2013 Volume 55 Issue 6 Pages 479-486
Objectives: While bedtime may influence circadian rhythms, potentially leading to depression, epidemiological data on this issue are limited. We cross-section-ally investigated the association between bedtime and depressive symptoms in Japanese workers, taking sleep duration into consideration. Methods: The participants were 1,197 workers who participated in a health survey during a periodic checkup and had no history of psychiatric disease. Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression (CES-D) scale. Usual bedtime and wake time were inquired about using a self-administered questionnaire, and sleep duration was calculated based on the difference between these two values. Multiple logistic regression was used to estimate the odds ratio of depressive symptoms for bedtime or sleep duration categories. Results: Short sleep duration (<6 hours) was significantly associated with an increased prevalence of depressive symptoms. Late bedtime was also significantly associated with an increased prevalence of depressive symptoms (CES-D score of £19); the multi-variable-adjusted odds ratio of depressive symptoms for a bedtime of 1:00 or later versus 23:00 to 23:59 was 1.90 (95% confidence interval, 1.16–3.12). After additional adjustment for sleep duration, however, the association was largely attenuated (odds ratio, 1.17; 95% confidence interval, 0.66–2.06). Conclusions: Late bedtime was associated with increased prevalence of depressive symptoms, but this association could be largely accounted for by short sleep duration. Avoiding a late bedtime and obtaining a sufficient sleep duration may prevent depressive mood among workers.
(J Occup Health 2013; 55: 479–486)
Depression is one of the most common diseases1, 2) and is a highly prevalent psychiatric disorder with a lifetime risk close to 20%3). Depression lowers quality of life, reduces work productivity and increases risk of suicide4). In Japan, the number of patients with depression has been increasing in recent years5); according to a nationwide survey on worker health in Japan, 58% of workers report feeling stressed by work6), and many Japanese companies have cited issues with maintaining workers) mental health7). Depression has been linked to sleep-related factors, including sleep duration, sleep quality and chronotype. An excessively long or short sleep duration has been shown to be associated with an increased prevalence or incidence of depressive symptoms8–19). Subjects with insomnia, defined as difficulty in initiating or maintaining sleep or experiencing non-restorative sleep, are known to have an increased risk of depression compared with those without insomnia20, 21).
Regular patterns of social behaviors—including bedtime and wake time, social interactions and meal time—have been suggested to affect circadian rhythms22), and any disruption of these rhythms could cause depressive symptoms23, 24). However, relatively few studies have investigated the association between bedtime and depressive symptoms. In studies assessing chronotype using Morningness-Eveningness
Questionnaire25–29), evening-type subjects, who tend to go to bed late, had a higher prevalence of depressive symptom than morning-type subjects. Given that bedtime is strongly correlated with sleep duration among workers30) and that short sleep duration has been shown to be associated with increased prevalence or incidence of depressive symptoms8, 9–13, 15–19), however, the observed association between chronotype and depressive symptoms may be attributable to sleep duration and not bedtime itself. In fact, a Japanese study showed that a relatively late bedtime (after 0:00) was not associated with an increased prevalence of depressive symptoms after adjustment for sleep duration28), denying the possibility of an independent role of bedtime in the pathogenesis of depression. However, given that the study also included family members of workers (who might have had different daily schedules compared with the workers) and shift workers (who likely had different sleep patterns compared with daytime workers) and did not adjust for work-related factors, these previous findings should be interpreted with caution, and further investigation should be conducted.
Here, we cross-sectionally investigated the association of bedtime and sleep duration with depressive symptoms among daytime workers, with adjustment for work-related factors. In assessing the association between depressive symptoms and bedtime, we also considered sleep duration as a factor mediating the association between bedtime and depressive symptoms.
The subjects were participants in a health survey conducted in April 2011 among workers of a manufacturing company and affiliated companies in Chiba Prefecture, Japan. At the time of the periodic health examination, all full-time workers were asked to complete a survey questionnaire. Of the 1,723 subjects who received the health examination, 1,668 agreed to participate in the survey. Also obtained were data routinely collected in the health examination, including anthropometric measurements, biochemical data and information on medical history and smoking and drinking habits. Of the respondents to the survey, we excluded 62 subjects with a history of psychiatric disorders such as depression, psycho-neurosis and dysautonomia. After further exclusion of 13 subjects who had missing data for 5 or more items on the Center for Epidemiologic Studies Depression (CES-D) scale, 31 subjects with missing or implausible data on bedtime or wake time and 365 shift workers, a total of 1,197 daytime workers (1,029 men and 168 women) remained for analysis. The protocol of the study was approved by the ethics committee of the National Center for Global Health and Medicine.
Depressive symptomsDepressive symptoms were assessed using a Japanese version31) of the CES-D scale32), which was incorporated into the questionnaire. This scale consists of 20 questions addressing 6 symptoms of depression, including depressed mood, guilt or worth-lessness, helplessness or hopelessness, psychomotor retardation, loss of appetite and sleep disturbance experienced during the preceding week. Each question is scored on a scale of 0–3 according to the frequency of the symptom, and the total CES-D score ranges from 0 to 60. The criterion validity of the CES-D scale has been well established both in Western32) and Japanese31) subjects. To investigate the association with depression, we excluded 1 sleep question (“my sleep was restless”) from the 20 items and calculated the CES-D score. In addition, we adjusted the CES-D scores to correct them as a conventional scale of 0 to 60 using the following formula: CES-D score=(sum of scores for 19 items) x (20/19) x (19/ number of answered questions)9, 13). Depressive symptoms were defined as present when subjects had a CES-D score of ≥16. A CES-D cutoff of ≥19, which may be suitable for Japanese33), was also used.
Sleep measurementUsual bedtime and wake time were ascertained using a self-administered questionnaire. Sleep duration was calculated from bedtime and wake time. Participants were categorized into four categories according to bedtime and sleep duration: before 23:00, 23:00 to 23:59, 0:00 to 0:59 and 1:00 or later for bedtime; <6 hours, 6 to <7 hours, 7 to <8 hours and ≥8 hours for sleep duration. Subjective sleep sufficiency was assessed based on three potential responses (everyday, sometimes and almost never) to the question, “How often do you feel your sleep is sufficient when you get up in the morning?”
Other variablesLifestyle and work-related factors such as physical activity, employment type, job type, amount of overtime work and commuting time were asked in the survey questionnaire. For leisure-time physical activity, the frequency and amount of time per occasion were ascertained for each of three categories of exercise intensity (light, moderate and heavy activity). The total amount of time spent engaged in leisure-time physical activities was expressed in hours per week. For employment type, participants were asked to choose from three options (regular, temporary or part-time staff). Overtime work during the previous month and one-way commuting time to work were inquired about with 5 (not engaged, 0< to <10, 10 to <30, 30 to <50 or ≥50 hours/month) and 6 (<15, 15 to <30, 30 to <45, 45 to <60, 60 to <90 or ≥90 minutes) potential responses, respectively.
Statistical analysisData were expressed as means (standard deviation [SD]) and percentages for continuous and categorical variables. Differences in mean values and proportions of characteristics between subjects with and without depressive symptoms were assessed using an independent t-test (continuous variables) and the chi-squared test (categorical variables). Differences in mean values and proportions of the characteristics according to bedtime were assessed using analysis of variance (continuous variables) and the chi-squared test (categorical variables). Multiple logistic regression was performed to estimate the odds ratio (OR) and 95% confidence interval (CI) of depressive symptoms for each category of bedtime or sleep duration, taking 23:00 to <0:00 for bedtime or 6 to <7 hours for sleep duration as the reference. The first model was adjusted for age (years) and sex, and the second model was further adjusted for marital status (married or other), employment type (regular or other), job type (managerial and clerical or technical work), job position (low or middle and high), amount of overtime work (<10, 10 to <30 or ≥30 hours/month), one-way commuting time (<30, 30 to <60 or ≥60 minutes), alcohol consumption habit (nondrinker, occasional drinker, drinker consuming <23 g of ethanol/day or drinker consuming ≥23 g of ethanol/day), smoking status (lifetime nonsmoker, former smoker or current smoker), leisure-time physical activity (<120 or ≥120 minutes/week), history of serious diseases (including cancer, ischemic heart disease or cerebrovascular disease; yes or no) and history of common diseases (including hypertension, diabetes or dyslipidemia; yes or no). For analysis incorporating bedtime, sleep duration (<6, 6 to <7, 7 to <8 or ≥8 hours) was added to the second model (third model). In our sensitivity analysis, we adjusted for subjective sleep sufficiency, which has been shown to be associated with depres-sion9, 13). Trend association was assessed by assigning ordinal numbers 1–4 to categories of bedtime or sleep duration. Two-side p-values less than 0.05 were considered statistically significant. All analyses were performed using STATA version 11.2 (StataCorp, College Station, TX, USA).
ResultsThe characteristics of study subjects according to depressive symptoms are shown in Table 1. Of 1,197 subjects, 252 men (24.5%) and 57 women (33.9%) were identified as having depressive symptoms. Compared with subjects without depressive symptoms, those with depressive symptoms tended to be younger and were more likely to be female and unmarried, to be engaged in technical work and to have short sleep duration. The characteristics of study subjects according to bedtime are shown in Table 2. More than half (55.9%) went to bed before 0:00. Subjects with a late bedtime tended to be younger and were more likely to be unmarried and regular workers as well as overtime workers but less likely to be a daily drinker, to be engaged in technical work, to report a history of serious diseases and to have a history of hypertension, diabetes or dyslipidemia than those with an early bedtime. In addition, late bedtime was associated with a short sleep duration.
The ORs of depressive symptoms according to sleep duration are shown in Table 3. Short sleep duration (<6 hours) was significantly associated with an increased prevalence of depressive symptoms; the multivariable-adjusted OR for depressive symptoms of <6 hours versus 6 to <7 hours of sleep was 1.57 (95% CI: 1.08–2.29). Longer sleep duration (7 to <8 hours and ≥8 hours) was not associated with depressive symptoms; the corresponding values were 0.93 (95% CI: 0.67–1.30) for the 7 to <8 hours category and 0.96 (95% CI: 0.58–1.58) for ≥8 hours category. When a higher cutoff (CES-D score of ≥19) was used for the definition of depressive symptoms, the association between short sleep duration and depressive symptoms was strengthened. The multivariable-adjusted OR for depressive symptoms of <6 hours of sleep was 2.17 (95% CI: 1.42–3.32) compared with 6 to <7 hours. This result did not materially change after adjusting for subjective sleep sufficiency; the multivariable-adjusted OR for depressive symptoms (CES-D score of ≥19) of <6 hours was 1.90 (95% CI: 1.22–2.95).
The ORs of depressive symptoms according to bedtime categories are shown in Table 4. No association was noted between bedtime and depressive symptoms, defined as a CES-D score of ≥16. When a relatively high cutoff was used for the definition of depressive symptoms (CES-D score of ≥19), late bedtime was significantly associated with increased prevalence of depressive symptoms even after adjustment for potential confounders (except sleep duration); the multivariable-adjusted OR of depressive symptoms for bedtime of 1:00 or later versus 23:00 to 23:59 was 1.90 (95% CI: 1.16–3.12). However, this association was largely attenuated after additional adjustment for sleep duration (OR: 1.17; 95% CI: 0.66–2.06). Further adjustment for subjective sleep sufficiency did not appreciably alter the result (OR: 1.13; 95% CI: 0.63–2.04). For early bedtime (before 23:00), the multivariable-adjusted OR (except sleep duration) of depressive symptoms was 1.28 (95% CI: 0.81–2.04), which was increased after further adjustment for sleep duration (OR: 1.68; 95% CI: 0.99–2.85).
| Subjects without depressive symptoms | Subjects with depressive symptoms | p value2 | |
|---|---|---|---|
| No. of subjects | 888 | 309 | |
| Age (year, mean ± SD) | 45.0 ± 11.0 | 42.4 ± 9.9 | <0.001 |
| Sex (male, %) | 87.5 | 81.6 | 0.01 |
| Marital status (married, %) | 70.7 | 61.7 | 0.004 |
| Employment type (regular, %) | 86.0 | 85.8 | 0.94 |
| Job type (technical work, %) | 30.6 | 42.4 | <0.001 |
| Job position (low, %) | 72.7 | 78.1 | 0.07 |
| Overtime work (≥30 hours/month, %) | 24.1 | 26.8 | 0.34 |
| One-way commuting time (≥60 minutes, %) | 15.2 | 12.0 | 0.16 |
| Alcohol consumption (daily drinker, %) | 28.6 | 27.5 | 0.71 |
| Smoking status (current smoker, %) | 23.9 | 24.9 | 0.71 |
| Leisure-time physical activity (≥120 minutes/week, %) | 32.7 | 26.7 | 0.09 |
| History of serious diseases3 (yes, %) | 1.9 | 2.9 | 0.30 |
| History of common diseases4 (yes, %) | 17.7 | 16.5 | 0.64 |
| Subjective sleep sufficiency (seldom, %) | 15.0 | 32.0 | <0.001 |
| Sleep duration (%) | 0.04 | ||
| <6 hours | 14.5 | 21.4 | |
| 6 to <7 hours | 44.0 | 42.1 | |
| 7 to <8 hours | 31.5 | 27.5 | |
| ≥8 hours | 9.9 | 9.1 | |
| Bedtime (hour:minute, mean ± SD) | 23:29 ± 1:06 | 23:34 ± 1:08 | 0.21 |
| Bedtime (%) | 0.85 | ||
| Before 23:00 | 19.4 | 19.7 | |
| 23:00 to <0:00 | 36.6 | 35.9 | |
| 0:00 to <1:00 | 30.1 | 28.5 | |
| 1:00 or later | 14.0 | 15.9 |
In this cross-sectional study among Japanese workers, late bedtime (1:00 or later) was significantly associated with an increased prevalence of depressive symptoms (CES-D score of ≥19). However, this association was largely attenuated after adjustment for sleep duration, a factor associated with both bedtime and depressive symptoms. To our knowledge, this is one of the few studies addressing the association between bedtime and depressive symptoms among workers.
We observed increased prevalence of depressive symptoms associated with late bedtime (1:00 or later). Similarly, chronotype studies have shown that individuals with the extreme evening type, who tend to go to bed relatively late, had increased prevalence of depressive states25–29) compared with other types. These findings suggest that late bedtime is indeed associated with depressive mood. Given that sleep duration was shorter among participants with the extreme evening type than among participants with the extreme morning type (6.57 vs. 7.66 hours, respectively)28), however, the observed associations in chronotype studies might be attributable to short sleep duration. In the present study population, bedtime was closely associated with sleep duration, and adjustment for sleep duration largely attenuated the association between late bedtime and depressive symptoms. Likewise, in a Japanese study among 1,170 workers or members of their families aged 20−59 years28), late bedtime (after 0:00) was not significantly associated with depressive symptoms after adjustment for sleep duration. Taken together, bedtime may not be associated with depressive mood independently of sleep duration.
| Before 23:00 | 23:00 to 23:59 | 0:00 to 0:59 | 1:00 or later | p value1 | |
|---|---|---|---|---|---|
| No. of subjects | 233 | 436 | 355 | 173 | |
| Age (year, mean ± SD) | 50.9 ± 11.9 | 45.3 ± 10.3 | 41.7 ± 9.3 | 38.3 ± 8.1 | <0.001 |
| Sex (male, %) | 87.6 | 84.9 | 86.5 | 85.5 | 0.79 |
| Marital status (married, %) | 81.0 | 74.7 | 64.2 | 45.4 | <0.001 |
| Employment type (regular, %) | 73.2 | 84.5 | 91.4 | 94.8 | <0.001 |
| Job type (technical work, %) | 55.3 | 38.4 | 22.8 | 16.3 | <0.001 |
| Job position (low, %) | 79.2 | 72.9 | 70.5 | 78.1 | 0.07 |
| Overtime work (>30 hours/month, %) | 14.1 | 19.6 | 33.8 | 33.5 | <0.001 |
| One-way commuting time (>60 minutes, %) | 13.8 | 16.3 | 13.6 | 12.1 | 0.52 |
| Alcohol consumption (daily drinker, %) | 43.4 | 31.7 | 21.8 | 13.3 | <0.001 |
| Smoking status (current smoker, %) | 23.2 | 22.9 | 26.2 | 24.3 | 0.73 |
| Leisure-time physical activity (>120 minutes/week, %) | 26.7 | 33.1 | 29.4 | 35.2 | 0.33 |
| History of serious diseases2 (yes, %) | 5.2 | 1.8 | 1.1 | 1.2 | 0.006 |
| History of common diseases3 (yes, %) | 31.3 | 17.7 | 12.4 | 8.1 | <0.001 |
| Subjective sleep sufficiency (seldom, %) | 17.3 | 16.1 | 21.4 | 26.6 | 0.016 |
| Sleep duration (hour, mean ± SD) | 7.52 ± 0.75 | 6.65 ± 0.67 | 6.13 ± 0.69 | 5.62 ± 0.81 | <0.001 |
| <6 hours | 6 to <7 hours | 7 to <8 hours | >8 hours | p for trend1 | |
|---|---|---|---|---|---|
| No. of subjects | 195 | 521 | 365 | 116 | |
| CES-D (≥16) | |||||
| No. of cases | 66 | 130 | 85 | 28 | |
| Age- and sex-adjusted OR (95% CI) | 1.53 (1.06–2.19) | 1.00 (reference) | 0.98 (0.71–1.34) | 1.09 (0.67–1.75) | 0.12 |
| Multivariable-adjusted OR (95% CI)2 | 1.57 (1.08–2.29) | 1.00 (reference) | 0.93 (0.67–1.30) | 0.96 (0.58–1.58) | 0.045 |
| CES-D (≥19) | |||||
| No. of cases | 51 | 75 | 50 | 15 | |
| Age- and sex-adjusted OR (95% CI) | 2.10 (1.40–3.15) | 1.00 (reference) | 1.02 (0.69–1.51) | 1.02 (0.56–1.86) | 0.008 |
| Multivariable-adjusted OR (95% CI)2 | 2.17 (1.42–3.32) | 1.00 (reference) | 0.96 (0.64–1.44) | 0.87 (0.46–1.63) | 0.003 |
Abbreviations: CES-D, Center for Epidemiologic Studies Depression scale; CI, confidence interval; OR, odds ratio.
Despite the lack of suggestion of an independent role of bedtime in the pathogenesis of depressive symptoms, bedtime is important because it is a major determinant of sleep duration, especially for employees who are expected to show up for work on time in the morning. In fact, late bedtime was closely associated with short sleep duration in the present study population (Table 2). Therefore, we may reasonably assume that sleep duration is a mediator, rather than a confounder, linking late bedtime to depressive symptoms. If so, analysis without adjustment for sleep duration would provide a better estimate of the true association between late bedtime and depressive symptoms than analysis with such adjustment.
| Before 23:00 | 23:00 to 23:59 | 0:00 to 0:59 | 1:00 or later | p for trend1 | |
|---|---|---|---|---|---|
| No. of subjects | 233 | 436 | 355 | 173 | |
| CES-D (>16) | |||||
| No. of cases | 61 | 111 | 88 | 49 | |
| Age- and sex-adjusted OR (95% CI) | 1.19 (0.82–1.73) | 1.00 (reference) | 0.90 (0.65–1.25) | 1.00 (0.66–1.50) | 0.34 |
| Multivariable-adjusted OR (95% CI)2 | 1.05 (0.71–1.55) | 1.00 (reference) | 1.04 (0.73–1.46) | 1.19 (0.77–1.83) | 0.64 |
| Multivariable-adjusted OR (95% CI)3 | 1.20 (0.77–1.87) | 1.00 (reference) | 0.90 (0.62–1.30) | 0.85 (0.52–1.40) | 0.30 |
| CES-D (>19) | |||||
| No. of cases | 40 | 62 | 50 | 39 | |
| Age- and sex-adjusted OR (95% CI) | 1.46 (0.93–2.27) | 1.00 (reference) | 0.91 (0.61–1.37) | 1.49 (0.94–2.37) | 0.94 |
| Multivariable-adjusted OR (95% CI)2 | 1.28 (0.81–2.04) | 1.00 (reference) | 1.07 (0.70–1.63) | 1.90 (1.16–3.12) | 0.20 |
| Multivariable-adjusted OR (95% CI)3 | 1.68 (0.99–2.85) | 1.00 (reference) | 0.86 (0.54–1.35) | 1.17 (0.66–2.06) | 0.43 |
Abbreviations: CES-D, Center for Epidemiologic Studies Depression scale; CI, confidence interval; OR, odds ratio.
In the present study, several work-related factors— including employment type, job type and overtime work—were found to be associated with bedtime (Table 2). Previously, extensive overtime, long commuting and work overload were identified as associated with short sleep duration34). From an occupational health perspective, it would be advisable not only to encourage workers to go to bed earlier but also to identify factors (either occupational or personal ones) that could potentially cause short sleep duration and take measures to reduce their levels for the prevention of depressive mood among workers.
We found an increased prevalence of depressive symptoms among workers with a short sleep duration (<6 hours), and several previous studies have also reported that a short sleep duration (<5 hours11, 16–18), <6 hours9, 10, 12, 13, 15, 19) or <7 hours8) was associated with increased prevalence or incidence of depression. Long sleep duration (≥8 hours12) or ≥9 hours11, 14, 15, 18) has also been shown to be associated with a higher prevalence of depression. However, we observed no association between long sleep duration and depressive symptoms, likely due to the small number of subjects who slept ≥8 hours (9.7%) in our study population. In addition, we were unable to examine a relatively long sleep duration (≥9 or ≥10 hours) in relation to depressive symptoms.
The association between bedtime or sleep duration and depressive symptoms became more obvious when a higher cutoff (CES-D score of ≥19) was used compared with a conventional cutoff (CES-D score of ≥16). Although a cutoff for the CES-D of ≥16 is widely recommended, a cutoff of ≥19 is suggested as suitable for Japanese populations; according to a validation study among Japanese workers, the sensitivity and specificity were 92.7% and 91.8%, respectively33. High sensitivity and specificity for a cutoff of ≥19 may be one reason for the relatively strong association between bedtime or sleep duration and depressive symptoms observed in the present study.
Major strengths of this study include the relatively high participation rate and adjustment for known and suspected risk factors (including work-related ones) of depressive symptoms. However, several limitations to the present study warrant mention. First, an association derived from a cross-sectional study does not necessarily indicate causality. Participants with milder forms of depressive symptoms may also fall asleep later and for shorter durations than others due to their depressive symptoms, leading to a spurious association. Further, we did not obtain specific information on history of other psychiatry diseases such as anxiety disorders or use of medication such as antidepressive, hypnotic or anti-mania drugs. However, we excluded subjects with a history of psychiatric disorders such as depression and psychoneurosis or
dysautonomia to minimize the possibility of reverse causality. Second, we did not collect information on sleep latency. Because subjects with depressive symptoms tend to have longer sleep latency35, 36), they might actually fall asleep at later time points after going to bed than those without symptoms, even if their bedtimes are the same. Accurate sleep latency can be measured using a polysomnograph or electroencepha-logram37), but use of such equipment is not feasible in a large sample survey. Third, although stress is associated with depressive symptoms38), we did not collect information concerning troubles in relationships with others, work-related burdens or worries outside of work. Finally, the present findings may not be generalized to a population with a different background from that examined here. The study subjects were daytime workers of a large-scale manufacturing company in Japan. The mandatory working time of managerial and clerical workers is from 8:45 to 17:30 and that of technical workers is from 7:45 to 16:30. Some employees (26.7%) worked a varied schedule and were largely involved in managerial and clerical work (94.0%). We adjusted for job type to minimize the effect of work schedule.
In conclusion, the present study found that late bedtime was associated with increased prevalence of depressive symptoms and that the association was largely attenuated after adjustment for sleep duration. Given the close correlation between late bedtime and short sleep duration in this working population, attention should be paid to occupational and nonoc-cupational factors associated with bedtime to obtain sufficient sleep, which could prevent depressive symptoms. Prospective studies are required to confirm the findings from this cross-sectional study.
Acknowledgments: The present study was partially supported by a research fund of the Industrial Health Foundation. The authors thank Kae Saito and Kayoko Washizuka (National Center for Global Health and Medicine) for data processing.
