%), followed by “at least 40 hours but less than 60 hours” (88 participants; 41.1%), and a total of 19 participants (18.4%) indicated that they worked “at least 80 hours” The average number of hours of sleep per night in the past month was 6.2 ± 0.8 hours, and 135 participants (63.4%) reported satisfactory sleep quality. There were 88 habitual drinkers (40.7%) who consumed alcoholic beverages almost every day and 99 habitual smokers (46.7%). A total of 185 participants (86.4%) did not exercise regularly, and 125 participants (57.9%) reported that they were in good health.

Back to Top | Article Outline
Longitudinal Data

The total number of workers at this business was 225. There were 218 respondents in the baseline survey conducted in 1999 (time 1; 96.9% response rate). After excluding some respondents because of missing values, data from 210 respondents were available for analyses. In the survey conducted in 2000 (time 2), 220 employees responded (97.8% response rate), including those who were newly hired (N = 225). Of the participants with complete surveys at time 1, which constituted the parent population for the next year (N = 210), 15 dropped out of the study (N = 195). Thus, complete data at time 2 originated from 194 participants. The longitudinal data from 155 participants who continued follow-up until 2002 (time 4) were analyzed in this study (Fig. 1). The mean age of the 155 participants (127 men and 28 women) at baseline was 41.5 years (standard deviation 10.1 years). Individuals who dropped out of the study between time 1 and time 2 were women with high job demand (P = 0.002, P = 0.060, respectively), although there were no differences in the other variables analyzed (eg, days off, the number of hours at work, the number of days with overtime, job control, and depressive mood). Those who dropped out between time 2 and time 3 were more depressed and had lower job control (P = 0.017, P = 0.029, respectively). In analyzing the nonresponses from time 1 to time 4, the baseline number of employees who worked long hours was high (P = 0.004). Nevertheless, there were no differences in terms of sex, age, days off, days worked overtime, job demand, job control, or depressive mood.

Figure 1
Figure 1
Image Tools
Back to Top | Article Outline
Correlation Analysis

The results of correlation analysis are shown in Table 1. With the exception of three pairs, the expected directions are indicated for correlations between variables. A negative correlation between baseline age and depressive mood was observed at all four time points (times 1 to 4; P < 0.05). For women, positive correlations were observed between baseline age and depressive mood at times 1 and 3 (P < 0.01) and time 3 (P < 0.05). Correlations between the values of a variable at two measurement time points incorporated into the model ranged from 0.343 for the number of hours worked per week to 0.626 for job demand (P < 0.01) at times 1 and 2. At times 2 and 3, correlation values ranged from 0.279 for the number of hours worked per week to 0.603 for job demand (P < 0.01). At times 3 and 4, the values ranged from 0.465 for the number of hours worked per week to 0.635 for depressive mood (P < 0.01). The paths drawn from each of these four variables (hours worked per week to depressive mood) at times 1 to 3 to the same variable at times 2 to 4, times 3 to 4, and time 4, respectively, were incorporated into all the competing models.

Table 1
Table 1
Image Tools
Back to Top | Article Outline
SEM Regarding Number of Hours Worked per Week and Depression

The goodness-of-fit indices for each model, including chi-squared value, degrees of freedom [df], normed fit index, goodness-of-fit index, and root-mean-square error of approximation, were 98.284 [80], 0.895, 0.927, and 0.039, respectively, for Model 1; 87.708 [74], 0.913, 0.939, and 0.026 for Model 2; 83.683 [74], 0.911, 0.938, and 0.039 for Model 3; and 66.888 [64], 0.929, 0.950, and 0.023 for Model 4. The best fit was obtained with Model 4, the reciprocal causation model (Fig. 2). For Model 4, the estimates and significance probabilities of the path coefficients between two variables separated by 0-, 1-, 2-, or 3-year intervals or with reverse causality are shown in Table 2. Nearly all paths between two variables at 0- or 1-year intervals were statistically significant at an α criterion of 0.05. Nevertheless, at 2- or 3-year intervals or with reverse causality, almost none of the paths between two variables were significant (values in bold). Therefore, the “only within model,” which was suggested by our previous results23 and only consists of simultaneous data, was set as the “basic skeletal model” (bold solid borders in Fig. 2), and the final model was constructed in incremental steps. This was determined by successively removing the paths with the largest significance probabilities in Table 2 and conducting repeated SEM analyses. Consequently, a model that satisfied the established criteria and only consisted of statistically significant path coefficient estimates was obtained. This was determined as the final model (Fig. 3).

Figure 2
Figure 2
Image Tools
Figure 3
Figure 3
Image Tools
TABLE 2-a. Standardi...
TABLE 2-a. Standardi...
Image Tools

The standardized total effect of the number of hours worked per week on depressive state in the final model shows that hours worked per week increased the depressive state 0 to 3 years after baseline (Table 3).

TABLE 2-b. Standardi...
TABLE 2-b. Standardi...
Image Tools
Table 3
Table 3
Image Tools
Back to Top | Article Outline
Multiple Logistic Regression Analysis

For the 210 participants at time 1 (Fig. 1), the analysis (C1 model) revealed an OR of 3.01 (Table 4). For the 155 participants included in the C2 model, the OR was 0.60. Nevertheless, none of these ORs were considered statistically significant. The analysis that used LHO status at time 1 as the predictor and depression at times 2 to 4 as the outcome variable (L1 model) revealed that the ORs relative to the occurrence of de novo depression were 0.36 and 1.11, and not estimated at times 2, 3, and 4, respectively. The same result was observed for the 155 participants who completed the follow-up (data not shown), in that none of the ORs were statistically significant. On the basis of the final model (Fig. 3), analyses were performed using LHO status at time 1 and times 2 to 4 as the predictors and de novo depression at times 2 to 4 as the outcome variable (L2 model). An omnibus test of the model coefficients between time 1 and time 2 (N = 185) was statistically significant. The OR of LHO status at time 1 to de novo depression at time 2 was 0.11 and was estimated to be 14.7 at time 2. The results were not statistically significant at time 3 and not estimated at time 4 because of a sparse cell problem. In addition, analysis results between three or more points (L3 model) were also found to be not statistically significant or not estimated (Table 4). Finally, the OR of female–male to depression was 10.3 (1.85 to 57.7; P = 0.008), and the OR of low job control was 5.63 (1.18 to 26.9; P = 0.03).

Table 4
Table 4
Image Tools
Back to Top | Article Outline
Sensitivity Analysis

When the cutoff was set to the upper 3rd or 10th percentiles for depressive mood scoring, the results were very similar to those found for the main analyses, even though the statistical significance differed somewhat.

Back to Top | Article Outline

DISCUSSION

The present longitudinal analyses of all data, which involved four repeated measurements over 3 years after the baseline, showed that long working hours increased the risk of depression at the time as well as 1 to 3 years later. The SEM results also supported the hypothesis that when “job demand” is considered an intermediate variable rather than a confounding factor, a positive relationship is revealed between long working hours and the risk of depression.23 The multiple logistic analyses showed that moving from a status of not LHO to one of LHO increased the risk of depression 1 year later and that change in the opposite direction decreased this risk.

The Whitehall II study of public servants21,22 has revealed that long working hours increase the future risk of depressive symptoms or major depressive episodes. The SEM findings of this study suggest that long working hours represent a risk factor for future depression, even for employees who work in the private sector.

Interestingly, the logistic analyses in this study showed that those who were LHO at baseline did not increase the risk of new depressive symptoms (time 2 for the C2 and L1 models in Table 4). This may be due to differences in the characteristics or the overworked state of the target population. Compared with the public servants targeted in the Whitehall II study, the office workers at small to medium private businesses in this study worked far more hours. The longitudinal analysis (ie, longitudinal data in the Results) revealed that many of the dropouts had experienced adverse effects of being LHO. Therefore, we surmise that the target populations for both the logistic analyses for which at least 1 year was set as the outcome and the SEM included those who could continue to work without dropping out even when forced to overwork long hours. People who can work even in such an overworked state are also at a low risk of depression (time 2 for L1 model). Nevertheless, although not statistically significant, the point estimate of the OR for the risk of depression 2 years (rather than 1 year) after baseline increased and exceeded 1 (L1 model in Table 4). In a population for which data at all four time points were available (N = 155), the point estimates of the OR 1 and 2 years after baseline were successively greater than the OR point estimate in the baseline cross-sectional analysis (C2 model in Table 4; data not shown). This suggests that the risk of developing depression in those who are LHO may increase with time. If so, it is understandable that the risk of depression is detected approximately 5 years later, as noted by previous studies.21,22 Furthermore, our results suggest that a change in status from not LHO to LHO increases the risk of depression, but the change in the opposite direction reduces it. The reason for this is clear from the results at time 2 for the L2 model in Table 4, which project the point estimate of the OR for depression as 1 × 14.7 in the case who was not LHO at time 1 but was LHO at time 2, and 0.11 × 1 in the converse case.

In a previous study that examined the relationship between long working hours and depressive symptoms,21 job demand and working hours were not simultaneously incorporated into the analysis. This was the case in many previous studies that elucidated the relationship between work stress factors and depression. In contrast, in a study that examined the relationship between long working hours and major depressive episodes,22 job strain, which is defined as high demand and low job control, was incorporated into the analysis as a predictor. Unlike the observations from previous studies,6–12,35 however, no positive association between job strain and major depressive disorder could be detected. One possible reason for this is that because job demand and job strain were treated as confounding factors, overadjustment occurred and they could not be supplemented as risk factors.20,23 As our SEM results suggest (Fig. 3), job demand should probably be treated as an intermediate factor.

Few studies have investigated the relationship between changes in work stress factors and depression by repeatedly measuring work stress factors as predictive factors for depression. Stansfeld et al36 have repeatedly measured job strain and incorporated changes in this variable to analyze future risk for developing major depressive disorder. During the follow-up period of this study, the portion of the population that maintained low job strain served as the reference group; changes from high to low or from low to high, as well as maintenance of high job strain, in this order, all showed increasingly higher risks of depression. From this study, one can anticipate that a change from excessive overwork to nonexcessive overwork will reduce the risk of depression. Nevertheless, our results after 2 years did not show a statistically significant change (L2 model, N = 177). The effects of changing one's status of long working hours/LHO may only produce a short-lived reduction in risk of depression (1 year, according to our results).

The men–women ratio recently reported for employees in Japan is 6:4.37 Approximately 30% of employees worked 50 hours or more a week in 2000,38 although the proportion of women at the business in this study was small and most of them worked long hours. Given that sex was controlled in this study, the relationship between LHO and depression is unaffected by sex. Incidentally, a sex-dependent difference in the relationship between working hours and health, especially with regard to the possibility of greater adverse effects for women, has been noted repeatedly,13,14,15,18,21,22 and our logistic analysis results were consistent with this. Nevertheless, because long working hours were binarized at 60 hours per week, it is possible that many workers could be classified as working long hours compared with national averages, even if they were classified in the nonoverworked group. Consequently, the magnitude of the effects of long working hours may have been underestimated.

In Japan, the Industrial Safety and Health Act was revised in 2005. After 2008, employers who received a request from an employee who worked more than 40 hours per week and 100 hours per month (corresponding to approximately 65 hours per week) with observable fatigue were required to provide in-person guidance from a physician.38 Nevertheless, the results of this study suggest the necessity of stricter regulations against both long working hours and job demand for maintaining the mental health of workers.

Although this study is longitudinal and observational, causality could be assessed if confounding variables are removed when changes in the predictors correlate with subsequent changes in the outcome variable.39 Intervention studies could determine the causal relationship between long working hours and depression, except that acknowledgment of long working hours as an interventional factor is not ethical. A more practical approach is to carefully interpret the findings of observational studies to assist in social decision-making.

There are some limitations to this study. Although this study population was a typical working population in Japan, it primarily comprised a homogeneous sample of full-time regular clerical workers. Thus, caution should be exerted when extrapolating our results to other working populations. Because of the limited number of participants, careful interpretation is required, particularly for the cases with wide confidence intervals for ORs. In addition, “long working hours” in this study were self-reported and not based on objective data. Nevertheless, even if a time card system is used, long working hours may be underreported because accurate reporting of long working hours is a sensitive issue from a legal standpoint for both the employee and the employer. If a workplace is excessively focused on work, employees may overreport their hours to show how devoted they are to their job. This investigation was performed only after fully informing respondents that their responses would not be disclosed to the business owner and that the results would be used to contribute to improving mental health measures for employees. Consequently, we expected to receive responses with a low level of bias in a single direction (over- or underestimation). Furthermore, although the sensitivity analysis compensated for the shortcomings of our results, this study assessed depression as a proxy of clinical depression.

The practical implications for this study are to develop more effective mental health measures to reduce depression by focusing on LHO status and controlling factors in the work environment. With regard to implications for research, more multiple longitudinal studies that clarify the long-term risk of long working hours on depression using all variables at all measurement points should be performed in occupational fields with various work patterns and at different scales.15 Careful determination of the definition of long working hours and measurement methods will be essential. Other issues should also be considered, such as the unit used to represent long working hours (eg, days, weeks, months, or years), the cutoff to determine long and short working hours (eg, debate on whether anything other than normal working hours is considered overtime), overtime without pay or nonstandard working arrangements, and the use of self-reported surveys and objective measurement methods, including behavioral records and time cards or both.17 In this study, overwork was defined as job demand above the national average value for the target population, although this definition may not be applicable for every occupation. The Job Content Questionnaire is used in numerous countries,40 and the reliability and validity of the Japanese version have already been established.41 If standardized questionnaires are used, interpretation of results may become more generalizable.

Back to Top | Article Outline

CONCLUSIONS

In this longitudinal observational study, we have shown that working hours can affect the long-term risk of depression and that being LHO is a risk factor for future de novo depression. By targeting LHO, especially changes in LHO status, mental health measures that effectively reduce the occurrence of major depressive disorder will become possible by controlling factors in the occupational environment.

Back to Top | Article Outline

REFERENCES

1. World Health Organization and International Labour Organization. Mental Health and Work: Impact, Issues and Good Practices. Geneva, Switzerland: World Health Organization and International Labour Organization; 2000:56.

2. Amagasa T, Nakayama T, Takahashi Y. Karojisatsu in Japan: characteristics of 22 cases of work-related suicide. J Occup Health. 2005;47:157–164.

3. Karasek RA. Job demands, job decision latitude, and mental strain: implications for job redesign. Admin Sci Quart. 1979;24:285–308.

4. Kivimäki M, Leino-Arjas P, Luukkonen R, Riihimäki H, Vahtera J, Kirjonen J. Work stress and risk of cardiovascular mortality: prospective cohort study of industrial employees. BMJ. 2002;19;325:857–860.

5. Kivimäki M, Virtanen M, Elovainio M, Kouvonen A, Väänänen A, Vahtera J. Work stress in the etiology of coronary heart disease—a meta-analysis. Scand J Work Environ Health. 2006;32:431–442.

6. Michie S, Williams S. Reducing work related psychological ill health and sickness absence: a systematic literature review. Occup Environ Med. 2003;60:3–9.

7. Stansfeld SA, Fuhrer R, Shipley MJ, Marmot MG. Work characteristics predict psychiatric disorder: prospective results from the Whitehall II Study. Occup Environ Med. 1999;56:302–307.

8. Niedhammer I, Goldberg M, Leclerc A, Bugel I, David S. Psychosocial factors at work and subsequent depressive symptoms in the Gazel cohort. Scand J Work Environ Health. 1998;24:197–205.

9. Mausner-Dorsch H, Eaton WW. Psychosocial work environment and depression: epidemiologic assessment of the demand-control model. Am J Public Health. 2000;90:1765–1770.

10. de Lange AH, Taris TW, Kompier MA, Houtman IL, Bongers PM. “The very best of the millennium”: longitudinal research and the demand-control-(support) model. J Occup Health Psychol. 2003;8:282–305.

11. de Lange AH, Taris TW, Kompier MA, Houtman IL, Bongers PM. The relationships between work characteristics and mental health: examining normal, reserved, and reciprocal relationships in a 4-wave study. Work Stress. 2004;18:149–166.

12. Kawakami N, Haratani T. Epidemiology of job stress and health in Japan: review of current evidence and future direction. Ind Health. 1999;37:174–186.

13. Spurgeon A, Harrington MJ, Cooper LC. Health and safety problems associated with long work hours: a review of the current position. Occup Environ Med. 1997;54:367–375.

14. Sparks K, Cooper C. The effects of hours of work on health: a meta-analytic review. J Occup Organ Psychol. 1997;70:391–408.

15. van der Hulst M. Long work hours and health. Scand J Work Environ Health. 2003;29:171–188.

16. van der Hulst M, Geurts S. Associations between overtime and psychological health in high and low reward jobs. Work Stress. 2001;15:227–240.

17. International Labour Office. Working Time Its Impact on Safety and Health. Geneva, Switzerland: International Labour Organization; 2003:42–44.

18. Shields M. Long work hours and health. Health Rep. 1999;11:33–48.

19. Fujino Y, Horie S, Hoshuyama T, Tsutsui T, Tanaka Y. A systematic review of working hours and mental health burden. Sangyo Eiseigaku Zasshi. 2006;48:87–97.

20. Hobson J, Beach JR. An investigation of the relationship between psychological health and workload among managers. Occup Med (Lond). 2000;50:518–522.

21. Virtanen M, Ferrie JE, Singh-Manoux A, et al. Long working hours and symptoms of anxiety and depression: a 5-year follow-up of the Whitehall II study. Psychol Med. 2011;18:1–10.

22. Virtanen M, Stansfeld SA, Fuhrer R, Ferrie JE, Kivimäki M. Overtime work as a predictor of major depressive episode: a 5-year follow-up of the Whitehall II study. PLoS One. 2012;7:e30719. doi: 10.1371/journal.pone.0030719.

23. Amagasa T, Nakayama T. Relationship between long working hours and depression in two working populations: a structural equation model approach. J Occup Environ Med. 2012;54:868–874.

24. Nakayama T, Amagasa T. Special reference to employee knowledge about depression and suicide: baseline results of a workplace-based mental health support program. Psychiat Clin Neurosci. 2004;58:280–284.

25. Stress Disease Worker's Compensation Insurance Study Group. Stress and Health General Survey Research Group. Stress and Health in Japanese Industrial Workers General Survey Report. Tokyo, Japan: Stress Disease Worker's Compensation Insurance Study Group; 1991.

26. Sugisawa A, Uehata T, Sekiya E, et al. Mental health and its related factors among middle-aged male workers: an 18-month follow-up study. Jpn J Ind Health. 1994;36:91–101.

27. Nakayama T, Toyoda H, Ohno K, Yoshiike N, Futagami T. Validity, reliability and acceptability of the Japanese version of the General Well-Being Schedule (GWBS). Qual Life Res. 2000;9:529–539.

28. Toyoda H ed. Covariance Structure Analysis [Amos Ed.] Structural Equation Modeling. Tokyo, Japan: Tokyotosho; 2007.

29. Toyoda H. Covariance Structure Analysis [Introduction] Structural Equation Modeling. Tokyo, Japan: Asakura Publishing; 1998.

30. Ministry of Health, Labour, and Welfare. Amended Industrial Safety and Health Act. Enacted April 1, 2006 (in Japanese). Available at: http://law.e-gov.go.jp/htmldata/S47/S47HO057.html. Published 2006. Accessed May 2011.

31. Shima S, Haratani T. Epidemiology of mental disorders in the workplace. Japanese Society of Occupational Mental Health Volume. Occupational Mental Health Handbook 1. Tokyo, Japan: Nakayama Shoten; 1998:310–323.

32. Nakao K, Iwase M. Relationship between anxiety and depression. Osaka, Japan: The Mental Health Okamoto Memorial Foundation Research Support Report. 1999;11:107–111.

33. Takeuchi T, Nakao M, Yano E. Relationship between smoking and major depression in a Japanese workplace. J Occup Health. 2004;46:489–492.

34. IBM SPSS. Statistics 2288. Available at: http://www-06.ibm.com/software/jp/analytics/spss/. Accessed January 2013.

35. Bonde JP. Psychosocial factors at work and risk of depression: a systematic review of the epidemiological evidence. Occup Environ Med. 2008;65:438–445.

36. Stansfeld SA, Shipley MJ, Head J, Fuhrer R. Repeated job strain and the risk of depression: longitudinal analyses from the Whitehall II study. Am J Public Health. 2012;102:2360–2366.

37. Ministry of Internal Affairs and Communication, Statistics Bureau Home Page. World statistics. Chapter 12. Labour and Wages. Ministry of Internal Affairs and Communication Statistics Bureau. Available at: http://www.stat.go.jp/data/sekai/12.htm#h12-06. Accessed May 2011.

38. Lee S. Working-hour gaps: trends and issues. In: Messenger JC, ed. Working Time and Workers' Preference in Industrialized Countries: Finding the Balance. London and New York: Routledge Taylor & Francis Group; 2004:29–59.

39. Ichige Y. “Statistically Correct” Facts You Shouldn't Believe. Tokyo, Japan: Gentosharunessansu; 2012.

40. Karasek R, Brisson C, Kawakami N, Houtman I, Bongers P, Amick B. The Job Content Questionnaire (JCQ): an instrument for internationally comparative assessments of psychosocial job characteristics. J Occup Health Psychol. 1998;3:322–355.

41. Kawakami N, Fujigaki Y. Reliability and validity of the Japanese version of Job Content Questionnaire: replication and extension in computer company employees. Ind Health. 1996;34:295–306.

Copyright © 2013 by the American College of Occupational and Environmental Medicine

Login

Article Tools

Images

Share